Add files using upload-large-folder tool

008b5c9aca9316214d43b070fec9de15a85a502f4d4f74861052175cf679e7cb/preprint/images_list.json

@@ -0,0 +1,58 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Resistance responses of leaf rust resistance gene Lr47. a Infection types in the introgression line KernLr47 and its recurrent parent Kern in response to Pt race THDB at 2, 4, 6, and 8 dpi. Scale bars represent 0.5 cm. b Pt infection areas visualized by WGA-FITC staining. Leaves were collected at 2, 4, 6, and 8 dpi, cleared with KOH, and stained with WGA-FITC. Scale bars represent 100 \u03bcm. c Interaction graphs showing the average size of individual fungal infection areas of Kern Lr47 (blue) and Kern (orange) estimated by fluorescence microscopy at 2, 4, 6, and 8 dpi (n = 40). ***, P < 0.001. Error bars are standard errors of the mean. d Infection types of Lr47 NILs and their recurrent parents in response to selected Pt races FHJL, PHQS, FHJR, THDB, PHRT, PHTT, THTT, HCJR, and FHHM. +, resistant Lr47 allele present; -, no resistant Lr47 allele. R, Resistant; S, Susceptible.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Fine mapping of the leaf rust resistance gene Lr47. a Schematic representation of chromosome 7A in Kern Lr47 carrying the introgressed Ae. speltoides segment 7S#1S (indicated by a black rectangle). b Fine mapping of Lr47 using 7A/7S homoeologous recombinants induced by the ph1b mutant. The genome-specific markers (Supplementary Table 4) distributed along the introgressed 7S#1S chromosome segment and their physical locations based on the Chinese Spring reference genome Refseq v1.1. c Recombinant haplotypes (L1\u2013L7) with Ae. speltoides segments of different lengths determined by marker analysis. White rectangles represent chromosome 7A; gray rectangles indicate heterozygous for 7S/7A; Sus., susceptible; Res., resistant. d Genetic map of Lr47 based on 1,141 F2 plants from cross m118 \u00d7 Kern Lr47. e Physical map of Lr47 in the sequenced reference genome of Ae. speltoides TS01. Yellow arrows represent NLR genes.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Identification of Lr47 using the EMS mutagenesis and transcript assembly (EMTA) method. a Susceptible EMS mutants used to isolate theLr47 candidate gene. Infection types for Kern Lr47 (positive control), ten independent EMS mutants, and Kern (negative control) inoculated with Pt race THDB. R, resistant; S, susceptible. b Schematic representation of the EMTA method. RNA-seq reads from the resistant parent (wild type) are de novo assembled. A search of the transcriptome database of the resistant parent using the sequences of genes within the mapping interval in sequenced reference genomes yields transcripts with high similarity. Next, RNA-seq readsfrom susceptible EMS mutants are mapped to the obtained transcripts. Transcript 1 in this example, which has a preponderance of single-nucleotide variants (red lines) across the susceptible mutants, is considered a good candidate. c Gene structure of the Lr47candidate gene. The positions of the EMS-derived loss-of-function mutations are indicated by blue arrows. Gray boxes indicate untranslated regions, blue boxes represent coding exons, and dotted lines represent introns.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Functional validation of Lr47 by BSMV-mediated gene editing and transgenic complementation. a Sequencing chromatogram showing the induced polymorphisms between WT and the selected BSMV-Cas9-induced editing mutants. The mutated nucleotides are underlined and the CCT PAM sequence is highlighted in red. b Reactions to Pt race THDB. 1) Kern Lr47; 2-3) mutant line mut-1; 4-5) mutant line mut-2; and 6) Cas9-transgenic Bobwhite. c The 7,234-bp genomic DNA fragment including Lr47 that was used for transformation. The black arrow indicates the Lr47 gene (from ATG to TGA). d Seedling infection types of Kern Lr47, two transgenic families T1CS401-1 and T1CS401-2, and Fielder control in response to Pt races PHQS and FHJL. Copy number of transgenes was estimated using TaqMan assays (Supplementary Table 9). R, resistant; S, susceptible.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Reduction in the length of the introgressed Ae. speltoides chromosome segment carrying Lr47. a Infection types of homozygous BC1F2 plants from the introgression line YM21+Lr47-2 (+) and the plants lacking Lr47 (-). Six Pt races (THDB, PHQS, FHJL, HCJR, PHRT, and PHTT) were used for evaluation. R, Resistant; S, Susceptible. b, c, d GISH images of wheat lines Kern Lr47, L2 (BC2F2), and L8 (BC1F2). The magnified images show the Ae. speltoides chromosome segments introgressed into wheat chromosome 7A (Ae. speltoides chromatin is painted in green and marked with red arrowheads). e, f, g FISH images from wheat lines Kern Lr47, L2 (BC2F2), and L8 (BC1F2). Probes pSc119.2 (green), pTa535 (red), and pTa713 (yellow) were used. Yellow arrows indicate the wheat-Ae. speltoides translocated 7A chromosomes, and magnified images of these chromosomes are shown in the insets. Scale bar = 10 \u03bcm.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Statistical analysis of agronomic and quality traits in YM21+Lr47-1 and its sister control line. a-p Plants grown in a controlled walk-in growth chamber at 24 \u2103 day/22 \u2103 night with a 16 h light/8 h dark photoperiod were phenotyped for the following traits: a plant height (PH); b tillers number (TN); c spike length (SL); d spikelet number per spike (SNS); e grain number per spike (GNS); f grain yield per plant (GYP); g thousand-seed weight (TSW); h grain length (GL); i grain width (GW); j shearing force (SF) in Newtons (N); k grain moisture content (GMC); l grain protein content (GPC); m flour water absorption (FWA); n flour ash (FA); o flour yield (FY); and p grain hardness (GH). q-u Plants grown in the field. q Visual phenotypes of YM21+Lr47-1 and its sister line in the field. r Infection types of YM21+Lr47-1 and the sister control line. s Close-up views of spikes. t Spike length (SL). uGrain number per spike (GNS). Error bars are standard errors of the means. ns = not significant (P >0.05), * = P < 0.05. Lr47, resistant Lr47 allele present (YM21+Lr47-1); lr47, sister control line lacking the alien chromosome segment.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_7.png",
+        "caption": "Characterization of Lr47. a Transcript levels of Lr47 in mock-inoculated and Pt-inoculated plants. Leaves were collected at five time points: 0 h, 1 dpi, 2 dpi, 4 dpi, and 6 dpi. Transcript levels were expressed as fold-Actin (n = 7). Error bars are standard errors of the mean. ns = not significant. b \u00a0Schematic diagram of conserved domains in the Lr47 protein. c Subcellular localization of the GFP- Lr47 fusion protein in tobacco leaves. Scale bars represent 50 or 100 \u03bcm. BF = bright field; GFP = green fluorescent protein. d GFP-horseradish peroxidase Western blot showing proteins expressed at expected sizes for all constructs. Target bands are highlighted by blue arrowheads. ev = empty vector. e Macroscopic cell death in N. benthamiana leaves 48 hpi with Agrobacterium tumefaciens carrying Lr47 constructs. GFP, negative control; BAX, positive control.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

008b5c9aca9316214d43b070fec9de15a85a502f4d4f74861052175cf679e7cb/preprint/preprint.md

@@ -0,0 +1,240 @@
+# Abstract
+
+Leaf rust, caused by *Puccinia triticina* Eriksson (*Pt*), is one of the most severe foliar diseases of wheat. Breeding for leaf rust resistance is a practical and sustainable method to control this devastating disease. Here, we report the identification of *Lr47*, a broad-spectrum leaf rust resistance gene introgressed into wheat from *Aegilops speltoides*. The *Lr47* gene encodes a coiled-coil nucleotide-binding leucine-rich repeat protein that is both necessary and sufficient to confer *Pt* resistance, as demonstrated by loss-of-function mutations and transgenic complementation. New *Lr47* introgression lines with no or reduced linkage drag were generated using the *ph1b* mutation, and a diagnostic molecular marker for *Lr47* was developed. The CC domain of the Lr47 protein was not able to induce cell death, nor did it have self-protein interaction. The cloning of *Lr47* expands the number of *Pt*-resistance genes that can be incorporated into multigene transgenic cassettes to control this devastating disease.
+
+Biological sciences/Genetics/Plant breeding  
+Biological sciences/Plant sciences/Plant immunity/Virulence  
+wheat  
+leaf rust  
+resistance genes  
+CC-NBS-LRR  
+*Aegilops speltoides*  
+introgression
+
+# Introduction
+
+Wheat is one of the leading food crops providing one-fifth of the total food calories and proteins consumed by humans. Reducing yield losses inflicted by fungal pathogens is an effective measure to increase wheat production. *Puccinia triticina* Eriksson (*Pt*), the causal agent of wheat leaf rust (or brown rust), is one of the most devastating fungal pathogens in wheat. This disease occurs in most wheat growing regions<sup>1,2</sup> and can cause significant yield losses in susceptible wheat varieties under favorable climatic conditions<sup>3</sup>. According to recent studies, global annual wheat yield losses due to this pathogen could be approximately 25 million tons<sup>4,5</sup>. Breeding for leaf rust resistance is considered the most practical and sustainable way to control this devastating disease.
+
+To date, over 80 leaf rust resistance (*Lr*) genes have been officially designated in wheat and its wild relatives<sup>6</sup>. However, due to the size and complexity of the wheat genome, only ten *Lr* genes have been cloned to date either by classical map-based cloning (*Lr1*, *Lr10*, *Lr21*, *Lr34*, *Lr42*, and *Lr67*)<sup>7-12</sup> or by rapid gene-cloning approaches including MutRenSeq (*Lr13*), TACCA (*Lr22a*), MutChromSeq (*Lr14a*), and MutIsoSeq (*Lr9*)<sup>13-16</sup>. Among the cloned *Lr* genes, *Lr34* and *Lr67* are known as slow rusting genes encoding a putative ATP-binding cassette transporter and a hexose transporter, respectively<sup>10,11</sup>. *Lr14a* encodes a protein containing 12 ankyrin repeats, and *Lr9* encodes a protein with an N-terminal tandem kinase domain followed by vWA/Vwaint domains at its C-terminus<sup>15,16</sup>. The other six isolated genes have been shown to encode typical coiled-coil nucleotide-binding leucine-rich repeat (NLR) proteins<sup>7-9,12-14</sup>.
+
+Cloning additional *Lr* genes is desirable to diversify the combinations of *Pt* resistance genes used in transgenic cassettes or gene pyramiding to achieve durable resistance. The diploid wheat species *Aegilops speltoides* Tausch (2n = 2x = 14, SS), the closest extant donor of the B-genome to bread wheat<sup>17,18</sup>, harbors valuable *Lr* genes, such as *Lr28*, *Lr35*, *Lr36*, *Lr47*, *Lr51*, and *Lr66<sup>18-20</sup>*. These genes have been successfully transferred from *Ae. speltoides* to common wheat, but none of them has been cloned so far due to the limited recombination observed between wheat and *Ae. speltoides* chromosomes.
+
+The resistance gene *Lr47* was transferred from chromosome 7S#1 of *Ae. speltoides* to the short arm of chromosome 7A of hexaploid wheat translocation line T7AS-7S#1S-7AS·7AL using homoeologous recombination in the presence of the *Pairing homeologous1* mutation (*ph1b*)<sup>18</sup>, but the ancestral origin of this *Ae. speltoides* segment is not clear<sup>21,22</sup>. Using C-banding and restriction fragment length polymorphism (RFLP) markers, it was determined that the genetic length of the translocated segment 7S#1S was approximately 20–30 centimorgans (cM) long and that it was located 2–10 cM from the centromere<sup>18</sup>. In a recent study, the length of the *Ae. speltoides* segment 7S#1S was estimated to be between 157 and 174 Mb based on a set of simple sequence repeat (SSR) markers<sup>23</sup>. This interstitial translocation segment carrying *Lr47* was subsequently backcrossed into several spring wheat cultivars, such as Pavon, Express, Kern, RSI5, Yecora Rojo, and UC1041<sup>22,24</sup>. However, the presence of the *Lr47* introgression was found to be associated with several detrimental effects on agronomic and quality traits, including reduced grain yield, decreased flour yield, and increased flour ash<sup>22</sup>. Therefore, new rounds of homoeologous recombination using the *ph1b* mutation are needed to reduce the size of the *Ae. speltoides* chromosome segment containing *Lr47* and minimize linkage drag. The linkage between *Lr47* and these negative effects can also be broken by cloning the gene and generating transgenic plants. This is a worthwhile endeavor because *Lr47* is one of a few genes known to confer strong levels of resistance against a wide range of *Pt* isolates<sup>1,18,23,25-30</sup>, and may play an important role in the improvement of wheat resistance to leaf rust.
+
+Here, we report the identification of the *Lr47* gene through a combination of map-based cloning, ethyl methanesulfonate (EMS) mutagenesis and transcript assembly (EMTA) approaches. We also validate the function of the cloned candidate gene using independent susceptible EMS mutants, barley stripe mosaic virus (BSMV)-mediated gene editing<sup>31</sup>, and stable transgenic complementation. To minimize potential linkage drag, we generated smaller *Ae. speltoides* chromosome segments carrying *Lr47* using the *ph1b* mutation. Finally, we developed a diagnostic molecular marker to accelerate the deployment of *Lr47* in wheat breeding programs.
+
+# Results
+
+Lr47 provides broadly effective resistance to diverse leaf rust pathotypes. The bread wheat line Kern Lr47 (PI 603918*7/Kern) is near isogenic to wheat cultivar cv. Kern and carries the introgressed Ae. speltoides chromosome segment 7S#1S. In the seedling tests, all Kern Lr47 plants showed strong resistance to Pt race THDB, whereas all Kern plants were susceptible (Fig. 1a). In a subset of 128 F₂ plants from the cross Kern Lr47 × ZZ5389 evaluated with race THDB, 93 plants were resistant and 35 were susceptible, which fits a 3:1 segregation ratio expected for a single dominant genetic locus (χ² = 0.38, P = 0.54). To quantify the growth of Pt pathogen at 2, 4, 6, and 8 dpi in Kern Lr47 and its recurrent parent Kern, we measured the average infection areas using fluorescence microscopy and the fluorescent dye WGA-FITC, which stains the Pt pathogen. At all four time points, the average infection areas observed microscopically were significantly smaller (P < 0.001) in plants with Lr47 than in those without the gene (Fig. 1b, c). Fluorescent images of Pt growth in Kern showed a diffuse network of fungal hyphae at the edges of the infected sites, indicating that Pt appeared to spread unimpeded. In contrast, pathogen growth in Kern Lr47 was restricted and no significant differences were observed in the fungal infection areas from 4 to 8 dpi (Fig. 1c).
+
+Seedlings of Lr47 NILs (Express Lr47, Yecora Rojo Lr47, UC1041 Lr47, and RSI5 Lr47) and their recurrent parents (Express, Yecora Rojo, UC1041, and RSI5) were challenged with the other 23 Pt pathotypes collected in China (Supplementary Table 3). All Lr47 NILs exhibited strong levels of resistance (ITs = 0;) to the tested races (Fig. 1d, Supplementary Table 3). In contrast, the recurrent parents showed a wide range of responses to different Pt races, probably due to other Lr resistance genes present in these genetic backgrounds. Nevertheless, inoculation of these recurrent parents revealed that at least one accession was fully susceptible to each Pt race (Fig. 1d), confirming that Lr47 confers resistance to these Pt pathotypes.
+
+Characterization of the Ae. speltoides segment introgressed into common wheat. To determine the physical location and size of the Ae. speltoides chromosome segment introgressed into hexaploid wheat, we compared the SNPs identified in the RNA-seq of Lr47 NILs (Kern Lr47, Yecora Rojo Lr47, and UC1041 Lr47) with those from another 10 sequenced hexaploid wheat varieties and three Ae. speltoides accessions (AE915, AE1590, and PI 554292). We focused on the SNPs that were present in the three Ae. speltoides accessions, but absent in the 10 sequenced hexaploid wheat varieties, and that are hereafter referred to as Ae. speltoides-specific SNPs. Based on this rule, we identified 3,169 Ae. speltoides-specific SNPs (Supplementary Table 5) that were shared with Lr47 NILs in the proximal region of chromosome arm 7AS starting from 40.4 Mb to 190.5 Mb (CS RefSeq v1.1 coordinates, Supplementary Fig. 1). The translocation breakpoints ranged from 40,050,116 bp to 40,398,277 bp at the terminal end, and from 190,476,279 bp to 191,833,539 bp at the proximal end (Supplementary Table 5). Thus, the Ae. speltoides segment in the three Lr47 NILs ranges from 150.1 Mb to 151.8 Mb.
+
+To better define the translocation breakpoints, we designed 16 7A/7S genome-specific primers across the ~150 Mb introgressed Ae. speltoides segment (Supplementary Table 4), and used them to genotype the Lr47 NILs and their recurrent parents. We found that translocation breakpoints in Lr47 NILs were located between markers pku0738 and pku0745 at the terminal end, and between markers pku2216 and pku2233 at the proximal end (Supplementary Fig. 2).
+
+Fine mapping of Lr47 using 7A/7S homoeologous recombinants induced by the ph1b mutant. No recombination was expected between wheat chromosome 7A and the introgressed 7S chromosome segment in the presence of the Ph1 gene. Genotyping of 128 F₂ plants from the cross Kern Lr47 × ZZ5389 using the flanking markers pku0745 and pku2216 (Supplementary Table 4) revealed no recombination between these two markers, confirming that recombination between Ae. speltoides and T. aestivum chromosomes is suppressed. To induce recombination between the 7A/7S homoeologous chromosomes, the introgression line Kern Lr47 was crossed with CS ph1b. Supplementary Fig. 3 describes the procedure for identifying recombinants between the 7S and 7A chromosomes. A total of 15 F₂ plants heterozygous for Lr47 and homozygous for phlb were obtained and self-pollinated to generate F₃ seeds for the recombination screening. We genotyped 2,654 plants from eight selected segregating F₃ families with the flanking markers pku0745 and pku2216, and identified 542 plants with recombination events.
+
+The recombinants were genotyped using another 13 7A/7S-genome specific markers across the introgressed Ae. speltoides chromosome segment (Fig. 2a, b). Among the recombinants, we only selected plants heterozygous for a smaller Ae. speltoides chromosome segment for phenotyping. Finally, we identified Ae. speltoides segments of seven different lengths, which were designated as L1 to L7 (Fig. 2c). The selected recombinants were evaluated for resistance to race THDB at the three-leaf stage in growth chambers. Plants carrying introgressions L1 and L4 were susceptible, so Lr47 was mapped to a 3.5-Mb region (CS RefSeq v1.1 coordinates) flanked by markers pku1104 and pku1152 (Fig. 2c). The candidate gene region includes 49 annotated high-confidence genes in Chinese Spring (TraesCS7A01G110400 – TraesCS7A01G115200, Supplementary Table 6). The functional annotation of these genes revealed six typical NLR genes (Supplementary Table 6), a gene class frequently associated with disease resistance in plants³²–³⁷.
+
+To refine the mapping of Lr47 in the Kern Lr47 × CS ph1b population, we crossed the susceptible EMS mutant m118 (in the Kern Lr47 background) with the non-mutagenized resistant Kern Lr47, and generated an F₂ population consisting of 1,141 individuals. We performed whole genome re-sequencing of the parental lines, identified EMS-induced SNPs, and generated sequence-based markers in the Lr47 candidate region. We used these markers to genotype lines with recombination events in the target region and phenotyped these lines for resistance to Pt. Based on these results, we mapped Lr47 between markers pkus675 and pkus175 (Fig. 2d), within the same interval as in the Kern Lr47 × CS ph1b population. The reduced candidate region corresponds to a physical interval of 2.5 Mb in the reference genome of the Ae. speltoides accession ‘TS01’ (Fig. 2e). Within the candidate gene region in the TS01 genome, we found six typical NLR genes (Fig. 2e, Supplementary Table 7).
+
+Identification of the Lr47 candidate gene using EMS mutagenesis and transcript assembly (EMTA). Among the progenies of 662 M₂ mutant families screened with Pt race THDB, we identified ten independent families with susceptible plants and confirmed their susceptibility to race THDB by evaluting M₃ seeds derived from the susceptible plants (Fig. 3a). Genotyping using six 7A/7S-genome specific markers confirmed that all the susceptible mutant lines carried the complete Ae. speltopides chromosome segment.
+
+To identify the candidate gene of Lr47, we first generated RNA-seq reads (≥69.8 million 150-bp paired-end reads per sample) from Pt-inoculated leaves of Kern Lr47 and ten independent susceptible M₃ mutants (Supplementary Table 8). We used inoculated leaves because we did not know if the Lr47 gene is induced by the pathogen or not. Assembly of the Kern Lr47 sequenced reads yielded 146,715 high confidence transcript contigs (≥500 bp). We performed BLASTN searches of the Kern Lr47 transcriptome database using the sequences of the candidate NLR genes in CS and TS01 as queries, and obtained 45 transcript contigs with a BLAST e-value = 0. We mapped the sequenced reads of ten susceptible mutants against the 45 selected Kern Lr47 transcript contigs and found one contig named KN638873_g379_i12 with EMS-type (G/C-to-A/T) point mutations in all ten mutants (Fig. 3b and Supplementary Fig. 4a). Using the primer pairs EMS8054F2R1, EMS8054F1R4, and EMS8054F7R7 (Supplementary Table 4) developed from this contig, we performed PCRs to amplify the region containing the mutations and confirmed the presence of nucleotide transitions in these susceptible mutants (Supplementary Fig. 4b). In addition, genotyping of the critical recombinants derived from the cross m118 × Kern Lr47 with marker EMS8054F1R4 confirmed that the C-to-T SNP in m118 co-segregated with the disease phenotype.
+
+To determine the structure of the Lr47 candidate gene in contig KN638873_g379_i12, we (i) assembled the whole genome re-sequencing reads to obtain a genomic sequence containing the Lr47 candidate; (ii) confirmed the transcript and genomic sequences of the Lr47 candidate by PCR and Sanger sequencing; and (iii) compared the transcript with the corresponding genomic sequence. These analyses revealed that the Lr47 candidate gene has six exons encoding a typical NLR protein of 928 amino acids, which is hereafter referred to as CNL2 (Fig. 3c; GenBank accession number OQ919262). Using 5’ and 3’ RACE (Supplementary Fig. 5), we determined that the 5’-untranslated region (UTR) of CNL2 is 1,142 bp long with three introns, and the 3’UTR is 607 bp long with only one intron (Fig. 3c). The predicted 2,787 bp coding sequence of CNL2 is 95.6% and 91.9% identical to its closest homologs within the candidate regions in Ae. speltoides TS01 and CS, respectively. Based on the predicted coding sequence, we found that four of the mutations introduced premature stop codons and the others introduced nonsynonymous amino acid changes (Fig. 3c). AlphaFold prediction of the full length CNL2 protein yielded a structural model with the expected structures for the CC and NB domains as well as an LRR domain containing multiple repeat units forming a typical α/β horseshoe fold (Supplementary Fig. 6). Four amino acid substitutions were located in the CC domain, three in the NB domain, and four in the LRR domain. The amino acid substitutions are associated with large changes in the predicted CNL2 structure (Supplementary Fig. 7).
+
+Validation of CNL2 using BSMV-sgRNA-based gene editing. Because the BSMV-sgRNA-based gene editing method requires a wheat line expressing the Cas9 gene, we first crossed the introgression line Kern Lr47 with the cultivar Bobwhite carrying a highly expressed Cas9 allele³¹. The resulting F₁ plants, carrying both the Lr47 and Cas9 genes, were inoculated with BSMV-sgRNA constructs targeting the candidate gene CNL2. Leaves of the inoculated F₁ plants showed chlorotic spots and white streaks, consistent with BSMV-induced viral symptoms (Supplementary Fig. 8). Next-generation sequencing of amplicons prepared with DNA isolated from the infected leaves of F₁ plants showed a somatic editing efficiency of ~16%.
+
+Infected F₁ (M₀) plants were self-pollinated to produce M₁ seeds. Genotyping of 216 M₁ plants identified 24 plants (11.1%) with heterozygous or homozygous mutations. Among them, two plants (mut-1 and mut-2) that were homozygous for “T” or “TT” deletions at position 2,280 downstream of the ATG in the cDNA (Fig. 4a) were selected for further characterization. These frameshift deletions alter 18% of the protein sequence, resulting in loss of function of the CNL2 protein. Evaluations of the progeny of the two selected plants with race THDB showed that the plants homozygous for the deletions displayed susceptible reactions similar to those of Cas9-transgenic Bobwhite, whereas Kern Lr47 and its sister line without editing were resistant (Fig. 4b). These results indicate that CNL2 is required for Lr47-mediated resistance to Pt.
+
+Wheat plants transformed with CNL2 were resistant to leaf rust. To determine whether CNL2 is sufficient to confer resistance to leaf rust, a 7,234 bp genomic DNA fragment from the introgression line Kern Lr47, including the complete coding region, introns, and regulatory sequences (Fig. 4c), was transformed into the spring wheat cultivar Fielder via A. tumefaciens-mediated transformation. A total of 80 independent transgenic T₀ plants were generated, and the presence of the transgene was confirmed with the primer pairs EMS8054F7R7 and Lr47speF5R5 (Supplementary Table 4), which amplify the transcribed region of CNL2. All transgenic plants containing the transgene and Kern Lr47 (positive control) showed high levels of resistance to Pt race PHQS, whereas the untransformed control Fielder was completely susceptible (Supplementary Figs. 9a and 10). Some of the resistant transgenic plants showed lower or even higher levels of resistance than Kern Lr47 (Supplementary Figs. 9a and 10), suggesting that these transgenic plants may have different numbers of CNL2 insertions or expression levels.
+
+To test this hypothesis, the qRT-PCR primer pair Lr47qPCRF2R3 were used to assess the transcript levels of CNL2 in 11 randomly selected transgenic T₀ plants (T₀ CS401-1 to T₀ CS401-11) using ACTIN as an endogenous control. CNL2 transcript levels were significantly higher in all selected transgenic T₀ plants than in the Fielder control (P < 0.01), and eight of them showed higher CNL2 transcript levels than Kern Lr47 (P < 0.05; Supplementary Fig. 9b). Approximately 25 T₁ plants from each selected transgenic family were genotyped with the primer pair Lr47speF5R5, and all except three showed significant departures from the expected segregation ratio of 3:1 (transgenic/non-transgenic), with an excess of transgenic plants (Supplementary Table 9). The number of CNL2 insertions was determined by TaqMan copy number assays (Supplementary Table 9). Overall, plants derived from three of the transgenic events (T₁ CS401-2, T₁ CS401-7, and T₁ CS401-8) were estimated to have only a single copy of the transgene, whereas the other eight transgenic families were estimated to have between two and five CNL2 copies (Supplementary Table 9). Estimated CNL2 copy number based on TaqMan assays correlated significantly with the transcript levels in the selected transgenic plants (R = 0.90, P < 0.001).
+
+Transgenic T₁ plants from the 11 selected transgenic events were challenged with races PHQS and FHJL, which are virulent on Fielder. All plants from the T₁ transgenic families T₁ CS401-1 and T₁ CS401-10, which were fixed for the transgene, exhibited a high level of resistance, and resistance in the transgenic families T₁ CS401-2 and T₁ CS401-7 perfectly co-segregated with the presence of the transgene (Fig. 4d and Supplementary Fig. 11). These results demonstrated that CNL2 is sufficient to confer resistance to Pt races PHQS and FHJL. Inoculation of transgenic plants and the Fielder control with race THDB did not provide useful information because Fielder was highly resistant (ITs = 1-) to this Pt race (Supplementary Fig. 12).
+
+Taken together, the high-resolution genetic map, loss-of-function EMS and BSMV-Cas9-induced mutations, and the transgenic results demonstrated that the candidate CNL2 is Lr47.
+
+The Lr47 translocation segment originated from Ae. speltoides var. speltoides. A search of the released reference genomes identified the closest homologs of Lr47 on chromosomes 7A, 7B, and 7D of diploid, tetraploid, and hexaploid wheat, as well as on chromosome 7 of Sitopsis species (including Ae. speltoides, Ae. longissimi, Ae. sharonensis, and Ae. searsii), which had 83.29–92.78% protein similarity to Lr47. The phylogenetic tree in Supplementary Fig. 13 shows that Lr47 and its homologs from Sitopsis species are in the same phylogenetic clade. Multiple alignment of the protein sequences revealed that most of the variation was present in the LRR domain (Supplementary Fig. 13).
+
+A dominant marker, Lr47mas, was designed based on the DNA polymorphism C1059G (resulting in amino acid change D353E), which distinguishes Lr47 from all other homologs found so far (Supplementary Fig. 14). The forward primer was designed in the first intron and the reverse primer contained the critical nucleotide 1059C at its terminal 3′ end. PCR amplification with the marker Lr47mas (Supplementary Table 4) at an annealing temperature of 54 ℃ yielded an amplicon of 488 bp only when Lr47 was present (Supplementary Fig. 15). Using this marker, we examined a large collection of wheat genotypes, including 144 accessions of T. aestivum, 78 of T. turgidum, 24 of T. monococcum, and 118 of Ae. speltoides (Supplementary Table 10). PCR amplicons of the predicted size were present in only three of the Ae. speltoides accessions but were absent in all diploid, tetraploid, and hexaploid wheat accessions (except for six Lr47 introgression lines) tested in this study (Supplementary Table 10). To confirm the presence of Lr47 in the three Ae. speltoides lines (T2140002, Y162, and Y397), we designed two pairs of gene-specific primers, Lr47SHF3/R7 and Lr47SHF2/R1 (Supplementary Table 4), that amplify the complete coding region of Lr47. Sanger sequencing confirmed that these three Ae. speltoides accessions carry a functional Lr47 allele identical to that of Kern Lr47.
+
+Ae. speltoides includes two subspecies: Ae. speltoides var. speltoides and Ae. speltoides var. ligustica. Since the taxonomic subspecies and origin of the Lr47 introgressed segment are not clear²¹,²², we compared the SNPs identified from Kern Lr47 RNA-seq data with those of eight Ae. speltoides var. speltoides accessions, six Ae. speltoides var. ligustica lines, and two Ae. speltoides lines for which the subspecies is unknown but carry Lr47 (T2140002 and Y162). We focused only on the polymorphisms that were within the ~150 Mb introgressed segment 7S#1S and were polymorphic among the 16 Ae. speltoides accessions described above, but were absent in the recurrent parent Kern. Based on this approach, we obtained 199 polymorphic sites (Supplementary Table 11). A Neighbor-Joining tree based on these polymorphisms showed that Kern Lr47 is in a branch encompassing multiple Ae. speltoides var. speltoides accessions (Supplementary Fig. 16), suggesting that the translocated segment originated from this subspecies. The two accessions that most closely resemble to Kern Lr47 are T2140002 and Y162, which carry Lr47. The geographical origins of T2140002 and Y397 (carrying Lr47) are unknown, but the most closely related accession Y162 was collected in Iraq, suggesting that Iraq and its neighboring countries are a likely region where Lr47 originated.
+
+Reducing the length of the introgressed Ae. speltoides chromosome segment carrying Lr47. Because the large introgressed Ae. speltoides segment 7S#1S was associated with several undesirable traits, we reduced the size of the introgressed segment carrying Lr47 using the ph1b mutation. Supplementary Fig. 17 describes the crosses that led to the shortening of the alien chromosome segment 7S#1S. We started from recombinant L2 (Fig. 2c), which has a smaller introgression segment of approximately 40 Mb with a recombination event between 76.1 and 80.1 Mb (Fig. 2c and Supplementary Fig. 18).
+
+Subsequently, the L2 plant was self-pollinated and its progeny was used for a second round of recombination with the ph1b mutant. Using the markers pku0745 and pku1176, we identified 20 recombinants from 313 segregating individuals. These recombinants were genotyped with the 7A/7S genome-specific markers from this region (Supplementary Table 4). Among them, we identified one critical recombinant, which we hereafter refer to as F284-128 (L8). The crossover breakpoint in L8 was between markers pku0957 (58.8 Mb) and pku1026 (63.0 Mb), so the 7S chromosome segment in this resistant line is reduced to between 13.1 Mb and 21.3 Mb (Supplementary Fig. 18d).
+
+Recombinant L8 with the truncated 7S segment was crossed and backcrossed once with the Chinese common wheat variety Yangmai21 (YM21) (Supplementary Fig. 17), which is susceptible to multiple Pt races, including THDB, PHQS, FHJL, HCJR, PHRT, and PHTT. Four PCR markers pku1026, Lr47speF5R5, Lr47mas, and pku1176 (Supplementary Table 4), were used to confirm the presence of the truncated 7S segment in the selected BC₁ F₂ plants (hereafter referred to as YM21+Lr47-2). Homozygous BC₁ F₂ plants from the introgression line YM21+Lr47-2 challenged with six different Chinese Pt races showed high levels of resistance (ITs = 0 to 1), whereas the recurrent parent YM21 and the sister line lacking the smaller alien chromosome segment exhibited susceptible infection types (ITs = 3 to 4) in response to the same races (Fig. 5a).
+
+We then performed GISH experiments to validate the results of the molecular marker analysis described above. GISH analyses confirmed the presence of the interstitial Ae. speltoides segment in Kern Lr47 on the recombinant chromosome arm 7A/7S (Fig. 5b), whereas recombinants L2 and L8 had significantly smaller alien chromatin segments (Fig. 5c, d).
+
+FISH-based karyotype analysis revealed strong Oligo-pSc119.2 and Oligo-pTa713 signals on the introgressed Ae. speltoides chromosome segment 7S#1S in Kern Lr47 (Fig. 5e), whereas only the pSc119.2 signal was detected in L2 (Fig. 5f), suggesting that the proximal region of the introgressed segment 7S#1S was eliminated by the recombination event in L2 (Fig. 2c). In L8, neither Oligo-pSc119.2 nor Oligo-pTa713 signal was detected (Fig. 5g), suggesting that the distal region of the alien chromatin was also replaced by wheat chromatin after the second round of homoeologous recombination. These results are consistent with the analysis with PCR markers (Supplementary Fig. 18) and with the GISH results (Fig. 5b-d).
+
+Agronomic and quality evaluation of the wheat line carrying the shortened alien chromosome segment. To evaluate the effects of the truncated alien chromosome segment containing Lr47 (35.9–40.0 Mb, Supplementary Fig. 18c) on agronomic and quality traits, recombinant L2 was crossed and backcrossed three times with the hexaploid wheat variety YM21 and self-pollinated to generate the BC₃ F₃ seeds (Supplementary Fig. 17). This line was selected because it was available earlier than L8. We planted BC₃ F₃ sister lines homozygous for the presence (YM21+Lr47-1) or absence of the alien chromosome segment in both the greenhouse and growth chamber (Supplementary Fig. 19a, b) and measured the phenotypic changes. Under disease-free conditions, no significant differences were observed between the new introgression line YM21+Lr47-1 and its sister line for 15 morphological and quality traits examined (Fig. 6). The only exception was the spike length (SL), which was significantly shorter in YM21+Lr47-1 than in its sister line (P < 0.05; Fig. 6c). The results were consistent between the growth chamber and the greenhouse experiments (Supplementary Fig. 19).
+
+In a field experiment in Shandong Province, China, the introgression line YM21+Lr47-1 was highly resistant to Pt, whereas the sister control line was fully susceptible (Fig. 6q, r). Analysis of morphological traits showed that SL was also significantly shorter in YM21+Lr47-1 than in the isogenic sister line without the Lr47 introgression (Fig. 6s, t). There was no significant difference in SNS, PH, or TN between YM21+Lr47-1 and the sister control line (Fig. 6u and Supplementary Fig. 20).
+
+Characterization of Lr47, which encodes a typical NLR immune receptor protein. Transcript levels of Lr47 relative to ACTIN were measured in Kern Lr47 by qRT-PCR. We observed no significant transcriptional differences between plants inoculated with Pt race PHQS and mock inoculated with water at 1, 2, 4, and 6 dpi (Fig. 7a), indicating that Lr47 is not induced by the presence of the Pt pathogen.
+
+The predicted Lr47 protein includes an N-terminal CC domain, a central NB site, and a C-terminal LRR region, with two predicted monopartite nuclear localization signals determined by the cNLS Mapper program (amino acids 22–54 and 533–539; Fig. 7b). To determine the subcellular localization of Lr47, a GFP tag was used to visualize Lr47 in tobacco (N. benthamiana) leaves and wheat protoplasts. Both cytoplasmic and nuclear fluorescence were observed in tobacco leaves for all constructs, namely GFP-Lr47_CDS, Lr47_CDS-GFP, GFP-Lr47_CC, and Lr47_CC-GFP constructs (Fig. 7c). Anti-GFP immunoblots confirmed that the proteins were expressed at the expected sizes for all constructs (Fig. 7d). In wheat protoplasts, cytoplasmic and nuclear signals were also detected in GFP-Lr47_CDS, Lr47_CC-GFP, and GFP-Lr47_CC constructs (Supplementary Fig. 21).
+
+To test whether the full length Lr47 and/or the CC domain alone are capable of triggering cell death in N. benthamiana, we investigated signaling after Agrobacterium-mediated transient expression in leaves of N. benthamiana. We did not observe cell death or obvious yellowing in leaf regions transiently overexpressing Lr47 and its protein domains individually. In contrast, robust cell death was observed in leaf regions expressing BAX and the CC domain of Sr13 (Fig. 7e and Supplementary Fig. 22). Moreover, the self-interaction ability of Lr47 was initially assessed using yeast two-hybrid assays. We did not detect any direct interaction between the Lr47 protein itself and its protein domains CC, NB, and LRR in vitro (Supplementary Fig. 23).
+
+# Discussion
+
+Wild relatives of wheat are valuable sources of disease resistance genes, and they have been used previously for introgressing *Lr* genes into common wheat varieties <sup>6</sup>. However, cloning resistance genes in wheat relatives is often hampered by the suppressed recombination in the regions carrying alien introgressions. Several newly developed technologies, such as MutRenSeq and MutChromSeq, have facilitated the isolation of resistance (*R*) genes in wheat and its wild relatives <sup>38</sup>. Here, we report the cloning of the broadly effective leaf rust resistance gene *Lr47* using a combination of *ph1b*-induced recombination and the EMTA method. Using this approach, the *CNL2* candidate gene for *Lr47* was successfully identified. The susceptibility to *Pt* races of the 12 independent EMS or BSMV-based editing mutants and the resistance of the *CNL2* transgenic Fielder lines (Figs. <span class="InternalRef" refid="Fig3">3</span> and <span class="InternalRef" refid="Fig4">4</span>) confirmed that this candidate gene is both necessary and sufficient to confer resistance to leaf rust, and therefore that *CNL2* is *Lr47*.
+
+The EMTA method developed in this study (Fig. <span class="InternalRef" refid="Fig3">3</span> b) does not require the generation of a physical map/reference sequence across the map interval. This feature is useful and may spark the interest of researchers in identifying target genes in recombination-sparse regions, especially in plant species with large genomes. However, the effectiveness of the EMTA method relies on several factors: (1) the delimitation of the target gene within a physical interval; (2) the isolation of multiple independent mutants carrying loss-of-function mutations; (3) the presence of orthologs or homologs or paralogs in the colinear regions in sequenced reference genomes; and (4) the comparison of the selected mutants with wild type using RNA-seq data. It is worth noting that RNA-seq might be biased by tissue or sampling time, and therefore other sequencing strategies, such as whole genome re-sequencing or exome capture sequencing, could be good options. In this study, the wild-type Kern *Lr47* was sequenced at very high sequencing depth (> 250 million reads, Supplementary Table 8) on the Illumina NovaSeq 6000 platform to ensure a high-quality transcriptome assembly. *Lr47* was found within a complex locus comprising a cluster of NLR genes that exhibit both copy number and structural variations (Supplementary Tables 6 and 7). Since NLR genes are the most common class of genes associated with disease resistance in wheat and other plants <sup>32,34,35,39−41</sup>, we hypothesized that one of the NLR genes might be *Lr47* and therefore prioritized NLR genes within the candidate region (Fig. <span class="InternalRef" refid="Fig2">2</span> d, e) for further analysis. This significantly reduced the amount of sequencing data to be analyzed.
+
+Recently, virus-based sgRNA delivery systems have been developed and tested in wheat and other plant species <sup>31,42</sup>. Compared with the biolistic-based or *Agrobacterium*-mediated gene editing systems, the virus-based delivery systems do not require conventional plant genetic transformation and regeneration procedures <sup>31</sup>. In our study, the BSMV-based sgRNA delivery system was able to generate both somatic and heritable mutations in wheat via infection of previously obtained Cas9-transgenic wheat plants <sup>31,43</sup>. We demonstrated that the Cas9-transgenic wheat plants can be crossed with a line carrying the targeted alien chromosome (Kern *Lr47*) to generate edits in the F<sub>1</sub> plants (Fig. <span class="InternalRef" refid="Fig3">3</span> b). The results suggest that the BSMV-based sgRNA delivery system could reduce genotype dependency for targeted gene editing by transferring the high-expressing Cas9 locus into different wheat cultivars. Considering the simplicity of the BSMV-sgRNA infection procedure, this genome editing tool has great potential for applications in wheat.
+
+Sequence alignments of the Lr47 protein and representative homologs revealed critical amino acid polymorphisms, which we used to develop a diagnostic marker for the presence of Lr47 (Supplementary Figs. 14 and 15). We validated this diagnostic marker in a larger number of accessions of *T. monococcum*, *T. turgidum*, *T. aestivum*, and *Ae. speltoides* (Supplementary Table 10). We detected *Lr47* in only 2.5% of the *Ae. speltoides* genotypes and in none of the diploid, tetraploid, and hexaploid wheat accessions tested, except for six *Lr47* NILs. This result indicates that the incorporation of *Lr47* has the potential to improve leaf rust resistance in a wide range of modern wheat varieties and highlights the importance of mining new *R* genes from wild progenitors.
+
+One limitation for the incorporation of alien segments by homologous recombination is that the introgressed chromosome segments are usually large and can carry linked genes with negative agronomic or quality effects. Examples of linkage drag of alien introgressions carrying resistance genes include *Lr19* from *Ag. elongatum*, *Pch1* from *Ae. ventricosa*, and *Pm16* from *Ae. speltoides* <sup>44,45</sup>. Similarly, the introgression of the *Ae. speltoides* segment 7S#1S carrying *Lr47* was reported to be associated with several negative effects on agronomic and quality traits <sup>24</sup>. To enhance the utility of *Lr47* in wheat breeding, we attempted to eliminate the deleterious linkage drag by two rounds of *ph1b*-induced homoeologous recombination. Using a combination of marker analysis, cytology, and *Lr* phenotypic screening (Fig. <span class="InternalRef" refid="Fig5">5</span>, Supplementary Fig. 18), two recombinants with much smaller *Ae. speltoides* chromosome segments (L2: ~36 Mb and L8: ~13 Mb) were identified.
+
+A preliminary characterization of the introgression line YM21 + *Lr47*-1 in greenhouse and chamber experiments showed no significant adverse effects on morphological and quality traits in the absence of disease. The only exception was SL, which was significantly shorter in YM21 + *Lr47*-1 than in its isogenic sister line (Fig. <span class="InternalRef" refid="Fig6">6</span> c). Shorter spikes were also observed in the field experiment (Fig. <span class="InternalRef" refid="Fig6">6</span> s, t). Despite reducing SL, the presence of the *Lr47* introgression was not associated with reduction in spike grain yield (Fig. <span class="InternalRef" refid="Fig6">6</span> and Supplementary Fig. 19). We are currently introgressing the shorter ~ 13 Mb *Ae. speltoides* chromosome segment into the Chinese common wheat cultivar YM21 to determine whether the shorter spikes observed in YM21 + *Lr47*-1 is due to linkage with other alien genes or pleiotropic effects of the *Lr47* gene. The latter is unlikely because no significant differences in SL were observed between transgenic plants homozygous for the transgene (1 copy) and the untransformed control Fielder under disease-free conditions (Supplementary Fig. 24). If necessary, the flanking markers and the diagnostic marker for *Lr47* (Supplementary Table 4) can be used to develop wheat lines with even smaller introgressed segments carrying the *Lr47* gene.
+
+For NLR genes in plants, the CC domain is thought to play a crucial role in downstream signaling. Previous studies reported that the CC domains of numerous NLR proteins (such as MLA10, Pm21, Sr33, and Sr50) are sufficient to cause cell death after transient expression in *N. benthamiana* leaves <sup>46</sup>. However, other NLR proteins, including RPM1, RPS5, Rx, and Sr35, have CC domains that do not trigger cell death in leaves of *N. benthamiana* <sup>47</sup>. In this study, we found that the CC domain of the Lr47 protein cannot directly induce cell death (Fig. <span class="InternalRef" refid="Fig7">7</span> e and Supplementary Fig. 22), nor does it have self-protein interaction (Supplementary Fig. 23), indicating that an unknown activation mechanism is essential for Lr47 function. Although NLR proteins have been shown to mediate hypersensitive response by forming pentameric disease resistosomes in *Arabidopsis* (ZAR1) <sup>48</sup> and wheat (Sr35) <sup>49</sup>, the activation and downstream pathways of resistance for Lr47 require further investigation.
+
+*Lr47* provides strong resistance to virulent *Pt* pathotypes in a wide range of regions, including the United States <sup>23</sup>, Mexico <sup>1</sup>, Russia <sup>50</sup>, South Africa <sup>50</sup>, the Indian sub-continent <sup>51</sup>, and China (the current study), indicating broad-spectrum resistance to leaf rust. However, elucidation of the typical NLR structure of *Lr47* indicates that its resistance may not be durable in the field because of rapid changes in leaf rust pathogen populations. The rust pathogen may evolve through the mutation or deletion of the corresponding *Avr* genes, leading to the breakdown of resistance conferred by NLR genes <sup>52,53</sup>. For instance, the Ug99 race group of the wheat stem rust pathogen (*Puccinia graminis* f. sp. *tritici*) has successfully broken the resistance of wheat varieties carrying the resistance genes *Sr24*, *Sr31*, *Sr36*, *Sr38*, and *SrTmp* <sup>39,54−56</sup>. Therefore, a combination of *Lr47* with other slow-rusting multi-pathogen resistance genes, such as *Lr34*/*Sr57* <sup>10</sup> and *Lr67*/*Sr55* <sup>11</sup>, is a preferred strategy for developing wheat cultivars with durable resistance against this devastating rust pathogen.
+
+In conclusion, the identification of *Lr47*, the new introgression lines with eliminated or reduced deleterious linkage drag, and the available diagnostic marker developed in this study provide useful tools to diversify the deployed *Lr* genes and accelerate the use of *Lr47* in modern wheat breeding programs.
+
+# Methods
+
+**Plant materials and mapping populations.** The six pairs of hard red spring wheat near-isogenic lines (NILs) with and without the leaf rust resistance gene *Lr47*<sup>18,22</sup> used in this study are listed in Supplementary Table 1. A total of 7,590 recombinant gametes from two mapping populations were used to construct a high-density genetic map of *Lr47*. The first one included 2,654 plants from selected F<sub>3</sub> families homozygous for the *ph1b* mutation and segregating for the *Ae. speltoides* chromosome segment (population Kern *Lr47* × CS *ph1b*). The second population, included 1,141 F<sub>2</sub> individuals, was generated from a cross between a Kern *Lr47* EMS mutant susceptible to *Pt* designated m118 and the wild type (Kern *Lr47*). Since both parents (m118 and Kern *Lr47*) carry the same *Ae. speltoides* chromosome, recombination is normal in the presence of the *Ph1b* wild-type allele. Moreover, Kern *Lr47* was crossed with the susceptible wheat line ZhengZhou 5389 (ZZ5389) to characterize leaf rust resistance in the presence of the *Ph1* gene. A collection of 118 accessions of *Ae. speltoides* (Supplementary Table 2), 24 of *Triticum monococcum*, 78 of *T. turgidum*, and 144 of *T. aestivum* (including six *T. aestivum* genetic stocks carrying the *Lr47 Ae. speltoides* segment 7S#1S) was used to determine the value of the diagnostic marker developed in this study for marker assisted selection.
+
+**Leaf rust assays and pathogen growth in infected leaves.** Leaf rust seedling evaluations were performed at the Peking University Institute of Advanced Agricultural Sciences, Weifang, China and Hebei Agricultural University, Baoding, China. A total of 24 *Pt* isolates and their avirulence/virulence profiles are presented in Supplementary Table 3. Three-leaf stage seedlings were inoculated with fresh *Pt* urediniospores mixed with talcum powder at a ratio of 1:20 using the shaking off method<sup><span citationid="CR57" class="CitationRef">57</span></sup>. The inoculated plants were placed in a dark dew chamber set at 22 ℃ for approximately 24 h and then maintained at 22–24 ℃ with a 16-h photoperiod. Infection types (ITs) of plants were scored at ~12 days post inoculation (dpi) using a 0–4 scale as described previously<sup><span citationid="CR58" class="CitationRef">58</span></sup>.
+
+To estimate the growth of *Pt* pathogen in NILs, leaf segments (~6 cm long) from Kern *Lr47* and its recurrent parent Kern inoculated with race THDB were sampled at 2, 4, 6, and 8 dpi. The collected leaves were autoclaved in 1 M KOH, stained with wheat germ agglutinin labeled with fluorescein isothiocyanate (WGA-FITC; Cat No. L4895-10MG, Sigma-Aldrich, USA) and visualized by a Zeiss Discovery V20 fluorescence dissecting microscope (Zeiss, Jena, Germany) according to previously described methods<sup><span citationid="CR32" class="CitationRef">32</span>, <span citationid="CR59" class="CitationRef">59</span></sup>.
+
+**EMS mutant screening.** Seeds of wheat line Kern *Lr47* were treated with 250 mL of 0.8% EMS (Cat No. M0880-25G, Sigma-Aldrich, USA) and incubated for 18 h at 25 ℃ on a shaker at 150 rpm. Next, the treated seeds were washed five times with 300 mL tap water and then placed under running water for 3 h. All surviving M<sub>1</sub> plants were grown in a greenhouse, and a total of 4,756 independent M<sub>2</sub> families were obtained. Approximately 25 M<sub>2</sub> seeds per family were challenged with *Pt* race THDB in growth chambers at Peking University Institute of Advanced Agricultural Sciences, Weifang, China. The M<sub>3</sub> seeds derived from susceptible M<sub>2</sub> plants were re-evaluated with races PHQS and THDB. Finally, all the identified mutants were genotyped with 7S-specific primers to rule out the possibility of seed contamination.
+
+**Sequences, markers, and bioinformatics analysis.** RNA sequencing (RNA-seq) of *Lr47* NILs, *Ae. speltoides* accessions (AE915, AE1590, and PI 554292), and the susceptible EMS mutant lines (accession number CRA011051) were performed at Beijing Novogene Bioinformatics Technology Co., Ltd. (Beijing, China). The released reference genomes of hexaploid wheat Chinese Spring (CS; RefSeq v1.1) and other 10 wheat genotypes from the Wheat Pan Genome Project were used in our analyses<sup><span citationid="CR60" class="CitationRef">60</span>, <span citationid="CR61" class="CitationRef">61</span></sup>. RNA-seq data of Yecora Rojo *Lr47*, UC1037 *Lr47*, and White Yecora *Lr47* were from a previous study<sup><span citationid="CR26" class="CitationRef">26</span></sup>. Raw RNA-seq reads were trimmed using Trimmomatic software version 0.32 to remove low-quality reads and adaptors<sup><span citationid="CR62" class="CitationRef">62</span></sup>. Trimmed reads were aligned to the CS reference sequence using STAR software v2.7.10a<sup><span citationid="CR63" class="CitationRef">63</span></sup>. Variant calls were made using Freebayes v1.3.6<sup>64</sup> and filtered using BCFtools v1.14 (<span class="ExternalRef"><span class="RefSource">https://github.com/samtools/bcftools</span><span address="https://github.com/samtools/bcftools" class="RefTarget" targettype="URL"></span></span>).
+
+EMS mutagenesis and transcript assembly (EMTA) was performed as follows: RNA-seq reads from Kern *Lr47* were assembled *de novo* using Spades version 3.14.1<sup>65</sup>. We performed BLASTN searches of the transcriptome database of Kern *Lr47* using sequences of typical NLR genes within the *Lr47* candidate regions from the reference genomes of CS and TS01<sup>66</sup> as queries, and obtained transcript contigs with similarity values greater than 85% to the homologs or paralogs within the interval that *Lr47* was mapped to. Next, the RNA-seq reads of the ten susceptible EMS mutants were mapped to these transcript contigs as reported previously<sup><span citationid="CR67" class="CitationRef">67</span></sup>.
+
+To develop markers in the m118 × Kern *Lr47* population, we performed whole genome re-sequencing for both parents (accession number CRA011051). We aligned the m118 and Kern *Lr47* sequences of the genes in the candidate region and identified the polymorphic sites. Genome-specific primers were designed using the Primer 3 online software (v0.4.0, <span class="ExternalRef"><span class="RefSource">https://bioinfo.ut.ee/primer3-0.4.0/primer3/</span><span address="https://bioinfo.ut.ee/primer3-0.4.0/primer3/" class="RefTarget" targettype="URL"></span></span>) and used to amplify regions carrying putative *Ae. speltoides*-specific single nucleotide polymorphisms (SNPs). The procedures for developing Insertion-Deletion (InDel) and cleaved amplified polymorphic sequence (CAPS) markers were reported previously<sup><span citationid="CR68" class="CitationRef">68</span>, <span citationid="CR69" class="CitationRef">69</span></sup>. The other published reference genome of *Ae. speltoides* AEG-9674-1<sup>70</sup> was used to support marker development.
+
+**Rapid amplification of cDNA ends (RACE).** 5’- and 3’-RACE reactions were carried out using total RNA extracted from leaves of the introgression line Kern *Lr47*. RACE reactions were performed using the Invitrogen FirstChoice RLM-RACE Kit (Cat. no. AM1700, ThermoFisher Scientific, MA) according to the manufacturer’s instructions. PCR products from RACE reactions were cloned into the pMD18-T vector (TaKaRa Biotechnology, Kyoto, Japan) using the TA cloning method. The selected positive clones were sequenced using the Sanger method.
+
+**BSMV-sgRNA-based gene editing.** The BSMV-based guide RNA delivery system<sup><span citationid="CR31" class="CitationRef">31</span></sup> was used to validate the *Lr47* candidate gene. The methods and procedures for plasmid construction and virus infection were reported previously<sup><span citationid="CR31" class="CitationRef">31</span></sup>. The wheat transgenic line exogenously expressing Cas9 gene in genetic background of Bobwhite (Cas9-transgenic Bobwhite) was crossed with the introgression line Kern *Lr47*, and the resulting F<sub>1</sub> plants were infected with BSMV vectors targeting the *Lr47* candidate gene. The infected M<sub>0</sub> plants were self-pollinated to produce M<sub>1</sub> seeds. We identified mutations in the M<sub>1</sub> plants using the primer pair *Lr47BSMVF2R1* (Supplementary Table 4). Finally, the edited plants and their progenies were challenged with *Pt* race THDB.
+
+**Wheat transformation and copy number assays.** A 7,234-bp genomic DNA fragment consisting of the entire coding region and introns (3,132 bp) of *Lr47*, 2,097 bp upstream of the start codon, and 2,005 bp downstream from the stop codon was amplified from the introgression line Kern *Lr47* by PCR using the PrimeStar Max DNA Polymerase (TaKaRa, Kyoto, Japan). The fragment was inserted into the linearized binary vector pCAMBIA1300 using an In-Fusion® HD Cloning Kit (Clontech, CA, USA). The resulting plasmid pCAMBIA1300-*Lr47* was introduced into the susceptible hexaploid wheat line Fielder via *Agrobacterium tumefaciens* (strain EHA105)-mediated transformation. Primer pairs *EMS8054F7R7* and *Lr47speF5R5* (Supplementary Table 4) were used to confirm the presence of the transgene, and the qRT-PCR primer pair *Lr47qPCRF2R3* (Supplementary Table 4) was used to estimate the transcript levels of the transgene in selected transgenic T<sub>0</sub> plants. The T<sub>0</sub> and T<sub>1</sub> transgenic plants were challenged with three Chinese *Pt* races PHQS, FHJL, and THDB. The number of integrated *Lr47* transgenes in each transgenic line was estimated based on the segregation ratio of T<sub>1</sub> transgenic plants and the results of a TaqMan copy number assay<sup><span citationid="CR32" class="CitationRef">32</span>, <span citationid="CR71" class="CitationRef">71</span></sup>.
+
+**qRT-PCR analysis.** At the two-leaf stage, plants from the introgression line Kern *Lr47* were mock or *Pt* inoculated in two independent growth chambers under the same temperature and photoperiod (24°C day/22°C night and 16 h light/8 h darkness). Leaf samples from different plants were collected immediately before inoculation (0 h) and at 1, 2, 4, and 6 dpi. Total RNAs were isolated using the Spectrum Plant Total RNA Kit (MilliporeSigma, MA, USA) and purified using the Direct-zol RNA MiniPrep Plus Kit (Zymo Research, CA, USA). qRT-PCR reactions were carried out on an ABI QuantStudio 5 Real-Time PCR System (Applied Biosystems, CA, USA) using Fast SYBR GREEN Master Mix. qRT-PCR primer pair *Lr47qPCRF2R3* (Supplementary Table 4) was used to evaluate the transcript levels of *Lr47*. Transcript levels were determined in seven biological replicates and presented as fold-*ACTIN* levels as described previously<sup><span citationid="CR72" class="CitationRef">72</span></sup>.
+
+**Phylogenetic analysis.** Sequences of Lr47 protein homologs were obtained from the National Center for Biotechnology Information (NCBI) database (<span class="ExternalRef"><span class="RefSource">https://www.ncbi.nlm.nih.gov</span><span address="https://www.ncbi.nlm.nih.gov" class="RefTarget" targettype="URL"></span></span><span class="Underline" name="Emphasis" type="Underline">)</span> and the published reference genomes of wheat and its wild relatives (<span class="ExternalRef"><span class="RefSource">https://wheat.pw.usda.gov/blast/</span><span address="https://wheat.pw.usda.gov/blast/" class="RefTarget" targettype="URL"></span></span>). To explore the origin of the introgressed segment containing *Lr47*, we performed RNA-seq and variant calling for seven *Ae. speltoides* genotypes, including T2140002, Y162, PI 542258, PI 542251, PI 560751, PI 542260, and PI 542270 (accession number CRA011051). Moreover, SNPs of four *Ae. speltoides* Tausch var. *speltoides* accessions (KU-2208A, KU-14601, KU-14605, and KU-12963a) and three *Ae. speltoides* Tausch var. *ligustica* accessions (KU-2236, KU-7716, and KU-7848) were obtained from RNA-seq data<sup><span citationid="CR17" class="CitationRef">17</span></sup>. Mulitple sequence alignment was conducted using MUSCLE method implemented in MEGA software v7.0. The phylogenetic tree was generated using the pairwise deletion method (bootstrap values based on 1000 iterations) and visualized using the Interactive Tree Of Life (iTOL) v5.0 (<span class="ExternalRef"><span class="RefSource">https://itol.embl.de/</span><span address="https://itol.embl.de/" class="RefTarget" targettype="URL"></span></span>).
+
+**Reduction of *Ae. speltoides* chromosome segment.** The introgression line Kern *Lr47* was crossed with the CS *ph1b* mutant (CS *ph1b*)<sup><span citationid="CR73" class="CitationRef">73</span></sup> to induce homoeologous recombination and reduce the size of the *Lr47* introgressed chromosome segment 7S#1S. Markers *Xwgc2111* and *Xwgc2049* were used to validate the presence of the *ph1b* mutation<sup><span citationid="CR74" class="CitationRef">74</span></sup>. The resulting F<sub>1</sub> plants from the cross Kern *Lr47* × CS *ph1b* were self-pollinated. The derived F<sub>2</sub> plants which were heterozygous for the introgressed segment 7S#1S and homozygous for *ph1b* were self-crossed to generate F<sub>3</sub> families. Using 7A/7S genome-specific markers, we identified recombination events within the introgressed alien chromosome segment and tested these plants for leaf rust resistance. The obtained recombinant plants heterozygous for *Lr47* and homozygous for *ph1b* (residual heterozygous lines) were self-crossed to generate F<sub>4</sub> progeny for another round of screening of new recombinants. The identified critical recombinants carrying *Lr47* were crossed and backcrossed to the Chinese hexaploid wheat variety Yangmai21 (YM21). Finally, plants homozygous for the shortened alien segments were selected using molecular markers and evaluated for resistance to multiple *Pt* races.
+
+**Cytogenetic assays.** Genomic *in situ* hybridization (GISH) and fluorescence *in situ* hybridization (FISH) were conducted using the methods described previously<sup><span citationid="CR75" class="CitationRef">75</span>, <span citationid="CR76" class="CitationRef">76</span></sup>. The FISH probes oligo-pSc119.2, oligo-pTa535, and oligo-pTa713 were suitable for the identification of chromosomes of common wheat and *Aegilops* species<sup><span citationid="CR77" class="CitationRef">77</span></sup>. Oligo-pSc119.2 preferentially paints tandem repeats on B- (or S-) genome chromosomes, and oligo-pTa535 hybridized well to wheat D-genome chromosomes<sup><span citationid="CR78" class="CitationRef">78</span></sup>. A combination of oligo-pSc119.2 and oligo-pTa535 was used to distinguish the 42 wheat chromosomes<sup><span citationid="CR79" class="CitationRef">79</span></sup>. The specific hybridization pattern of probe oligo-pTa713 makes it easy to differentiate chromosomes<sup><span citationid="CR75" class="CitationRef">75</span></sup>. The probes were labeled with either FAM or TAMRA and were synthesized by TsingKe Biological Technology Co., Ltd. (Chengdu, China).
+
+**Evaluation of agronomic and quality traits.** Homozygous BC<sub>3</sub>F<sub>3</sub> plants from the introgression line YM21-*Lr47*-1 and its sister line lacking the reduced alien chromosome introgression were grown in a greenhouse (22–30 ℃ day and 16–24 ℃ night with a 16 h photoperiod) and in a controlled walk-in growth chamber (24 ℃ day/22 ℃ night with a 16 h photoperiod). For the greenhouse and growth chamber experiments, a single plant was grown per pot (~3.8 L) and agronomic traits were measured for each plant. The number of productive tillers (TN) was counted at the time of physiological maturity. Plant height (PH) was calculated by measuring the height of the main tiller of each plant from the ground level to the tip of spike excluding awns. Spike length (SL), spikelet number per spike (SNS), and grain number per spike (GNS) were measured as the mean of the main spikes (2–3 main spikes per plant). Grain yield per plant (GYP), thousand-seed weight (TSW), grain length (GL), and grain width (GW) were automatically calculated in the laboratory using a crop scanning test system (Wanshen SC-G, Hangzhou, China)<sup><span citationid="CR80" class="CitationRef">80</span></sup>. At the maturity stage, the penultimate internodes were cut into segments of equal length (~10 cm), dried at 32 ℃ for 10 days, and then used for measurement. The shearing force (SF) of stems was measured using a universal testing machine (Instron 5848 Microtester, Instron, USA) as described previously<sup><span citationid="CR81" class="CitationRef">81</span></sup>. Grain moisture content (GMC), grain protein content (GPC), flour water absorption (FWA), and flour ash (FA) were estimated using near-infrared reflectance spectroscopy (Model SpectraStar 2600 XT-R, Unity Scientific, USA). Grain hardness (GH) was determined in a TA-XT-plus Texture Analyzer (Stable Micro Systems, United Kingdom). Flour yield (FY) was expressed as grams of flour per 100 g of grain (conditioned to 15% moisture content, AACC 26−10). The field experiment was conducted during the 2022–2023 growing season at Peking University Institute of Advanced Agricultural Sciences, Weifang, China (36°26′04.0″N, 119°26′42.6″E). The introgression line YM21-*Lr47*-1 and its sister line were planted in 10 rows, each one meter long. The significance of differences in agronomic and quality traits between YM21-*Lr47*-1 and its sister line was estimated using Student’s *t*-test.
+
+**Yeast two-hybrid assay.** The coding regions of the full-length Lr47 protein (CDS, amino acids 1–928), coiled-coil (CC) domain (amino acids 1–172), nucleotide-binding site (NB) domain (amino acids 173–608), and leucine-rich repeat (LRR) domain (amino acids 609–928) were cloned into the yeast two-hybrid system vectors pGBKT7 (bait) and pGADT7 (prey). The recombinant vectors were co-transformed into yeast strain AH109 using the lithium acetate and polyethylene glycol 3350 method. Co-transformed yeast was spotted onto synthetic dropout medium lacking leucine and tryptophan (SD-Leu-Trp) for selection of transformed colonies and then leucine, tryptophan, histidine, and adenine (SD-Leu-Trp-His-Ade) to detect interactions.
+
+*In planta* expression of GFP-fused Lr47 protein. The coding regions of the full-length Lr47 protein (amino acids 1–928) and CC domain (amino acids 1–172aa) were cloned into vectors pJIM19-GFP and pJIT163-Ubi-GFP<sup><span citationid="CR82" class="CitationRef">82</span>, <span citationid="CR83" class="CitationRef">83</span></sup>. The recombinant constructs were expressed in *Nicotiana benthamiana* leaves using transformed *A. tumefaciens* strain GV3101. Wheat protoplasts were isolated from the susceptible common wheat variety Fielder and were transformed as previously described<sup><span citationid="CR84" class="CitationRef">84</span></sup>. Fluorescence was checked using a confocal microscope (A1 HD25 Nikon, Tokyo, Japan). Total proteins were extracted from *A. tumefaciens*-transformed tobacco leaves. Protein samples were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred onto polyvinylidene difluoride (PVDF) membranes. Immunoblots were performed by using anti-GFP antibody (Solarbio, Beijing, China).
+
+For the cell death induction assay, the coding regions of the CC domain of a previously identified stem rust resistance protein Sr13<sup>32</sup> and a mammalian cell death inducer BAX were cloned into pJIM19-GFP as positive controls. Highly concentrated (OD = 1.0) *A. tumefaciens* carrying Lr47_CDS-GFP, GFP-Lr47_CDS, Lr47_CC-GFP, GFP-Lr47_CC, Lr47_NB-GFP, GFP-Lr47_NB, Lr47_LRR-GFP, GFP-Lr47_LRR, Sr13_CC-GFP, BAX, and GFP were infiltrated into *N. benthamiana* leaves. Necrosis induced by BAX was observed at 48 h post inoculation (hpi), while yellowing caused by Sr13_CC-GFP was detected at 120 hpi.
+
+# References
+
+1. Huerta-Espino, J. et al. Global status of wheat leaf rust caused by *Puccinia triticina*. Euphytica 179, 143–160 (2011).
+2. Prasad, P., Savadi, S., Bhardwaj, S. & Gupta, P. The progress of leaf rust research in wheat. Fungal Biol 124, 537–550 (2020).
+3. Knott, D. R. The wheat rusts-breeding for resistance. 7 (In: Monograph on Theoretical and Applied Genetics. Chapter 12. Springer-Verlag Berlin Heidelberg. p. 7, 1989).
+4. Savary, S. et al. The global burden of pathogens and pests on major food crops. Nat Ecol Evol 3, 430–439 (2019).
+5. Kolmer, J. Leaf rust of wheat: pathogen biology, variation and host resistance. Forests 4, 70–84 (2013).
+6. Ren, X. et al. Genetics of resistance to leaf rust in wheat: an overview in a genome-wide level. Sustainability 15, 3247 (2023).
+7. Feuillet, C. et al. Map-based isolation of the leaf rust disease resistance gene *Lr10* from the hexaploid wheat (*Triticum aestivum* L.) genome. Proc Natl Acad Sci USA 100, 15253–15258 (2003).
+8. Huang, L. et al. Map-based cloning of leaf rust resistance gene *Lr21* from the large and polyploid genome of bread wheat. Genetics 164, 655–664 (2003).
+9. Cloutier, S. et al. Leaf rust resistance gene *Lr1*, isolated from bread wheat (*Triticum aestivum* L.) is a member of the large psr567 gene family. Plant Mol Biol 65, 93–106 (2007).
+10. Krattinger, S. G. et al. A putative ABC transporter confers durable resistance to multiple fungal pathogens in wheat. Science 323, 1360–1363 (2009).
+11. Moore, J. W. et al. A recently evolved hexose transporter variant confers resistance to multiple pathogens in wheat. Nat Genet 47, 1494–1498 (2015).
+12. Lin, G. et al. Cloning of the broadly effective wheat leaf rust resistance gene *Lr42* transferred from *Aegilops tauschii*. Nat Commun 13, 1–12 (2022).
+13. Hewitt, T. et al. Wheat leaf rust resistance gene *Lr13* is a specific *Ne2* allele for hybrid necrosis. Mol Plant 14, 1025–1028 (2021).
+14. Thind, A. K. et al. Rapid cloning of genes in hexaploid wheat using cultivar-specific long-range chromosome assembly. Nat Biotechnol 35, 793–796 (2017).
+15. Kolodziej, M. C. et al. A membrane-bound ankyrin repeat protein confers race-specific leaf rust disease resistance in wheat. Nat Commun 12, 1–12 (2021).
+16. Wang, Y. et al. An unusual tandem kinase fusion protein confers leaf rust resistance in wheat. available at Research Square https://doi.org/10.21203/rs.3.rs-1807889/v1 (2022).
+17. Miki, Y. et al. Origin of wheat B-genome chromosomes inferred from RNA sequencing analysis of leaf transcripts from section *Sitopsis* species of *Aegilops*. DNA Res 26, 171–182 (2019).
+18. Dubcovsky, J., Lukaszewski, A., Echaide, M., Antonelli, E. & Porter, D. Molecular characterization of two *Triticum speltoides* interstitial translocations carrying leaf rust and greenbug resistance genes. Crop Sci 38, 1655–1660 (1998).
+19. Schneider, A., Molnár, I. & Molnár-Láng, M. Utilisation of *Aegilops* (goatgrass) species to widen the genetic diversity of cultivated wheat. Euphytica 163, 1–19 (2008).
+20. Marais, G. et al. Attempts to remove gametocidal genes co-transferred to common wheat with rust resistance from *Aegilops speltoides*. Euphytica 171, 71–85 (2010).
+21. Wells, D. G., Kota, R., Sandhu, H., Gardner, W. & Finney, K. Registration of one disomic substitution line and five translocation Lines of winter wheat germplasm resistant to wheat streak mosaic virus 1 (Reg. No. GP 199 to GP 204). Crop Sci 22, 1277–1278 (1982).
+22. Brevis, J. C. et al. Agronomic and quality evaluation of common wheat near-isogenic lines carrying the leaf rust resistance gene *Lr47*. Crop Sci 48, 1441–1451 (2008).
+23. Xu, X. et al. Characterization of an incomplete leaf rust resistance gene on chromosome 1RS and development of KASP markers for *Lr47* in wheat. Phytopathology 111, 649–658 (2021).
+24. Helguera, M., Khan, I. & Dubcovsky, J. Development of PCR markers for the wheat leaf rust resistance gene *Lr47*. Theor Appl Genet 100, 1137–1143 (2000).
+25. Yuan, L., Gebrewahid, T. W., Zhang, P., Li, Z. & Liu, D. Identification of leaf rust resistance genes in common wheat varieties from China and foreign countries. J Integr Agr 20, 1302–1313 (2021).
+26. Wu, J. et al. Genome-wide expression profiling of genes associated with the *Lr47*-mediated wheat resistance to leaf rust (*Puccinia triticina*). Int J Mol Sci 20, 4498 (2019).
+27. Figlan, S. et al. Breeding wheat for durable leaf rust resistance in Southern Africa: variability, distribution, current control strategies, challenges and future prospects. Front Plant Sci 11, 549 (2020).
+28. Omara, R. & Abdelaal, K. Molecular and genetic analysis of leaf rust resistance genes in two new Egyptian wheat cultivars. Egypt J Phytopathol 45, 33–52 (2017).
+29. Gultyaeva, E., Shaydayuk, E. & Gannibal, P. Leaf rust resistance genes in wheat cultivars registered in Russia and their influence on adaptation processes in pathogen populations. Agriculture 11, 319 (2021).
+30. Bhardwaj, S. et al. Physiologic specialization and genetic differentiation of *Puccinia triticina* causing leaf rust of wheat on the Indian subcontinent during 2016 to 2019. Plant Dis 105, 1992–2000 (2021).
+31. Li, T. et al. Highly efficient heritable genome editing in wheat using an RNA virus and bypassing tissue culture. Mol Plant 14, 1787–1798 (2021).
+32. Zhang, W. et al. Identification and characterization of *Sr13*, a tetraploid wheat gene that confers resistance to the Ug99 stem rust race group. Proc Natl Acad Sci USA 114, E9483–9492 (2017).
+33. Zhang, J. et al. A single amino acid change can alter the specificity of the multi-allelic wheat stem rust resistance locus *SR9*. available at Research Square https://doi.org/10.21203/rs.21203.rs-2208386/v2208381 (2022).
+34. Luo, J. et al. Identification and characterization of *Sr22b*, a new allele of the wheat stem rust resistance gene *Sr22* effective against the Ug99 race group. Plant Biotechnol J 20, 554–563 (2022).
+35. Li, H. et al. Mapping and characterization of a wheat stem rust resistance gene in durum wheat “Kronos”. Front Plant Sci 12 (2021).
+36. Saintenac, C. et al. Identifcation of wheat gene *Sr35* that confers resistance to Ug99 stem rust race group. Science 341, 783–786 (2013).
+37. Wang, S. et al. *RppM*, encoding a typical CC-NBS-LRR protein, confers resistance to southern corn rust in maize. *Front Plant Sci* 13, 951318 (2022).
+38. Hafeez, A. N. et al. Creation and judicious application of a wheat resistance gene atlas. Mol Plant 14, 1053–1070 (2021).
+39. Chen, S., Zhang, W., Bolus, S., Rouse, M. N. & Dubcovsky, J. Identification and characterization of wheat stem rust resistance gene *Sr21* effective against the Ug99 race group at high temperature. PLoS Genet 14, e1007287 (2018).
+40. Saintenac, C. et al. Identification of wheat gene *Sr35* that confers resistance to Ug99 stem rust race group. Science 341, 783–786 (2013).
+41. Gassmann, W., Hinsch, M. E. & Staskawicz, B. J. The *Arabidopsis RPS4* bacterial-resistance gene is a member of the TIR-NBS-LRR family of disease-resistance genes. Plant J 20, 265–277 (1999).
+42. Hu, J. et al. A barley stripe mosaic virus-based guide RNA delivery system for targeted mutagenesis in wheat and maize. Mol Plant Pathol 20, 1463–1474 (2019).
+43. Wang, W. et al. Multiplexed promoter and gene editing in wheat using a virus-based guide RNA delivery system. Plant Biotechnol J 20, 2332–2341 (2022).
+44. Knott, D. Mutation of a gene for yellow pigment linked to *Lr19* in wheat. Can J Genet Cytol 22, 651–654 (1980).
+45. Bohra, A. et al. Reap the crop wild relatives for breeding future crops. Trends Biotechnol 40, 412–431 (2022).
+46. Gao, A. et al. Pm21 CC domain activity modulated by intramolecular interactions is implicated in cell death and disease resistance. Mol Plant Pathol 21, 975–984 (2020).
+47. Bolus, S., Akhunov, E., Coaker, G. & Dubcovsky, J. Dissection of cell death induction by wheat stem rust resistance protein Sr35 and its matching effector AvrSr35. Mol Plant Microbe In 33, 308–319 (2020).
+48. Wang, J. et al. Reconstitution and structure of a plant NLR resistosome conferring immunity. Science 364, eaav5870 (2019).
+49. Förderer, A. et al. A wheat resistosome defines common principles of immune receptor channels. Nature 610, 532–539 (2022).
+50. Volkova, G., Kudinova, O., Vaganova, O. & Agapova, V. Effectiveness of leaf rust resistance genes in the adult and juvenile stages in southern Russia in 2011–2020. Plants 11, 793 (2022).
+51. Kumar, S. et al. Characterization of five new pathotypes of *Puccinia triticina* identified from Northeast India, Nepal, and Bangladesh. Australas Plant Path 51, 315–325 (2022).
+52. Jones, J. D. G. & Dang, J. L. The plant immune system. Nature 444, 323–329, doi: 10.1038/nature05286 (2006).
+53. Chen, G. et al. Cloning southern corn rust resistant gene *RppK* and its cognate gene *AvrRppK* from *Puccinia polysora*. Nat Commun 13, 4392 (2022).
+54. Jin, Y. et al. Detection of virulence to resistance gene *Sr24* within race TTKS of *Puccinia graminis* f. sp *tritici*. Plant Dis 92, 923–926 (2008).
+55. Jin, Y. et al. Detection of virulence to resistance gene *Sr36* within the TTKS race lineage of *Puccinia graminis* f. sp *tritici*. Plant Dis 93, 367–370 (2009).
+56. Patpour, M. et al. First report of the Ug99 race group of wheat stem rust, *Puccinia graminis* f. sp. *tritici*, in Egypt in 2014. Plant Dis 100, 863–863 (2016).
+57. Chen, S. et al. Stripe rust resistance gene *Yr34* (synonym *Yr48*) is located within a distal translocation of *Triticum monococcum* chromosome 5A^m into common wheat. *Theor Appl Genet* 134, 2197–2211 (2021).
+58. Stakman, E. C., Stewart, D. M. & Loegering, W. Q. Identification of physiologic races of *Puccinia graminis* var. *tritici*. (Washington DC, 1962).
+59. Dawson, A. M. et al. The development of quick, robust, quantitative phenotypic assays for describing the host–nonhost landscape to stripe rust. Front Plant Sci 6, 876 (2015).
+60. The International Wheat Genome Sequencing Consortium. Shifting the limits in wheat research and breeding using a fully annotated reference genome. Science 361, eaar7191 (2018).
+61. Walkowiak, S. et al. Multiple wheat genomes reveal global variation in modern breeding. Nature 588, 277–283 (2020).
+62. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics (Oxford, England) 30, 2114–2120 (2014).
+63. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics (Oxford, England) 29, 15–21 (2013).
+64. Garrison, E. & Marth, G. Haplotype-based variant detection from short-read sequencing. Preprint at http://arxiv.org/abs/1207.3907 (2012).
+65. Bankevich, A. et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J Comput Biol 19, 455–477 (2012).
+66. Li, L.-F. et al. Genome sequences of five *Sitopsis* species of *Aegilops* and the origin of polyploid wheat B subgenome. Mol Plant 15, 488–503 (2022).
+67. Yu, G. et al. *Aegilops sharonensis* genome-assisted identification of stem rust resistance gene *Sr62*. Nat Commun 13, 1607 (2022).
+68. Konieczny, A. & Ausubel, F. M. A procedure for mapping *Arabidopsis* mutations using co-dominant ecotype-specific PCR-based markers. Plant J 4, 403–410 (1993).
+69. Bhattramakki, D. et al. Insertion-deletion polymorphisms in 3′ regions of maize genes occur frequently and can be used as highly informative genetic markers. Plant Mol Biol 48, 539–547 (2002).
+70. Avni, R. et al. Genome sequences of three *Aegilops* species of the section Sitopsis reveal phylogenetic relationships and provide resources for wheat improvement. Plant J 110, 179–192 (2022).
+71. Diaz, A., Zikhali, M., Turner, A. S., Isaac, P. & Laurie, D. A. Copy number variation affecting the *Photoperiod-B1* and *Vernalization-A1* genes is associated with altered flowering time in wheat (*Triticum aestivum*). PloS One 7, e33234 (2012).
+72. Pearce, S., Vanzetti, L. S. & Dubcovsky, J. Exogenous gibberellins induce wheat spike development under short days only in the presence of *VERNALIZATION 1*. Plant Physiol 163, 1433–1445 (2013).
+73. Sears, E. R. An induced mutant with homoeologous pairing in common wheat. Can J Genet Cytol 19, 585–593 (1977).
+74. Gyawali, Y., Zhang, W., Chao, S., Xu, S. & Cai, X. Delimitation of wheat *ph1b* deletion and development of *ph1b*-specific DNA markers. Theor Appl Genet 132, 195–204 (2019).
+75. Komuro, S., Endo, R., Shikata, K. & Kato, A. Genomic and chromosomal distribution patterns of various repeated DNA sequences in wheat revealed by a fluorescence *in situ* hybridization procedure. Genome 56, 131–137 (2013).
+76. Hao, M. et al. Comparison of homoeologous chromosome pairing between hybrids of wheat genotypes Chinese Spring *ph1b* and Kaixian-luohanmai with rye. Genome 54, 959–964 (2011).
+77. Song, Z. et al. Analysis of structural genomic diversity in *Aegilops umbellulata*, *Ae. markgrafii*, *Ae. comosa*, and *Ae. uniaristata* by fluorescence *in situ* hybridization karyotyping. Front Plant Sci 11, 710 (2020).
+78. Tang, Z. et al. New types of wheat chromosomal structural variations in derivatives of wheat-rye hybrids. PloS One 9, e110282 (2014).
+79. Zhang, A. et al. Molecular cytogenetics identification of a wheat-*Leymus mollis* double disomic addition line with stripe rust resistance. Genome 60, 375–383 (2017).
+80. Liang, Y. et al. Drought-responsive genes, late embryogenesis abundant group3 (*LEA 3*) and vicinal oxygen chelate, function in lipid accumulation in *Brassica napus* and *Arabidopsis* mainly via enhancing photosynthetic efficiency and reducing ROS. Plant Biotechnol J 17, 2123–2142 (2019).
+81. Li, Y. et al. *BRITTLE CULM1*, which encodes a COBRA-like protein, affects the mechanical properties of rice plants. The Plant Cell 15, 2020–2031 (2003).
+82. Wang, J. et al. SAUR17 and SAUR50 differentially regulate PP2C-D1 during apical hook development and cotyledon opening in Arabidopsis. Plant Cell 32, 3792–3811 (2020).
+83. Li, C. et al. Expanded base editing in rice and wheat using a Cas9-adenosine deaminase fusion. Genome Biol 19, 1–9 (2018).
+84. Shan, Q., Wang, Y., Li, J. & Gao, C. Genome editing in rice and wheat using the CRISPR/Cas system. Nat Protoc 9, 2395–2410 (2014).
+
+# Supplementary Files
+
+- [rs3.pdf](https://assets-eu.researchsquare.com/files/rs-2944166/v1/3aff61b1e60915c56b324bc4.pdf)  
+  Reporting Summary
+
+- [Lr47SupplementaryFiguresandTables.pdf](https://assets-eu.researchsquare.com/files/rs-2944166/v1/595fc8c716c88b63ccf8678b.pdf)
+
+- [SupplementaryTable5.xlsx](https://assets-eu.researchsquare.com/files/rs-2944166/v1/14e997617b47e6bf63024df8.xlsx)
+
+- [SupplementaryTable11.xlsx](https://assets-eu.researchsquare.com/files/rs-2944166/v1/0f86e1046a6088424ad677a6.xlsx)

0296ee99dbe763ceaabec7ea58acb94c6ac7108f8403e7a567d55fccc0d3b8c9/preprint/images_list.json

@@ -0,0 +1,74 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "A) Depiction of MSAs generated by AF2 and the paired version matched using organism information. Both AF and paired representations are sections containing 10% of the sequences aligned in the original MSA. Concatenated chains are separated by a vertical line (magenta). The visualisations were made using Jalview version 2.11.1.441 B) Docking visualisations for PDB ID 5D1M with the model/native chains A in blue/grey and B in green/magenta using the three different MSAs in A. The DockQ scores are 0.01, 0.02 and 0.90 for AF2, paired, and AF2+paired MSAs, respectively.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Distribution of DockQ scores as boxplots for different modelling strategies on the test set. The boxes encompass the quartiles of the data, while the notches and horizontal lines mark the medians. The success rates (SR) and medians (M) are reported below the name of each method. All AF2 models have been run with the same neural network configuration (m1-10-1). Outlier points are not displayed here. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "A) ROC curve as a function of different metrics for the development dataset (first run). C\u03b2s within 8 \u00c5 from each other from different chains are used to define the interface. B) Impact of different initialisations on the modelling outcome in terms of DockQ score on the development dataset. The maximal and minimal scores are plotted against the top-ranked models using the average plDDT in the interface for the AF2+paired MSAs, m1-10-1. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "A) Distribution of DockQ scores for three sets of interfaces with the majority of Helix, Sheet and Coil secondary structures. B) Distribution of DockQ scores for tertiles derived from the distribution of contact counts in docking model interfaces. C) Distribution of DockQ scores for tertiles derived from the distribution of Paired MSAs Neff scores. D) Distribution of DockQ scores for the top three organisms Homo Sapiens, S. cerevisiae and E. coli. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Predicted and native structures from the set of novel proteins without templates. The native structures are represented as grey ribbons A) Docking of 7EIV chains A (blue) and C (green) (DockQ=0.76). B) Docking of 7MEZ chains A (blue) and B (green) (DockQ=0.53). C) Prediction of structure 7EL1 chains A (blue) and E (green) (DockQ=0.01). The DNA going through chain A is coloured in orange. D) Docking of 7LF7 chains A (blue) and M (magenta) (DockQ=0.02) and chains B (green) and M (magenta) (DockQ=0.02).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "A) The ROC curve as a function of different metrics for discriminating between interacting and non-interacting proteins. B) Distribution of the top three discriminating features for interacting (coloured) and non-interacting proteins (grey). ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/[IMAGE_MATERIAL_AND_METHODS_1].png",
+        "caption": "",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/[IMAGE_MATERIAL_AND_METHODS_2].png",
+        "caption": "",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/[IMAGE_MATERIAL_AND_METHODS_3].png",
+        "caption": "",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0296ee99dbe763ceaabec7ea58acb94c6ac7108f8403e7a567d55fccc0d3b8c9/preprint/preprint.md

@@ -0,0 +1,327 @@
+# Abstract
+
+Predicting the structure of interacting protein chains is fundamental for understanding the function of proteins. Here, we examine the use of AlphaFold2 (AF2) for predicting the structure of heterodimeric protein complexes. We find that using the default AF2 protocol, 44% of the models in a test set can be predicted accurately. However, by optimising the multiple sequence alignment, we can increase the accuracy to 59%. In comparison, the alternative fold-and-dock method RoseTTAFold is only successful in 10% of the cases on this set, template-based docking 35% and traditional docking methods 22%. We can distinguish acceptable (DockQ>0.23) from incorrect models with an AUC of 0.85 on the test set by analysing the predicted interfaces. The success is higher for bacterial protein pairs, pairs with large interaction areas consisting of helices or sheets, and many homologous sequences. Further, we test the possibility to distinguish interacting from non-interacting proteins and find that by analysing the predicted interfaces, we can separate truly interacting from non-interacting proteins with an AUC of 0.82 in the ROC curve, compared to 0.76 with a recently published method. In addition, when using a more realistic negative set, including mammalian proteins, the identification rate remains (AUC=0.83), resulting in that 27% of interactions can be identified at a 1% FPR. All scripts and tools to run our protocol are freely available at: https://gitlab.com/ElofssonLab/FoldDock.
+
+General Biochemistry  
+Structural Biology  
+Computational Biology  
+Bioinformatics  
+heterodimeric protein complexes  
+interacting protein chains  
+AlphaFold2
+
+# Introduction
+
+Protein-protein interactions are central mediators in biological processes. Most interactions are governed by the three-dimensional arrangement and the dynamics of the interacting proteins<sup>1</sup>. Such interactions vary from being permanent to transient<sup>2, 3</sup>. Some protein-protein interactions are specific for a pair of proteins, while some proteins are promiscuous and interact with many partners. This complexity of interactions is a challenge both for experimental and computational methods.
+
+Often, studies of protein-protein interactions can be divided into two categories, the identification of what proteins interact and the identification of how they interact. Although these problems are distinguished, some methods have been applied to both problems<sup>4, 5</sup>. Protein docking methodologies refer to how proteins interact and can be divided into two categories; those based on shape complementarity<sup>6</sup> and those based on alignments (both sequence and structure) to structural templates<sup>7</sup>. Shape complementary approaches rely on protein structures or models of the monomers<sup>8, 9</sup>, while template-based docking needs suitable templates. However, flexibility has often to be considered in protein docking to account for interaction-induced structural rearrangements<sup>10, 11</sup>. Therefore, flexibility limits the accuracy achievable by rigid-body docking<sup>12</sup>, and flexible docking is traditionally too slow and inaccurate for large scale applications.
+
+Regardless of different strategies, docking remains a challenging problem. In the CASP13-CAPRI experiments, human group predictors achieved up to 50% success rate for top-ranked docking solutions<sup>13</sup>. Alternatively, a recent benchmark study<sup>8</sup> reports success rates of different web-servers reaching up to 16% on the well known Benchmark 5 dataset<sup>14</sup>.
+
+Recently, in the CASP14 experiment, AlphaFold2 (AF2) reached an unprecedented performance level in structure prediction of single-chain proteins<sup>15</sup>. Thanks to an advanced deep learning model that efficiently utilises evolutionary and structural information, this method consistently outperformed all competitors, reaching an average GDT_TS score of 90<sup>15</sup>. Recently, RoseTTAFold was developed, trying to implement similar principles<sup>16</sup>. Since then, other end-to-end structure predictors have emerged using different principles such as fast MSA processing in DMPFold2<sup>17</sup> and language model representations<sup>18</sup>.
+
+As an alternative to other docking methods it is possible to utilise co-evolution to predict the interaction between two protein chains. Initially, direct coupling analysis was used to predict the interaction of bacterial two-component signalling proteins<sup>19, 20</sup>. Later, these methods were improved using machine learning<sup>21</sup>.
+
+In a Fold and Dock approach, two proteins are folded and docked simultaneously. We recently developed a Fold and Dock pipeline using another distance prediction method focused on protein folding (trRosetta<sup>22</sup>). In this pipeline, the interaction between two chains from a heterodimeric protein complex and their structures were predicted using distance and angle constraints from trRosetta<sup>23, 24</sup>. This study demonstrated that a pipeline focused on intra-chain structural feature extraction can be successfully extended to derive inter-chain features as well. Still, only 7% of the tested proteins were successfully folded and docked.
+
+In that study, we found that generating the optimal MSA is crucial for obtaining accurate Fold and Dock solutions, but this is not always trivial due to the necessity to identify the exact set of interacting protein pairs<sup>25</sup>, see Figure 1. Given the existence of multiple paralogs for most eukaryotic proteins, this is difficult. We also found that this process requires an optimal MSA depth to optimise inter-chain information extraction. Too deep MSAs might contain false positives (i.e. protein pairs that interact differently), resulting in noise masking the sought after co-evolutionary signal, while too shallow alignments do not provide sufficient co-evolutionary signals.
+
+We systematically applied the AF2 pipeline on two different datasets to Fold and Dock protein-protein pairs simultaneously. We explore the docking success using the MSAs generated by AF2 and combine them with MSAs paired on the organism level to study the dependence of AF2 on the input MSAs. We find that the results in terms of successful docking using AF2 are superior to all other docking methods. In addition, we analyse the ability to distinguish truly interacting from non-interacting proteins using the created pipeline. AF2 outperforms a recent state-of-the-art method<sup>26</sup> developed using the same data at this task as well.
+
+# Material And Methods
+
+## Data
+
+### Development set
+
+A set of heterodimeric complexes from Dockground benchmark 4 <a href="https://paperpile.com/c/8zYQoL/0NQw"><sup>27</sup></a> is used to develop the pipeline, focusing on the AF2 configuration presented here. This set contains protein pairs, with each chain having at least 50 residues, sharing less than 30% sequence identity and no crystal packing artefacts. There are 219 protein interactions for which both unbound (single-chain) and bound (interacting chains) structures are available. Unbound chains share at least 97% sequence identity with the bound counterpart and, to facilitate comparisons, non-matching residues are deleted and renumbered to become identical to the unbound counterpart. AF2 MSAs could not be generated for three of the complexes due to memory limitations (1gg2, 2nqd and 2xwb) using a computational node with 128 Gb RAM for the MSA generation and were thus disregarded, resulting in a total of 216 complexes.  The dataset consists of 54% Eukaryotic proteins, 38% Bacterial and 8% from mixed kingdoms,  e.g. one bacterial protein interacting with one eukaryotic.
+
+### Test set
+
+We used 1,661 protein complexes with known interfaces from a recent study <a href="https://paperpile.com/c/8zYQoL/1kCw"><sup>26</sup></a> to test the developed pipeline. Here, three large biological assemblies were excluded. These complexes share less than 30% sequence identity, have a resolution between 1-5 Å and constitute unique pairs of PFAM domains (no single protein pair have PFAM domains matching that of any other pair). Some structures failed to be modelled for various reasons (see limitations of data generation), resulting in a total of 1481 structures. These proteins are mainly from H. Sapiens (25%), <em>S. Cerevisiae</em> (10%), <em>E.coli</em> (5%) and other Eukarya (30%).
+
+107 of the complexes in the test set lack beta carbons (Cβs), and 50 have overlapping PDB codes with the development set and were therefore excluded. In the MSA generation from AF2, 20 MSAs report MergeMasterSlave errors regarding discrepancies in the number of match states, resulting in a total of 1484 AF2 MSAs. When folding, three of these (5AWF_D-5AWF_B, 2ZXE_B-2ZXE_A and 2ZXE_A-2ZXE_G) report “ValueError: Cannot create a tensor proto whose content is larger than 2GB”, leading to a final set of 1481 complexes. DSSP could only be run successfully for 1391 out of the 1481 protein complexes, and we ignored the rest in the analysis.
+
+For RF, 26 complexes produced out of memory exceptions during prediction using a GPU with 40 Gb RAM and were excluded from the RF analyses, leaving 1455 complexes.
+
+For the mammalian proteins from Negatome, seven out of 1733 single chains were redundant according to Uniprot (C4ZQ83, I0LJR4, I0LL25, K4CRX6, P62988, Q8NI70, Q8T3B2), 34 had no matching species in the MSA pairing, 106 produced out of memory exceptions during prediction using a GPU with 40 Gb RAM, 35 gave a tensor reshape error, and 65 complexes were homodimers, leaving 1715 complexes for this set.
+
+### CASP14 set and novel protein complexes
+
+As an additional test set, we used a set of six heterodimers from the CASP14 experiment. In addition, we extracted eight novel protein complexes deposited in PDB after 15 June 2021, which produced no results for at least one chain in each complex when submitted to the HHPRED web server (version 01-09-2021) <a href="https://paperpile.com/c/8zYQoL/KtBD+0HAd"><sup>28,29</sup></a>, see Table S1. We selected this small set to test the performance on data AF2 is guaranteed not to have seen.
+
+### Non-interacting proteins
+
+Two datasets of known non-interacting proteins were used, one from the same study as the positive test set <a href="https://paperpile.com/c/8zYQoL/1kCw"><sup>26</sup></a>. Here, all proteins are from <em>E.coli.</em> Two methods were used to identify non-interacting proteins, first a set of proteins with no reported interaction signal in Yeast Two-Hybrid Experiments <a href="https://paperpile.com/c/8zYQoL/SAWC"><sup>30</sup></a> and secondly complexes whose individual proteins were found in different APMS benchmark complexes <a href="https://paperpile.com/c/8zYQoL/O90B"><sup>31</sup></a>. This dataset contains in total 3989 non-interacting pairs.
+
+The second set contains 1964 unique mammalian protein complexes filtered against the IntAct <a href="https://paperpile.com/c/8zYQoL/TUhs"><sup>32</sup></a> dataset from Negatome <a href="https://paperpile.com/c/8zYQoL/xpqV"><sup>33</sup></a>. This data deemed “the manual stringent set” contains proteins annotated from the literature with experimental support describing the lack of protein interaction. Some structures in this dataset are homodimers (65) and are therefore excluded, resulting in 1705 structures. Together there are 5694 non-interacting protein complexes.
+
+### Methods to generate MSAs
+
+#### AlphaFold2 default methodology
+
+The input to AlphaFold2 (AF2) consists of several MSAs. We used the AF2 MSA generation <a href="https://paperpile.com/c/8zYQoL/2GbD"><sup>15</sup></a>, which builds three different MSAs generated by searching the Big Fantastic Database <a href="https://paperpile.com/c/8zYQoL/yGBM"><sup>34</sup></a> (BFD) with HHBlits <a href="https://paperpile.com/c/8zYQoL/uvEk"><sup>35</sup></a> (from hh-suite v.3.0-beta.3 version 14/07/2017) and both MGnify v.2018_12 <a href="https://paperpile.com/c/8zYQoL/ju8S"><sup>36</sup></a> and Uniref90 v.2020_01 <a href="https://paperpile.com/c/8zYQoL/p0HI"><sup>37</sup></a> with jackhmmer from HMMER3 <a href="https://paperpile.com/c/8zYQoL/rxBf"><sup>38</sup></a>. The AF2 MSAs were generated by supplying a concatenated protein sequence of the entire complex to the AF2 MSA generating pipeline in FASTA format. The resulting MSAs will thus mainly contain gaps for one of the two query proteins in each row, as only single chains can obtain hits in the searched databases (Figure 1). No trimming or gap removal was performed on these MSAs.
+
+#### Fused HHblits MSAs
+
+In addition to the default AF2 MSA, we generated an additional MSA by simply “fusing” MSAs generated independently from each of the two chains. These MSAs were constructed by running HHblits <a href="https://paperpile.com/c/8zYQoL/uvEk"><sup>35</sup></a> version 3.1.0 against uniclust30_2018_08 <a href="https://paperpile.com/c/8zYQoL/5Ra2"><sup>39</sup></a> with these options:
+
+hhblits -E 0.001 -all -oa3m -n 2
+
+The “fusing” is done by writing gaps for the length of the interacting chain for each sequence in both individual chain MSAs.
+
+#### Paired MSAs
+
+In addition to the fused MSAs, we used a “paired MSA”, constructed using organism information, as described before <a href="https://paperpile.com/c/8zYQoL/O5iz+qFPB+8sBd"><sup>4,20,23</sup></a> (Figure 1). The rationale behind using a paired MSA is to identify inter-chain coevolutionary information. An unpaired MSA has a limited inter-chain signal since the chains are treated in isolation (Figure 1).
+
+The organism information was, using the OX identifier, was extracted from the two HHblits MSAs <a href="https://paperpile.com/c/8zYQoL/lVC3"><sup>40</sup></a>. Next, all hits with more than 90% gaps were removed. From all remaining hits in the two MSAs, the highest-ranked hit from one organism was paired with the highest-ranked hit of the interacting chain from the same organism. Pairing the correct sequences should result in MSAs containing inter-chain co-evolutionary information <a href="https://paperpile.com/c/8zYQoL/1kCw"><sup>26</sup></a>.
+
+### Number of effective sequences (Neff)
+
+To estimate the information in each MSA, we calculated the Neff score by clustering sequences at 62% identity, as used in a previous study <a href="https://paperpile.com/c/8zYQoL/JQPJ"><sup>42</sup></a>. Unaligned FASTA sequences were extracted from the three AF2 default MSAs. Obtained sequences were processed with the CD-HIT software <a href="https://paperpile.com/c/8zYQoL/DAas"><sup>43</sup></a> version 4.7 (<a href="http://weizhong-lab.ucsd.edu/cd-hit/">http://weizhong-lab.ucsd.edu/cd-hit/</a>) using the options:
+
+-c 0.62 -G 0 -n 3 -aS 0.9
+
+We calculated the Neff scores separately for the paired and the AF2 MSAs.
+
+## Prediction of protein-protein complexes
+
+### AlphaFold2
+
+We modelled complexes using AlphaFold2 <a href="https://paperpile.com/c/8zYQoL/2GbD"><sup>15</sup></a> (AF2) by modifying the script <a href="https://github.com/deepmind/alphafold/blob/main/run_alphafold.py">https://github.com/deepmind/alphafold/blob/main/run_alphafold.py</a> to insert a chain break of 200 residues - as suggested in the development of RoseTTAFold <a href="https://paperpile.com/c/8zYQoL/4Uvt"><sup>16</sup></a> (RF). During modelling, relaxation was turned off, and only the atoms generated in RF (N, CA, C) were used in subsequent analyses. Sidechains were thus not used to score interfaces. We note that performing model relaxation did not increase performance in the AF2 paper <a href="https://paperpile.com/c/8zYQoL/2GbD"><sup>15</sup></a> and was, therefore, ignored to save computational cost. No templates were used to build structures, as this would not assess the prediction accuracy of unknown structures or structures without sufficient matching templates. Further, AF2 has been shown to perform well for single chains without templates and has reported higher accuracy than template-based methods even when robust templates are available <a href="https://paperpile.com/c/8zYQoL/2GbD"><sup>15</sup></a>.
+
+We supplied three different types of MSAs to AF2: the MSAs generated by using the default AF2 settings, the top paired MSAs constructed using HHblits, described above, and finally, a concatenation of these both alignments. AF2 was run with two different network models, AF2 model_1 (used in CASP14) and AF2 model_1_ptm, for each MSA. The second model, model_1_ptm, is a fine-tuned version of model_1 that predicts the TMscore <a href="https://paperpile.com/c/8zYQoL/QHoV"><sup>44</sup></a> and alignment errors <a href="https://paperpile.com/c/8zYQoL/2GbD"><sup>15</sup></a>. We ran these two different models by using two different configurations. The configurations utilise a varying amount of recycles and ensemble structures. Recycle refers to the number of times iterative refinement is applied by feeding the intermediate outputs recursively back into the same neural network modules. At each recycling, the MSAs are resampled, allowing for new information to be passed through the network. The number of ensembles refers to how many times information is passed through the neural network before it is averaged <a href="https://paperpile.com/c/8zYQoL/2GbD"><sup>15</sup></a>. The two configurations used are; the CASP14 configuration (three recycles, eight ensembles) and an increased number of recycles (ten) but only one ensemble.
+
+Since structure prediction with AF2 is a non-deterministic process, we generate five models initiated with different seeds. To save computational cost, this was only performed for the best modelling strategy. We rank the five models for each complex by the number of residues in the interface, giving the best result.
+
+### RoseTTAFold
+
+For comparison, the RoseTTAFold (RF) end-to-end version <a href="https://paperpile.com/c/8zYQoL/4Uvt"><sup>16</sup></a> was run using the paired MSAs with the top hits. The RoseTTAFold pipeline for complex modelling only generates MSAs for bacterial protein complexes, while the proteins in our test set are mainly Eukaryotic. Therefore, we use the paired alignments here. We compare RF with AF2 using the same inputs (the paired MSAs) for both the development and test datasets to provide a more fair comparison, as AF2 searches many different databases to obtain as much evolutionary information as possible when generating its MSAs. To predict the complexes, we use the “chain break modelling” as suggested in RF ( <a href="https://github.com/RosettaCommons/RoseTTAFold/tree/main/example/complex_modeling">https://github.com/RosettaCommons/RoseTTAFold/tree/main/example/complex_modeling</a> ) using the following command:
+
+predict_complex.py -i msa.a3m -o complex -Ls chain1_length chain2_length
+
+### GRAMM
+
+For comparison, a rigid-body docking method, GRAMM <a href="https://paperpile.com/c/8zYQoL/e5Nd"><sup>45</sup></a>, was used. Here, two protein models are docked using a Fast Fourier Transform (FFT) procedure to generate 340’000 docking poses for each complex. The bound structures extracted from complexes in the test set were used as inputs. This docking generation stage mainly considers the geometric surface properties of the two interacting structures, allowing minor clashes to leave some space for conformational flexibility adjustment. As the bound form of the proteins is used, this should represent an easy case for GRAMM based docking, and performance drops significantly when unbound structures or models are used <a href="https://paperpile.com/c/8zYQoL/c7WL"><sup>46</sup></a>. The atom-atom contact energy AACE18 is used to score and rank all poses, as this has been shown to provide better results than shape-complementarity alone <a href="https://paperpile.com/c/8zYQoL/Mi5F"><sup>47</sup></a>.
+
+### Template-based docking
+
+For comparison, a template-based docking protocol <a href="https://paperpile.com/c/8zYQoL/W02n"><sup>7</sup></a> referred to as “TMdock’’ is also adopted. The adopted template library includes 11756 protein complexes obtained from the Dockground database <a href="https://paperpile.com/c/8zYQoL/0NQw"><sup>27</sup></a> (release 28-10-2020). Target complexes are structurally aligned with the supplied template structures (depleted of the target structure PDB ID). TM-scores resulting from the alignment of target proteins to each template are averaged and used to score obtained docking models. Alternatively, we refer to “TMdock Interfaces” when targets are structurally aligned only to the template interfaces, defined as every residue with a Cβ atom closer than 12 Å from any Cβ atom in the other chain.
+
+### Scoring
+
+The backbone atoms (N, CA and C) were extracted from the predicted AF2 structures (as these are the only predicted atoms in the end-to-end version of RF). The interface scoring program DockQ <a href="https://paperpile.com/c/8zYQoL/QXiB"><sup>48</sup></a> was then run to compare the predicted and actual interfaces. This program compares interfaces using a combination of three different CAPRI <a href="https://paperpile.com/c/8zYQoL/gmik"><sup>49</sup></a> quality measures (F<sub>nat</sub>, LRMS, and iRMS) converted to a continuous scale, where an acceptable model comprises a DockQ score of at least 0.23.
+
+### Ranking and scoring models
+
+To analyse the ability of AF2 to distinguish correct models as well as interacting from non-interacting proteins, we analyse the separation between acceptable and incorrect models as a function of different metrics on the development set: the number of unique interacting residues (Cβs from different chains within 8 Å from each other), the total number of interactions between Cβs from different chains, average predicted lDDT (plDDT) score from AF2 for the interface, the minimum of the average plDDT for both chains and the average plDDT over the whole heterodimer.
+
+We use these metrics as a threshold to build a confusion matrix, where True/False Positives (TP and FP respectively) are correct/incorrect docking models which places above the threshold and False/True Negatives (FN and TN respectively) are correct/incorrect docking models which scores below the threshold. From the built confusion matrix, we derive the True Positive Rate (TPR), False Positive Rate (FPR) defined as:
+
+[IMAGE_MATERIAL_AND_METHODS_1]
+
+Then, we calculate TPR and FPR for each possible value assumed by the set of dockings given a single metric and plot TPR as a function of FPR in order to obtain a Receiver Operating Characteristic (ROC) curve. We compute the Area Under Curve (AUC) for ROC curves obtained for each metric to compare different metrics. The AUC is defined as:
+
+[IMAGE_MATERIAL_AND_METHODS_2]
+
+The TPR and FPR for different thresholds are used to calculate the fraction of models that can be called correct out of all models and the Positive Predictive Value (PPV). The fraction of acceptable and incorrect models are obtained by multiplying the TPR and FPR with the success rate (SR). Multiplying the FPR with the SR results in the False Discovery rate(FDR) and the PPV can be calculated by dividing the fraction of acceptable models by the sum of the acceptable and incorrect models. The PPV, FDR and SR are defined as:
+
+[IMAGE_MATERIAL_AND_METHODS_3]
+
+### Analysis of models
+
+To analyse the possibility of determining when AF2 can model a complex correctly, we analyse the structures and the multiple sequence alignments. We investigated: the Number of effective sequences (Neff), the secondary structure in the interface annotated using DSSP <a href="https://paperpile.com/c/8zYQoL/Zhav"><sup>50</sup></a>, the length of the shortest chain, the number of residues in the interface and the number of contacts in the interface.
+
+DSSP was run on the entire complexes, and the resulting annotations were grouped into three categories; helix (3-turn helix (3<sub>10</sub> helix), 4-turn helix (α helix) and 5-turn helix (π helix)), sheet (extended strand in parallel or antiparallel β-sheet conformation and residues in isolated β-bridges) and loop (residues which are not in any known conformation).
+
+## Computational cost
+
+To compare the computation required for each MSA, we compared the time it took to generate MSAs for three protein pairs (PDB: 4G4S_P-O, 5XJL_A-2 and 5XJL_2-M), using either the fused or AF2 protocol. The tests were performed on a computer using 16 CPU cores from an Intel Xeon E5-2690v4.
+
+Fusing the MSAs took 3 seconds on average per tested complex. It took 7884 seconds for generating the AF2 MSAs, the single-chain searches took 338 seconds on average and the pairing 2 seconds. The pairing and fusing are thereby negligible compared to searching, resulting in a speedup of 24 times for the hhblits searches. In comparison, folding using the m1-10-1 strategy took 191 seconds on average for these pairs.
+
+# Results And Discussion
+
+## Identifying the best AlphaFold2 model
+
+The fraction of acceptable models (DockQ>0.23), the success rate (SR) is used to measure performance for each AF2 model using the different MSAs. The best performance is 32.4% for the AF2 MSAs and 38.4% for the AF2+paired MSAs (Table 1). It is thereby evident that combining both paired and AF2 MSAs is superior to using them separately. The average performance of the AF2 and the paired MSAs is similar, but for individual protein pairs, frequently one of the two MSAs is superior to the other, as seen from that the Pearson correlation coefficient for the DockQ scores between AF2 vs paired MSAs is 0.48 (Table S2). Therefore, combining AF2 and paired MSAs improves the results.
+
+Next, we compared the default AF2 model (model_1) with the fine-tuned versions of (model_1_ptm). Surprisingly, the original AF2 model_1 outperforms AF2 model_1_ptm in most cases (Table 1). Further, the difference between 10 recycles-one ensemble and three recycles-eight ensembles is minor across all MSAs and AF2 models. Therefore, the input information and the AF2 model appears to impact the outcome the most.
+
+### Table 1.
+Results from AF2 using different MSAs and neural network configurations.
+
+| NN model | model_1 | model_1 | model_1_ptm | model_1_ptm |
+| --- | --- | --- | --- | --- |
+| Recycles | 10 | 3 | 10 | 3 |
+| Ensembles | 1 | 8 | 1 | 8 |
+| MSA short name | m1-10-1 | m1-3-8 | mp-10-1 | mp-3-8 |
+| Paired | **26.4** | **26.4** | **26.4** | 24.5 |
+| AF2 | 31.0 | **32.4** | 24.1 | 23.1 |
+| AF2+Paired | **38.4** | 36.6 | 30.6 | 30.1 |
+
+## Test set analysis
+
+### Test set performance
+
+The best model and configuration for AF2 (m1-10-1) was used for further studies on the test set. The best outcome using this modelling strategy results in an SR of 55.9% (828 out of 1481 correctly modelled complexes) for the AF2+paired MSAs compared with 43.9% using the AF2 MSAs alone (Figure 2, Table S3). The results using the fused+paired MSAs are almost identical (SR=56.0%,median=0.302). Further, running five initialisations with random seeds and ranking the models using the average plDDT in the interface increases the SR to 57.8% and 58.7% for the AF2+paired and fused+paired MSAs, respectively (model variation and ranking, Figure 3). Using the combination of AF2 and paired MSAs increases performance, suggests that AF2 gains both from larger and paired MSAs, although it often can manage with less information.
+
+What is most striking is that AF2 outperforms all other methods by a large margin.
+
+RF is better than AF2 only for 14 pairs in the test set, while GRAMM and template-based docking (TMdock interface) outperform AF2 for 188 and 225 pairs, respectively. The reason for GRAMM’s good performance is likely due to the use of the bound form of the proteins, resulting in very high shape complementarity and therefore having the “answer” provided in a way.
+
+### Distinguishing acceptable from incorrect models
+
+It is not only essential to obtain improved predictions but also to be able to identify acceptable predictions. We measure the separation between acceptable and incorrect models using a receiver operating characteristic (ROC) curve. Different criteria were examined, including (i) the number of unique interacting residues (Cβ atoms from different chains within 8 Å from each other) in the interface, (ii) the total number of interactions between Cβ atoms in the interface, (iii) the average plDDT for the interface, (iv) the lowest plDDT of each single chain average, and (v) the average plDDT over the whole protein heterodimer (Figure 3A). Three criteria result in very similar areas under the curve (AUC) measures. The total number of interactions between Cβs and the number of residues in the interface can separate the correct/incorrect models with an AUC of 0.86, while the average interface plDDT results in an AUC of 0.85. However, pLDDT results in higher TPRs at lower FPRs; therefore, it is better for model ranking.
+
+Interestingly, the average plDDT of the entire complex only results in an AUC of 0.68, suggesting that both single chains in a complex are often predicted very well, while their relative orientation is wrong.
+
+### Model variation and ranking
+
+Five models were generated using the best strategy (m1-10-1 with AF2+paired MSAs) with different initialisation (random seeds). The average SR (55.2% ± 0.0%) was similar for all five runs. However, the average deviation for individual models is DockQ=0.08 when comparing the best and worst models for a target (Figure 3B), i.e. there is some randomness to the success for an individual pair. If the maximal DockQ score across all models is used, the SR would be 61.0%. Although this is unachievable, ranking the models using the total number of interactions in the interface results in an SR of 57.8%. The AUC using the average plDDT in the interface for the ranked test set is 0.82, which means that 16% of all models are acceptable at an error rate of 1% and 37% at an error rate of 10% (Table S4).
+
+### Bacterial protein pairs with large interfaces and many homologs are easier to predict
+
+In the test set, about 60% of the complexes can be modelled correctly. We tried to answer what distinguishes the successful and unsuccessful cases by analysing different subsets of the test set. First, we divided the proteins by taxa, interface characteristics, and finally by examining the alignments.
+
+The Success Rates (SRs) for each kingdom is; Eukarya 57%, Bacteria 72%, Archaea 80%, and Virus 55% (Figure S1B). Further, the SRs for *Homo Sapiens* and *S.cerevisiae* are similar (58% vs 59%). The better performance in prokaryotes is consistent with previous observations regarding the availability of evolutionary information in prokaryotes compared to Eukarya (Figure S2A).
+
+Next, we examined the interfaces. First, different secondary structural content of the native interfaces was investigated (Figure 4A). The highest SR is obtained for mainly helix interfaces (62%), followed by interfaces containing mainly sheets (59%). The loop interface SR of 53% is substantially lower than the others, suggesting that interfaces with more flexible structures are harder to predict. We divided the dataset by the size of the interface, and it is clear that pairs with larger interfaces are easier to predict, as the SR increases from 47 to 74% between the smallest and biggest tertiles (Figure 4B).
+
+Next, we examined how the size of the MSA (both paired and AF2) influences the results. It is clear that the fraction of correctly modelled sequences increases with larger MSAs (Figure 4C), and the size of the paired MSA (Figure 4C) has a more considerable influence on the outcome than the size of the AF2 MSA (Figure S1A).
+
+## CASP14 and novel proteins without templates
+
+Chains derived from CASP14 heteromeric targets and chains from PDB complexes with no templates have been folded in pairs using the presented AF2 pipeline (default AF2+paired MSAs, ten recycles, m1-10-1 and five differently seeded runs).
+
+For the CASP14 chains, four out of six pairs display a DockQ score larger than 0.23 (SR of 67%). No ranking is necessary in this case, given that all produced docking models for the same chain pair are very similar (the average standard deviation is 0.01 between each set of DockQ scores). An interesting unsuccessful docking is obtained modelling chains from the complex with PDB ID 6TMM (Figure S3), which are known to form a heterotetramer. In this structure, each chain A is in contact with its partner chain B at two different sites. Both docking configurations (6TMM_A-B and 6TMM_A-D) put the chain in between the two binding sites. The other unsuccessful docking (6VN1_A-H) has an interface of just 19 residue pairs.
+
+The SR for docking the proteins without templates is 50%. Between the five different initialisations, the average difference in the DockQ score is 0.03, and there is no deviation in SR, i.e. ranking did not improve the SR. Two acceptable models are displayed in Figures 5A (7EIV_A-C) and B (7MEZ_A-B). More interesting, in one of the incorrect models (7NJ0_A-C, Figure S4), the predictions get the location of both chains correct, but their orientations wrong, resulting in DockQ scores close to 0. For 7EL1_A-E (Figure 5C), the shorter chain E is not folded correctly, and instead of folding to a defined shape, it is stretched out and inserted within chain A. It occupies the shape of the DNA in the native structure. In the two remaining incorrect models (7LF7_A-M and 7LF7_B-M), Figure 5D, the chains only interact with a short loop of the M chain, making the docking very difficult and possibly biologically meaningless.
+
+## Identifying interacting proteins
+
+Using the best separator from the model ranking the interface plDDT, it is possible to distinguish the 3989 non-interacting proteins from *E.coli* and the truly interacting proteins from the test set with an AUC of 0.82. Another recently published method obtains AUC 0.76 on this set . However, these results are probably overstated since the negative set only contains bacterial proteins, while the positive set is mainly eukaryotic.
+
+To obtain a more realistic estimate, we also include a set of non-interacting proteins from mammalian organisms combined with the non-interacting proteins from *E.coli*. On this set, we obtain an AUC of 0.82 for the average interface plDDT and slightly higher (0.84 and 0.85) for the number of interface contacts and residues, respectively (Figure 6A). Here, the average interface plDDT provides a better separation at low FPRs, enabling a TPR of 27% at FPR of 1% compared to 18 and 13% for the number of interface contacts and residues, respectively. At FPR 5%, the reverse is true, with the number of interface contacts and residues reporting TPRs of 49 and 42%, respectively, compared to 43% for the average interface plDDT. The distribution of the three top separators can be seen in Figure 6B.
+
+## Limitations
+
+Here, we only consider the structures of protein complexes in their heterodimeric state, although each protein chain in these complexes may have homodimer configurations or other higher-order states. It is also possible that the complex itself exists as part of larger biological units, in potentially more complex conformations. Investigating alternative oligomeric states and larger biological assemblies is outside of the scope of this analysis and left for future work.
+
+The study of AF2s ability to separate interacting and non-interacting proteins here contains more extensive data than recent studies . However, to test this separation thoroughly, the data studied here needs to be extended to compare interactions within individual organisms. We leave this extensive analysis to further studies.
+
+# Conclusions
+
+Here we show that AlphaFold2 (AF2) can predict the structure of many heterodimeric protein complexes, although it is trained to predict the structure of individual protein chains. Even using the default settings, it is clear that AF2 is superior to all other docking methods, including other Fold and Dock methods<sup>16,23</sup>, methods based on shape complementarity<sup>45</sup> and template-based docking. Using optimised multiple sequence alignments with AF2, we can accurately predict the structure of heterodimeric complexes for an unprecedented success rate of 59.0% on a large test set. The success rate is higher in E.Coli (75%) than in *Homo Sapiens* or *S. cerevisiae* (58 %).
+
+Further, by examining the average interface plDDT, we can separate acceptable and incorrect models with an AUC of 0.85, resulting in that 14% of the models can be called acceptable at a specificity of 99% (or 38% at 90% specificity). Interestingly, no additional constraints are implemented in AF2 to pull two chains in contact, meaning chain interactions (and subsequently interface sizes) are exclusively determined by the amount of inter-chain signals extracted by the predictor. Assuming that all residues in an interface contribute to the interaction energy could explain why larger interfaces are more likely to be correctly predicted.
+
+We find that the MSA generation process can be sped up substantially at no performance loss by simply fusing MSAs from two HHblits runs on Uniclust30 instead of using the MSAs from AF2. Fast MSA generation circumvents the main computational bottleneck in the pipeline. Analysing the interfaces of predicted complexes makes it possible to separate truly interacting from non-interacting proteins with an AUC of 0.82, making it possible to identify 27% of interacting proteins at an error rate of 1%. Features of the predicted interfaces discriminate between model quality and binary interactions. Therefore the same pipeline can identify if two proteins interact and the accuracy of their structure. Never before has the potential for expanding the known structural understanding of protein interactions been this large, at such a small cost. There are currently 11.9 million pairwise human protein interactions in the String DB<sup>51</sup>. If 14% of these can be predicted at an error rate of 1%, this results in the structure of 1.5 million human heterodimeric protein structures. The computational cost to run all of this would take approximately three months on an Nvidia A100 system.
+
+# References
+
+1. Liddington, R. C. Structural Basis of Protein–Protein Interactions. Protein-Protein Interactions 003–014 doi:10.1385/1-59259-762-9:003.
+
+2. Keskin, O., Gursoy, A., Ma, B. & Nussinov, R. Principles of protein-protein interactions: what are the preferred ways for proteins to interact? Chem. Rev. 108, 1225–1244 (2008).
+
+3. Nooren, I. M. A. NEW EMBO MEMBER’S REVIEW: Diversity of protein-protein interactions. The EMBO Journal vol. 22 3486–3492 (2003).
+
+4. Cong, Q., Anishchenko, I., Ovchinnikov, S. & Baker, D. Protein interaction networks revealed by proteome coevolution. Science 365, 185–189 (2019).
+
+5. Zhang, Q. C. et al. Structure-based prediction of protein-protein interactions on a genome-wide scale. Nature 490, 556–560 (2012).
+
+6. Marshall, G. R. & Vakser, I. A. Protein-Protein Docking Methods. Proteomics and Protein-Protein Interactions 115–146 doi:10.1007/0-387-24532-4_6.
+
+7. Kundrotas, P. J., Zhu, Z., Janin, J. & Vakser, I. A. Templates are available to model nearly all complexes of structurally characterized proteins. Proc. Natl. Acad. Sci. U. S. A. 109, 9438–9441 (2012).
+
+8. Porter, K. A., Desta, I., Kozakov, D. & Vajda, S. What method to use for protein–protein docking? Current Opinion in Structural Biology vol. 55 1–7 (2019).
+
+9. Halperin, I., Ma, B., Wolfson, H. & Nussinov, R. Principles of docking: An overview of search algorithms and a guide to scoring functions. Proteins 47, 409–443 (2002).
+
+10. Clarke, J. Mechanisms of Folding upon Binding. The FASEB Journal vol. 29 (2015).
+
+11. Eginton, C., Naganathan, S. & Beckett, D. Sequence-function relationships in folding upon binding. Protein Science vol. 24 200–211 (2015).
+
+12. Andrusier, N., Mashiach, E., Nussinov, R. & Wolfson, H. J. Principles of flexible protein-protein docking. Proteins 73, 271–289 (2008).
+
+13. Lensink, M. F. et al. Blind prediction of homo- and hetero-protein complexes: The CASP13-CAPRI experiment. Proteins 87, 1200–1221 (2019).
+
+14. Vreven, T. et al. Updates to the Integrated Protein-Protein Interaction Benchmarks: Docking Benchmark Version 5 and Affinity Benchmark Version 2. J. Mol. Biol. 427, 3031–3041 (2015).
+
+15. Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
+
+16. Baek, M. et al. Accurate prediction of protein structures and interactions using a three-track neural network. Science 373, 871–876 (2021).
+
+17. Kandathil, S. M., Greener, J. G., Lau, A. M. & Jones, D. T. Ultrafast end-to-end protein structure prediction enables high-throughput exploration of uncharacterised proteins. doi:10.1101/2020.11.27.401232.
+
+18. Chowdhury, R. et al. Single-sequence protein structure prediction using language models from deep learning. bioRxiv 2021.08.02.454840 (2021) doi:10.1101/2021.08.02.454840.
+
+19. Procaccini, A., Lunt, B., Szurmant, H., Hwa, T. & Weigt, M. Dissecting the Specificity of Protein-Protein Interaction in Bacterial Two-Component Signaling: Orphans and Crosstalks. PLoS ONE vol. 6 e19729 (2011).
+
+20. Weigt, M., White, R. A., Szurmant, H., Hoch, J. A. & Hwa, T. Identification of direct residue contacts in protein-protein interaction by message passing. Proc. Natl. Acad. Sci. U. S. A. 106, 67–72 (2009).
+
+21. Hashemifar, S., Neyshabur, B., Khan, A. A. & Xu, J. Predicting protein–protein interactions through sequence-based deep learning. Bioinformatics vol. 34 i802–i810 (2018).
+
+22. Yang, J. et al. Improved protein structure prediction using predicted inter-residue orientations. doi:10.1101/846279.
+
+23. Pozzati, G. et al. Limits and potential of combined folding and docking using PconsDock. doi:10.1101/2021.06.04.446442.
+
+24. Lamb, J. & Elofsson, A. pyconsFold: a fast and easy tool for modelling and docking using distance predictions. Bioinformatics (2021) doi:10.1093/bioinformatics/btab353.
+
+25. Szurmant, H. & Weigt, M. Inter-residue, inter-protein and inter-family coevolution: bridging the scales. Curr. Opin. Struct. Biol. 50, 26–32 (2018).
+
+26. Green, A. G. et al. Large-scale discovery of protein interactions at residue resolution using co-evolution calculated from genomic sequences. Nat. Commun. 12, 1–12 (2021).
+
+27. Kundrotas, P. J. et al. Dockground: A comprehensive data resource for modeling of protein complexes. Protein Sci. 27, 172–181 (2018).
+
+28. Gabler, F. et al. Protein Sequence Analysis Using the MPI Bioinformatics Toolkit. Curr. Protoc. Bioinformatics 72, e108 (2020).
+
+29. Zimmermann, L. et al. A Completely Reimplemented MPI Bioinformatics Toolkit with a New HHpred Server at its Core. J. Mol. Biol. 430, 2237–2243 (2018).
+
+30. Rajagopala, S. V. et al. The binary protein-protein interaction landscape of Escherichia coli. Nat. Biotechnol. 32, 285–290 (2014).
+
+31. Kuhlbrandt, W. The Resolution Revolution. Science vol. 343 1443–1444 (2014).
+
+32. Orchard, S. et al. The MIntAct project--IntAct as a common curation platform for 11 molecular interaction databases. Nucleic Acids Res. 42, D358–63 (2014).
+
+33. Blohm, P. et al. Negatome 2.0: a database of non-interacting proteins derived by literature mining, manual annotation and protein structure analysis. Nucleic Acids Research vol. 42 D396–D400 (2014).
+
+34. BFD. https://bfd.mmseqs.com/.
+
+35. Steinegger, M. et al. HH-suite3 for fast remote homology detection and deep protein annotation. BMC Bioinformatics 20, 473 (2019).
+
+36. Mitchell, A. L. et al. MGnify: the microbiome analysis resource in 2020. Nucleic Acids Res. 48, D570–D578 (2020).
+
+37. Suzek, B. E., Huang, H., McGarvey, P., Mazumder, R. & Wu, C. H. UniRef: comprehensive and non-redundant UniProt reference clusters. Bioinformatics 23, 1282–1288 (2007).
+
+38. Eddy, S. R. Accelerated Profile HMM Searches. PLoS Computational Biology vol. 7 e1002195 (2011).
+
+39. Mirdita, M. et al. Uniclust databases of clustered and deeply annotated protein sequences and alignments. Nucleic Acids Res. 45, D170–D176 (2017).
+
+40. UniProt Consortium. UniProt: the universal protein knowledgebase in 2021. Nucleic Acids Res. 49, D480–D489 (2021).
+
+41. Waterhouse, A. M., Procter, J. B., Martin, D. M. A., Clamp, M. & Barton, G. J. Jalview Version 2--a multiple sequence alignment editor and analysis workbench. Bioinformatics 25, 1189–1191 (2009).
+
+42. Kosciolek, T. & Jones, D. T. Accurate contact predictions using covariation techniques and machine learning. Proteins 84 Suppl 1, 145–151 (2016).
+
+43. Li, W., Jaroszewski, L. & Godzik, A. Clustering of highly homologous sequences to reduce the size of large protein databases. Bioinformatics 17, 282–283 (2001).
+
+44. Zhang, Y. & Skolnick, J. Scoring function for automated assessment of protein structure template quality. Proteins 57, (2004).
+
+45. Vakser, I. A. Evaluation of GRAMM low-resolution docking methodology on the hemagglutinin-antibody complex. Proteins Suppl 1, 226–230 (1997).
+
+46. Singh, A., Dauzhenka, T., Kundrotas, P. J., Sternberg, M. J. E. & Vakser, I. A. Application of docking methodologies to modeled proteins. Proteins 88, 1180–1188 (2020).
+
+47. Anishchenko, I., Kundrotas, P. J. & Vakser, I. A. Contact Potential for Structure Prediction of Proteins and Protein Complexes from Potts Model. Biophys. J. 115, 809–821 (2018).
+
+48. Basu, S. & Wallner, B. DockQ: A Quality Measure for Protein-Protein Docking Models. PLoS One 11, e0161879 (2016).
+
+49. Lensink, M. F. & Wodak, S. J. Docking and scoring protein interactions: CAPRI 2009. Proteins 78, 3073–3084 (2010).
+
+50. Kabsch, W. & Sander, C. Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers 22, 2577–2637 (1983).
+
+51. Szklarczyk, D. et al. The STRING database in 2021: customizable protein–protein networks, and functional characterization of user-uploaded gene/measurement sets. Nucleic Acids Res. 49, D605–D612 (2020).
+
+# Supplementary Files
+
+- [SupplementaryMaterial.docx](https://assets-eu.researchsquare.com/files/rs-951605/v1/8a6e1c7abe9dca6dbe3bfcfd.docx)

08c99ee7ca75d50af4d6934f96f90513a2dc9b59d917f22d4ce6f57eb222af23/preprint/images_list.json

@@ -0,0 +1,58 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpg",
+        "caption": "Community biogas production and distribution system (CBPD) design for a demand-driven biogas supply. a. Schematic flow diagram of biogas production, adjustment, and supply. Biogas slippage occurs when the temporal redundant value exceeds the storage capacity. b. Planning framework to establish a CBPD for a demand-driven biogas supply. Data on a rural community\u2019s energy consumption is initially collected to estimate the biogas consumption rate. The biogas supply rate is designed to equal the consumption/utilization rate in the community. (I), (II) and (III) are sources of greenhouse gas emission or mitigation. Dotted arrows indicate information flow, and solid arrows indicate biogas flow.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpg",
+        "caption": "Biogas production and consumption curve fitting for the five community biogas production and distribution systems (CBPD) and their greenhouse gas (GHG) emissions under different scenarios. a. Hourly biogas production with a feeding interval of five days. Continuous quality data were selected between Nov. 1\u201320, 2019 to investigate the characteristics of the biogas production curves. Fermentation temperature was 27\u00b0C, and the feeding time was 09:00 am. The different color points represent the data from the corresponding feeding, and Arabic numerals show the average values at each respective time point. The rate of biogas production was relatively steady during the initial stage (hours 0\u201338). A sharp increase was observed during the middle stage (hours 39\u201348). After this peak, it declined gradually until the next feeding point (hours 49\u2013120). b. Customers\u2019 hourly biogas consumption in the five communities during the whole observation period. The range of each box shows the interquartile range (IQR) of the distribution; the horizontal line inside the box shows the median value; and the Tukey whiskers extend to the farthest points of the distribution that are not outliers (i.e., those more than 1.5\u2009\u00d7\u2009IQR from the edge of the box). The outliers are denoted by asterisks. c. Six operational strategies to satisfy biogas demand over time (colored bubble plots). The bubble size represents the energy-related and biogas leakage-related GHG emissions for various CBPD operational parameters. The consumption-to-production ratio (CPR) of the current strategy is 67.5% based on fitting the curves, and represents the baseline condition, shown in black. Both accurate operational training for feed optimization on the plant-side and the amount of storage capacity could provide effective means of increasing the CPR. Operational training for feed point optimization alone could further raise customer consumption on the user-side, shown in light blue. In addition, two scenarios with lowered substrate loading on the plant-side to achieve CPR = 1 are shown: i) storage capacity optimization, shown in red; and ii) a combination of storage capacity optimization and operational training, shown in green. Furthermore, two scenarios with raised customer consumption on the user-side to achieve CPR = 1 are shown: i) storage capacity optimization, shown in blue; and ii) a combination of storage capacity optimization and operational training, shown in pink. The number at the top of each circle shows the value of the GHG emissions.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpg",
+        "caption": "Ten cities\u2019 outdoor solar-air temperatures and the decarbonization contributions for the upgraded community biogas production and distribution system (CBPD) deployment. The bulb temperatures and solar radiation intensities recorded in 2015 (for detailed calculations see Supplementary Note 3) were used to calculate daily outdoor solar-air temperature and upgraded CBPD GHG emissions, the average annual value being taken as representative. Under zero biogas loss scenarios, the amount of biogas supplied was equal to the biogas produced minus the heat used to maintain fermentation temperature. The calorific value of biogas containing 60% methane was 21.54 MJ/m3, and the heat efficiency of biogas conversion and heat exchange was 70% 28.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpg",
+        "caption": "The availability of animal manure for the national deployment of community biogas production and distribution systems (CBPD) in the 31 Chinese mainland provinces and its potential to achieve greenhouse gas (GHG) emission targets. a. The availability of manure (pigs, chickens, and cattle) in each province (omitting Taiwan, Macao, and Hong Kong). The detailed values of rate of methane production potential from manure and domestic gas demand in the rural community (RPD) see Supplementary Table 4. b. Nationally upgraded CBPD deployment for carbon mitigation. Without a biogas collection and utilization system, methane conversion factors were assumed to be 0.2\u20130.8 for different manure treatment processes 10; the mean value of 0.5 was assumed as the baseline scenario. Chengdu city\u2019s climate parameters were taken as the overall average value for China, with 11.3% of the energy produced in biogas generation being used to maintain the CBPD fermentation temperature. The dot-dash line represents the calculated result based on these assumed baseline values. The pink area shows the possible reduction in GHG emissions, with the upper and lower limits based on a methane conversion factor of 0.8 and 0.2 for manure treatment, respectively. The red line represents the IPCC 1.5\u00b0C global warming target for methane emission reduction of the nationwide manure treatment, resulting from a 45% methane reduction. The blue line represents the carbon neutralization of national manure treatment. The green line shows the carbon mitigation contribution of upgraded CBPD deployment rates to meet the Chinese 1.5\u00b0C global warming target.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/[IMAGE_METHODS_1].png",
+        "caption": "",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/[IMAGE_METHODS_2].png",
+        "caption": "",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/[IMAGE_METHODS_3].png",
+        "caption": "",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

08c99ee7ca75d50af4d6934f96f90513a2dc9b59d917f22d4ce6f57eb222af23/preprint/preprint.md

@@ -0,0 +1,214 @@
+# Abstract
+
+On-site conversion of organic waste into biogas to satisfy consumer demand for direct primary energy usage has the potential to mitigate climate change in a highly trustworthy manner. Existing approaches usually ignore either the targeted supply of biogas in dynamic situations or methane slippage, especially where the goal is to achieve energy equality in developing areas, which has become the main obstacle to achieving the full co-benefits of organic waste conversion. Here, we describe an upgraded community biogas production and distribution system (CBPD) to achieve a biogas consumption-to-production ratio of close to 1 in rural or remote areas, compared with the actual performance of five current CBPD systems. Improvements in the practical operation of CBPD systems are proposed to better align out-of-step biogas flow rates on the plant-side with user-side demand. We also demonstrate that upgraded CBPDs can achieve universally high and competitive benefits under the prevailing climate conditions, and that national deployment of such systems in China would contribute a 3.77% reduction in carbon emissions towards meeting the 1.5°C global warming target set by the Paris Agreement.
+
+Earth and environmental sciences/Environmental social sciences/Climate-change mitigation  
+Scientific community and society/Energy and society/Energy supply and demand  
+Scientific community and society/Developing world
+
+# Introduction
+
+Under many scenarios, fossil fuels will be the dominant energy source until at least 2050<sup>1</sup>, and their growing use is bringing increased carbon emissions due to the lack of convenient, reliable, and renewable energy supply systems<sup>2</sup>. To address this problem, the use of biogas (55–65% methane content) as a fossil fuel replacement presents a viable emission reduction strategy, based on its waste-to-energy-produced and methane control characteristics<sup>3</sup>. Besides, biogas also has the potential to deliver the same energy-return-on-investment ratio as fossil gas, considering the ecological cost<sup>1</sup>.
+
+Although biogas and biogas-based natural gas generation have a long history of use for cooking, heating, and power<sup>4</sup>, their usage is relatively low in developing rural areas<sup>5,6</sup>, and the climate benefits are limited. That is because the current designs of conventional biogas systems often fail to realize energy equality and mitigate climate change, owing to their intermittent energy supply and methane slippage. Taking China as an example, approximately 800 million people, most of whom live in rural areas, have no access to natural gas or biogas<sup>7</sup>. In spite of this, the methane production potential of the available manure and crop straw is 73.6 billion m<sup>3</sup>/yr, which is several times greater than the amount needed to cover civil gas demands<sup>8</sup>. On the other hand, the current methane emissions due to Chinese animal waste treatment alone could reach 6.4×10<sup>7</sup> tones CO<sub>2</sub>-eq/yr<sup>9</sup>. The need to control the methane generated by organic waste management, has led to the recognition of high quality biogas systems as the most efficient strategy<sup>10</sup>.
+
+The best kinds of biogas system and how to deploy them in developing areas has become a hot topic<sup>11</sup>. On-site biogas generation and direct supply to consumers would result in significant advantages over the traditional means of exploitation, conversion and distribution<sup>12</sup>. They could make a much greater contribution to climate change mitigation than biogas conversion/utilization using centralized distribution grids, where the energy produced is delivered to the electricity grid or gas distribution network<sup>13</sup>. Thus, on-demand biogas supply and close-to-zero methane leakages are prerequisites to achieving the best co-benefits for achieving energy equality and greenhouse gas (GHG) emission mitigation<sup>14</sup>.
+
+A community biogas production and distribution system (CBPD) for improving organic matter availability and alleviating clean fuel poverty in rural areas represents the best feasible strategy over a large range of applications<sup>15</sup>, providing a timely, adjustable, and flexible biogas supply flow<sup>16</sup>. In so far as possible, the amount of biogas consumed on the user-side should be equal or close to that of biogas production on the plant-side, resulting in a consumption-to-production ratio (CPR) of 1. A higher CPR indicates not only less methane discharged into the atmosphere<sup>17</sup>, but also a greater proportion of methane collected and utilized to substitute fossil fuels. Sufficient feedstock for biogas conversion and an adequate number of consumers are the prerequisite factors for achieving the highest cost-benefit ratio<sup>18</sup>. In this case, the proposed CBPD would provide dominant advantages over conventional fossil fuel systems<sup>19</sup>, omitting the complicated chain of primary energy exploitation, transformation, distribution, commercial energy production, and final energy use<sup>20</sup>.
+
+To realize this goal, the best CBPD designs and their corresponding operational strategies are essential to establish complete, self-contained units, which can operate independently without external assistance, and provide a feasible and efficient innovation compared with conventional generalized infrastructures<sup>21</sup>. Based on a case study in China, we therefore propose an upgraded CBPD for maximum co-benefits and broad deployment, to provide a concrete example of how dynamically-variable conflicts between biogas flows on the user- and plant-sides can influence a system’s performance. The principal ingredients include fitted biogas supply rate curves based on user-side consumption characteristics, building self-adjusting platforms, and modulating the biogas flow using simple and practical management strategies. The particular research questions addressed here are what should the general form of an upgraded CBDP be, and how can an upgraded CBPD provide the maximum contribution to carbon mitigation. In addition, we also show the feasibility of universal deployment, and the high worth of Chinese national deployment towards meeting China’s 1.5°C global warming target.
+
+# CBPD optimization for demand-dependent biogas production and distribution
+
+The proposed general form of a self-adjusting biogas flow between the biogas plant- and user-sides is shown in Fig. 1. Biogas flow planning is categorized into four analytical steps: data-driven identification of biogas demand rate; quantification of the operational parameters of biogas production; design of the temporal biogas storage capacity to avoid discontinuous supply and biogas slippage; and determination of the operational strategy for coordinating biogas flows on the plant- and user-sides.
+
+Figure 1  
+∣ Community biogas production and distribution system (CBPD) design for a demand-driven biogas supply.
+
+In the first step, the biogas consumption rate curve is simulated to conceptualize the production rate and buffer system. The status-quo characteristics are identified based on analysis of the communities’ usage data, incorporating both routine activities, such as cooking, and intermittent activities, such as heating in cold weather. The biogas users include households, restaurants, local businesses, and so on. Once customers completely understand the characteristics of CBPD, the systematic benefits can also be appreciated to increase their consumption of biogas, i.e., biogas is a prior good<sup>22</sup>, and carbon tax free. Consequently, biogas usage will grow further, both temporally and spatially, as time goes on.
+
+In the second step, the amount of biogas production is quantified according to the commonly accepted parameters for fermentation temperature and organic load rate (OLR)<sup>23</sup>. The approximate shape of the biogas production curve can be calculated using variables related to the reactor configuration and operational parameters, and existing mathematical models or improved equations derived from fitting related operational data<sup>17</sup>. Accurate assessment of the amplitude and duration of biogas production rates depend on actual data-driven modeling and step four, since the feeding interval and feeding time points can play essential roles in determining the preferred pairing of set-points on the plant- and user-sides<sup>24</sup>.
+
+In the third step, biogas storage facilities are built as a buffer system to store excess biogas output to meet biogas production shortages in times of various conflict scenarios. The determination of biogas storage capacity is based on precise estimates of biogas consumption, accurate feeding process controls, and efficient margin design. Among these, margin design should pay adequate attention to the various operational scenarios key to achieving the optimal CPR value of 1, including factors relating to the feeding process, deviations in fermentation temperature, irregular changes in temporal and quantitative biogas usage, and so on.
+
+In the fourth step, a management scheme is determined to formulate the scheduled biogas flows within an allowable fluctuation range<sup>17</sup>. On the plant-side, the detailed operational strategy should be actively adjusted, based on the site-specific data-driven analysis. On the user-side, the quality of the biogas supply service will influence individuals’ willingness to use biogas and the amount used. As the patterns of production or consumption will inevitably vary to some degree<sup>25</sup>, we argue that additional flexibility in biogas flow, such as biogas storage optimization and professional operation, is also necessary to avoid unwanted biogas emissions.
+
+# Biogas flow fitting and the carbon mitigation potential
+
+The status-quo for CBPD operation is analyzed before applying the proposed general application (Fig. 1). We present five actual CBPD projects in Chinese rural areas (see Supplementary Table 1 for detailed parameters), which supply biogas directly for household usage¹⁷. Their operational performance is used to provide insights into how to further realize the expected carbon mitigation potential. The biogas production capacities were designed according to the number of customers in each settlement community, with a provision of 1 m³ biogas per customer. We investigated the regularity of biogas generation, the characteristics of biogas consumption, and the carbon mitigation potential under different operational flexibility scenarios (Fig. 2).
+
+Figure 2 | Biogas production and consumption curve fitting for the five community biogas production and distribution systems (CBPD) and their greenhouse gas (GHG) emissions under different scenarios.
+
+The biogas production rates had a lag phase of approximately 1.5 days after each feeding point (Fig. 2a), while the biogas consumption rate followed a daily repeating cycle (Fig. 2b). Currently, each community’s customers tend to use biogas during normal mealtimes and, all CBPD have the same biogas consumption characteristics and similar rates of use. Although the status quo of methane slippage for the five CBPD pilots studied was 32.5%, reliable storage capacity adjustments and accurate operation of CBPD could help to achieve CPR = 1 (see Fig. 2c). On one hand, optimal storage capacity is the prerequisite to efficiently preset temporary excess biogas output. On the other hand, a feeding process with optimal amounts and time points can synchronize production rate on the plant-side and usage rate on the user-side (see Supplementary Note 1). In the framework for proposed deployment, current empirical operation with GHG emissions of 1.58 CO₂ eq /(d·customer) would be converted to a negative GHG emission of -0.73 CO₂ eq /(d·customer), through positive changes to more elaborate CBPD designs and sophisticated operation. This also indicates that the replacement of fossil fuel alone is not sufficient to realize carbon mitigation, and that biogas emission control is the most important factor in further broad deployment of CBPD. In addition to the exact estimation of customer’s use of biogas, vocational skills training for operators to establish management strategies in biogas supply is also a critical requirement for the reduction of the investment. Under current production and consumption curves, scenarios to raise the number of customers for more biogas consumption on the user-side and decreasing OLR on the plant-side could reduce the requirement for biogas storage capacity by 11.6% and 9.5%, respectively, to achieve CPR = 1.
+
+# State-of-the-art carbon mitigation at prevailing climate conditions
+
+Metrological and climatic conditions for various ambient temperatures and solar radiation intensities would play an critical role in upgraded CBPDs’ GHG emissions, and their combined influence could be calculated based on the outdoor solar-air temperature<sup>26</sup>. Once it falls below the fermentation temperature, part of the biogas generated should be used to compensate for the heat lost (thermal calculation equation is provided in Supplementary Note 2)<sup>27</sup>. We considered 10 cities in different developing areas to evaluate the performance of an ungraded CBPD to reflect its universality; the decarbonization contributions are shown in Fig.<span class="InternalRef">3</span> (calculations and detailed parameters are shown in Supplementary Note 3).
+
+Figure<span class="InternalRef">3</span><strong>| Ten cities’ outdoor solar-air temperatures and the decarbonization contributions for the upgraded community biogas production and distribution system (CBPD) deployment.</strong>
+
+Upgraded CBPD deployment could achieve the highest decarbonization contribution in the most isohyperthermic regions, such as Bangkok, Thailand, and Kuala Lumpur, Malaysia. Nevertheless, the additional heating required for biogas production in Nairobi, Kenya would decrease the carbon mitigation by 9%. In frigid zones, extremely low temperatures in winter would result in low levels of carbon mitigation, such as in Harbin, China, -0.28 CO<sub>2</sub> eq /(d·customer) during the day in January; nevertheless, an upgraded CBPD could still provide a greater contribution to carbon mitigation and realize negative emissions, compared to the widely applied combined heat and power generation technologies used in developed areas (Supplementary Table 3) or other sources of renewable energy<sup>29</sup>, such as solar energy. For the other six cities, GHG emissions are − 0.64 to -0.69 CO<sub>2</sub> eq/(customer·d). The results indicate that an upgraded CBPD had a significant advantage on carbon mitigation achievement in most developing areas. It also is a positive sign that an upgraded CBPD can be widely used as a scalable decarbonization solution. For example, national deployment of upgraded CBPD in China could have the potential to remove 62.4% of the carbon emissions of rural residential energy use<sup>30</sup>.
+
+# National deployment to meet the Chinese 1.5℃ warming target
+
+In respect of the intensive management of animal breeding and village inhabitants living in scattered Chinese communities<sup>31</sup>, decentralized deployment upgraded CBPD represents a viable system for feedstock collection on the plant-side, and biogas provision for a relatively fixed customer group on the user-side (for detailed instructions on implementation, see Supplementary Note 4). Because animal waste is an excellent, cheap, and highly accessible feedstock material<sup>32</sup>, its local availability is the most essential limiting factor for broad upgraded CBPD deployment based on the community’s energy consumption. Local availability of the most common kinds of manure (from pigs, chickens, and cattle), which are also the main sources of methane emission, dictate the feasibility of upgraded CBPD deployment at the provincial level (Fig.<span class="InternalRef">4</span> a). Most of the Chinese mainland areas, with their high rates of methane production from manure and high domestic gas demand in the rural community (RPD), are likely candidates for the broad deployment of upgraded CBPD. Although Shanghai is the only province without a supply of manure, and a RPD of 0.88, feedstock could be obtained from other organic sources, such as crop straw. However, Shanghai is a relatively small population province with a low rural population proportion, accounting for 0.7%, its impact on the achievable targets will probably be negligible.
+
+Figure<span class="InternalRef">4</span><strong>| The availability of animal manure for the national deployment of community biogas production and distribution systems (CBPD) in the 31 Chinese mainland provinces and its potential to achieve greenhouse gas (GHG) emission targets.</strong>
+
+We were interested in the effect of the national deployment of upgraded CBPD (with CPR = 1) on the mitigation of methane emissions from manure, and its contribution towards meeting the national climate target (Fig.<span class="InternalRef">4</span> b). In 2020, there were still 186.7 million rural households living in China<sup>33</sup>, most of which had no access to natural gas<sup>34</sup>. National deployment of upgraded CBPDs could provide an alternative path to the establishment of energy equality, and achieve a 190.0 million tones reduction in CO<sub>2</sub> eq/yr due to methane control and fossil fuel substitution, under the assumed manure management baseline and current biogas consumption rate<sup>35</sup>.
+
+According to the IPCC’s sixth report<sup>36</sup>, an upgraded CBPD deployment rate of 68.2% would achieve the goal of 45% methane emission reduction for nationwide manure management optimization. Based on the recent Chinese report submitted to the United Nations<sup>30</sup>, national upgraded CBPD deployment could contribute 3.77% to the Chinese carbon mitigation commitment to the Paris Agreement of a 1.5°C increase in global warming<sup>36</sup>.
+
+# Discussion and policy implications
+
+We have illustrated a simple and practical path to upgrade CBPD to achieve CPR = 1 by adjusting the biogas flow in order to maximize the energy and climate benefits in developing rural areas, rather than relying on centralized grids for fluctuated biogas integration<sup>37</sup>. Regarding Chinese village inhabitants’ energy consumption, cooking represents the largest use of energy, accounting for 41.6% of total usage<sup>38</sup>, and this approach could definitely enable a transition towards ever wider development of rural biogas industries. On the other hand, energy consumption in rural communities will certainly increase until 2030 due to Chinese demographic changes<sup>2,39</sup>, and greater upgraded CBPD deployment with accompanying methane capture and utilization could further increase biogas utilization for fossil fuel replacement and climate benefits, which have the potential to reverse the rural energy development tendency towards increasing carbon emissions<sup>2</sup>. Such policy options must be evaluated in more detail to explore a pathway for increases in deployment and the levels of operational improvement.
+
+The amount of poultry and livestock farming in China is rising and strategies to control methane emission from manure management are varied<sup>40</sup>. National upgraded CBPD deployment can provide two further advantages. First, it enables optimal manure treatment to reduce methane emissions<sup>41</sup>. Second, decentralized deployment of CBPD’s in rural communities facilitates ecological utilization of anaerobic digested fertilizer in nearby farmlands<sup>42</sup>, as fermented manure is the best organic fertilizer, replaces chemical products and reduces nitrous oxide emissions from farmland<sup>43</sup>, and fixes carbon into soils<sup>10</sup>. If all elements can be integrated regionally, further co-benefits could be achieved to result in a more intrinsic synergistic interaction<sup>44</sup>. Thus, the recommended CBPD is a fundamental, indispensable, and valuable investment to realize energy equality and low carbon emission agriculture in rural communities in developing areas, because of the wide available of manure based feedstock<sup>45</sup>.
+
+Rural communities in different regions have a variety of biogas consumption patterns. For example, rural manufacturing industries can exponentially increase during the active production periods, and space heating requires a lot of biogas, usually during the night. Further research is needed to discover the best strategies to manage energy load in different kinds of community. The policy of time-of-use pricing for peak and trough usage of biogas and strategies to supply biogas selectively (cooking is a priority usage during periods of extremely high biogas demand) are feasible practical auxiliary measures to modulate biogas consumption patterns<sup>46</sup>, and reduce the likelihood of biogas slippage.
+
+To resolve the energy challenges both in quantitative and operational terms, policy makers in developing rural areas should not only apply the new technologies, but also improve managerial skills to find the best protocols to achieve CPR close to 1. Further research could provide efficient and simple methods to evaluate CPR, and identify effective subsidy strategies for different CPR operational performances based on institutional designs from Common Pool Resource theory<sup>47</sup>, according to a rural community’s situation. Last but not least, combinations of upgraded CBPD and other renewable energy systems, such as solar energy, heat pumps, etc., would develop synergistic systems for stand-alone energy supply with high primary energy ratio and zero carbon emissions in rural areas<sup>48</sup>, in order to solve the rural energy challenge.
+
+# Methods
+
+## Biogas flow
+
+If the amount of biogas held in a storage facility is $ Q(t) $, then at time $ t+\Delta t $ it is $ Q(t+\Delta t) $, where $ \Delta t $ is time interval. On the basis of mass balance, $ Q(t+\Delta t) $ should theoretically be equal to the sum of $ Q(t) $ and the net change in the amount of biogas between production, consumption, emission, and digester heating, during the time interval $ \Delta t $.
+
+[IMAGE_METHODS_1]
+
+Where: $ q(t) $, $ g(t) $, $ e(t) $, and $ h(t) $ are the biogas production rate, biogas consumption rate, biogas emission rate, and the biogas required for heating rate in cold weather, respectively (the thermal balance calculation is described in Supplementary Note 2).
+
+## Data analysis of CBPD
+
+Hourly biogas production, hourly biogas consumption, and methane content of the five selected CBPDs were measured using multipath ultrasonic gas flow meters (TY1030, TianYu, Wuhan, China). All of the hourly data ($ X_1 $, $ X_2 $,…, $ X_n $) were restructured to achieve schema integration of the feeding interval data, including steps such as splitting, merging, folding, and unfolding, in order to resolve and overlap the conflicting representations. The measured biogas production set and biogas consumption set were represented by sets $ S_p $ and $ S_c $, respectively; $ S = \{X_1, X_2, \ldots, X_n\} $. The daily biogas production set and biogas consumption set were represented by sets $ X_p $ and $ X_c $, respectively; $ S = X(n, t) = \{x_1, x_2, \ldots, x_{24}\} $. The data collected were used as the respective $ X $ at day $ n $ and hour $ t $.
+
+In order to detect the kinds of errors and inconsistencies to be removed, a detailed data analysis was performed. The cleaning process on the given dataset made the following assumptions:
+
+1. If any value of $ X_p $ was less than 0.1 or more than five times the daily average value, the datum was either considered to be an outlier or biogas production did not follow the normal distribution, and day $ X_p $ was removed from the dataset.
+2. If any value of $ X_c $ between 1 am–4 am was more than 0.5 m³/(h•customer), it meant that biogas leakage or an inaccurate measurement may have occurred. Furthermore, if the daily average value of any $ X_c $ was two times higher or 0.5 times lower than that of the previous or following day, the data for the daily biogas consumption rate was considered to be an outlier, and the day $ X_c $ was removed from the dataset.
+3. Finally, only when day $ X_p $ and $ X_c $ were both in the dataset, could the values be considered to be quality data; otherwise, the single $ X $ value was deleted.
+
+## GHG emissions of CBPD
+
+In this study, the GHG emissions of CBPD ($ E_{CBPD} $) were calculated by only accounting for the substitution of fossil fuel and the slipping biogas emissions due to methane emissions into the open air.
+
+[IMAGE_METHODS_2]
+
+Where: 25 is the conversion factor of a methane emission to CO₂ equivalents; $ r $ is the methane content of biogas (%); and $ ef $ is the energy-specific CO₂ emission factor (kg CO₂ e/m³ CH₄). The substitution of fossil fuels refers to natural gas in this study.
+
+## RPD
+
+RPD is equal to the methane production potential of available manure divided by the total biogas demand in an assumed rural community.
+
+[IMAGE_METHODS_3]
+
+Where: $ N_n $ is the daily manure excreted by a livestock species/category $ n $ in the region (tone/d); $ VS_n $ is the volatile solid of livestock's manure in category $ n $ (tone/tone); $ B_{0,n} $ is the maximum methane production capacity using manure produced by livestock in category $ n $ (m³ CH₄/tone); $ S_{\text{average}} $ is the average daily biogas consumption in the region (m³ biogas/d); 0.6 is the recognized methane content of biogas (m³ CH₄/m³ biogas). $ N_n $ was obtained from the China Agriculture Yearbook⁴⁹, $ VS_n $ and $ B_{0,n} $ were obtained from the 2019 Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories¹⁰, and $ S_{\text{average}} $ was calculated from the data for the five CBPDs examined in this study.
+
+## Methane emissions in different manure treatments
+
+The methane conversion factor was determined for each specific manure management system, and represents the degree to which $ B_{0,n} $ was achieved. In this study, anaerobic digestion and the biogas utilization system were excluded. The amount of methane emitted was equal to the methane conversion factor multiplied by the methane production potential of the available manure.
+
+# References
+
+1. Brockway, P. E., Owen, A., Brand-Correa, L. I. & Hardt, L. Estimation of global final-stage energy-return-on-investment for fossil fuels with comparison to renewable energy sources. *Nature Energy* **4**, 612–621 (2019).
+
+2. Yu, B., Wei, Y. M., Kei, G. & Matsuoka, Y. Future scenarios for energy consumption and carbon emissions due to demographic transitions in Chinese households. *Nature Energy* 3, 109–118 (2018).
+
+3. Pettersson, M., Olofsson, J., Borjesson, P. & Bjornsson, L. Reductions in greenhouse gas emissions through innovative co-production of bio-oil in combined heat and power plants. *Applied Energy* **324**, 119637 (2022).
+
+4. Deng, Y., Xu, J., Liu, Y. & Mancl, K. Biogas as a sustainable energy source in China: Regional development strategy application and decision making. *Renewable and Sustainable Energy Reviews* **35**, 294–303 (2014).
+
+5. MOA. *China Rural Energy Yearbook*. (China Agriculture Press, 2022).
+
+6. Luo, T., Khoshnevisan B., Huang, R. Chen, Q., Mei, Z., Pan J. & Liu, H. Analysis of revolution in decentralized biogas facilities caused by transition in Chinese rural areas. *Renewable and Sustainable Energy Reviews* **133**, 110133 (2020).
+
+7. Tan, X., Lai, H., Gu, B., Zeng, Y. & Li, H. Carbon emission and abatement potential outlook in China's building sector through 2050. *Energy Policy* **118**, 429–439 (2018).
+
+8. Xue, S., Song, J., Wang, X., Shang, Z., Sheng, C., Li, C., Zhu, Y. & Liu, J. A systematic comparison of biogas development and related policies between China and Europe and corresponding insights. *Renewable and Sustainable Energy Reviews* **117**, 109474 (2020).
+
+9. Zhang, Y. et al. Observed changes in China’s methane emissions linked to policy drivers. *PNAS* **119**, 2202742119 (2022).
+
+10. IPCC. *2019 Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories*, Prepared by the National Greenhouse Gas Inventories Programme (eds Maksyutov, S., Eggleston, S., Woo, J. H., Fang, S. & Weitz, M. M) (Institute for Global Enrionmental strategies, Japan, 2019).
+
+11. Wang, P., Wang, J., Jin, R., Li, G., Zhou, M. & Xia, Q. Integrating biogas in regional energy systems to achieve near-zero carbon emissions. *Applied Energy* **322**, 119515 (2022).
+
+12. Tao, S. et al. Quantifying the rural residential energy transition in China from 1992 to 2012 through a representative national survey. *Nature Energy* **3**, 567 (2018).
+
+13. Patrizio, P., Leduc, S., Chinese, D., Dotzauer, E. & Kraxner, F. Biomethane as transport fuel–A comparison with other biogas utilization pathways in northern Italy. *Applied Energy* **157**, 25–34 (2015).
+
+14. Reinelt, T. & Liebetrau, J. Monitoring and mitigation of methane emissions from pressure relief valves of a biogas plant. *Chemical Engineering & Technology* **43**, 1–11 (2020).
+
+15. Liu, B. & Rajagopal, D. Life-cycle energy and climate benefits of energy recovery from wastes and biomass residues in the United States. *Nature Energy* **4**, 700–708 (2019).
+
+16. Alanne, K. & Saari, A. Distributed energy generation and sustainable development. *Renewable and sustainable energy reviews* **10**, 539–558 (2006).
+
+17. Luo, T., Pan, J., Fu, L., Mei, Z., Kong, C. & Huang, H. Reducing biogas emissions from village-scale plant with optimal floating-drum biogas storage tank and operation parameters. *Applied Energy* **208**, 312–318 (2017).
+
+18. Richard, T. L. Challenges in scaling up biofuels infrastructure. *Science* **329**, 793–796 (2010).
+
+19. Grubler, A. et al. A low energy demand scenario for meeting the 1.5 C target and sustainable development goals without negative emission technologies. *Nature Energy* **3**, 515–527 (2018).
+
+20. Surendra, K. C., Takara, D., Hashimoto, A. G. & Khanal, S. K. Biogas as a sustainable energy source for developing countries: Opportunities and challenges. *Renewable & Sustainable Energy Reviews* **31**, 846–859 (2014).
+
+21. Richard, T. L. Challenges in Scaling Up Biofuels Infrastructure. *Science* **329**, 793–796 (2014).
+
+22. Bluemling, B. & Visser, I. D. Overcoming the “club dilemma” of village-scale bioenergy projects—The case of India. *Energy Policy* **63**, 18–25 (2013).
+
+23. Rehman, L., U., M., Iqbal, A., Chang, C. C., Li, W. & Ju, M. Anaerobic digestion. *Water Environment Research* **91**, 1253–1271 (2019).
+
+24. Manser, N. D., Mihelcic, J. R. & Ergas, S. J. Semi-continuous mesophilic anaerobic digester performance under variations in solids retention time and feeding frequency. *Bioresource Technology* **190**, 359–366 (2015).
+
+25. Thomas, L., Zhou, Y., Long, C., Wu, J. & Jenkins, N. A general form of smart contract for decentralized energy systems management. *Nature Energy* **4**, 140 (2019).
+
+26. Wang, J., Liu, S. H., Meng, X., Gao, W. J. & Yuan, J. H. Application of retro-reflective materials in urban buildings: A comprehensive review. *Energy and Buildings* **247**, 111137 (2021).
+
+27. Wu, B. & Bibeau, E. Development of 3-D anaerobic digester heat transfer model for cold weather applications. *Transactions - American Society of Agricultural Engineers* **49**, 749–758 (2006).
+
+28. Liu, J., Zhou, X., Wu, J., Gao, W. & Qian, X. Heat transfer analysis of cylindrical anaerobic reactors with different sizes: a heat transfer model. *Environmental Science and Pollution Research* **24**, 23508–23517 (2017).
+
+29. Trommler, M., Barchmann, T., Dotzauer, M. & Cieleit, A. Can biogas plants contribute to lower the demand for power grid expansion? *Chemical Engineering & Technology* **40**, 359–366 (2017).
+
+30. Second China's national assessment report of climate change Committee. Second China's national assessment report on climate change. (Science Press, 2011).
+
+31. Huang, J. & Ding, J. Institutional innovation and policy support to facilitate small‐scale farming transformation in China. *Agricultural Economics* **47**, 227–237 (2016).
+
+32. Gonzalez, R., Garcia-Cascallana, J. & Gomez, X. Energetic valorization of biogas. A comparison between centralized and decentralized approach. *Renewable Energy* **215**, 119013 (2023).
+
+33. The Office of the Leading Group for the 7th National Population Census of the State Council. China Population Statistics Yearbook -2020. (China Statistics Press, 2022).
+
+34. National energy administration institute for resources and environmental policies. China Natural Gas Development Report. (Petroleum Industry Press, 2022).
+
+35. Lu, Y., Cohen, F., Smith, S. M. & Pfeiffer, A. Plant conversions and abatement technologies cannot prevent stranding of power plant assets in 2 degrees C scenarios. *Nature Communications* **13**, 806–816 (2022).
+
+36. IPCC. Climate Change 2021: The Physical Science Basis. (Cambridge University Press, 2021).
+
+37. Cucchiella, F., D'Adamo, I., Gastaldi, M. & Miliacca, M. A profitability analysis of small-scale plants for biomethane injection into the gas grid. *Journal of Cleaner Production* **184**, 179–187 (2018).
+
+38. Wu, S., Zheng, X. & Wei, C. Measurement of inequality using household energy consumption data in rural China. *Nature Energy* **2**, 795–803 (2017).
+
+39. Hu, B., Zhou, P. & Zhang, L. P. A digital business model for accelerating distributed renewable energy expansion in rural China. *Applied Energy* **316**, 119084 (2022).
+
+40. Ren, C., Zhou, X., Wang, C., Guo, Y., Diao, Y. Shen, S., Reis S., Li W., Xu J. & Gu B. Ageing Threatens Sustainability of Smallholder Farming in China. *Nature* **616**, 96–103 (2023).
+
+41. Trimmer, J. T., Miller, D. C. & Guest, J. S. Resource recovery from sanitation to enhance ecosystem services. *Nature Sustainability* **2**, 681–690 (2019).
+
+42. Möller, K. Effects of anaerobic digestion on soil carbon and nitrogen turnover, N emissions, and soil biological activity. A review. *Agronomy for sustainable development* **35**, 1021–1041 (2015).
+
+43. Rasapoor, M., Young, B., Asadov, A., Brar, R., Sarmah, A. K. Zhuang, W. Q. & Baroutian, S. Effects of biochar and activated carbon on biogas generation: A thermogravimetric and chemical analysis approach. *Energy Conversion & Management* **203**, 112221 (2020).
+
+44. Sepulveda, N. A., Jenkins, J. D., de Sisternes, F. J. & Lester, R. K. The role of firm low-carbon electricity resources in deep decarbonization of power generation. *Joule* **2**, 2403–2420 (2018).
+
+45. Cohen, A. u. a., Nzioka, J., Muthama, Mwanthi, A., Mutuku & Mutembei, H., M’IKiugu Analysis of energy poverty in Kenya and its implications for human health. *Energy Policy* **176**, 113506 (2023).
+
+46. Li, L., Gong, C., Tian, S. & Jiao, J. The peak-shaving efficiency analysis of natural gas time-of-use pricing for residential consumers: Evidence from multi-agent simulation. *Energy* **96**, 48–58 (2016).
+
+47. Rodela, R., Tucker, C. M., Šmid-Hribar, M., Sigura, M., Bogataj, N., Urbanc, M. & Gunya, A. Intersections of ecosystem services and common-pool resources literature: An interdisciplinary encounter. *Environmental science & policy* **94**, 72–81 (2019).
+
+48. Li, J., Wei, S., Dong, Y., Liu, X. & Novakovic, V. Technical and economic performance study on winter heating system of air source heat pump assisted solar evacuated tube water heater. *Applied Thermal Engineering* **221**, 119851 (2023).
+
+49. MOA. China Agriculture Yearbook. (China Agriculture Press, 2021).
+
+# Supplementary Files
+
+- [2.Supplementary.doc](https://assets-eu.researchsquare.com/files/rs-3496956/v1/b822e8fb8b06c44e51aa7af3.doc)

0a1b75e49f103cdf8d52a64e9c6d04b4fe6a41db65bf68a4ee14e89a66f058c4/preprint/images_list.json

@@ -0,0 +1,26 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Metabolite trafficking pathway for selective terpene secretion. By combining a squalene-binding protein, a lipid binding domain of supernatant protein factor (tSPF), and the export signal peptide of sucrose transport protein (Suc2), we systematically designed a fusion protein, Suc2-tSPF. Lipophilic squalene could be specifically loaded into the Suc2-tSPF and transported into the extracellular milieu, across the otherwise impermeable membrane. To achieve selective terpene secretion, Suc2-tSPF performs a series of programmed actions. During Suc2-tSPF mRNA translation, the N-terminal signal peptide is bound to a signal recognition particle for co-translational translocation, and the complex of ribosome with the partially translated Suc2-tSPF is transferred to the endoplasmic reticulum (ER) membrane (step a). A nascent polypeptide is synthesized, and the signal peptide is cleaved by signal peptidases in the ER lumen, leaving behind a tSPF polypeptide (step b). The tSPF polypeptide is further modified by ER chaperones to form a mature tSPF protein, which specifically captures squalene, the target metabolite (step c). Once squalene succeeds in hitchhiking, the tSPF protein, encapsulated by a transport vesicle, is subsequently transported to a Golgi apparatus (step d). To finalize squalene trafficking, a secretory vesicle exports the squalene-carrying tSPF into the extracellular space (step e).\u2003",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Signal peptide-guided tSPFs selectively transport loaded squalene to the extracellular medium. A. Schematic illustration of squalene production and recovery by our metabolite trafficking strategy. A SQ03-INO2 yeast strain (SQ), capable of squalene overproduction, was engineered to overexpress tSPF derivatives tagged with export signal peptides (left). Unlike conventional methods, such as cell disruption, which rely on multiple steps for squalene extraction (right, top), the signal peptide-tagged tSPF derivatives enable continuous flow production of squalene (right bottom) in an energy-efficient and cost-effective manner. B. High-performance liquid chromatograms of squalene secreted by tSPF derivatives. At the same retention time of the squalene standard (red), signal peptide-tagged tSPFs (Suc2-tSPF: navy; Pho5-tSPF: green; and MF\u03b1-tSPF: purple) yielded distinct peaks due to extracellular squalene secretion. In contrast, when tSPF was neither tagged with signal peptides (pink) nor expressed in yeast (blue), only negligible peaks were detected. C. Quantitative measurements of secreted and accumulated squalene. SQ strains included the tSPF derivative gene wherein each export signal peptide was added to the N-terminal of tSPF (left). At 72 and 144 h of shake flask cultivation, extracellular and intracellular squalene (middle and right) was quantified by collection of the dodecane phase and disruption of the harvested cells, respectively. D. Western blot analysis of cell lysates (top) and culture supernatants (bottom). When tagged with signal peptides, most tSPFs did not stay within the cell, but instead migrated into the culture medium. Without the guidance of signal peptides, the tSPF could not be secreted from the cell; the band of tSPF was only detected from cell lysates. Intracellular actin served as a marker of cell lysis. E. Differential interference contrast (DIC) and confocal fluorescence microscopy images of yeast cells that overexpressed tSPF derivatives fused with green fluorescent protein (GFP). Owing to extracellular secretion, fluorescent GFP-fused Suc2-tSPF and Pho5-tSPF were not accumulated inside the cells. In comparison, MF\u03b1-tSPF was not efficiently secreted from the cell; however, the non-tagged tSPF was completely reserved inside the cells. Scale bar: 5 \u03bcm. F. Semi-continuous fermentation system using Suc2-tSPF/SQ. With continuous nutrition supply, yeast cells with Suc2-tSPF enable sustainable squalene production and secretion. During semi-continuous culture, the growth medium was replenished every 3 days (left), and the levels of intra- and extracellular squalene were monitored for 15 days (right). Secreted squalene titers increased steadily with time, whereas intracellular squalene contents did not change notably, thereby indicating a time-independent behavior. All data represent the mean of biological triplicates, and error bars indicate the standard deviation.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Extension of our metabolite trafficking strategy for the secretion of other terpenes. A. The biosynthetic pathway of \u03b2-carotene production. Compared to the squalene biosynthesis pathway, this pathway further involves the expression of three constitutive genes, namely crtE, crtYB, and crtI. B. Quantitative measurements of secreted and accumulated \u03b2-carotene produced by the engineered strains that overexpress Suc2-tSPF. Yeast cells were grown in a defined minimal medium with 2% (v/w) glucose and 10% (v/v) dodecane at 30 \u00b0C for 144 h. All data represent the mean of biological triplicates, and error bars indicate the standard deviation. C. High-performance liquid chromatograms of \u03b2-carotene secreted by Suc2-tSPF. Compared to the \u03b2-carotene standard (red), overexpression of Suc2-tSPF (navy) resulted in a distinct peak at the same retention time, thus validating the extracellular secretion of \u03b2-carotene. In contrast, non-tagged tSPF (pink) failed to transport \u03b2-carotene outside cells. D. Colorimetric determination of secreted \u03b2-carotene, an orange-colored terpenoid. After 144 h of cultivation, the expression of Suc2-tSPF resulted in a visible color change of dodecane due to extracellular \u03b2-carotene secretion. However, when only tSPF or no carrier protein was expressed, the dodecane phase showed faint coloration. E. The binding promiscuity of tSPF towards a wide range of hydrophobic terpenes produced by microbial cells. The binding energy between SPF and each terpene predicted by molecular docking with AutoDock Vina in PyRx varied from -10.7 to -\u20096.1 kcal/mol, which was comparable to the binding energy between SPF and squalene.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0a1b75e49f103cdf8d52a64e9c6d04b4fe6a41db65bf68a4ee14e89a66f058c4/preprint/preprint.md

@@ -0,0 +1,143 @@
+# Abstract
+
+Metabolites are often unable to permeate cell membranes and are thus accumulated inside cells<sup>1</sup>. We investigated whether engineered microbes could exclusively secrete intracellular metabolites because sustainable metabolite secretion holds a great potential for mass-production of high-value chemicals in an efficient and continuous manner<sup>2,3</sup>. In this study, we demonstrated a synthetic pathway for a metabolite trafficking system that enables lipophilic terpene secretion by yeast cells. When metabolite-binding proteins are tagged with signal peptides, metabolite trafficking becomes programmable; loaded metabolites can be precisely delivered to a desired location within or outside the cell. As a proof of concept, we systematically coupled a terpene-specific protein with an export signal peptide and subsequently demonstrated exceptionally efficient, yet selective terpene secretion by yeast (~225 mg/L for squalene and ~1.6 mg/L for β-carotene). Other carrier proteins could also be readily fused with desired signal peptides, thereby tailoring different metabolite trafficking pathways in different microbes. To the best of our knowledge, this is the first and most efficient cognate pathway for metabolite secretion by microorganisms.
+
+**Biotechnology and Bioengineering**  
+**intracellular metabolites**  
+**engineered microbes**  
+**lipophilic terpene**
+
+# Main Text
+
+Microorganisms exocytose certain metabolites for maintaining homeostasis as well as for regulating overflow metabolism<sup>1</sup>. When metabolic pathways are highly imbalanced, intracellular metabolites can diffuse, sometimes assisted by membrane transport proteins, across membranes into an extracellular medium. Secretion of toxic metabolic products offers an effective way for host cells to relieve the feedback inhibition of metabolic pathways by replenishing intracellular reservoirs<sup>4-6</sup>. This indicates that regardless of toxicity, value-added metabolites can be consistently produced and transported out of genetically modified microbial cells<sup>7</sup>. The desired compounds can then be easily recovered from an extracellular liquid medium, without any harvesting or cell disruption. Thus, metabolite secretion engineering holds great potential for many applications in synthetic biology, including continuous flow production of valuable metabolites in an efficient and cost-effective manner<sup>2,8</sup>.
+
+Despite industrial and pharmaceutical importance, microbial engineering for selective metabolite secretion is exceptionally challenging. Recently, the use of ATP-binding cassette transporters, known to export lipids or drugs, was explored to mediate biofuel efflux in *Escherichia coli*<sup>9-12</sup>. However, since noncognate transmembrane proteins are polyspecific, exocytosis of isoprenoids was indiscriminate, and the degree of efflux was insignificant (~25 mg/L)<sup>13</sup>. In this regard, interorganellar protein trafficking in eukaryotes is unique. Thousands of cytoplasmic proteins are encrypted with sorting signals, and due to these address labels, each protein can accurately be delivered to intracellular compartments or extracellular milieu. This is the most significant systemic difference between secretion of proteins and metabolites.
+
+In this study, we investigated whether metabolite secretion could be selectively guided by discrete sorting signals, similar to those of the elaborate protein trafficking system. It is well known that all metabolites interact with proteins. They serve as either substrates or products in enzymatic reactions and are required by allosteric proteins as specific ligands. Molecular recognition can appropriately be exploited for metabolite trafficking; specific proteins that bind to desired metabolites can be chosen and readily tagged with signal peptides for further sorting. Thus, the fusion proteins, capable of carrying the target metabolites, are designated certain destinations, within or outside cells. By integrating a metabolite-binding protein with its sorting tag, we successfully engineered a metabolite trafficking system. As a proof of concept, we systematically designed an exclusive pathway for secretion of medicinal terpenes, large and hydrophobic high-value chemicals, and demonstrated that terpene production was substantially boosted in *Saccharomyces cerevisiae* (e.g., ~226 mg/L for 5-day batch fermentation and ~700 mg/L for 15-day semi-continuous culture). To the best of our knowledge, this is the first programmed cognate pathway for selective metabolite secretion in microorganisms, thus enabling the intracellularly accumulated target compounds to pass through otherwise impermeable membranes.
+
+Our fusion protein design involved coupling of a terpene-binding protein with an export signal peptide. Since supernatant protein factor (SPF) is involved in the regulation of cholesterol biosynthesis in the human liver<sup>14</sup>, we first identified this cytosolic lipid-binding protein as a terpene carrier. To facilitate the hydrophobic interaction of SPF with lipophilic metabolites, we only employed the lipid-binding domain of SPF (tSPF, amino acids 1-278) by eliminating the C-terminal Golgi dynamics domain (Supplementary Fig. 1)<sup>15</sup>. In parallel, we examined the ability of sucrose transport protein (invertase, Suc2), derived from yeast, for export signaling via co-translational translocation<sup>16</sup>. When the signal peptide of Suc2 is cleaved from a nascent protein in the endoplasmic reticulum (ER) lumen by signal peptidases, the remaining secretory protein is immediately translocated into the ER membrane for subsequent secretion. Taking advantage of this mechanism, we fused the Suc2 signal peptide to the tSPF N-terminal to obtain Suc2-tSPF. This tailored fusion protein could perform a series of programmed actions for selective terpene secretion (Fig. 1 and Supplementary Fig. 2). Briefly, once Suc2-tSPF was translated in the ER (step a), signal peptidases subsequently cleaved the Suc2 signal peptide (step b). The mature tSPF was specifically loaded with a terpene (step c) and then transported to the Golgi (step d). To complete terpene trafficking, the terpene-loaded tSPF was exported into the extracellular medium (step e).
+
+We successfully expressed Suc2-tSPF in an engineered *S. cerevisiae* strain for squalene overproduction and achieved extracellular secretion of hitherto membrane-impermeable terpenes (Fig. 2). Briefly, the Suc2-tSPF gene was systematically incorporated into our original squalene producer (SQ03-INO2; subsequently referred to as SQ), which is known to the most productive yeast strain, capable of producing up to 69 mg/g dry cell weight (DCW) squalene in 144 h<sup>17</sup>. Unlike conventional methods, a two-phase cultivation system comprising culture medium and 10% (v/v) immiscible dodecane was adopted (Fig. 2A)<sup>13,18</sup>. Hydrophobic metabolites are soluble in the organic phase, but not in the aqueous medium; thus, the secreted squalene can be fully recovered by simply collecting the dodecane (see Methods). High-performance liquid chromatography (HPLC) revealed that successful overexpression of Suc2-tSPF resulted in remarkably high levels of squalene in the dodecane phase (Fig. 2B (navy) and Supplementary Fig. 3). In contrast, no squalene secretion was observed when the control strain SQ was used (blue).
+
+Surprisingly, the most efficient production of squalene was demonstrated by the newly engineered yeast strain (Suc2-tSPF-expressing SQ, Suc2-tSPF/SQ) to date. Time-dependent quantification of squalene secretion (Fig. 2C and Supplementary Table 1) revealed a significant increase in extracellular squalene secretion by Suc2-tSPF/SQ compared with the original SQ. After 144 h of cultivation, the titer of extracellular squalene produced by Suc2-tSPF/SQ was 226 mg/L, which was ~26-fold more than that by SQ (8 mg/L). We noted that long-term cultivation (6 d) was accompanied by the natural death of old cells, which were inevitably included in the dodecane phase. Thus, SQ yielded markedly low levels of extracellular squalene via inadequate squalene secretion. Along with unprecedented squalene secretion, Suc2-tSPF/SQ also exhibited increased intracellular squalene accumulation (Supplementary Table 1); it produced 97 mg/g DCW of intracellular squalene after 144 h of fermentation, and this value showed a ~40% increase over that produced by the original SQ (69 mg/g DCW). Without the signal peptide, tSPF alone could not secrete squalene (Fig. 2B – C and Supplementary Table 1). The tSPF-expressing SQ strain (tSPF/SQ) showed negligible squalene secretion (13 mg/L), similar to that by the control SQ strain (8 mg/L), even after fermentation for 144 h.
+
+Selective metabolite secretion depends on the level of each intracellular product, and only surplus squalene can be exclusively secreted. Unlike squalene, 2,3-oxidosqualene, an important metabolic intermediate in the squalene biosynthesis pathway, was not secreted. Considering that SPF has no significant preference for either squalene or 2,3-oxidosqualene, we speculated that their intracellular accumulation would determine whether they could be loaded onto the export carrier, Suc2-tSPF. Since 2,3-oxidosqualene was rapidly consumed in our engineered yeast cells, its intracellular concentration was not sufficient to trigger Suc2-tSPF capture for subsequent secretion (Supplementary Table 2). Conversely, the extracellular transport of the desired product, squalene, increased in a time-dependent manner. On comparing the results of Suc2-tSPF/SQ cultivation for 72 and 144 h, we found that extracellular squalene was markedly increased by ~36%, confirming that squalene titers gradually increased over time (Fig. 2C and Supplementary Table 1).
+
+Even when other export signal peptides were coupled with tSPF, the selective squalene secretory pathway remained active. Two different proteins, viz., acid phosphatase (Pho5) and yeast α-mating factor (MFα) (Supplementary Table 3), were chosen on the basis of translocation modes<sup>19-21</sup>. Pho5 uses the same co-translational translocation pathway as Suc2, but its signal peptide is cleaved by different ER peptidases. In contrast, MFα is transported to the ER membrane via post-translational translocation, which involves two-step cleavage of the signal peptide (Supplementary Fig. 2). Following the removal of the export signal peptide of MFα by signal peptidases in the ER lumen, Kex2 and Ste13 proteases further cleave the remnant of the signal peptide in the Golgi. Owing to the same translocation mode, squalene secretion by Pho5-tSPF was comparable to Suc2-tSPF (Fig. 2B); however, total squalene secretion after 144 h of cultivation (190 mg/L) was lower (Fig. 2C and Supplementary Table 1). Interestingly, MFα-tSPF, which is trafficked via a route distinct from that of Suc2-tSPF, showed significantly lower squalene secretion (Fig. 2B – C and Supplementary Table 1). Although the fundamental mechanism underlying this is unclear, we speculated that the inherent complexity of signal peptide cleavage might have affected the loading efficiency of squalene onto tSPF before secretion<sup>22</sup>.
+
+To clarify further whether the presence of squalene in the dodecane phase was caused by programmed secretion or cell lysis-dependent release, we performed a western blot assay (Fig. 2D, see Methods) using actin as a loading control. In principle, with negligible cell lysis, cytosolic actin should not be detected in the extracellular medium, but it must be possible to detect tSPF by signal peptide-guided secretion. Based on the actin bands of the culture supernatant, we verified that the presence of tSPF in the extracellular space was not caused by cell lysis. Without signal peptides, tSPF was enriched only inside the cells (Fig. 2D, upper). However, when the cells overexpressed tSPF that was systematically integrated with signal peptides, tSPF was successfully secreted out of the cells (Fig. 2D, lower). Among the three different signal peptides, Suc2 exported tSPF most efficiently, which was consistent with the finding that Suc2-tSPF/SQ secreted squalene most productively.
+
+We further validated the role of signal peptide-guided tSPF in squalene secretion by fusing green fluorescent protein (GFP) to the C-terminal of tSPF and observed its localization using confocal fluorescence microscopy (Fig. 2E and Supplementary Fig. 4). Corroborating the western blot results, the intracellular GFP signal was only intense for the signal peptide-lacking tSPF, whereas no fluorescence inside the cells were observed for the signal peptide-tagged tSPF. From this observation, we concluded that GFP-fused Suc2-tSPF and Pho5-tSPF (Suc2-tSPF-GFP and Phot-tSPF-GFP, respectively) enabled the secretion of tSPF-GFP after cleavage of the signal peptide. We noted that MFα-tSPF-GFP could not be completely transported outside the cells owing to its poor secretion capability. However, without the export signaling tag, the tSPF-GFP was destined for intracellular retention.
+
+Indeed, the selective capture of tSPF was responsible for squalene secretion, which was validated by the use of control proteins incapable of squalene binding. For this purpose, we prepared Suc2-tagged GFP (Suc2-GFP), since GFP can neither perform extracellular secretion nor capture terpene (Supplementary Table 1)<sup>23</sup>. Overexpression of Suc2-GFP in the SQ strain clearly showed its extracellular export (Supplementary Fig. 5). However, there was no evidence of squalene secretion, thereby indicating that the engineered squalene secretion pathway was the exclusive result of synchronizing two different molecular functions, namely protein hitchhiking and metabolite trafficking.
+
+Additionally, we implemented a semi-continuous fermentation using Suc2-tSPF/SQ, the best squalene-secreting strain, to prove that our metabolite trafficking system could be used to produce and secrete the target metabolite in a continuous and efficient manner (Fig. 2F, Supplementary Fig. 6, and Supplementary Table 4; see Methods). Briefly, we sampled yeast cells and collected dodecane, replenishing with fresh culture medium once every 3 days. During five repeated cycles, intracellular squalene remained at ~60 mg/g DCW, indicating no further increase in production. However, extracellular squalene accumulated consistently, and by the fifth cycle, its titer exceeded 700 mg/L, the highest titer of squalene production reported to date. Moreover, the graph for total squalene production over time was linear, thereby indicating the potential for continuous flow production.
+
+Given the binding promiscuity of SPF towards other hydrophobic terpenes<sup>15</sup>, we further expanded our innovative concept of metabolite secretion by applying it for production of other terpenes. Thus, β-carotene, a tetra-terpene, was chosen to demonstrate the engineered secretion owing to its simplistic colorimetric detection<sup>13</sup>. For production of β-carotene in *S. cerevisiae* (Fig. 3A), we used the p415-BC plasmid containing four different β-carotene biosynthetic genes: *tHMG1* (truncated 3-hydroxy-3-methylglutaryl-CoA reductase 1) from *S. cerevisiae*, and *crtE* (geranyl diphosphate synthase), *crtYB* (phytoene synthase), and *crtI* (carotene desaturase) from *Xanthophyllomyces dendrorhous*. As ascertained by squalene secretion, we overexpressed Suc2-tSPF using yeast cells containing p415-BC for β-carotene production. Interestingly, we observed a ~23-fold increase (1.4 mg/L) in β-carotene secretion compared to that by the control, which was incapable of secreting tSPF (0.06 mg/L) (Fig. 3B – D, Supplementary Fig. 3, 7, and Supplementary Table 5).
+
+Finally, we identified the metabolites that could benefit from our SPF-driven programmed secretion, using molecular docking with AutoDock Vina in PyRx<sup>24,25</sup>. We scrutinized 43 different terpenes that microbial cells have produced as production hosts<sup>26,27</sup>, and the predicted binding energy was calculated for each compound (Fig. 3E and Supplementary Table 6, see Methods). The molecular interaction energy between SPF and each terpene varied from -10.7 to -6.1 kcal/mol. Squalene and 2,3-oxidosqualene interaction energies were -10.6 and -10.5 kcal/mol, respectively, confirming their comparable binding capabilities to the SPF. In contrast, glutamate and pyruvate, key intermediates of metabolic pathways in yeast, exhibited much lower affinities (-4.4 and -3.6 kcal/mol, respectively), suggesting that the SPF carrier protein is terpene-specific. Furthermore, we thoroughly analyzed the pattern of SPF-terpene interactions after the docking simulation; the positions of all terpenes overlapped within the SPF binding pocket (Supplementary Fig. 8), although end groups of several tetra-terpenes (e.g., an isophorone group of zeaxanthin) protruded from the SPF protein surface. These simulations validated that SPF could serve as a suitable vehicle for various terpenes. Thus, the signal peptide-guided SPF would enable precise delivery of target terpenes to a desired location, including the extracellular space, thereby actualizing the synthetic pathway of metabolite trafficking.
+
+In summary, we first demonstrated that a common multiplexed protein secretion pathway could mediate sustainable and efficient extracellular transport of target metabolites. The integration of protein hitchhiking with metabolite trafficking was highly synergistic. Membrane-impermeable terpenes could be rescued from the extracellular medium without cell disruption and subsequent extraction, thereby demonstrating a potential cost-effective, high-yielding continuous flow process for production of valuable chemicals on an industrial scale. Unlike previous noncognate transmembrane engineering<sup>13</sup>, our cognate secretion pathway programming achieved exceptionally efficient terpene secretion (~226 mg/L for squalene and ~1.6 mg/L for β-carotene), the highest values reported in microorganisms for the first time in this study.
+
+To maximize terpene production and secretion by optimal cell growth, our metabolite trafficking pathway can be fine-tuned by employing various synthetic biology tools and strategies<sup>3,28,29</sup>. For this proof-of-concept study, we overexpressed Suc2-tSPF using the strongest yeast TDH3 promoter on a high-copy number plasmid, likely causing the protein burden effect<sup>23,30</sup>. However, this could be mitigated by balancing or tuning the expression levels of relevant proteins. Furthermore, synthetic regulatory systems with diverse operational modes<sup>31,32</sup>, such as auto-regulatory feedback loops, toggle switches, and engineered riboswitches, may permit the decoupling or integration of cellular growth and terpene secretion.
+
+Importantly, our metabolite trafficking strategy is versatile. Despite the promiscuity of SPF, its concentration-dependent binding characteristic enables the desired metabolic product to be loaded only onto the carrier protein, thereby indicating that metabolic pathway engineering can allow us to readily choose the terpene to be secreted exclusively. Furthermore, by systematically combining other carrier proteins and signal peptides, various high-value biomolecules of economic, environmental, or therapeutic importance could be precisely delivered to desired intracellular locations, as well as to extracellular spaces, thereby pioneering the unprecedented avenues of synthetic biology in controlling cellular life.
+
+# Methods
+
+Plasmid and strain construction. All plasmids, strains, and primers used in this study are listed in Supplementary Tables 8 and 9. The *tSPF* gene was codon-optimized for *S. cerevisiae*, synthesized and cloned into pD1214-FAKS (ATUM, formerly DNA 2.0; Newark, CA, USA), generating pSEC-MFα-tSPF. All other tSPF derivatives tagged with different export signal peptides in this study were generated with a set of primers replacing the MFα signal sequence with that of Suc2 or Pho5, resulting in pSEC-Suc2-tSPF and pSEC-Pho5-tSPF, respectively. For the expression of tSPF without any export signal peptides, the MFα signal sequence was deleted in pSEC-MFα-tSPF, generating pSEC-tSPF. We evaluated the signal peptide-tagged tSPF derivatives for selective squalene secretion by transforming corresponding tSPF-expressing plasmids into SQ strain, our squalene producer<sup><span citationid="CR19" class="CitationRef">19</span></sup>. For β-carotene production in yeast, the plasmid pLM494, encoding β-carotene biosynthetic genes *crtE*, *crtI*, *crtYB*, and *tHMG1*, was obtained from Addgene (plasmid #100539). This pLM494 plasmid was digested with the restriction enzymes BamHI and SalI and then the resulting 11.3-kb fragment harboring expression cassettes for *crtE*, *crtI*, *crtYB*, and *tHMG1* was cloned into p415-GPD, yielding p415-BC. The p415-BC for β-carotene production and pSEC-tSPF or pSEC-Suc2-tSPF for β-carotene secretion were co-transformed into the wild-type CEN.PK2-1D. To visualize locations of tSPF, pSEC-tSPF-GFP, pSEC-MFα-tSPF-GFP, pSEC-Pho5-tSPF-GFP and pSEC-Suc2-tSPF-GFP were constructed by insertion of a PCR fragment of tSPF, MFα-tSPF, Pho5-tSPF or Suc2-tSFP, respectively, and the PCR fragment flanks SpeI and XhoI restriction enzyme sites into p426-TEF1 vector, followed by the GFP gene insertion at the C-terminus of tSPF. The resulting plasmids were transformed into SQ. The standard LiAc/ssDNA/PEG method was used for yeast transformation<sup><span citationid="CR37" class="CitationRef">37</span></sup>.
+
+Terpene production by two-phase flask fermentation. To examine terpene production and secretion by our metabolite trafficking, a two-phase culture system was adopted for flask fermentation of yeast cells. Yeast strains were grown on yeast synthetic complete (YSC) agar plates for 2–3 days at 30°C, and then transferred to 50 mL conical tubes with vent caps (SPL Life Sciences, Korea) containing 10 mL of YSC medium. YSC medium was composed of 0.67% (w/v) yeast nitrogen base without amino acids (BD Difco, USA), 0.19% (w/v) yeast synthetic drop-out medium supplements without uracil (Sigma-Aldrich, USA), and 2% (w/v) glucose. For β-carotene production, the drop-out medium supplement was substituted with the complete supplement mixture without leucine and uracil (MP Biomedicals, USA). Saturated overnight yeast cultures were used to inoculate 45 mL of YSC medium gently overlaid with 5 mL of dodecane (Sigma-Aldrich, USA) as an extractive solvent in 250 mL flasks to give an initial OD<sub>600</sub> of 0.5. For batch fermentation, two-phase cultures were incubated at 30°C with shaking at 250 rpm for 6 days. For semi-continuous cultures, growing yeast cells and dodecane were collected and then transferred to freshly prepared YSC medium every 72 h for 15 days. After cultivation, cells, culture media, dodecane layers were collected for quantitative analysis of terpene production by HPLC. All flask fermentations were performed in three independent experiments.
+
+Terpene extraction from yeast cells and quantification. Yeast cells were harvested by centrifugation at 13,000 × *g* for 5 min. The harvested cells were resuspended in 0.6 mL of a 1:1 methanol-acetone solution. The mixture was transferred to a tube containing lysing matrix C, then disrupted mechanically using a FastPrep-24 5G homogenizer (MP Biomedicals, USA) according to the manufacturer’s instructions. For the quantification of terpenes, intracellular amounts were measured from the cell lysates, and extracellular (secreted) amounts were measured in the collected culture media or dodecane layers. Culture media and dodecane layers were prepared from cell cultures separated by centrifugation at 4,000 × *g* for 10 min. All samples, including cell lysates, culture media, and collected dodecane phases, were again centrifuged at 13,000 × *g* for 5 min, filtered using 0.2 µm syringe filters, and analyzed using an Agilent HPLC system equipped with Kinetex 5 µm EVO C18 column (Phenomenex, Aschaffenburg, Germany) and Agilent UV detector at 203 nm. Metabolites were separated by an isocratic elution with a flow rate of 1.0 mL/min at 30°C for 30 min<sup><span citationid="CR19" class="CitationRef">19</span>, <span citationid="CR38" class="CitationRef">38</span></sup>.
+
+Protein preparation for western blot analysis. Overnight yeast cultures were inoculated into 50 mL YSC medium and grown at 30°C, 250 rpm, for 2 days. Cell cultures were harvested to collect cells equivalent to an OD<sub>600</sub> of 100 for intracellular protein extraction, and supernatants were used to analyze extracellular proteins. Collected cell pellets were resuspended in 1 mL RIPA buffer (Thermo Scientific, MA, USA) with 1 × protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany), then the resuspensions were homogenized on ice using a Vibra-Cell VCX750 ultrasonic processor (SONICS, CT, USA) at 20% amplitude with 3 s intervals for 6 min. Extracellular proteins in supernatants were precipitated with 25% of ice-cold trichloroacetic acid (TCA) for 30 min, followed by centrifugation at 4,000 × *g* for 40 min at 4°C. The precipitated protein pellets were washed with 1 mL ice-cold acetone three times and resuspended in 50 µl sterile water. The precipitated proteins were further concentrated using 30 kDa Amicon Ultra Centrifugal Filters (MilliPore, Tullagreen, Ireland).
+
+Western blot analysis. All protein samples were mixed with 5 × SDS sample buffer (containing 250 mM Tris-HCl pH 6.8, 5% β-mercaptoethanol, 10% SDS, 0.5% bromophenol blue and 50% glycerol), and then separated on 10% SDS-PAGE gels with Precision Plus Protein Standards (Bio-Rad, CA, USA). The gels were transferred onto PVDF membranes (Bio-Rad, CA, USA) and then blocked overnight with 4% (w/v) skim milk dissolved in TBST (Tris buffer saline with Tween-20). An anti-actin antibody (Chemicon International Inc., MA, USA, 1:5000) or an anti-His antibody (Santa Cruz Biotechnology, Inc., CA, USA, 1:1000) was used as a primary antibody. The primary antibodies were diluted in TBST for 1 h at room temperature. After washing with TBST three times, a peroxidase-conjugated Goat anti-Mouse IgG Antibody (Jackson Immun. Lab., USA, 1:5000) was used as a secondary antibody. The secondary antibody was incubated for 1 h at room temperature. After washing with TBST three times, the signals were visualized with Pierce ECL Western Blotting Substrate (Thermo Scientific, MA, USA) using the ChemiDoc MP imaging system (Bio-Rad, USA).
+
+Confocal fluorescence microscopy. Yeast cells were cultured in the YSC medium supplemented with all amino acids without uracil to maintain plasmids expressing GFP-fused tSPF derivatives. After growing cells for 24 h at 30°C with 250 rpm of shaking, cells were fixed for 20 min with 3.7% formaldehyde in PBS (pH 7.4) and observed using a multiphoton confocal microscope (Zeiss-LSM 780, Germany) equipped with a Plan-Apochromat 63 × oil immersion objective. Confocal images were processed and analyzed using ZEN imaging software (Zeiss) and Image-Pro (Media Cybernetics).
+
+Identification of ligand binding sites of SPF. The three-dimensional crystallography structure of SPF (PDB ID: 4OMK) was acquired from the RCSB protein data bank (pdb). For molecular docking using Discovery Studio 2020, the A chain of 4OMK was used, in which the bound squalene and water molecules in the crystal structure were eliminated before simulation. In the crystallography structure, several amino acids (R4, K66, F113, C176, and Y202) are vibrating, but we intentionally chose one possible structure for each amino acid to construct a rigid protein structure. The active site for squalene binding of A chain was determined using Computer Atlas of Surface Topology of protein (CASTp): identified residues for squalene binding were L84, I103, L106, A108, L111, L112, L120, L121, K124, I151, Y153, C155, L158, H162, A167, V168, A170, Y171, F174, L175, L186, L189, F198, A201, Y202, I205, L209, T213, and I217.
+
+Ligand selection. 43 terpenes as sdf format were selected to molecular docking from PubChem compound database.; it includes 2,3-oxidosqualene (CID: 5366020), myrcen (CID: 5319723), (z)-ocimene (CID: 5320250), linalool (CID: 6549), genariol (CID: 637566), nerol (CID: 643820), limonene (CID: 22311), menthol (CID: 1254), α-pinene (CID: 6654), camphor (CID: 2537), carvacrol (CID: 10364), carvone (CID: 7439), β-farnesene (CID: 5281517), nerolidol (CID: 5284507), β-bisabolene (CID: 10104370), valencen (CID: 9855795), amorphadiene (CID: 11052747), capsidiol (CID: 161937), α-cedrane (CID: 9548702), artemisinin (CID: 68827), phytane (CID: 12523), phytol (CID: 5280435), camphorene (CID: 101750), retionol (CID: 445354), labdane (CID: 9548711), clerodane (CID: 182677), pimarane (CID: 9548698), abietane (CID: 6857485), tigliane (CID: 154992), kaurane (CID: 9548699), taxane (CID: 9548828), squalene (CID: 638072), lanosterol (CID: 246983), α-amyrin (CID: 73170), β-amyrin (CID: 73145), hopane (CID: 10115), oleanolic acid (CID: 10494), lycopene (CID: 446925), rubixanthin (CID: 5281252), β-carotene (CID: 5280489), lutein (CID: 5281243), zeaxanthin (CID: 5280899), and astaxanthin (CID: 5281224). Furthermore, hydrophilic ligands such as glutamate (CID: 33032) and pyruvate (CID: 107735) were added as control molecules. Downloaded sdf files were converted to pdbqt format by OpenBabel toolbox in PyRx virtual tool.
+
+Molecular docking. The molecular docking between SPF and potential ligands (43 terpenes) was implemented by AutoDock Vina option of PyRx virtual screening tool. Based on aforementioned binding sites, the grid box for posing was assigned as 24.6222 x 29.3125 x 21.3921 Å, and the docking poses were scored with exhaustiveness 8. The A chain of 4OMK was considered as a rigid body and docked with flexible terpene ligands. After the docking process, each ligand yielded several docking poses (up to 9 poses) on the basis of root mean square deviation (RMSD) values. The first pose of each terpene, which has the highest score, were selected to compare the docked structures of ligands and binding affinities with SPF.
+
+Mass spectrometry. Electrospray ionization-mass spectrometry (ESI-MS) experiment was performed by using Synapt G2-Si (Waters Corporation, MA, USA) mass spectrometer equipped with nano-electrospray source. The sample solution of tSPF (~ 2 µM) and squalene (100 µM) prepared in water/methanol (1:1, v/v) was continuously transferred to the nano-electrospray emitter with a flow rate of 300 nL/min and nebulized with 2 kV electrospray voltage. The mass spectrum containing tSPF-squalene complex ions was obtained by averaging signals for 20 mins.
+
+Figure preparation. Figures were prepared using BioRender.Com for scientific illustrations.
+
+# References
+
+1. Pinu, F. R. et al. Metabolite secretion in microorganisms: the theory of metabolic overflow put to the test. *Metabolomics* **14**, 43 (2018).
+
+2. Montano Lopez, J. & Avalos, J. L. Genetically engineered yeast makes medicinal plant products. *Nature* (2020).
+
+3. Xu, X., Liu, Y., Du, G., Ledesma-Amaro, R. & Liu, L. Microbial Chassis Development for Natural Product Biosynthesis. *Trends Biotechnol.* **38**, 779-796 (2020).
+
+4. Borodina, I. Understanding metabolite transport gives an upper hand in strain development. *Microb. Biotechnol.* **12**, 69-70 (2019).
+
+5. Kell, D. B., Swainston, N., Pir, P. & Oliver, S. G. Membrane transporter engineering in industrial biotechnology and whole cell biocatalysis. *Trends Biotechnol.* **33**, 237-246 (2015).
+
+6. Kramer, R. Analysis and modeling of substrate uptake and product release by prokaryotic and eukaryotic cells. *Adv. Biochem. Eng. Biotechnol.* **54**, 31-74 (1996).
+
+7. Darbani, B., Stovicek, V., van der Hoek, S. A. & Borodina, I. Engineering energetically efficient transport of dicarboxylic acids in yeast *Saccharomyces cerevisiae*. *Proc. Natl Acad. Sci. USA* **116**, 19415-19420 (2019).
+
+8. Srinivasan, P. & Smolke, C. D. Biosynthesis of medicinal tropane alkaloids in yeast. *Nature* (2020).
+
+9. During-Olsen, L., Regenberg, B., Gjermansen, C., Kielland-Brandt, M. C. & Hansen, J. Cysteine uptake by *Saccharomyces cerevisiae* is accomplished by multiple permeases. *Curr. Genet.* **35**, 609-617 (1999).
+
+10. Magasanik, B. & Kaiser, C. A. Nitrogen regulation in *Saccharomyces cerevisiae*. *Gene* **290**, 1-18 (2002).
+
+11. Nehls, U., Mikolajewski, S., Magel, E. & Hampp, R. Carbohydrate metabolism in ectomycorrhizas: gene expression, monosaccharide transport and metabolic control. *New Phytol.* **150**, 533-541 (2001).
+
+12. Schweikhard, E. S. & Ziegler, C. M. Amino acid secondary transporters: toward a common transport mechanism. *Curr. Top. Membr.* **70**, 1-28 (2012).
+
+13. Doshi, R., Nguyen, T. & Chang, G. Transporter-mediated biofuel secretion. *Proc. Natl Acad. Sci. USA* **110**, 7642-7647 (2013).
+
+14. Shibata, N. et al. Supernatant protein factor, which stimulates the conversion of squalene to lanosterol, is a cytosolic squalene transfer protein and enhances cholesterol biosynthesis. *Proc. Natl Acad. Sci. USA* **98**, 2244-2249 (2001).
+
+15. Christen, M. et al. Structural insights on cholesterol endosynthesis: Binding of squalene and 2,3-oxidosqualene to supernatant protein factor. *J. Struct. Biol.* **190**, 261-270 (2015).
+
+16. Hou, J., Tyo, K. E., Liu, Z., Petranovic, D. & Nielsen, J. Metabolic engineering of recombinant protein secretion by *Saccharomyces cerevisiae*. *FEMS Yeast Res.* **12**, 491-510 (2012).
+
+17. Kim, J. E. et al. Tailoring the *Saccharomyces cerevisiae* endoplasmic reticulum for functional assembly of terpene synthesis pathway. *Metab. Eng.* **56**, 50-59 (2019).
+
+18. Tadros, T., Izquierdo, P., Esquena, J. & Solans, C. Formation and stability of nano-emulsions. *Adv. Colloid. Interface. Sci.* **108**, 303-318 (2004).
+
+19. Martoglio, B. & Dobberstein, B. Signal sequences: more than just greasy peptides. *Trends Cell Biol.* **8**, 410-415 (1998).
+
+20. Barrero, J. J., Casler, J. C., Valero, F., Ferrer, P. & Glick, B. S. An improved secretion signal enhances the secretion of model proteins from *Pichia pastoris*. *Microb. Cell Fact.* **17**, 161 (2018).
+
+21. Brake, A. J. et al. Alpha-factor-directed synthesis and secretion of mature foreign proteins in *Saccharomyces cerevisiae*. *Proc. Natl Acad. Sci. USA* **81**, 4642-4646 (1984).
+
+22. Fitzgerald, I. & Glick, B. S. Secretion of a foreign protein from budding yeasts is enhanced by cotranslational translocation and by suppression of vacuolar targeting. *Microb. Cell Fact.* **13**, 125 (2014).
+
+23. Kafri, M., Metzl-Raz, E., Jona, G. & Barkai, N. The Cost of Protein Production. *Cell Rep.* **14**, 22-31 (2016).
+
+24. Trott, O. & Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. *J. Comput. Chem.* **31**, 455-461 (2010).
+
+25. Dallakyan, S. & Olson, A. J. Small-molecule library screening by docking with PyRx. *Methods Mol. Biol.* **1263**, 243-250 (2015).
+
+26. Vickers, C. E., Williams, T. C., Peng, B. & Cherry, J. Recent advances in synthetic biology for engineering isoprenoid production in yeast. *Curr. Opin. Chem. Biol.* **40**, 47-56 (2017).
+
+27. Belcher, M. S., Mahinthakumar, J. & Keasling, J. D. New frontiers: harnessing pivotal advances in microbial engineering for the biosynthesis of plant-derived terpenoids. *Curr. Opin. Biotechnol.* **65**, 88-93 (2020).
+
+28. Chen, B. et al. Synthetic biology toolkits and applications in *Saccharomyces cerevisiae*. *Biotechnol. Adv.* **36**, 1870-1881 (2018).
+
+29. Gibson, D. G., Hutchison, C. A., Smith, H. O. & Venter, J. C. *Synthetic Biology: Tools for Engineering Biological Systems*. (Cold Spring Harbor Laboratory Press, 2017).
+
+30. Eguchi, Y. et al. Estimating the protein burden limit of yeast cells by measuring the expression limits of glycolytic proteins. *Elife* **7**, e34595 (2018).
+
+31. Belcher, M. S. et al. Design of orthogonal regulatory systems for modulating gene expression in plants. *Nat. Chem. Biol.* **16**, 857-865 (2020).
+
+32. Chen, Y. et al. Genetic circuit design automation for yeast. *Nat. Microbiol.* (2020).
+
+33. Gietz, R. D. & Schiestl, R. H. High-efficiency yeast transformation using the LiAc/SS carrier DNA/PEG method. *Nat. Protoc.* **2**, 31-34 (2007).
+
+34. Kandar, R., Drabkova, P., Myslikova, K. & Hampl, R. Determination of retinol and alpha-tocopherol in human seminal plasma using an HPLC with UV detection. *Andrologia* **46**, 472-478 (2014).
+
+# Supplementary Files
+
+- [SupplementaryinformationJYLXSSO.docx](https://assets-eu.researchsquare.com/files/rs-78484/v1/9cce9062102b237dfd61885e.docx)  
+  Supplementary figures and tables

0aba7a6c67bb32744ae9f23d32bc02ac36daf6e1b2cba411a28c3517122eae16/preprint/images_list.json

@@ -0,0 +1,34 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpg",
+        "caption": "Formation of Cl2-CF and its effect on the perovskite film. (a) Diagram of the formation process of the Cl2-CF. UV-vis absorption spectra of (b) CF and Cl2-CF and (c) FAI dissolved in CF and Cl2-CF. (d) Grain size distribution of PVSK, PVSK-CF, and PVSK-Cl2-CF. (e) Schematic illustration of the redox reaction between the Cl2-CF and perovskite film.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpg",
+        "caption": "The distribution of Cl and the synergistic effect of HABr/Cl2-CF on the perovskite film. (a) Cl 2p XPS spectra of PVSK, PVSK-CF, and PVSK-Cl2-CF. (b) Cl content at the bottom surface of perovskite films. (c) Grain size distribution of PVSK, PVSK-HABr/CF, and PVSK-HABr/Cl2-CF. (d-f) Grazing-incidence wide-angle X-ray scattering (GIWAXS) characterization with a grazing angle of 0.5\u00b0. (g) The space-charge limited-current (SCLC) measurements of the electron-only devices. (h) Time-resolved photoluminescence (TRPL) and (i) steady-state photoluminescence (PL) spectra of perovskite films.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpg",
+        "caption": "Defect physics and carrier transport in PSCs based on the PVSK, PVSK-HABr/CF, and PVSK-HABr/Cl2-CF. (a) Trap density of states (tDOS). (b) Electrochemical impedance spectroscopy (EIS) at a bias of 0.9 V under dark conditions. (c) Transient photovoltage measurements (TPV) at the open circuit. (d) Mott\u2013Schottky analysis. (e) Transient photocurrent (TPC) at the short circuit. (f) J\u2013V curves measured under dark conditions.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpg",
+        "caption": "Photovoltaic performance of the PSCs based on PVSK, PVSK-HABr/CF, and PVSK-HABr/Cl2-CF. (a) Cross-sectional SEM image of the n-i-p perovskite solar cells. (b) J\u2212V curves measured in the reverse scan direction (1.2 to 0 V, 250 mV s\u22121). (c) Statistical PCE. (d) Statistical VOC. (e) Maximum power point tracking of PSCs under a LED lamp with a light intensity of 100 mW cm\u22122.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0aba7a6c67bb32744ae9f23d32bc02ac36daf6e1b2cba411a28c3517122eae16/preprint/preprint.md

@@ -0,0 +1,205 @@
+# Abstract
+
+Constructing 2D/3D perovskite heterojunction has been proven to be an effective strategy for fabricating high-performance perovskite solar cells (PSCs). However, previous reports only physically deposit a 2D perovskite passivation layer on the 3D perovskite layer. These methods are limited to surface passivation only, and the bulk 3D perovskite remains defective. Herein, we propose Cl₂-dissolved chloroform as a multifunctional and reactive solvent for constructing the 2D/3D perovskite heterojunction. Specifically, the dissolved Cl₂ reacts with the 3D perovskite, leading to Cl/I ionic exchange. The generated Cl⁻ anions further diffuse to passivate the buried interface of PSCs, improving the bulk and interface quality. Additionally, hexylammonium bromide reacts with the residual PbI₂ to form 2D/3D heterojunctions on the surface. As a result, we achieved high-performance PSCs with a champion efficiency of over 24% and substantially improved operational stability, retaining 80% of the initial efficiency after working for 905 h at the maximum power point.
+
+Physical sciences/Materials science/Materials for energy and catalysis/Solar cells  
+Physical sciences/Chemistry/Energy  
+dissolved-Cl2  
+redox reaction  
+ion exchange  
+secondary growth  
+perovskite solar cells
+
+# 1. Introduction
+
+Organic-inorganic hybrid perovskite solar cells have attracted extensive attention due to their high power conversion efficiency (PCE), low cost, and facile processing. To date, the highest certified PCE of perovskite solar cells (PSCs) has achieved 25.7%. High-quality perovskite film is the key factor in achieving high PCE. Researchers have made great efforts to regulate perovskite growth through compositional engineering, processing optimization, solvent engineering, and additive engineering. In general, it is inevitable to generate defects during perovskite film growth, and surface passivation is an important and effective strategy to reduce the defects. In the past, various passivation agents like lewis acids and bases, halide salts, ionic liquids, and other organic molecules have been used to passivate the perovskite films. Among them, constructing 2D/3D perovskite heterostructures using alkyl ammonium halide possesses many advantages: (1) The defective 3D perovskite surface can be converted to high-quality 2D perovskite with reduced surface defects; (2) 2D/3D perovskite stacking facilitates the charge transfer/extraction due to the extra built-in field; (3) The modified interface adjusts the energy level arrangement of perovskite films with the adjacent charge transporting layer; (4) The bulky organic spacer prevents ion migration and water erosion, thus improving the stability of the material.
+
+The most widely used method for constructing 2D/3D heterojunctions is spin-coating bulky alkyl ammonium halides onto the 3D perovskite, followed by a thermal annealing process. The resultant 2D perovskite capping layer was most likely physically stacked onto the 3D perovskite, and only weak interfacial interactions existed between the 2D and 3D perovskite layers. Moreover, this method can’t modulate the chemical composition and crystal domains of the underlying 3D perovskite. Therefore, the inside part of the 3D perovskite film is still defective and needs further improvement.
+
+It was reported that introducing chloride compounds can improve the quality of perovskite film in previous studies. Ye *et al.* reported that methylammonium chloride (MACl) could adjust the intermediate-related perovskite crystallization and improve the crystal quality. Mahmud *et al.* demonstrated that octylammonium chloride (OACl) treatment could induce the diffusion of Cl⁻ into the bulk of the underlying 3D perovskite, which ensures effective passivation. Compared to the chloride compound, the more reactive and aggressive gaseous halogen may chemically and crystallographically reconstruct the bottom 3D perovskite film. Inspired by these reports, we assume that dissolved-Cl₂ is promising in renovating the quality of bulk perovskite, which is probably able to achieve chemically soldered 2D/3D perovskite heterojunction for high-performance PSCs.
+
+Herein, we report adopting chlorine-dissolved chloroform (Cl₂-CF) as a multifunctional solvent for dissolving hexylammonium hydrobromide (HABr) to construct 2D/3D perovskite heterojunction, as well as to induce secondary growth of perovskite grains and defect passivation through the redox reaction between Cl₂ and I⁻. During the redox reaction, Cl/I ionic exchange was achieved, and the secondary growth of perovskite grains occurred through the Ostwald ripening, forming a Cl-doped perovskite film with larger grains. Moreover, the incorporated Cl⁻ could diffuse into the bulk and the buried interface of PSCs, enabling defect passivation inside the device. On the surface of the 3D perovskite, HABr reacted with the defective surface and obtained a high-quality 2D capping layer. The generated Ruddlesden–Popper (RP-type) 2D/3D perovskite heterojunctions achieved both effects of perovskite crystal regrowth and surface passivation. As a result, the champion PSC delivered a high efficiency of 24.06% with negligible hysteresis. The optimized devices showed largely enhanced operational stability, which retained 80% of the initial efficiency after 905 h under continuous one-sun illumination at the maximum power point.
+
+# 2. Results And Discussion
+
+## 2.1. Formation of Cl₂-CF and its effect on the perovskite film
+
+In this work, the Cl₂-CF was obtained by illuminating the fresh chloroform (CHCl₃, CF) with a xenon lamp under ~ 35% relative humidity. The chemical equations in Fig. 1a describe the species’ transformation. The chloroform (CHCl₃) firstly reacted with oxygen (O₂) under light illumination to form phosgene (COCl₂) and hydrogen chloride (HCl), and COCl₂ further reacted with O₂ to form carbon dioxide (CO₂) and chlorine (Cl₂), leading to the desired Cl₂-CF with the strong oxidizing property. We performed chemical analysis experiments to confirm the composition of the Cl₂-CF. The barium hydroxide (Ba(OH)₂) test in Supplementary Fig. 1 confirmed the presence of CO₂ in the Cl₂-CF, while the silver nitrate (AgNO₃) test in Supplementary Fig. 2 proved the presence of Cl⁻. We also used the wet starch potassium iodide test paper to check the presence of Cl₂ in the Cl₂-CF. When the test paper was placed above the Cl₂-CF, it immediately changed to blue color as Cl₂ could oxidize I⁻ to get I₂ (Supplementary Fig. 3). We further found that the Cl₂-CF could make I-containing species turn purple-red and Br-containing ones turn orange-yellow, while Cl-containing species showed no color change (Supplementary Fig. 4). UV-vis absorption spectra in Fig. 1b showed that the Cl₂-CF presents a characteristic absorption peak of Cl₂ at 330 nm while the CF has no signal.³³,³⁴ These characterizations confirmed the formation of Cl₂-CF under light illumination in the humid air.
+
+To reveal the reactivity of the Cl₂-CF, we conducted a series of UV-vis absorption spectroscopy measurements for each precursor of perovskite dissolved in CF or Cl₂-CF. Lead iodide (PbI₂), formamidinium iodide (FAI), and methylammonium iodide (MAI) dissolved in Cl₂-CF showed UV-vis absorption peaks at 508, 511, and 510 nm, respectively (Supplementary Fig. 5a, Fig. 1c, and Supplementary Fig. 5b). Because of the redox reaction between Cl₂ and I⁻, I₂ can be formed and lead to a UV-vis absorption peak at 508 nm.³⁵ Similarly, the oxidative Cl₂ in the Cl₂-CF can oxidize Br⁻ ions to obtain Br₂. Supplementary Fig. 6a and 6b showed that methylammonium bromide (MABr) and HABr dissolved in Cl₂-CF exhibited UV-vis absorption peaks at 411 and 410 nm, respectively.³⁶ However, for the MACl dissolved in the Cl₂-CF, there is no color change except that a UV-vis absorption peak (330 nm) belonging to the Cl₂ (Supplementary Fig. 7) was observed.³⁷
+
+We studied the effect of the solvent (CF and Cl₂-CF) on the perovskite film with a composition of (Cs₀.₀₅FA₀.₇₅MA₀.₂₀)Pb(I₀.₉₆Br₀.₀₄)₃³⁸ (Supplementary Fig. 8a-c). The control perovskite, CF-treated perovskite, and Cl₂-CF treated perovskite are denoted as PVSK, PVSK-CF, and PVSK-Cl₂-CF, respectively. The morphology of PVSK-CF was similar to that of the PVSK, where abundant PbI₂ (white particles) spread on the surface of perovskite films. In contrast, the PVSK-Cl₂-CF presented enlarged perovskite grains with small white particles appearing at the grain boundaries. We ascribed this morphology change to the redox reaction induced by Cl₂, and it triggered the secondary growth of the perovskite grains via Ostwald ripening, increasing the crystal size to as large as 5 µm (Fig. 1d). Meanwhile, as shown in Supplementary Fig. 8d-f and g-i, the surface of PVSK-Cl₂-CF showed more smooth morphology with a roughness of 28.2 nm (32.9 nm for PVSK and 29.6 nm for PVSK-CF).
+
+We performed X-ray diffraction (XRD) measurements to study the effect of Cl₂-CF on the crystal structure of perovskite. Due to the ionic exchange of I⁻ and Cl⁻,²⁹ XRD patterns of α-FAPbI₃ located at 14.07° shifted to 14.09° in PVSK-Cl₂-CF after the Cl₂-CF treatment, indicating the formation of α-FAPbI₍₃₋ₓ₎Clₓ (Supplementary Fig. 9a). Similarly, the XRD peak of the residual PbI₂ revealed a peak shift from 12.71° to 12.73°, indicating the formation of PbI₍₂₋ₓ₎Clₓ (Supplementary Fig. 9b). Figure 1e illustrates the schematic diagram of the redox reaction between Cl₂ and perovskite. Cl₂ and I⁻ ions in perovskite undergo a redox reaction to form I₂, and a small amount of Cl⁻ ions enter the perovskite lattice. Therefore, the redox reaction enables Cl-doped perovskite film with larger grains.
+
+## 2.2. The distribution of Cl and the synergistic effect of HABr/Cl₂-CF on the perovskite film
+
+To reveal the spatial distribution of Cl in perovskite film, XPS was first performed to detect the composition of the perovskite surface (Fig. 2a). We noticed that the Cl signals could only be detected for PVSK-Cl₂-CF. Subsequently, we characterized the bottom interface of the perovskite films (Supplementary Fig. 10). Many small white particles appeared at the grain boundaries of the bottom interface of PVSK-Cl₂-CF, which was probably due to Cl⁻ entering the bottom interface, forming a Pb-Cl-I mixed compound. The corresponding energy dispersive spectroscopy (EDS) results in Fig. 2b revealed that the PVSK-Cl₂-CF exhibited the highest Cl content of 2.64% at the bottom interface (vs. 0.28% for PVSK, 0.25% for PVSK-CF). The XPS and EDS results together with the XRD results in Supplementary Fig. 9 confirmed that Cl₂-CF could introduce Cl⁻ anion into the bulk of perovskite and diffuse to the bottom side of the film.
+
+As we know, defects are inevitable in solution-processed polycrystalline perovskite films.³¹,³⁹ To reduce those defects and improve the film quality, we passivated the perovskite films with an organic salt HABr, which can be well dissolved in CF or Cl₂-CF. Supplementary Fig. 11a-c and Fig. 2c shows the SEM images and the statistical grain size for the control perovskite film (PVSK), HABr/CF-treated perovskite film (PVSK-HABr/CF), and HABr/Cl₂-CF-treated perovskite film (PVSK-HABr/Cl₂-CF). The PVSK exhibited small perovskite grains (black parties) with a lot of residual PbI₂ grains (white parties)⁴⁰ (Supplementary Fig. 11a). For the PVSK-HABr/CF in Supplementary Fig. 11b, a significant decrease in the amount of PbI₂ can be observed, which could be ascribed to the transformation to 2D perovskite. The similar average grain size for the PVSK and the PVSK-HABr/CF confirmed that the post-treatment didn’t affect the crystal of the 3D perovskite. Supplementary Fig. 11c presents the morphology of the PVSK-HABr/Cl₂-CF. Along with an obvious reduction of the surface PbI₂, the perovskite grains significantly increased to even up to ~ 5 µm (Fig. 2c). The effect of Cl₂-CF on the morphology of perovskite film changes was further revealed by the AFM results in Supplementary Fig. 11d-f. The roughness of the PVSK-HABr/Cl₂-CF is largely reduced to 26.2 nm, which was comparable with that of 25.7 nm for the PVSK-HABr/CF and much less than that of 32.9 nm for the PVSK (Supplementary Fig. 11g-i).
+
+According to XRD results in Supplementary Fig. 12, all three perovskite films indicated the obvious characteristic peaks at 12.7°, 14.1°, and 28.3°, which correspond to the (100) crystal plane of PbI₂, the (001) and (002) crystal plane of perovskite phase, respectively. Among them, PVSK-HABr/CF and PVSK-HABr/Cl₂-CF display the additional characteristic peaks at ~ 4.0° corresponding to the Ruddlesden-Popper perovskite HA₂FAPb₂Br₂I₅ with a 2D structure (n = 2).³¹ To confirm the formed 2D phase induced by HABr, we probed the crystalline structure of perovskite films using grazing-incidence wide-angle X-ray scattering (GIWAXS). As shown in Fig. 2d-f, all perovskite films indicated the diffraction rings of PbI₂ and 3D perovskite phases. Meanwhile, the diffraction ring of the 2D perovskite phase (n = 2) can be observed in Fig. 2e and 2f. The integrated profiles of the GIWAXS patterns are shown in Supplementary Fig. 13. No signal of 2D perovskite could be detected for the PVSK. In contrast, there is a diffraction peak of the 2D structure at q = 0.22 Å⁻¹ for PVSK-HABr/CF and at q = 0.25 Å⁻¹ for PVSK-HABr/Cl₂-CF, which is consistent with the XRD results in Supplementary Fig. 12. Especially, due to the oxidizing property of Cl₂-CF, a slight shift of the diffraction peak of the 2D phase to a higher angle for PVSK-HABr/Cl₂-CF can be ascribed to the partial replacement of I⁻ by Cl⁻.
+
+To study the changes in surface elements of perovskite films, we conducted XPS measurements. Only the PVSK-HABr/Cl₂-CF exhibits the Cl signal with a Cl 2p₁/₂ peak located at 199.8 eV and a Cl 2p₃/₂ peak located at 198.2 eV (Supplementary Fig. 14a), which suggests the incorporation of Cl. As shown in Supplementary Fig. 14b, the PVSK and PVSK-HABr/CF present a similar Pb signal with a Pb 4f₅/₂ peak located at 143.2 eV and a Pb 4f₇/₂ peak located at 138.3 eV. However, the Pb 4f peaks for PVSK-HABr/Cl₂-CF exhibited a shift to the high binding energy. This shift is probably due to the formation of the Pb-Cl bond in PVSK-HABr/Cl₂-CF.⁴¹
+
+To investigate the defect-related carrier transport dynamics, we quantitatively evaluated the trap density of the perovskite films. According to the space-charge limited-current (SCLC), trap density (Nₜ) is determined by the trap-filled voltage (V_TFL) following the equation:
+
+$$N_{t}=2{\epsilon }_{0}{\epsilon }_{r}{V}_{TFL}/q{L}^{2}$$
+
+where ε₀, εᵣ, q, and L are the vacuum permittivity, relative permittivity, elementary charge, and the thickness of the perovskite film, respectively.⁴²,⁴³ Based on the electron-only devices with a structure of ITO/SnO₂/perovskite/[6,6]-phenyl-C61-butyric acid methyl ester (PCBM)/Ag, the current-voltage curves were recorded in Fig. 2g. We found that the Nₜ decreased from 2.73 × 10¹⁵ for the PVSK to 2.59 × 10¹⁵ cm³ for the PVSK-HABr/CF, while the PVSK-HABr/Cl₂-CF exhibited the lowest Nₜ of 1.80×10¹⁵ cm³. Figure 2h shows the time-resolved photoluminescence (TRPL) for perovskite films. The curves were fitted with a biexponential function:
+
+$$y={y}_{0}+{A}_{1}{e}^{(-x/{\tau }_{1})}+{A}_{2}{e}^{(-x/{\tau }_{2})}$$
+
+the average carrier lifetime was determined by:
+
+$${\tau }_{avg}=\left({A}_{1}{\tau }_{1}^{2}+{A}_{2}{\tau }_{2}^{2}\right)/{A}_{1}{\tau }_{1}+{A}_{2}{\tau }_{2}$$
+
+where the fast decay process (τ₁) is related to the nonradiative recombination that occurred at the perovskite surface and grain boundaries, and the slow decay process (τ₂) is related to the nonradiative recombination of photo-generated free carriers in bulk perovskite.⁴⁴ The fitted results (Supplementary Table 1) showed that, compared with an average lifetime of 834.6 ns for the PVSK, the PVSK-HABr/CF exhibited a longer average lifetime of 1078.9 ns, while the PVSK-HABr/Cl₂-CF displayed the longest average lifetime of 1644.5 ns. Figure 2i shows the steady-state photoluminescence (PL) results of perovskite films. Among the three perovskite films, the peak intensity of PVSK-HABr/Cl₂-CF is the strongest, indicating the suppressed nonradiative recombination. In addition, the blue shift of the emission peak suggests a widening bandgap of the PVSK-HABr/Cl₂-CF due to the introduction of Cl⁻ into the perovskite lattice, which can be confirmed by the Tauc results in Supplementary Fig. 15. The bandgap of the PVSK and PVSK-HABr/CF is about 1.54 eV, while that of the PVSK-HABr/Cl₂-CF increased to 1.55 eV.
+
+Based on the abovementioned changes in morphology and crystal structure, the role of HABr/CF and HABr/Cl₂-CF post-treatment is illustrated in Supplementary Fig. 16. The PVSK contains many small perovskite grains that could generate many grain boundaries and numerous defects, trapping the photo-generated carriers (Supplementary Fig. 16a). The introduction of HABr dissolved in CF can form a 2D/3D structure to passivate the defects in perovskite films (Supplementary Fig. 16b). However, this enhancement is insufficient to substantially improve the quality of perovskite films. Encouragingly, post-treatment with HABr dissolved in Cl₂-CF can realize multiple functions (Supplementary Fig. 16c). Except for the formed 2D/3D structure, the secondary growth of perovskite grains induced by Cl₂-CF can largely reduce the number of grain boundaries and defects. Therefore, we believe that using HABr/Cl₂-CF to treat perovskite films can more effectively improve the perovskite film quality by a synergistic effect that HABr induces the 2D/3D structure, and Cl₂-CF results in the secondary growth and Cl-doping of the 3D perovskite film.
+
+## 2.3. Defect physics and carrier transport of PSCs based on PVSK-HABr/Cl₂-CF
+
+We employed the trap density of states (tDOS) measurement to evaluate defect states in perovskite films. Previous studies claimed that the deep trap states are mainly related to the surface defects of perovskite films and the shallow trap states are likely resulted from the bulk perovskite films.⁴⁵,⁴⁶ Fig. 3a shows that the PVSK-HABr/Cl₂-CF device displayed the lowest tDOS in the deep-trap region (0.40 ~ 0.55 eV), suggesting the high-quality PVSK-HABr/Cl₂-CF with much-suppressed defects.
+
+Electrochemical impedance spectroscopy (EIS) was measured at a bias of 0.9 V under dark conditions, and the results were fitted with an equivalent circuit shown in the inset of Fig. 3b. For the double arc shape in the Nyquist plot, the low-frequency arc is associated with the perovskite dielectric relaxation resistance (R_dr), whereas the high-frequency one is associated with the drift-diffusion and recombination processes.⁴⁷ The PVSK-HABr/CF device displayed a larger perovskite recombination resistance (R_rec) of 1800 Ω than that of PVSK device (1500 Ω), while PVSK-HABr/Cl₂-CF device shows the largest R_rec of over 2000 Ω, which can effectively hinder the recombination between electrons and holes. Transient photovoltage measurement (TPV) shows that the decay lifetime of PVSK-HABr/CF device (0.88 ms) was higher than that of PVSK device (0.16 ms), while PVSK-HABr/Cl₂-CF device showed the highest decay lifetime (0.91 ms), further indicating that the recombination in PVSK-HABr/Cl₂-CF device was greatly suppressed (Fig. 3c).
+
+From the Mott-Schottky analysis,⁴⁸ the obtained built-in potential (V_bi) of the PVSK, PVSK-HABr/CF, and PVSK-HABr/Cl₂-CF devices were 0.83, 0.91, and 0.96 V, respectively (Fig. 3d). It indicates that the PVSK-HABr/Cl₂-CF device has a stronger internal driving force for the separation and transport of charge carriers. Transient photocurrent (TPC) measurements were employed to study carrier transport. As shown in Fig. 3e, the charge transfer time (τ) was decreased from 2.15 µs to 1.79 and 1.52 µs at the short circuit, indicating that PVSK-HABr/Cl₂-CF device has the fastest carrier extraction capability, which is beneficial to reduce the hysteresis. Figure 3f shows the dark current of the three devices. The dark current of PVSK-HABr/Cl₂-CF device is much lower than PVSK and PVSK-HABr/CF devices, indicating suppressed leakage pathways and improved ideality.
+
+## 2.4. Enhancing the photovoltaic performance by HABr/Cl₂-CF post-treatment
+
+The HABr/Cl₂-CF post-treatment has a significant effect on the photovoltaic performance of PSCs. We fabricated photovoltaic devices with the structure of ITO/SnO₂/perovskite/2,2’,7,7’-tetrakis(N,N-di(4-methoxyphenyl)amino)-9,9-spirobifluorene (Spiro-OMeTAD)/Ag (Fig. 4a). The discontinuous white particles between the perovskite layer and hole transfer layer in the PVSK-based PSC are deemed as PbI₂ (Supplementary Fig. 17). Under long-term operating conditions, the unreacted PbI₂ at the carrier transport interface is essentially the catalytic site that triggers perovskite decomposition.⁴⁹ In the PVSK-HABr/CF and PVSK-HABr/Cl₂-CF device, the amount of PbI₂ was decreased due to the transformation to 2D perovskite. Moreover, the PVSK-HABr/Cl₂-CF device displayed vertically arranged perovskite grains, which was beneficial to charge transport.
+
+Fig. 4b showed the J-V curves under the reverse scan direction (from the open circuit to the short circuit). It demonstrated that the open-circuit voltage (V_OC) of PVSK-HABr/Cl₂-CF was increased from 1.108 of PVSK and 1.122 of PVSK-HABr/CF to 1.160 V, while the short-circuit current density (J_SC) didn’t distinctly change. The statistical photovoltaic performances are presented in Fig. 4c, d, and Supplementary Fig. 18. Compared to the PVSK devices, PVSK-HABr/CF devices indicate a minor increase in PCE, while the PVSK-HABr/Cl₂-CF devices indicated a large increment in PCE based on the improved V_OC. After optimization, we obtained a champion PCE of 24.06% (V_OC of 1.160 V, J_SC of 25.57 mA cm⁻², and FF of 81.30%) for the PVSK-HABr/Cl₂-CF device. Supplementary Fig. 19 shows the incident photon-to-current conversion efficiency (IPCE) in 300–900 nm. The integrated J_SC for the PVSK, PVSK-HABr/CF, and PVSK-HABr/Cl₂-CF was 24.95, 24.97, and 24.88 (mA cm⁻²), respectively. It is well matched with the value obtained from the J-V curves (< 3% discrepancy), proving the reliability of the J_SC results. As expected, along with the introduction of HABr/Cl₂-CF, the hysteresis became negligible, which suggested that the defects were greatly passivated by HABr/Cl₂-CF treatment (Supplementary Fig. 20). Furthermore, the steady-state PCE measured at the maximum power point (V_max) is shown in Supplementary Fig. 21. Compared with the other two devices, the PVSK-HABr/Cl₂-CF device exhibited the most stable PCE of 23.07% at 0.98 V after measuring for 600 s.
+
+Stability is critical to PSCs technology.⁵⁰–⁵³ We investigated the long-term operational stability of the PSCs. The unencapsulated PVSK-HABr/Cl₂-CF device presented excellent operational stability under the maximum power point (MPP) tracking at 1-sun illumination with a LED.⁴² As shown in Fig. 4e, the PVSK-HABr/Cl₂-CF device can obtain a T₈₀ lifetime (the time when the efficiency decreased to 80% of the initial PCE) of 905 h, which is much longer than the PVSK device (305 h) and PVSK-HABr/CF device (534 h), suggesting the dramatical enhancement in the operational stability.
+
+# 3. Conclusions
+
+In conclusion, we report in-situ formed oxidative Cl₂-dissolved chloroform as the post-processing solvent for bulky cations to construct 2D/3D perovskite heterojunctions, which allows us to obtain high-quality perovskite films with passivated surface and enlarged grains in the bulk. The introduced Cl⁻ further diffuses to passivate the buried interface of perovskite solar cells. These effects are enabled by the redox reaction between Cl₂ and I⁻. As a result, the defect density of the perovskite films can be reduced and the nonradiative recombination can be largely suppressed. Finally, we achieve a high PCE of 24.06% for the optimized HABr/Cl₂-CF device with negligible hysteresis effect. The optimized device also showed a substantially enhanced long-term operational stability with a T₈₀ lifetime of 905 h.
+
+# Methods
+
+**Materials.** Unless otherwise stated, all chemicals were purchased from Sigma-Aldrich and used as received.
+
+**Synthesis of chlorine-dissolved chloroform (Cl₂-CF).** A newly purchased chloroform (HPLC-grade) is stored in a nitrogen environment for isolating from H₂O and O₂. For obtaining the Cl₂-CF, fresh CF was transferred to a transparent bottle and illuminated with a xenon lamp (Abet Technologies’ model 11002 SunLite™ Solar Simulator, 100 mW cm⁻²) for 24 h in the ambient air with ~35% relative humidity. During the illumination process, we periodically detected the state of the CF with wet starch potassium iodide test paper. Once the wet starch potassium iodide test paper turns blue, it suggests that the fresh CF has been translated into Cl₂-CF.
+
+**Film deposition and device fabrication.** The ITO-coated glasses were washed in deionized water, acetone, IPA, and ethanol for 20 minutes respectively by ultrasonic treatment. After drying, the surface was treated with plasma (Harrick, PDC-002-HP) for 5 min. Then, the SnO₂ precursor (Alfa Aesar, 15% in H₂O colloidal dispersion, diluted to 5%) was deposited onto the ITO substrate by spin-coating (4,000 rpm for 20 s). Later, these samples were annealed at 150 ℃ for 15 min. After the ITO/SnO₂ substrates were cooled to room temperature, another plasma treatment process was performed for 5 min to clean the SnO₂ film surface. Subsequently, the SnO₂-coated substrates were transferred to a glovebox with an N₂ atmosphere, and perovskite films were prepared via a two-step sequential deposition method. Firstly, PbI₂ solution was prepared by dissolving PbI₂ (691.5 mg, TCI) and 5% mole ratio of CsI (relative to PbI₂, 19.5 mg) in 1 mL mixed solvent of DMF (N, N-dimethylformamide) and DMSO (dimethylsulfoxide) with a v/v ratio of 9:1. PbI₂ film was obtained by spin-coating (2,000 rpm for 30 s) the PbI₂ solution onto the substrates and annealing at 70 ℃ for 60 s. Secondly, the salts solution was mixed with FAI (118.6 mg, Dyesol), MACl (18 mg, Dyesol), MABr (5.6 mg, Dyesol), and MAI (10 mg, Dyesol) in 2 mL of IPA. It was spin-coated onto the PbI₂ film (1,800 rpm for 30 s) and annealed at 150 ℃ for 15 min in the air (relative humidity is about 35%). Based on the above steps, control perovskite films (PVSK) were obtained. After the preparation of the perovskite layer, HABr/CF (or HABr/Cl₂-CF) (10 mmol mL⁻¹) was spin-casted at 5,500 rpm for 30 s on top of the perovskite films and annealed at 100 ℃ for 10 min in N₂. The hole transporting layer (HTL) solution was prepared by dissolving 90 mg of Spiro-OMeTAD into 1 mL of chlorobenzene, followed by the addition of 4-tert-butylpyridine (28.8 µL), and bis(trifluoromethane) sulfonamide lithium salt (17.5 µL, 520 mg mL⁻¹ in acetonitrile). It was spin-coated onto the perovskite films (3,000 rpm for 30 s). Finally, about 60 nm of Ag was thermally evaporated on top of the Spiro-OMeTAD layer.
+
+**Characterization.** UV-vis spectrophotometer (Techcomp, UV2600) was used to study the chemical properties of the involved solution. Grazing incidence wide-angle X-ray scattering (GIWAXS) measurements were performed on the Xeuss Sax/WAXS system (Xenocs, France) with a Pilatus3R 300K detector (grazing angle of 0.5°). Steady-state photoluminescence (PL) spectra of perovskite films were obtained on an instrument provided by Xipu Electronics with integrated spheres installed in a glovebox. The scanning electron microscope (SEM) images of all samples were observed by JEOL, JSM-7610F. XRD was recorded by SmartLab X-ray diffractometer (Rigaku Corporation) with the Cu Kα radiation source. The surface morphology of perovskite films was studied using AFM on a multi-mode 8 SPM system (Bruker). Time-resolved photoluminescence (TRPL) was measured by FLS920 (Edinburgh Instruments Limited) with a 375 nm pulse excitation. The CHI660E electrochemical workstation was used to collect the dark J-V curves, electrochemical impedance spectroscopy (EIS), and space-charge-limited current (SCLC). Current density-voltage characteristics were measured using a source meter (Keithley 2400) under AM 1.5G conditions (EnliTech, AAA solar simulator). Light intensity was calibrated using an NREL-calibrated silicon solar cell equipped with an infrared cut-off filter (KG-5). A scan rate of 250 mV s⁻¹ (voltage step of 10 mV, delay time of 40 ms) was used. The active area of PSCs was 0.2 cm². A black shadow mask was used to define an effective area of 0.12 cm² for the measured PSCs. IPCE spectra were measured in DC mode using the QE-R666 system (Enli tech). The mott-Schottky analysis, trap density of states (tDOS), transient photocurrent (TPC), and transient photovoltage (TPV) were performed on Zahner's electrochemical workstation equipped with a transient electrochemical measurement unit (Fast CIMPS). The operational stability of the best-performing PSCs was tested on a solar cell stability test system (Suzhou D&R Instruments Co, Ltd.) under 100 mW cm⁻² illumination using a white LED light source. The temperature of the sample was about 60°C.
+
+# References
+
+1. Kojima, A., Teshima, K., Shirai, Y. & Miyasaka, T. Organometal Halide Perovskites as Visible-Light Sensitizers for Photovoltaic Cells. J. Am. Chem. Soc. **131**, 6050–6051, (2009).
+
+2. Zhao, Y. *et al.* Inactive (PbI₂)₂RbCl stabilizes perovskite films for efficient solar cells. Science **377**, 531–534, (2022).
+
+3. National Renewable Energy Laboratory (NREL). *Photovoltaic Research, Best Research-Cell Efficiency Chart.* https://www.nrel.gov/pv/cell-efficiency.html (July 2022).
+
+4. Heo, J. H. *et al.* Efficient inorganic–organic hybrid heterojunction solar cells containing perovskite compound and polymeric hole conductors. Nat. Photonics **7**, 486–491, (2013).
+
+5. Lee, J. W., Seol, D. J., Cho, A. N. & Park, N. G. High-Efficiency Perovskite Solar Cells Based on the Black Polymorph of HC(NH₂)₂PbI₃. Adv. Mater. **26**, 4991–4998, (2014).
+
+6. Jeon, N. J. *et al.* Compositional engineering of perovskite materials for high-performance solar cells. Nature **517**, 476–480, (2015).
+
+7. Liang, K., Mitzi, B. D. & Prikas, M. T. Synthesis and Characterization of Organic – Inorganic Perovskite Thin Films Prepared Using a Versatile Two-Step Dipping Technique. Chem. Mater. **10**, 403–411, (1998).
+
+8. Xiao, M. *et al.* A Fast Deposition-Crystallization Procedure for Highly Efficient Lead Iodide Perovskite Thin-Film Solar Cells. Angew. Chem. Int. Ed. **53**, 9898–9903, (2014).
+
+9. Zhao, P. *et al.* Antisolvent with an Ultrawide Processing Window for the One-Step Fabrication of Efficient and Large-Area Perovskite Solar Cells. Adv. Mater. **30**, 1802763, (2018).
+
+10. Deng, Y. *et al.* Surfactant-controlled ink drying enables high-speed deposition of perovskite films for efficient photovoltaic modules. Nat. Energy **3**, 560–566, (2018).
+
+11. Lee Michael, M., Teuscher, J., Miyasaka, T., Murakami Takurou, N. & Snaith Henry, J. Efficient Hybrid Solar Cells Based on Meso-Superstructured Organometal Halide Perovskites. Science **338**, 643–647, (2012).
+
+12. Conings, B. *et al.* Perovskite-Based Hybrid Solar Cells Exceeding 10% Efficiency with High Reproducibility Using a Thin Film Sandwich Approach. Adv. Mater. **26**, 2041–2046, (2014).
+
+13. Nie, W. *et al.* High-efficiency solution-processed perovskite solar cells with millimeter-scale grains. Science **347**, 522–525, (2015).
+
+14. Saliba, M. *et al.* Incorporation of rubidium cations into perovskite solar cells improves photovoltaic performance. Science **354**, 206–209, (2016).
+
+15. Rong, Y. *et al.* Challenges for commercializing perovskite solar cells. Science **361**, eaat8235, (2018).
+
+16. Subbiah, A. S. *et al.* High-Performance Perovskite Single-Junction and Textured Perovskite/Silicon Tandem Solar Cells via Slot-Die-Coating. ACS Energy Lett. **5**, 3034–3040, (2020).
+
+17. Bi, D. *et al.* Efficient luminescent solar cells based on tailored mixed-cation perovskites. Sci. Adv. **2**, e1501170, (2016).
+
+18. Jiang, Q. *et al.* Surface passivation of perovskite film for efficient solar cells. Nat. Photonics **13**, 460–466, (2019).
+
+19. Alanazi, A. Q. *et al.* Atomic-Level Microstructure of Efficient Formamidinium-Based Perovskite Solar Cells Stabilized by 5-Ammonium Valeric Acid Iodide Revealed by Multi-Nuclear and Two-Dimensional Solid-State NMR. J. Am. Chem. Soc. **141**, 17659–17669, (2019).
+
+20. Kim, J. Y., Lee, J.-W., Jung, H. S., Shin, H. & Park, N. G. High-Efficiency Perovskite Solar Cells. Chem. Rev. **120**, 7867–7918, (2020).
+
+21. Zhang, F. *et al.* Suppressing defects through the synergistic effect of a Lewis base and a Lewis acid for highly efficient and stable perovskite solar cells. Energy Environ. Sci. **11**, 3480–3490, (2018).
+
+22. Mahmud, M. A. *et al.* Combined Bulk and Surface Passivation in Dimensionally Engineered 2D-3D Perovskite Films via Chlorine Diffusion. Adv. Funct. Mater. **31**, 2104251, (2021).
+
+23. Hui, W. *et al.* Stabilizing black-phase formamidinium perovskite formation at room temperature and high humidity. Science **371**, 1359–1364, (2021).
+
+24. Liang, L., Luo, H., Hu, J., Li, H. & Gao, P. Efficient Perovskite Solar Cells by Reducing Interface-Mediated Recombination: a Bulky Amine Approach. Adv. Energy Mater. **10**, 2000197, (2020).
+
+25. Cho, K. T. *et al.* Selective growth of layered perovskites for stable and efficient photovoltaics. Energy Environ. Sci. **11**, 952–959, (2018).
+
+26. Kim, H. *et al.* Optimal Interfacial Engineering with Different Length of Alkylammonium Halide for Efficient and Stable Perovskite Solar Cells. Adv. Energy Mater. **9**, 1902740, (2019).
+
+27. Xia, J. *et al.* Deep surface passivation for efficient and hydrophobic perovskite solar cells. J. Mater. Chem. A **9**, 2919–2927, (2021).
+
+28. Wu, G. *et al.* Surface Passivation Using 2D Perovskites toward Efficient and Stable Perovskite Solar Cells. Adv. Mater. **34**, 2105635, (2022).
+
+29. Solis-Ibarra, D., Smith, I. C. & Karunadasa, H. I. Post-synthetic halide conversion and selective halogen capture in hybrid perovskites. Chem. Sci. **6**, 4054–4059, (2015).
+
+30. Min, H. *et al.* Perovskite solar cells with atomically coherent interlayers on SnO₂ electrodes. Nature **598**, 444–450, (2021).
+
+31. Yoo, J. J. *et al.* An interface stabilized perovskite solar cell with high stabilized efficiency and low voltage loss. Energy Environ. Sci. **12**, 2192–2199, (2019).
+
+32. Ye, F. *et al.* Roles of MACl in Sequentially Deposited Bromine-Free Perovskite Absorbers for Efficient Solar Cells. Adv. Mater. **33**, 2007126, (2021).
+
+33. Tan, H. *et al.* Efficient and stable solution-processed planar perovskite solar cells via contact passivation. Science **355**, 722–726, (2017).
+
+34. Sant’Anna, R. *et al.* Kinetics and Mechanism of Chlorate-Chloride Reaction. J. Braz. Chem. Soc. **23**, 1543–1550, (2012).
+
+35. Yang, S. *et al.* Tailoring Passivation Molecular Structures for Extremely Small Open-Circuit Voltage Loss in Perovskite Solar Cells. J. Am. Chem. Soc. **141**, 5781–5787, (2019).
+
+36. Lin, Y.-P. *et al.* Reversible Release and Fixation of Bromine in Vacancy-Ordered Bromide Perovskites. Energy Environ. Mater. **3**, 535–540, (2020).
+
+37. Sant'Anna, R. T. P. *et al.* Kinetics and mechanism of chlorate-chloride reaction. J. Braz. Chem. Soc. **23**, 1543–1550, (2012).
+
+38. Liu, K. *et al.* Moisture-triggered fast crystallization enables efficient and stable perovskite solar cells. Nat. Commun. **13**, 4891, (2022).
+
+39. Liu, M., Johnston, M. B. & Snaith, H. J. Efficient planar heterojunction perovskite solar cells by vapour deposition. Nature **501**, 395–398, (2013).
+
+40. Jiang, Q. *et al.* Planar-Structure Perovskite Solar Cells with Efficiency beyond 21%. Adv. Mater. **29**, 1703852, (2017).
+
+41. Liu, X. *et al.* Influence of Halide Choice on Formation of Low-Dimensional Perovskite Interlayer in Efficient Perovskite Solar Cells. Energy Environ. Mater. **5**, 670–682, (2022).
+
+42. Xie, L. *et al.* Efficient and Stable Low-Bandgap Perovskite Solar Cells Enabled by a CsPbBr₃-Cluster Assisted Bottom-up Crystallization Approach. J. Am. Chem. Soc. **141**, 20537–20546, (2019).
+
+43. Zhu, P. *et al.* Simultaneous Contact and Grain-Boundary Passivation in Planar Perovskite Solar Cells Using SnO₂-KCl Composite Electron Transport Layer. Adv. Energy Mater. **10**, 1903083, (2020).
+
+44. Kirchartz, T., Márquez, J. A., Stolterfoht, M. & Unold, T. Photoluminescence-Based Characterization of Halide Perovskites for Photovoltaics. Adv. Energy Mater. **10**, 1904134, (2020).
+
+45. Ni, Z. *et al.* Resolving spatial and energetic distributions of trap states in metal halide perovskite solar cells. Science **367**, 1352–1358, (2020).
+
+46. Li, N. *et al.* Cation and anion immobilization through chemical bonding enhancement with fluorides for stable halide perovskite solar cells. Nat. Energy **4**, 408–415, (2019).
+
+47. Li, J. *et al.* Highly Efficient Thermally Co-evaporated Perovskite Solar Cells and Mini-modules. Joule **4**, 1035–1053, (2020).
+
+48. Wang, P. *et al.* 2D perovskite or organic material matter? Targeted growth for efficient perovskite solar cells with efficiency exceeding 24%. Nano Energy **94**, 106914, (2022).
+
+49. Zhao, L. *et al.* Chemical Polishing of Perovskite Surface Enhances Photovoltaic Performances. J. Am. Chem. Soc. **144**, 1700–1708, (2022).
+
+50. Yang, G. *et al.* Stable and low-photovoltage-loss perovskite solar cells by multifunctional passivation. Nat. Photonics **15**, 681–689, (2021).
+
+51. Wang, Z. *et al.* Efficient ambient-air-stable solar cells with 2D–3D heterostructured butylammonium-caesium-formamidinium lead halide perovskites. Nat. Energy **2**, 17135, (2017).
+
+52. Domanski, K., Alharbi, E. A., Hagfeldt, A., Grätzel, M. & Tress, W. Systematic investigation of the impact of operation conditions on the degradation behaviour of perovskite solar cells. Nat. Energy **3**, 61–67, (2018).
+
+53. Nie, W. *et al.* Light-activated photocurrent degradation and self-healing in perovskite solar cells. Nat. Commun. **7**, 11574, (2016).
+
+# Supplementary Files
+
+- [TOC.jpg](https://assets-eu.researchsquare.com/files/rs-2275473/v1/9737a83c3c6fa8f680341b9a.jpg)  
+  Entry for the Table of Contents  
+  We adopt Cl₂-dissolved chloroform as a multifunctional and reactive solvent to construct 2D/3D perovskite heterojunctions. The redox reaction between Cl₂ and I⁻ leads to chloride doping, secondary growth of perovskite grains, and the formation of 2D/3D perovskite heterojunction. As a result, the device based on high-quality perovskite shows a champion efficiency of 24.06%.
+
+- [SINC.docx](https://assets-eu.researchsquare.com/files/rs-2275473/v1/a8a495210b244c4e3f6c4cb6.docx)

0d10610523ffb87560f0bdc7c934673d6cd81f4eb4a868e1e3843f250c5522aa/preprint/images_list.json

@@ -0,0 +1,34 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpg",
+        "caption": "Preparation and characterization of BixIy. a, A typical growth procedure for BixIy. b, Optical image of as-grown crystals. c, XRD patterns of BixIy and BiI3 (Aladdin, 98%). d, XPS spectrum of Bi 4f for BixIy. e, Side view of BixIy along [100] direction.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpg",
+        "caption": "Band structure and charge separation kinetics of BixIy. a, Diffuse reflectance spectra of BiI3 and BixIy. Inset: calculated optical bandgap using the Tauc method by assuming an indirect bandgap. b, UPS spectrum of BixIy. Inset: linear extrapolation in the low-binding-energy region. c, Band diagram of BiI3 and BixIy, the values of                     \u00a0and   \u00a0of BiI3 are extracted from ref. 22. d, Anisotropic bias dependent photoconductivity of the Cu/BixIy/Cu devices measured by intracavity configuration under dose rate with an x-ray tube current of 80 \u03bcA. e, Illumination of electron-hole pairs separated at the interface between BiI3 and BiI.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpg",
+        "caption": "Room temperature device performance. a, Illustration of x-ray detector (the Cu tapes cover the whole surface for full charge collection) and measurement configuration. b, Anisotropic on/off x-ray responses of BixIy at different dose rates measured by intracavity configuration under 1 V mm-1 bias. c, Anisotropic bias dependent SNR (average from values under different dose rates with an x-ray tube current of 5-100 \u03bcA) measured by intracavity configuration. d, Anisotropic bias dependent SNR measured by cavity-edge configuration. The blue dotted line represents a SNR of 3, so the detection limits are 34 nGy s\u22121 for lateral and 317 nGy s\u22121 for vertical devices, respectively. e, A typical leakage current-field relations of the Cu/BixIy/Cu devices, insert: the resistivity distribution.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpg",
+        "caption": "X-ray attenuation, sensitivity and stability of BixIy detector at room temperature, the absorption of copper conductive tape is subtracted. a, b, Attenuation efficiencies of BiI3, (NH4)3Bi2I9, Cs2AgBiBr6, MAPbBr3 and CdTe semiconductors versus thickness to 50 keV x-ray photons (a) and photon energy (b). c, Anisotropic x-ray photocurrent densities at different dose rates measured by cavity-edge configuration under 1 V mm-1 bias. d, Anisotropic x-ray sensitivities at different fields measured by cavity-edge configuration. e, Device stability under repeated and continuous x-ray radiation. f, Device stability under humidity. Inset: the images of primitive and stripped BixIy after 1 h water immersion. The primitive one exhibits surface hydrolysis, which vanished after stripping.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0d10610523ffb87560f0bdc7c934673d6cd81f4eb4a868e1e3843f250c5522aa/preprint/preprint.md

@@ -0,0 +1,134 @@
+# Abstract
+
+X-ray detectors need to be operated at as low as possible doses to reduce radiation health risks during x-ray security examination or medical inspection, which drives the efforts to explore novel detectors with high sensitivity and low detection limits. Organolead trihalide perovskite is the champion, with the highest sensitivity and lowest detection limit. Unfortunately, these materials threaten the safety of the human body and environment by the toxic issue of lead. Here we present a new environmentally friendly material with van der Waals heterostructure of Bi<sub>x</sub>I<sub>y</sub> for ultra-sensitive x-ray detection. The Bi<sub>x</sub>I<sub>y</sub> detector provides anisotropic x-ray detecting performance with a sensitivity up to 1.3×10<sup>5</sup> µC Gy<sup>− 1</sup> cm<sup>− 2</sup> and a detection limit as low as 34 nGy s<sup>− 1</sup>. Meanwhile, our Bi<sub>x</sub>I<sub>y</sub> detector demonstrates wonderful environmental and hard radiation stabilities. Our discovery inspires the search for new van der Waals heterostructure classes to realize high-performance x-ray detectors and other optoelectronic devices without employing toxic elements.
+
+# Full Text
+
+Highly sensitive x-ray detection requiring a low dose rate is of particular importance to reduce the risks of cancer caused by repeated exposure to ionizing radiation in the fields of physical examination such as medical diagnosis and security inspection<sup>1, 2</sup>. Therefore, it promotes the exploration of novel x-ray detectors to improve the sensitivity and reduce the detection limit. But high sensitivity and low detection limit require the x-ray detectors to possess high resistivity, high attenuation coefficient, low electron-hole formation energy (<em>ε</em><em><sub>pair</sub></em>) and excellent charge collection ability. Here, ‘high resistivity’ results in the selection of materials with a large bandgap to reduce the temperature-induced carrier excitation. Whereas ‘low <em>ε</em><em><sub>pair</sub></em>’ needs the target materials with a small bandgap to generate more electron-hole pairs by a single x-ray photon. Therefore, a medium bandgap between 1.5 and 3.0eV is considered appropriate to balance the <em>ε</em><em><sub>pair</sub></em> and resistivity<sup>3</sup>. Semiconductors such as metal halide perovskites with medium bandgap have been chosen for fabricating sensitive x-ray detectors and achieved both high sensitivity and low detection limit at room temperature<sup>4 - 9</sup>. However, there may exist other type of semiconductors which possessing both small band gap and high resistivity, and therefore being suitable for x-ray detection at room temperature.
+
+This article reports a novel van der Waals heterostructure of Bi<sub>x</sub>I<sub>y</sub> that served as a high sensitive x-ray detector. By analyzing the samples’ cross-sectional images by aberration-corrected scanning transmission electron microscopy (ac-STEM), we found the obtained Bi<sub>x</sub>I<sub>y</sub> is composed of thick BiI<sub>3</sub> layers (main) alternately stacked with new thin Bi-rich layers (minor) with a chemical formula of BiI. Namely, Bi<sub>x</sub>I<sub>y</sub> presents a heterostructure formed by the stackings of BiI/BiI<sub>3</sub>/BiI. Like many halide perovskite single crystals, macro size Bi<sub>x</sub>I<sub>y</sub> can be grown using a low-cost, handy low-temperature solution method. Although the BiI phase in the Bi<sub>x</sub>I<sub>y</sub> nearly has no impact on chemical composition and resistivity, they effectively reduce the energy bandgap and, therefore reduce the <em>ε</em><em><sub>pair</sub></em>. High resistivity with small band gap, together with other merits such as good x-ray attenuation efficiency and charge collection ability, environmental friendly, wonderful environmental and hard radiation stability make Bi<sub>x</sub>I<sub>y</sub> be attractive as a potential competitor for high sensitive room temperature x-ray detection.
+
+## Bi<sub>x</sub>I<sub>y</sub> grown in solution
+
+The Bi<sub>x</sub>I<sub>y</sub> were grown by a low-temperature solution technique, as shown in Fig. <span class="InternalRef">1</span> a. Firstly, Bi<sub>2</sub>O<sub>3</sub>, I<sub>2</sub>, and Au were dissolved in a mixed hydroiodic acid and ethanol solution to form precursor solutions. The excessive iodine is used to dissolve gold in hydroiodic acid fully. Then the solution refinement technique was employed to grow high-quality crystals<sup><span class="CitationRef">9</span></sup>. Namely, the precursor solutions were pretreated by solvothermal in 1,4-butyrolactone for 3 times and then refined by hydrothermal for 1-time. Bi<sup>3+</sup> was reduced by Au during the solution pretreatment, the chemical reactions were:
+
+$$
+{Au}^{+}+ {I}^{-} \underset{\iff }{\varDelta } Au+ \frac{1}{2}{I}_{2}\uparrow
+$$
+
+$$
+Au + \frac{1}{2}{Bi}^{3+}\underset{\iff }{\varDelta } {Au}^{+}+ \frac{1}{2}{Bi}^{+}
+$$
+
+The iodine diffused into 1,4-butyrolactone and make it darken. Bi<sub>x</sub>I<sub>y</sub> with regular hexagonal shape and size up to 6 × 6 × 1<em>mm</em><sup><em><span class="CitationRef">3</span></em></sup> were obtained from refined solution after 14 days of water bath growth at room temperature without any disturbance, as shown in Fig. <span class="InternalRef">1</span> b. More times pretreatment in 1,4-butyrolactone would promote massive nucleation. Smooth surfaces and Large flexible flakes of the grown Bi<sub>x</sub>I<sub>y</sub> can be obtained by mechanical exfoliation, as shown in Fig. S1. The morphology image (Fig. S1) collected by scanning electron microscopy (SEM) reveals that the studied Bi<sub>x</sub>I<sub>y</sub> flake possesses a high quality exfoliated surface with no bubbles, holes, and inclusions. The morphology profile (Fig. S1) of Bi<sub>x</sub>I<sub>y</sub> flakes observed by atomic force microscopy (AFM) shows a layered structure. A step height of 0.66<em>nm</em> was obtained and assigned to I-Bi-Bi-I monolayer (0.65<em>nm</em> obtained by ac-STEM, see below), as shown in Fig S1.
+
+Bi<sub>x</sub>I<sub>y</sub> has the same x-ray diffraction (XRD) pattern (Fig. <span class="InternalRef">1</span> c) as BiI<sub>3</sub>, but exhibits broader peaks and preferred orientation along [001], indicates a softer nature of Bi<sub>x</sub>I<sub>y</sub>. The inductively coupled plasma atomic emission spectra (ICP-AES) result (35.2 wt.% Bi of as-grown Bi<sub>x</sub>I<sub>y</sub>) confirms the chemical composition is BiI<sub>3</sub> (35.4 wt.% Bi in calculation). However, thermogravimetric analysis (Fig. S4) shows a different thermal behavior between Bi<sub>x</sub>I<sub>y</sub> and BiI<sub>3</sub>. As the temperature exceeding 410 ℃, Bi<sub>x</sub>I<sub>y</sub> remains 46.2% of its original mass, on the other hand BiI<sub>3</sub> lost nearly all the mass. Meanwhile, no detectable mass loss is observed in the Bi<sub>x</sub>I<sub>y</sub> even at a temperature of 300 ℃, indicating its high thermal stability.
+
+The x-ray photoelectron spectroscopy (XPS) survey of Bi 4f in the freshly stripped smooth surface of Bi<sub>x</sub>I<sub>y</sub> shows the expected Bi<sup>3+</sup> peaks<sup>10</sup> at binding energies (BE) of 164.4<em>eV</em> and 159.1<em>eV</em> together with a distinct additional component shifted by 1.5<em>eV</em> towards lower BE (Fig. <span class="InternalRef">1</span> d), assigned as Bi<sup>+</sup> (see below). The stripped surface examined by XRD shows a heterostructure character of Bi<sub>x</sub>I<sub>y</sub> with a major diffraction of BiI<sub>3</sub> (00k) and a regular minor secondary diffraction, as shown in Fig. S2. The dramatic change of valance band BE from BiI<sub>3</sub> to Bi<sub>x</sub>I<sub>y</sub> indicates a significant difference of energy band between BiI<sub>3</sub> and the Bi<sub>x</sub>I<sub>y</sub>, as shown in Fig. S3.
+
+Then we used aberration-corrected STEM to observe the stacking sequence of Bi<sub>x</sub>I<sub>y</sub> by a high-angle annular dark-field (HAADF) imaging mode, which provides directly interpretable Z-contrast images at the atomic level<sup><span class="CitationRef">11</span>, <span class="CitationRef">12</span></sup>. Fig. S5 shows several cross-sectional STEM images of flakes with different thickness exfoliated from a grown crystal. As seen from Fig. S5, the flakes show alternate stacking of thin bright layers and thick dark layers. The bright layers are assigned to a Bi-rich phase due to the Z-contrast HAADF image<sup><span class="CitationRef">11</span>, <span class="CitationRef">12</span></sup>. It should be noted here that the thickness of the exfoliated Bi<sub>x</sub>I<sub>y</sub> is mainly dependent on the middle part of BiI<sub>3</sub>. The STEM-EDS mapping results in Fig. S5 also confirm a higher concentration of Bi in the bright layers. Detailed Z-contrast images of Bi-rich phase and the dark layers are shown in Fig. S6. Bi-rich phase exhibits a layered van der Waals structure built by the stacking of I-Bi-Bi-I four atomic layers with a chemical composition of BiI. On the other hand, the dark layers have a BiI<sub>3</sub> structure charactered by the staking of I-Bi-I three atomic layers<sup><span class="CitationRef">13</span></sup>, which is consistent with the BiI<sub>3</sub> atomic structure model. BiI<sub>3</sub> and BiI layers are also held together by weak van der Waals force in the Bi<sub>x</sub>I<sub>y</sub> structure, as shown in Fig. <span class="InternalRef">1</span> e. The STEM images clearly confirm the van der Waals heterostructure of Bi<sub>x</sub>I<sub>y</sub> constructed by stacking of thick BiI<sub>3</sub> and thin BiI layers, as shown in Fig. <span class="InternalRef">1</span> f.
+
+The BiI layers could be separated from the Bi<sub>x</sub>I<sub>y</sub> by mechanical exfoliation. As a result, we successfully obtained the thinnest BiI film composed of 7 I-Bi-Bi-I layers, as shown in Fig. S7. Unlike other Bi-rich bismuth iodides such as Bi<sub>4</sub>I<sub>4</sub> with 1D structure<sup><span class="CitationRef">14</span></sup>, the BiI we obtained is a new member of 2D family and could be used to build the blocks of van der Waals heterostructures with other 2D atomic crystals. Moreover, considering the versatility of 1D Bi<sub>4</sub>I<sub>4</sub> in thermoelectric, topological insulator, and superconductivity<sup><span class="CitationRef">15</span></sup>, the 2D BiI may also exhibit similar exciting properties.
+
+## Dramatically reduced bandgap and enhanced charge collection ability
+
+The bandgap of the Bi<sub>x</sub>I<sub>y</sub> was measured by UV–Vis–NIR diffuse reflectance spectroscopy (DRS) and shown in Fig. <span class="InternalRef">2</span> a. The reflectance of Bi<sub>x</sub>I<sub>y</sub> exhibits an indirect band nature with a sharp increase at 670–810<em>nm</em>, assigned to the thick BiI<sub>3</sub> layers, and a gentle rise after 810<em>nm</em>, caused by the coupling of BiI and BiI<sub>3</sub> layers. Based on the DRS result, the bandgap energy (<em>E</em><sub><em>g</em></sub>) of Bi<sub>x</sub>I<sub>y</sub> was obtained by the Tauc method with 0.70<em>eV</em>, dramatically lower than that of BiI<sub>3</sub> (1.67<em>eV</em>). The absorption spectrum (Fig. S8) of a typical grown Bi<sub>x</sub>I<sub>y</sub> also confirmed the small bandgap. According to Klein’s relation <em>ε</em><sub><em>pair</em></sub> = (2.67<em>ΔE</em> + 0.87)<em>eV</em> (<em>ΔE</em>: energy band gap)<sup><span class="CitationRef">16</span></sup>, the reduced band gap lowers the value of <em>ε</em><sub><em>pair</em></sub> and benefits the electron-hole pairs generation under x-ray radiation. Moreover, the reduced bandgap also promotes photo responses up to 1800<em>nm</em> in the Bi<sub>x</sub>I<sub>y</sub>, as shown in Fig. S11.
+
+Figure <span class="InternalRef">2</span> b shows ultraviolet photoelectron spectroscopy (UPS) of the Bi<sub>x</sub>I<sub>y</sub>, which reveals the valence band energy (<span class="InlineEquation"><span class="mathinline">\\({E}_{v}\\)</span></span>) can be obtained by <em>E</em><sub><em>v</em></sub> = 21.22–15.11 + 0.31 = 6.42<em>eV</em> by using the cutoff energy (15.11<em>eV</em>) and the energy (0.31<em>eV</em>) extracted from linear extrapolation in the low-binding-energy region. The conduction band energy (<em>E</em><sub><em>c</em></sub>) is then calculated by <em>E</em><sub><em>c</em></sub> = <em>E</em><sub><em>v</em></sub> + <em>E</em><sub><em>g</em></sub> = 5.72<em>eV</em>. Then the band diagram can be plotted as shown in Fig. <span class="InternalRef">2</span> c using the above data.
+
+From the band diagram, we then derived the electron-hole separation at the BiI<sub>3</sub>-BiI interface of the Bi<sub>x</sub>I<sub>y</sub>, as shown in Fig. <span class="InternalRef">2</span> e. The charge separation is further confirmed by transient absorption (TA) measurements of BiI<sub>3</sub> (film on a quartz plate prepared by a solution method described in ref. 17) and Bi<sub>x</sub>I<sub>y</sub>. As seen from Fig. S12, BiI<sub>3</sub> exhibits an absorption bleaching at 645–675<em>nm</em>, but no bleaching was observed in the Bi<sub>x</sub>I<sub>y</sub>. Photocarriers generated by incident laser can form full-filled electrons in the conduction band of BiI<sub>3</sub>, resulting in filled state bleaching. However, the bleaching disappeared in the Bi<sub>x</sub>I<sub>y</sub> due to photoexcited electron transferring from BiI<sub>3</sub> to BiI layers at the BiI<sub>3</sub>-BiI interface (Fig. <span class="InternalRef">2</span> e) instead of maintaining a filled state in the conduction band of BiI<sub>3</sub>, which directly confirms the charge separation at the BiI<sub>3</sub>-BiI interface. The charge separation is also confirmed by a much weaker fluorescence around 700 nm of Bi<sub>x</sub>I<sub>y</sub> than BiI<sub>3</sub>, as shown in Fig. S9. The separation of electrons and holes into different layers would weaken their recombination and promote charge collection. Moreover, Bi bilayers were observed in the middle of I-Bi-Bi-I four atomic layers (Fig. S6) and further confirmed by the Raman spectrum of bismuth (Fig. S10), which is promising for ultrafast electron transport because of the ultra-high electron mobility of bismuth<sup><span class="CitationRef">18</span></sup>.
+
+The enhanced charge separation and transportation indicate an excellent charge collection ability of the Bi<sub>x</sub>I<sub>y</sub>. The charge collection ability of x-ray detector is charactered by a <em>µτ</em> product, where <em>µ</em> is the carrier mobility and <em>τ</em> the carrier lifetime. We derived the <em>µτ</em> products by fitting the photoconductivity using a modified Hecht Eq. 1<sup>9</sup>:
+
+$$
+I=\frac{{I}_{0}\mu \tau V}{{L}^{2}}\frac{1-exp\left(-\frac{{L}^{2}}{\mu \tau V}\right)}{1+\frac{L}{V}\frac{s}{\mu }}
+$$
+
+Where <em>I</em><sub><em>0</em></sub> is the saturated photocurrent, <em>L</em> the thickness between electrodes, <em>V</em> the applied bias, and <em>s</em> the surface recombination velocity. The fitting curves of hole and electron photoconductivity at vertical and lateral device configurations (Fig. <span class="InternalRef">3</span> a) are shown in Fig. <span class="InternalRef">2</span> d.
+
+The Bi<sub>x</sub>I<sub>y</sub> have <em>µτ</em> products of 9.8 × 10<sup>− 3</sup><em>cm</em><sup><em><span class="CitationRef">2</span></em></sup><em>V</em><sup><em>− 1</em></sup> (lateral) and 5.2 × 10<sup>− 4</sup><em>cm</em><sup><em><span class="CitationRef">2</span></em></sup><em>V</em><sup><em>− 1</em></sup> (vertical) for electron, 7.7 × 10<sup>− 3</sup><em>cm</em><sup><em><span class="CitationRef">2</span></em></sup><em>V</em><sup><em>− 1</em></sup> (lateral) and 3.8 × 10<sup>− 4</sup><em>cm</em><sup><em><span class="CitationRef">2</span></em></sup><em>V</em><sup><em>− 1</em></sup> (vertical) for hole. Nearly the same <em>µτ</em> products of electron and hole indicate a balanced charge collection ability of Bi<sub>x</sub>I<sub>y</sub>. Like other 2D materials<sup><span class="CitationRef">7</span></sup>, Bi<sub>x</sub>I<sub>y</sub> shows a large discrepancy of <em>µτ</em> products between vertical and lateral directions due to its strong anisotropic structure. According to the electronic dimensionality theory<sup><span class="CitationRef">20</span></sup>, the electronic bands are more dispersive in the (001) plane of Bi<sub>x</sub>I<sub>y</sub> owing to the strong in-plane chemical bonds interaction, while more localized perpendicular to the (001) plane induced by the weak out-of-plane van der Waals interaction. Therefore, Bi<sub>x</sub>I<sub>y</sub> shows a better charge collection ability in the lateral direction, and the corresponding <em>µτ</em> products are comparable to that of single-crystal MAPbBr<sub>3</sub> (1.4 × 10<sup>− 2</sup><em>cm</em><sup><em><span class="CitationRef">2</span></em></sup><em>V</em><sup><em>− 1</em></sup>)<sup>21</sup> and MAPbI<sub>3</sub> (1.3 × 10<sup>− 2</sup><em>cm</em><sup><em><span class="CitationRef">2</span></em></sup><em>V</em><sup><em>− 1</em></sup>)<sup>8</sup>, exhibiting an excellent charge collection ability suitable for x-ray detection.
+
+## High resistivity and low detection limit
+
+High resistivity is expected to reduce background current noise and is thus crucial for detecting weak signals. The resistivities of Bi<sub>x</sub>I<sub>y</sub> calculated from leakage current versus electric field relation (Fig. <span class="InternalRef">3</span> e) are obvious anisotropic with values of 3.8×10<sup>10</sup><em>Ω cm</em> for lateral configuration and 2.4×10<sup>11</sup><em>Ω cm</em> for vertical configuration, respectively. As mentioned above, Bi<sub>x</sub>I<sub>y</sub> has the same composition with BiI<sub>3</sub> but a dramatic reduced bandgap, which would promote the generation of electron-hole pairs and benefit the x-ray detection. However, a smaller bandgap generally brings lower resistivity and more prominent current noise due to temperature-induced carrier excitation. Surprisingly, there are no apparent differences between the resistivities of the Bi<sub>x</sub>I<sub>y</sub> and BiI<sub>3</sub> (10<sup>9</sup>−10<sup>13</sup><em>Ω cm</em>, values extracted from polycrystalline and single crystal samples prepared by different methods and measured at different crystal orientations)<sup><span class="CitationRef">17</span>, <span class="CitationRef">22</span> – <span class="CitationRef">36</span></sup> even the bandgap changes so much. The Bi<sub>x</sub>I<sub>y</sub> is expected to be suitable for weak x-ray signal detection, benefiting from the coexistence of excellent charge collection and high resistivity.
+
+Bulk (2.1×2.1×0.4<em>mm</em><sup><em><span class="CitationRef">3</span></em></sup>) x-ray detectors were fabricated with a device structure of Cu/ Bi<sub>x</sub>I<sub>y</sub>/Cu. The copper conductive tapes (0.06 mm thickness, contact resistance about dozens of ohms) were pasted on both sides of the surfaces parallel or perpendicular to (001) to fabricate the vertical or lateral devices, as shown in Fig. <span class="InternalRef">3</span> a. The devices were exposed to a source with x-ray photon energy up to 70<em>keV</em>. Then x-ray induced photocurrents were measured by a normal intracavity direct radiation configuration and a cavity-edge leakage radiation configuration, as shown in Fig. <span class="InternalRef">3</span> a.
+
+As seen from Fig. <span class="InternalRef">3</span> b, the photocurrents of lateral and vertical devices measured by intracavity configuration under various x-ray dose rates reveal clear on/off responses. The lateral device exhibits larger photocurrent responses than that of the vertical one, attributing to a better charge collection ability. The signal-to-noise ratios (SNR) calculated from the on/off responses (method described in ref. 9) show stable and large average values around 2000 of the lateral device and 1600 of the vertical device, under various x-ray dose rates and bias larger than 1<em>V mm</em><sup><em>− 1</em></sup>, as shown in Fig. <span class="InternalRef">3</span> c.
+
+We also performed a leakage x-ray radiation measurement using a cavity-edge configuration (Fig. <span class="InternalRef">3</span> a) to examine the responses of the Bi<sub>x</sub>I<sub>y</sub> detector in a simulated radiation leakage environment. As seen from Fig. S13, the photocurrents induced by small leakage radiations still exhibit clear on/off responses in lateral and vertical devices. According to IUPAC standard, the dose rate with an SNR value of 3 is defined as the lowest detection limit at a given electric field. The lowest detection limit, representing the minimum x-ray dose rate used for inspection, is an important parameter relevant to health risk during x-ray security examinations or x-ray medical inspections<sup><span class="CitationRef">1</span>, <span class="CitationRef">2</span></sup>. As seen from Fig. <span class="InternalRef">3</span> d, the lowest detection limit of the lateral device achieved a very small value of 34<em>nGy s</em><sup><em>− 1</em></sup>, which is comparable to the excellent x-ray detectors with low detection limit<sup><span class="CitationRef">4</span> – <span class="CitationRef">8</span></sup>. Moreover, the lateral device could achieve a small detection limit of 210<em>nGy s</em><sup><em>− 1</em></sup> under 0<em>V</em> bias (self-powered) when a Si/Bi<sub>x</sub>I<sub>y</sub>/Bi device configuration was employed, as shown in Fig. S14. The detection limit of the vertical device also achieves a small value of 317<em>nGy s</em><sup><em>− 1</em></sup>, much lower than that required for regular medical diagnostics (5.5<em>µGy s</em><sup><em>− 1</em></sup>)<sup>37</sup>.
+
+## Ultra-sensitive x-ray response with highly stability
+
+As mentioned above, thick BiI<sub>3</sub> layers constitute main body of Bi<sub>x</sub>I<sub>y</sub> with outstanding x-ray radiation attenuation efficiency attributed to its high atomic number (<em>Z</em><sub><em>Bi</em></sub> = 83, <em>Z</em><sub><em>I</em></sub> = 53), and high density (5.8<em>g cm</em><sup><em>− 3</em></sup>). As seen from Fig. <span class="InternalRef">4</span> a, BiI<sub>3</sub> showed a much better x-ray attenuation efficiency than MAPbBr<sub>3</sub>. For 50<em>keV</em> hard x-ray, BiI<sub>3</sub> would attenuate 99.82% of the incident photons, while MAPbBr<sub>3</sub> 88.41% at 1<em>mm</em> thickness. Therefore, high attenuation efficiency enables Bi<sub>x</sub>I<sub>y</sub> to adequately absorb x-ray with reduced thickness, accelerating the charge collection.
+
+Strong x-ray attenuation, high resistivity, excellent charge collection ability and the resulting apparent responses at weak radiation indicate the Bi<sub>x</sub>I<sub>y</sub> is highly sensitive to x-ray. The sensitivity of x-ray detectors is derived from the current-dose rate relations, as shown in Fig. <span class="InternalRef">4</span> c. The lateral device had a much larger photocurrent density than the vertical device and achieved an ultra-high sensitivity of 1.3×10<sup>5</sup><em>µC Gy</em><sup><em>− 1</em></sup><em>cm</em><sup><em>− 2</em></sup> (Fig. <span class="InternalRef">4</span> d), which is comparable to the newly reported high sensitive detectors<sup><span class="CitationRef">4</span> – <span class="CitationRef">9</span></sup>. As for the vertical device, it provides an anisotropic sensitivity of 2.4×10<sup>3</sup><em>µC Gy</em><sup><em>− 1</em></sup><em>cm</em><sup><em>− 2</em></sup>, which is also comparable to the excellent x-ray detectors shown in Fig. S15.
+
+The Bi<sub>x</sub>I<sub>y</sub> detector was exposed to repeated and continuous 120<em>keV</em> x-ray (used for CT) with a dose rate of 204<em>µGy s</em><sup><em>− 1</em></sup> to evaluate the anti-radiation stability. As seen from Fig. <span class="InternalRef">4</span> e, Stable x-ray photocurrent with a high SNR around 1000 was observed after 204 circles repeated radiation and followed continuous radiation more than 4 h, confirms the highly stability of Bi<sub>x</sub>I<sub>y</sub> detector under high energy x-ray radiation. Moreover, nearly unchanged (Fig. <span class="InternalRef">4</span> f) responses of the Bi<sub>x</sub>I<sub>y</sub> detectors after 1 h water (20℃) immersion were observed, confirms its superior environmental stability. Ultra-sensitive and highly stable x-ray response of Bi<sub>x</sub>I<sub>y</sub> detector offers its great prospects in real applications.
+
+# Conclusion
+
+In summary, we developed a handy and scalable solution method to first grow macro-size van der Waals heterostructure of Bi$_x$I$_y$ with regular shapes consisting of adjacent thick BiI$_3$ (main) and thin BiI (minor) 2D layers. The Bi$_x$I$_y$ heterostructure exhibits a surprising high resistivity with small bandgap. X-ray detectors fabricated on the Bi$_x$I$_y$ exhibit stable response and anisotropic properties at different crystal orientation. The lateral device realized an ultra-high sensitivity of 1.3×10$^5$ µC Gy$^{-1}$ cm$^{-2}$ with a very low detection limit of 34 nGy s$^{-1}$, meeting the demands of medical inspection to reduce the x-ray exposure to the human body. On the other hand, the Bi$_x$I$_y$ photodetectors are versatile and present a photo response ranges up to 1800 nm, revealing its potential for near-infrared detection. Generally speaking, our results inspire the exploration of novel van der Waals heterostructure materials for ultra-sensitive x-ray detection.
+
+# methods
+
+**Precursor solution preparation.** All the purchased chemical reagents except Au (99.999%) were of analytical reagent grade purity and used without further purification. Solution 1 was prepared by 5.5 g Bi₂O₃ it was dissolved in 20 ml 55% hydroiodic acid at room temperature. Solution 2 was prepared by 2 g Au and 10 g I₂. They were dissolved in 5 ml 55% hydroiodic acid for 3 days at room temperature. Solution 3 was prepared by 10% hydroiodic acid mixed with ethanol in a volume ratio of 1:1. The solution 1 and solution 2 were mixed and diluted by solution 3 to 40 ml. The prepared 40 ml diluted solution was used as a precursor solution.
+
+**Solution pretreatment and refinement**. The pretreatment and refinement procedures are schematically illustrated in Fig. 1a. Briefly, the obtained precursor solution was put into a 50 ml Ф20 mm conical flask. The flask was placed in a sealed beaker with 100 ml 1, 4-butyrolactone (GBL) for solvothermal treatment. 3 times treatments are needed. The solvothermal treatment was performed in an 80℃ oven. After the first-time treatment (3-5 days), the solution was concentrated to 35 ml. The concentrated solution was diluted by solution 3 to 40 ml again for the second time solvothermal treatment. After the second time treatment, the solution was concentrated to 30 ml. The second time concentrated solution was diluted by ethanol to 35 ml for the third time solvothermal treatment. After the third time treatment, the solution was concentrated to 25 ml. The third time concentrated solution was then refined by hydrothermal treatment in an 60℃ oven for three days. Some small bulks with an irregular shape formed at the bottom of the conical flask after hydrothermal. The upper portion of the supernatant was then carefully transferred into another clear container to grow high-quality crystals.
+
+**Crystal growth**. The solution after pretreatment and refinement was then handled by water bath (Fig. 1a) at room temperature to grow crystals. More than 7 days of growth without disturbance is needed to obtain millimeter crystals. The obtained crystals were washed by ethanol 1 time and dichloromethane 2 times followed. Bulks after washing dried naturally in the air and used for the following material characterization, device preparation, and test.
+
+**Characterization**. Powder x-ray diffraction was performed on a D8-DISCOVER diffractometer with Cu Kα (λ = 1.542 Å) radiation. The X-ray Photoelectron Spectroscopy (XPS) and Ultraviolet Photoelectron Spectroscopy (UPS) were performed on a Thermo Fisher ESCALAB XI+ photoelectron spectrometer. Freshly exfoliated surface after 30 s Ar ion sputtering was used for XPS and UPS measurements. Thermogravimetric analysis (TGA) was carried out under continuous nitrogen flow using a NETZSCH STA 449F3 thermal gravimetric analyzer. The sample was held on a platinum pan, and heated at a rate of 5 ℃ min⁻¹ up to 600 ℃. AFM measurements were carried out in an Oxford Instruments Asylum Research Cypher S atomic force microscope with a contact mode. An IT500 scanning electron microscope (SEM) with a maximum 30 kV electron beam accelerating voltage was employed to observe the surface morphology of BiₓIᵧ. STEM observations of the cross-section specimens were carried out in an aberration-corrected STEM microscope (Titan G2 60-300, Thermofisher equipped with a field emission gun) with 300 kV electron beam accelerating voltage. The probe convergence angle was 24.5 mrad, and the angular range of the HAADF detector was from 79.5 to 200 mrad. The cross-sectional TEM specimens were prepared by a dual-beam focused ion beam (FIB) nanofabrication platform (Helios 600i, Thermofisher). The UV–Vis–NIR diffuse reflectance spectroscopy (DRS) was measured by a HATACHI UH4150 spectrometer over the spectral range of 360−2000 nm. Room temperature photoluminescence and Raman spectra were collected by a Horiba LabRam HR Evolution microscopic confocal Raman spectrometer using a 6.8 mW, 532 nm CW Nd: YAG laser as an excitation source. The laser beam was focused to a spot size of about 0.7 μm in diameter. Transient absorption (TA) measurements were performed on a HARPIA-TA system (Light Conversion) at room temperature. A 1030 nm pulsed laser with 100 kHz repetition rate and 190 fs pulse duration was divided into two beams to generate pump laser and probe light, respectively. The pump laser of 480 nm was generated from an optical parametric amplifier system (OPA, Light Conversion) pumped by one beam of 1030 nm laser. The probe light was generated by exciting a sapphire plate by another beam of 1030 nm laser.
+
+**Photo and x-ray response measurement:** The device for photo detection was fabricated on the (001) surface of a BiₓIᵧ bulk, as shown in Fig. S10. A pair of Ag electrodes with an interval of 0.5 mm was formed by painting Ag paste on a freshly exfoliated surface and then dried at 100 ℃ in the air. The area between Ag electrodes formed the light absorption area of the photo detector. The I–V characteristic under ambient light and infrared irradiation was measured by a KEITHLEY 2450 source meter. A YSL SC-PRO 7 supercontinuum source was used to generate CW infrared laser. Devices with Cu tape pasted on a pair of adjacent (100) or (001) surfaces formed the Cu/ BiₓIᵧ/Cu structure (Fig. 3a), and were used for x-ray detection measurements. The x-ray detection performance was measured in a Pb cavity for intracavity mode and in aФ5 mm hole on the side of the cavity with a light-proof cover for cavity-edge mode, as shown in Fig. S13. A commercially available MOXTEX MagPro Mini-X tube with a tungsten target and 12 W maximum power output was used as the x-ray source. The x-ray tube was operated with a constant 50 kV voltage. The total x-ray dose was modulated by changing the current of the x-ray tube. The radiation dose rate was calibrated using a Radical ion chamber dosimeter. The x-ray photocurrent was measured by a KEITHLEY 2636B source meter. For the anti-radiation test, a 150 kV HAMAMATSU/L12161-07 microfocus x-ray source with 75 W maximum power was used.
+
+# References
+
+1. Brenner, D. J. et al. Estimated risks of radiation induced fatal cancer from pediatric CT. Am. J. Roentgenol **176**, 289–296 (2001).
+2. Polischuk, B. T. et al. Selenium direct converter structure for static and dynamic X-ray detection in medical imaging applications. Proc. Med. Imag. 1998, 494–504 (1998).
+3. He, Y. H. et al. Detecting ionizing radiation using halide perovskite semiconductors processed through solution and alternative methods. Nat. Photon. **16**, 14–26 (2022).
+4. Song, Y. L. et al. Atomistic Surface Passivation of CH₃NH₃PbI₃ Perovskite Single Crystals for Highly Sensitive Coplanar-Structure X-Ray Detectors. Research. 2020, 5958243.
+5. Sarah, D. et al. High-sensitivity high-resolution X-ray imaging with soft-sintered metal halide perovskites. Nat Electron **4**, 681–688 (2021).
+6. Zhang, P. et al. Ultrasensitive and Robust 120 keV Hard X-Ray Imaging Detector based on Mixed-Halide Perovskite CsPbBr₃ – n In Single Crystals. Advanced Materials. **34**, 2106562 (2022).
+7. Zhuang, R. Z. et al. Highly sensitive X-ray detector made of layered perovskite-like (NH₄)₃Bi₂I₉ single crystal with anisotropic response. Nat. Photon. **13**, 602–608 (2019).
+8. Liu, Y. et al. Ligand assisted growth of perovskite single crystals with low defect density. Nat. Commun. **12**, 1686 (2021).
+9. Zhang, Y. X. et al. Nucleation-controlled growth of superior lead-free perovskite Cs₃Bi₂I₉ single-crystals for high-performance X-ray detection. Nat. Commun. **11**, 2304 (2020).
+10. Jain, S. M. et al. An effective approach of vapour assisted morphological tailoring for reducing metal defect sites in lead-free, (CH₃NH₃)₃Bi₂I₉ bismuth based perovskite solar cells for improved performance and long-term stability. Nano Energy **49**, 614–624 (2018).
+11. Pennycook, S. J. & Jesson, D. E. High-resolution incoherent imaging of crystals. Phys. Rev. Lett. **64**, 938 (1990).
+12. Pennycook, S. J. & Jesson, D. E. High-resolution Z-contrast imaging of crystals. Ultramicroscopy **37**, 14–38 (1991).
+13. Cho, S. B. et al. Intrinsic point defects and intergrowths in layered bismuth triiodide. Phys. Rev. Materials **2**, 064602 (2018).
+14. Filatova, T. G. et al. Electronic structure, galvanomagnetic and magnetic properties of the bismuth subhalides Bi₄I₄ and Bi₄Br₄. Journal of Solid State Chemistry **180**, 1103–1109 (2007).
+15. Pisoni, A. et al. Pressure effect and superconductivity in the β – Bi₄I₄ topological insulator. Phys. Rev. B **95**, 235149 (2017).
+16. Klein, C. A. et al. Bandgap Dependence and Related Features of Radiation Ionization Energies in Semiconductors. J. Appl. Phys. **39**, 2029–2038 (1968).
+17. Hamdeh, U. H. et al. Solution-Processed BiI₃ Thin Films for Photovoltaic Applications: Improved Carrier Collection via Solvent Annealing. Chem.Mater. **28**, 6567–6574 (2016).
+18. Wang, Y. W. et al. Engineering Ultrafast Charge Transfer in Bismuthene/Perovskite Nanohybrid. Nanoscale **11**, 2637–2643 (2019).
+19. Many, A. High-field effects in photoconducting cadmium sulphide. J. Phys. Chem. Solids **26**, 575–578 (1965).
+20. Xiao, Z. W. et al. Searching for promising new perovskite-based photovoltaic absorbers: the importance of electronic dimensionality. Mater. Horiz. **4**, 206–216 (2017).
+21. Wei, H. et al. Sensitive X-ray detectors made of methylammonium lead tribromide perovskite single crystals. Nat. Photon. **10**, 333–339 (2016).
+22. Lehner, A. J. et al. Electronic structure and photovoltaic application of BiI₃. Appl. Phys. Lett. **107**, 131109 (2015).
+23. Tiwari, D. et al. Above 600 mV Open-Circuit Voltage BiI₃ Solar Cells. ACS Energy Lett. **3**, 1882–1886 (2018).
+24. Nikolas, J. P. et al. Band gap and structure of single crystal BiI₃: Resolving discrepancies in literature. J. Appl. Phys. **114**, 033110 (2013).
+25. Saito, T. et al. BiI₃ single crystal for room-temperature gamma ray detectors. Nucl. Instrum. Methods Phys. Res. A **806**, 395–400 (2016).
+26. Sun, H. et al. Preparation and characterization of free-standing BiI₃ single-crystal flakes for X-ray detection application. J. Mater. Sci.: Mater. Electron. **29**, 20003–20009 (2018).
+27. Gokhale, S. S. et al. Growth, fabrication, and testing of bismuth triiodide semiconductor radiation detectors. Radiation Measurements **74**, 47–52 (2015).
+28. Azaree, T. et al. Characterization of bismuth tri-iodide single crystals for wide bandgap semiconductor radiation detectors. Nucl. Instrum. Methods Phys. Res. A **652**, 166–169 (2011).
+29. Aguiar, I. et al. Bismuth tri-iodide polycrystalline films for X-ray direct and digital imagers. Nucl. Instrum. Methods Phys. Res. A **610**, 332–334 (2009).
+30. Fornaro, L. et al. Bismuth Tri-Iodide Polycrystalline Films for Digital X-Ray Radiography Applications. IEEE Transactions on Nuclear Science **51**, 96–100 (2004).
+31. Liu, Y. Z. et al. Electrical properties of x-ray detector based on bismuth tri-iodide single crystal with electrode configuration considering. Mater. Res. Express **6**, 055902 (2019).
+32. Chaudhari, R. et al. Bismuth tri-iodide-polystyrene composite for X-rays switching applications at room temperature. Radiation Physics and Chemistry **186**, 10953 (2021).
+33. Dmitriev, Y. et al. Bismuth iodide crystals as a detector material: Some optical and electrical properties. Proc. SPIE **3766**, 521–529 (1999).
+34. Matsumoto, M. et al. Bismuth tri-iodide crystals for nuclear radiation detectors. Trans. Nucl. Sci. **49**, 2517–2520 (2002).
+35. Cuña, A. et al. Growth of bismuth tri-iodide platelets by the physical vapor deposition method. Cryst. Res. Technol. **39**, 912–919 (2004).
+36. Nason, D. et al. The growth and crystallography of bismuth tri-iodide crystals grown by vapor transport. J. Crystal Growth **156**, 221–226 (1995).
+37. Shearer, D. R. et al. Dose rate limitations of integrating survey meters for diagnostic X-ray surveys. Health Phys. **79**, S20-S21 (2000).
+38. Kasap, S. O. et al. X-ray sensitivity of photoconductors: application to stabilized a-Se. J. Phys. D **33**, 2853–2865 (2000).
+39. Li, L. Q. et al. Enhanced X-ray Sensitivity of MAPbBr₃ Detector by Tailoring the Interface-States Density. ACS Appl. Mater. Interfaces **11**, 7522–7528 (2019).
+40. Wei, W. et al. Monolithic integration of hybrid perovskite single crystals with heterogenous substrate for highly sensitive X-ray imaging. Nat. Photon. **11**, 315–321 (2017).
+41. Pan, W. S. et al. Cs₂AgBiBr₆ single-crystal X-ray detectors with a low detection limit. Nat. Photon. **11**, 726–732 (2017).
+42. Shen, Y. et al. Centimeter-Sized Single Crystal of Two-Dimensional Halide Perovskites Incorporating Straight-Chain Symmetric Diammonium Ion for X-Ray Detection. Angewandte Chemie **59**, 14896–14902 (2020).
+43. Zhang, H. J. et al. High-sensitivity X-ray detectors based on solution-grown caesium lead bromide single crystals. J. Mater. Chem. C **8**, 1248–1256 (2020).
+
+# Supplementary Files
+
+- [VdwBiISI.docx](https://assets-eu.researchsquare.com/files/rs-1263325/v1/35fb456ea024c768295a50ba.docx)  
+  Van der Waals heterostructure of BixIy for ultra-sensitive x-ray detection

0d4b1446098459448337799d0a9c230142d07fcc2787ff549923493af08f21a6/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Topochemical synthetic conductive black TiO2 NF films at room-temperature. (a) The in-situ divergent intercalation model and the synchronous color changes of the TiO2 NF film. (b) The Li+-gated slow intercalation model and the color changes of the film along the Li+-diffusion from left to right. (c) The electron-gated rapid intercalation model and the color changes of the film along the electron conduction from top to down. (d) A large piece of white TiO2 NF film and the blue and black LixTiO2-\u03b4 NF films treated by model 3. (e) The evolution of color and conductivity of the TiO2 NF film with intercalation depth (time).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Mechanism illustration of the intercalation pathways and the synchronous build-up of electron conduction paths. (a) Schematic diagram of the movement of electrons between Li-metal and TiO2. (b) Li+-interaction kinetics in different solvents by forming solvated Li/Li+. (c-e) The experimental models 1-3 and the schematic digrams of resident charges on TiO2 NFs. (f) Synchronous resident electrons and Li+-ions resulted in explosive intercalation path in model 1. (g) Random Li+-ion diffusion led to roundabout and slow Li+-intercalation path in model 2. (h) High concentration of Li+-ion induced fast and directional Li+-intercalation path in model 3.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Characterizations of the intercalated structures. (a-c) SEM images and digital photos of the white TiO2, LixTiO2-\u03b4 (model 2) and D- LixTiO2-\u03b4 (model 3) NF films. (d-f)TEM images, (g-i) Inverse FFT images, and (j) XRD patterns of the white TiO2, and black LixTiO2-\u03b4 and D-LixTiO2-\u03b4 NF films. High-resolution XPS spectra of (k) O 1s and (l) Ti 2p3/2 of the three film samples.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Simulation and calculation characterizations. (a-b) Optimized geometric structures and charge difference density mappings for the initial TiO2 and the Li-adsorbed TiO2. (c-d) PDOS plots of the initial TiO2 and the intercalated LixTiO2-\u03b4. (e) EPR spectra and (f) Raman spectra of TiO2 before and after the intercalation at room temperature.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Structures and structure stability of the conductive blue and black LixTiO2-\u03b4. SEM images of the white TiO2, blue LixTiO2-\u03b4and black D-LixTiO2-\u03b4 NFs treated by model 3 for (a) 1 s, (b) 1 min, and (c) 30 min, and visual displays of the conductivity of these NF films. (d, e) XRD patterns of D-LixTiO2-\u03b4 after the different intercalation reaction time. XPS spectra of (f) Li 1s, (g) O 1s and (h) Ti 2p3/2 of D-LixTiO2-\u03b4 with different intercalation reaction time. (i) High-resolution XPS spectra of Li 1s of the D-LixTiO2-\u03b4 NFs after being deeply etched for 100 nm and 200 nm. (j) The generated metastable intercalated black LixTiO2-\u03b4 that contained lots of surface oxygen vacancies and intercalated Li+-ions. (k) UV\u2013vis diffuse re\ufb02ection spectroscopy of the four film samples.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0d4b1446098459448337799d0a9c230142d07fcc2787ff549923493af08f21a6/preprint/preprint.md

@@ -0,0 +1,200 @@
+# Abstract
+
+Cation intercalation is an effective method to optimize the electronic structures of metal oxides, but manipulating the intercalation structure and conductivity is difficult. Here, we report a visual topochemical synthesis strategy to control the intercalation pathways, structures, and conductivity, and realize rapid synthesis of flexible conductive metal oxide films in 1 min at room-temperature. This study uses a flexible topological TiO₂ nanofiber film as the prototype and designs three different charge-driven models to intercalate the preset Li⁺-ions into the TiO₂ lattice slowly (µm/s), rapidly (mm/s) or ultrafast (cm/s). The Li⁺-intercalation causes real-time color changes of the TiO₂ films from white to blue and then black, corresponding to the new structures of blue LiₓTiO₂ and black TiO₂₋δ derived in TiO₂, and the enhanced conductivity from 0 to 1 and 40 S/m. This work realizes the rapid synthesis of flexible TiO₂ nanofiber films with tunable conductivity on large-scale and has been extended to synthesize other conductive metal oxide films for appealing applications in fast-charge electrodes, electrochromic pattern designs, catalysis and so on.
+
+Physical sciences/Materials science/Nanoscale materials/Nanowires  
+Physical sciences/Chemistry/Materials chemistry/Electronic materials  
+Physical sciences/Nanoscience and technology/Nanoscale materials/Synthesis and processing  
+Topochemical synthetic strategy  
+Flexible metal oxide nanofibers  
+Visual intercalation pathways  
+Charge-driven models  
+Tunable intercalation-gated conductivity
+
+# 1. Introduction
+
+Intercalation chemistry-based transition metal oxides (ITMO), presenting rich redox chemical states and metastable polymorphisms, show important applications in electrical energy storage and catalysis.<sup>1–3</sup> However, most ITMO exhibit slow cationic diffusion and poor electronic conductivity due to large bandgaps that limit their performance.<sup>4,5</sup> Various methods have been developed to tune the electronic structures of ITMO.<sup>6–9</sup> Among them, cation doping is an effective strategy to synthesize new materials with metastable lattice environments and different valence states, which takes advantage of the unbalanced kinetic properties of different cations.<sup>10</sup> ITMO such as TiO<sub>2</sub>, niobium pentoxide (Nb<sub>2</sub>O<sub>5</sub>), vanadium pentoxide (V<sub>2</sub>O<sub>5</sub>), et al. have opening tunnel networks and abundant states of valence for the metal, which make them good platforms for guest species intercalation to achieve modulation of ITMO properties. Moreover, the degree of freedom of the electronic structures can be tuned by doping different amounts of cations.<sup>11</sup>
+
+Traditional strategies for cation doping to tune the ITMO structures include solid-state reaction, hydrothermal/solvothermal treatment, and ionic heat treatment, but it is difficult to design and obtain target electronic structures with these methods.<sup>12–15</sup> Recently, numerous studies have been reported on the intercalation of alkali metal (lithium, sodium, potassium) cations into the host ITMO materials to engineer their electronic properties.<sup>16–18</sup> For example, Li<sup>+</sup>-ions can be intercalated into TiO<sub>2</sub> to form metastable lithium titanate, an attractive anode that exhibits extraordinary rate capability and shows great prospect for designing fast-charging Li<sup>+</sup>-ion batteries.<sup>16</sup> These studies indicate that using electrochemical intercalation reactions is a feasible strategy to adjust the electronic structures of ITMO. However, there are still some open problems. For one thing, the intercalation chemical reaction is like a black box and due to the lack of ability to control the intercalation thermodynamics and kinetics, which are considered as the main factors controlling the intercalation reactions, it is extremely difficult to intuitively observe and manipulate the intercalation process and structures.<sup>19–21</sup> Moreover, the regulation mechanism of metal oxides’ electronic conductivity by intercalation structures is far to be understood.
+
+Here, we report a topochemical synthesis strategy (namely intercalation paths) to visualize the real-time synchronous cation transport and electron conduction pathways and realize the control of intercalation structures and metal oxides’ conductivity. This proof-of-concept is studied by using a flexible topological TiO<sub>2</sub> nanofiber (NF) film as the prototype and designing three types of Li<sup>+</sup>-gated, electron-gated and Li<sup>+</sup>/electron co-gated charge-driven models to advance the preset Li<sup>+</sup>-ions to intercalate into TiO<sub>2</sub> lattices slowly (µm/s), rapidly (mm/s) or explosively (cm/s). Both the experiments and calculations show that the initial concentrations of the resident charges on TiO<sub>2</sub> NFs determine the electron conduction paths, which then establish the Li<sup>+</sup>-intercalation paths. The Li<sup>+</sup>-intercalation in all models led to the reduction of Ti<sup>4+</sup> to Ti<sup>3+</sup>, lattice expansion and the creation of oxygen vacancy in TiO<sub>2</sub> crystals, resulting in real-time color changes of the film from white to blue and then black, and the synchronous intercalation-based build-up of electron conduction pathways. Both the color and conductivity are closely related to the intercalated structures, which contained a low stable but high conductive black TiO<sub>2−δ</sub> structure (> 40 S/m) and a high stable but low conductive blue Li<sub>x</sub>TiO<sub>2</sub> structure (1 ~ 40 S/m). The conductivity of TiO<sub>2</sub> can be thus facilely tuned by controlling the intercalation process. Unlike previous studies on ITMO crystals, the real-time intercalation pathways were observed and controlled for the first time. Importantly, this strategy has been extended to other cation intercalations and oxides to synthesize conductive metal oxide films on large-scale at room-temperature for many applications in fast-charge electrodes, electrochromic pattern designs, catalysis and so on.
+
+# 2. Results
+
+## 2.1 Topochemical synthetic conductive black TiO₂ NF films at room-temperature
+
+Figure 1 shows the schematic diagram of topochemical synthetic conductive black TiO₂ NF films at room-temperature based on three charge-driven models. The 1st model is shown in Fig. 1a. In this model, a piece of white TiO₂ NF film is put on a Li-metal sheet, and the interface is infiltrated by a drop of dimethylacetamide (DMAc) solvent. The highly active Li-metal can quickly react with the TiO₂ at the interface and create lots of active Li-nanoparticles and solvated Li⁺@DMAc, which rapidly transfer into the film due to the Siphoning effect of the nano-porous film structure (Fig. S1). The active nanoparticles then react with the other TiO₂ NFs layer-by-layer, while the solvated Li⁺ intercalate into TiO₂ to jointly form black LiₓTiO₂−δ (0 ≤ x ≤ 1; 0 ≤ δ ≤ 2), where x and δ represent the intercalation intensity and the concentration of oxygen vacancy. As a result, the white TiO₂ NF film turns blue and then black within 1 min along the direction perpendicular to the interface (movie S1) and its conductivity increases sharply from 0 to > 40 S/m. This model is named in-situ divergent intercalation reaction since both the electrons and Li⁺-ions transfer divergently from the interface and establish divergent intercalation paths. We found that the changes of color and conductivity of the white TiO₂ NFs were caused either by the deprivation of lattice oxygen by Li-metal or the Li⁺-intercalation induced reduction of Ti⁴⁺ into Ti³⁺. 22, <span citationid="CR23" class="CitationRef">23</span> However, the specific roles of these two mechanisms are not clear.
+
+To make it clear, another two models are designed. The 2nd model is shown in Fig. 1b. In this model, a white TiO₂ NF film (2 cm × 3 cm) and a circular Li-metal film (d = 2.1 cm) are vertically outward immersed into the DMAc solvent, and then connected by a wire on the top (Fig. S2a-b). The distance between these two films is set as 10 cm. The solvated Li⁺@DMAc will transfer from the left Li-metal side to the right NF film side due to the concentration difference, and a continuous Li⁺-ion diffusion channel and a battery circuit are formed after 9 h. At this time, electrons will flow to the same place where the Li⁺-ions has just reached and initiate the intercalation reaction. This reaction advances gradually along the Li⁺-diffusion from left to right. As a result, the white TiO₂ NF film began to turn blue from the left side after 9 h, and it took another 15 h to turn black. Interestingly, the left side of the TiO₂ NF film was darker, indicating that more Li⁺-ions had been intercalated into the left side and the intercalation intensity influenced the color. This model is named Li⁺-gated slow intercalation reaction due to the slow Li⁺-diffusion. The NF film still showed a blue color and a small conductivity of ~ 1 S/cm after 3 h’s intercalation (Fig. S2c), and a black color and a high conductivity of ~ 40 S/m after 15 h’s intercalation.
+
+Fantastically, if adding some salt (LiClO₄) into the DMAc solvent and maintaining the other conditions unchanged (here is the 3rd model, Fig. 1c and S3a), the white TiO₂ NF film quickly turned black from the top to the bottom within 3 min (movie S2). The detail for the color change process of the white TiO₂ NF film is shown in Fig. S3b. In this model, many Li⁺-ions resided on the NF surface beforehand since there were sufficient Li⁺-ions in DMAc, and the intercalation immediately started when connecting these two films with a wire. As a result, the intercalation reactions occurred along the electron conduction, thus forming a reaction channel gated by electrons. Since the electron conduction was far greater than Li⁺-ion diffusion, the film quickly turned blue and black along the electron conduction paths from top to down, and this model is named electron-gated rapid intercalation reaction. Correspondingly, the intercalated LiₓTiO₂−δ film showed a black color (Fig. S3c) with a high electronic conductivity of ~ 40 S/m after 3 mins’ intercalation.
+
+## 2.2 Li⁺-intercalation pathways and the synchronous build-up of intercalation-based electron conduction paths in the topological TiO₂ NF films
+
+As for the driving forces of Li⁺-intercalation, the general understanding lies in that the intercalation is forced by either the solid-state physic mechanical force or the liquid-state chemical potential energy. There is an interconnected 3D tunnel network in TiO₂ crystal, and Li⁺-ions can overcome the barrier energy to intercalate into the lattice. In model 1, the strong reductant Li-metal can directly reduce the TiO₂ in contact with it due to the difference in energy level barrier (Fig. 2a) and the created active Li/Li⁺ can synchronously intercalate into the TiO₂ lattice. <span citationid="CR24" class="CitationRef">24</span> We found that some solvents, such as DMAc that contains carbonyl-O functional group, had the ability to pull or lock electrons to form active solvated Li/Li⁺ and could form strong Coulombic interaction with these Li⁺-ions (Fig. 2b). <span citationid="CR25" class="CitationRef">25</span>
+
+From the experimental phenomena, we can conclude that the synchronous conduction of electrons and ions is a necessary condition for intercalation, and the initial concentration of electrons and Li⁺-ions that reside on the TiO₂ NFs first determine the electron transfer paths, which then establish the Li⁺-intercalation paths. In addition, the intercalation rates are closely related to the conduction pathways of electrons and Li⁺-ions along the TiO₂ NF films. The intercalation-based build-up of electron conduction pathways are shown in Fig. 2f-h, corresponding to the three different models. In model 1, the large resident electrons and Li⁺-ions established divergent intercalation paths with an explosive intercalation rate. In model 2, due to the random Li⁺-diffusion paths (reminiscent to the random walk-in graph theory) controlled by concentration difference, the electron conduction is roundabout, and the paths are not the optimal, leading to a slow intercalation rate. In model 3, many Li⁺-ions resided on the NF surface beforehand, which would form multiple parallel high-speed channels for electron conduction. Therefore, through different topological charge-driven models, the intercalation intensity and conductivity of TiO₂ can be adjusted.
+
+## 2.3 Characterization of the blue and black intercalation structures
+
+Different intercalation pathways not only affect the intercalation structures, but also affect the morphology of the TiO₂ NFs. As shown in Fig. 3a, the as-fabricated flexible white TiO₂ film is composed of countless intertwined smooth NFs, and its electronic conductivity is close to 0. The large-scale fabrication of flexible TiO₂ NF films by electrospinning and their fire-resistant property are described in Fig. S4 and Movie S3. The intercalated conductive LiₓTiO₂−δ NF films in models 2 (Fig. 3b) and 3 (Fig. 3c) were characterized by scanning electron microscopy (SEM). Unlike the traditional high-temperature reduction strategy that easily led to brittle fracture of TiO₂ NFs, this room-temperature intercalation reaction maintained the flexibility of the NFs well. The NF surface became rough and contained precipitated particles caused by the formation of Li₂O. There were more precipitated particles on the TiO₂ NFs in model 3 and this film showed a darker color change (named as D-LiₓTiO₂−δ), indicating a higher intensive intercalation reaction.
+
+To explore the correlation of the intercalated structures with the color and conductivity, high-resolution transmission electron microscopy (HRTEM) images were recorded. The white TiO₂ NFs exhibited clear lattice fringes with an in-plane characteristic d-spacing of 0.352 nm (Fig. 3d), corresponding to the (101) lattice plane of anatase TiO₂. <span citationid="CR26" class="CitationRef">26</span> The low conductive blue LiₓTiO₂−δ NFs showed a slight increase of lattice distance in comparison with the white TiO₂ (Fig. 3e), and the high conductive D-LiₓTiO₂−δ NFs exhibited a larger lattice distance of 0.361 nm (Fig. 3f). The changes of d-spacing can be observed more intuitively from the inverse fast Fourier transform (FFT) images (Fig. 3g-i). Both samples exhibited lattice distortions after the intercalation, but the D-LiₓTiO₂−δ showed a higher degree of distortion. Therefore, the Li⁺-intercalation changed the electronic structure of TiO₂, and the lattice expansion increased with increasing the intercalation intensity.
+
+However, according to X-ray diffraction (XRD) patterns, the diffraction peaks for both LiₓTiO₂−δ and D-LiₓTiO₂−δ corresponded to anatase TiO₂, indicating that no obvious phase transition was induced after the intercalation (Fig. 3j). <span citationid="CR27" class="CitationRef">27</span> Nevertheless, there were enlarged (101) peaks shift toward lower angles as compared with the white TiO₂, confirming that there was lattice expansion in TiO₂ crystals and the formation of oxygen vacancies. <span citationid="CR28" class="CitationRef">28</span> X-ray photoelectron spectra (XPS) were conducted to further evaluate the intercalated structures. As shown in Fig. 3k, the high-resolution spectra of O 1s revealed three characteristic peaks attributed to lattice oxygen (O1), defects (O2), and chemisorbed oxygen species (O3), respectively. <span citationid="CR29" class="CitationRef">29</span> Compared with the white TiO₂, the oxygen peak shifted to higher binding energy, and the oxygen vacancies determined by the area ratio of O2 peak increased after the intercalation, indicating a higher oxidation state of oxygen in Ti-O. <span citationid="CR30" class="CitationRef">30</span> In addition, there were chemisorbed oxygen species for both LiₓTiO₂−δ and D-LiₓTiO₂−δ samples due to the existence of DMAc. <span citationid="CR31" class="CitationRef">31</span> On the other hand, the Ti 2p₃/₂ spectra (Fig. 3l) confirmed that the intercalation caused the reduction of Ti⁴⁺ to Ti³⁺, and the D-LiₓTiO₂−δ contained a higher Ti³⁺ concentration than the LiₓTiO₂−δ sample.
+
+These results confirmed that the Li⁺-intercalation did not change the crystalline phase of TiO₂, but introduced lattice oxygen vacancies and Ti³⁺ sites that distorted the lattice. In addition, the concentration of Ti³⁺ increased with increasing the intercalation intensity, and the Ti³⁺ species come from two different ways that caused the color changes. First, the Li⁺-intercalation in TiO₂ can form Ti³⁺ and result in the white to blue transformation of the TiO₂ NF film. Second, with the Li⁺-intercalation reactions going on and once the intercalation structure is saturated, Li⁺-ions will obtain electrons and deposit on the TiO₂ surface. These metallic Li can deprive the lattice oxygen to form vacancies, and Ti³⁺ defects will thus be generated due to charge compensation, resulting in a transformation of the blue TiO₂ to black. Therefore, both the Ti³⁺ and lattice oxygen vacancy can induce the color change, and the oxygen vacancy induced Ti³⁺ species can significantly enhance the film conductivity. <span citationid="CR32" class="CitationRef">32</span>
+
+## 2.4 Conductive mechanisms of the blue and black LiₓTiO₂−δ NF films
+
+The direct manifestation of Li⁺-intercalation is that the intercalated structures showed significantly enhanced electronic conductivity. According to the classical theory of electron conduction in crystals, electrons can conduct along the Ti-O bond or lattice oxygen vacancy. For the former, electrons on the highest band first jump to the adjacent empty band under an electric field and tunnel the Ti-O bond once a threshold concentration is reached. <span citationid="CR33" class="CitationRef">33</span> Accompanying changes include the conversion of Ti⁴⁺ into Ti³⁺, and the move forward of the remaining electrons to the next Ti⁴⁺ center to reside. At the same time, Li⁺-ions begin to intercalate and form conductive LiₓTiO₂−δ phases, which serve to pave a highway channel for the subsequent intercalation. For the latter, the oxygen vacancy with stoichiometric deviation from hypoxia of TiO₂−δ can form donor energy level to promote carrier migration. Due to the introduction of oxygen vacancies defects, the Fermi energy of TiO₂−δ can move into the conduction band, thus exhibiting higher conductivity. <span citationid="CR34" class="CitationRef">34</span> Therefore, the black and blue LiₓTiO₂−δ NFs with vacancy defects show high conductivity.
+
+On the other hand, the Li⁺-intercalation intensity and oxygen vacancy concentration of the topochemical synthetic intercalated D-LiₓTiO₂−δ materials were characterized. As shown in Fig. 4e, the room-temperature electron paramagnetic resonance (EPR) spectra confirmed the high contents of Ti³⁺ species and oxygen vacancies in the black intercalated structures. The g value of 2.003 in the black D-LiₓTiO₂−δ NFs was caused by oxygen vacancies, which was related to the formation of Ti³⁺-O⁻· radical. <span citationid="CR37" class="CitationRef">37</span> Besides, the Raman spectra of D-LiₓTiO₂−δ showed a similar configuration with the white TiO₂, but a clear blue-shift of peak and a broadened Eg peak were observed, further confirming the robust intercalated D-LiₓTiO₂−δ structure and the structural evolution by nonstoichiometric measurements (Fig. 4f and S5). <span citationid="CR38" class="CitationRef">38</span>
+
+## 2.5 Structures and structure stability of the conductive blue and black LiₓTiO₂−δ
+
+From the above results, the conductivity of the intercalated NF film can be manipulated by controlling the intercalation pathways and intensity, but the intercalation structure is still unclear. To determine the intercalation structure, intercalated films after being treated with different time in model 3 were prepared, and their electronic conductivities were visually displayed by using as wires to light the bubbles. As shown in Fig. 5a-c, with increasing the intercalation time, the TiO₂ NF film changed from white to blue and then to black, and more particles were gradually generated on the NFs. At the same time, the bulb became brighter, indicating a higher conductivity. Figure 5d-e showed the XRD patterns of the three samples with different intercalation intensity. The peaks of D-LiₓTiO₂−δ were obviously weakened and had a shift toward lower angles, implying that more Li⁺-ions intercalated into the TiO₂ lattices with increasing the time, which led to lattice distortion and rearrangement. <span citationid="CR39" class="CitationRef">39</span>
+
+Then, the intercalation depth was characterized by using XPS depth profiling tests. As shown in Fig. 5f, an increase of Li content was observed according to the high-resolution spectra of Li 1s with the progress of Li⁺-intercalation. Since lithium species might exist on the surface of D-LiₓTiO₂−δ, the intercalated samples were washed with hydrochloric acid three times before the XPS tests. Furthermore, the high-resolution spectra of O 1s and Ti 2p were provided to evaluate the oxygen defects of D-LiₓTiO₂−δ. As shown in Fig. 5g, the oxygen defect (cyan line) content of the sample that reacted for 30 min was significantly increased compared with the samples that reacted for 1 min and 1 s, indicating more oxygen defects generated in the TiO₂ lattice. The spectra of Ti 2p of the three samples showed two fitted peaks that ascribed to Ti⁴⁺ and Ti³⁺ species, and the Ti³⁺ (cyan line) contents were gradually increased with the reaction time (Fig. 5h). Although the Li⁺-intercalation created lots of vacancy defects, most of them existed on the TiO₂ surface. <span citationid="CR40" class="CitationRef">40</span> As shown in Fig. 5i, it could not detect lithium element in the D-LiₓTiO₂−δ NFs that were etched for 100 nm, indicating that the Li⁺-intercalation mostly occurred on the NF surface.
+
+In both model 1 and 3, the white TiO₂ NF films could turn black rapidly and both films exhibited high electronic conductivity of > 40 S/m, but the black films showed different sensitivity to air. The black LiₓTiO₂−δ NF film prepared by model 3 gradually turned into blue that could be kept for several months without further fading when exposed to the air. By contrary, the black LiₓTiO₂−δ NF film prepared by model 1 could be stable only when it was stored in dry air (Fig. S6a), and it would gradually change from black to grey white when exposed to the air (Fig. S6b), indicating that the as-prepared black LiₓTiO₂−δ was metastable, and it achieved self-repair of surface defects. Simultaneously, the conductivity of the faded NF film decreased to 0. In model 1, when contact with the TiO₂ NF film, Li-metal quickly deprives O-atom on TiO₂ crystal surface to form surface oxygen vacancies, thus leading to the rapid color change from white to black. However, these surface defects can introduce excessive charges to absorb water molecules in the air, then form hydroxyl species and fill the oxygen vacancy defects, resulting in the dynamic change of LiₓTiO₂−δ NF film from black to grey (Fig. S6c). <span citationid="CR41" class="CitationRef">41</span> These results indicated that the intercalated LiₓTiO₂−δ was metastable, and it contained a low stable but high conductive black TiO₂−δ structure and a high stable but low conductive blue LiₓTiO₂ structure (Fig. 5j).
+
+Interestingly, if heated the fresh metastable LiₓTiO₂−δ NF film in an inert atmosphere at 100 ℃ for 2 h, it exhibited excellent air stability and could be exposed to air for months without fading. The thermodynamics causes the transfer of surface oxygen vacancies into the TiO₂ lattice and forms stable defect structures. <span citationid="CR42" class="CitationRef">42</span> From the UV–vis diffuse reflection spectroscopy (Fig. 5k), the black LiₓTiO₂−δ film shows the highest light absorption capacity, while both the grey LiₓTiO₂−δ and white TiO₂ films can hardly absorb light. By contrast, in model 3, Li⁺-ions have priority to intercalate into TiO₂ lattice with a low energy under an electric field, rather than depriving the O-atom on the TiO₂ surface to form oxygen vacancies. This is due to that Li⁺-ions are fixed under the action of intermolecular forces, thus forming stable blue LiₓTiO₂−δ NF structures. But with more Li⁺-ions intercalation reactions, oxygen vacancies will also be formed and stable blue LiₓTiO₂−δ NF will become metastable black LiₓTiO₂−δ NF. It is worth mentioning that the Li⁺-intercalation could be precisely designed by controlling the intercalation time and paths, and both targeted metastable and stable LiₓTiO₂−δ materials could be synthesized with different models.
+
+# 3. Discussion
+
+In this work, we have reported a new topochemical synthesis strategy to study the intercalation chemical reactions by using TiO₂ NF as a prototype and realized controllable intercalation pathways and structures with tunable conductivity at room temperature. The energy barrier for Li⁺-intercalation into TiO₂ is generally manipulated by thermodynamics and kinetics. However, few studies have reported the intercalation pathways and the stability of the intercalated structures, which were found to be closely related to the lattice oxygen vacancies, the concentration of Ti³⁺-species, and the different Ti³⁺-species that created from different mechanisms.¹⁴³,¹⁴⁴ This proof-of-concept was confirmed by designing a series of topological intercalation reactions using three different charge-driven models to control the concentrations and transfer pathways of the resident Li⁺-ions and electrons, which realized designable Li⁺-intercalation into TiO₂ NFs accompanying with a clear color change from white to blue and black of the intercalated LiₓTiO₂−δ NFs. The intercalation reaction would occur along the Li⁺-ion diffusion when the resident electrons were sufficient, and along the electron conduction if there were enough resident Li⁺-ions. Therefore, Li⁺- and electron-gated intercalation reaction mechanisms were proposed by controlling the intercalation conditions. With these models, target LiₓTiO₂−δ containing unstable but high conductive black TiO₂−δ and stable but low conductive blue LiₓTiO₂ structures can be precisely designed.
+
+It is worth noting that both electrons and Li⁺-ions are necessary for the intercalation. For example, the white TiO₂ NF films did not change if there was no wire connection or replacing the Li-metal sheet with a graphite rod in model 3 (Fig. S7a), but the film quickly turned black if adding a 3.7 V DC power for the graphite electrode (Fig. S7b), confirming that electron is necessary for the color change. In addition, if only immersed the bottom half of the TiO₂ NF film in the solvent, the soaked part changed from white to black, but the upper part in the air was still white (Fig. S8), confirming that Li⁺-intercalation contributed to the color change. On the other hand, another two models 4 and 5 were built by adding a 3.7 V DC power for models 2 and 3 (Fig. S9) to confirm the effect of electron concentration. For model 4, it still took ~24 h for the color change from white to black, confirming that the Li⁺-gating is the main limiting factor. By contrast, it only took 1 min for the color change from white to black in model 5 (movie S4). The reduced time from 3 min to 1 min confirmed that the high electron concentration accelerated the intercalation reaction. The TiO₂ NFs obtained from model 1, 4 and 5 exhibited similar rough structures as models 2 and 3, but the surface roughness of these intercalated LiₓTiO₂−δ was different (Fig. S10). Moreover, from the XRD patterns (Fig. S11), it is obvious that the peak shift of LiₓTiO₂−δ obtained from model 1 was the largest, suggesting that the divergent intercalation was rapid and intense. According to the TEM and FFT images (Fig. S12), the lattice fringes in all models exhibited different degrees of distortions and the Raman spectra showed obvious blue shift and broadened Eg peaks after the Li⁺-intercalation (Fig. S13) due to the lattice distortions.¹⁴⁵ The O 1s spectra also confirmed that the intercalation caused the generation of oxygen vacancies, and the model with deep intercalation contained a higher oxygen defect (Fig. S14).
+
+Moreover, discussing in a broader sense, the proposed current-driven topological chemical synthesis strategy is universal and scalable. For one thing, replacing the Li-metal with other conductive materials such as graphite in model 5, the intercalation reactions occurred normally, and the white TiO₂ NF film also quickly changed into black. However, under the same Li⁺-ion concentration, the intercalation intensity corresponding to different opposite electrodes were different (Fig. S15), indicating that the potential difference between TiO₂ and the counter electrodes would affect the intercalation reactions. In addition, replacing the DMAC solvent with other solvents with similar electron-pulling ability, such as toluene, also does not affect this experimental verification.¹⁴⁶ For another, in addition to Li⁺-ions, Zn²⁺-ions or other cation ions with higher valence could also be intercalated into TiO₂ NFs with the model 4 prototype (Fig. S16). In addition to the fundamental interests around the new mechanisms of intercalation chemistry, the reported electrospinning method for the fabrication of flexible ITMO NF films is also scalable, and we have prepared many flexible ITMO fabrics on large scale, such as TiO₂, BaTiO₃, SnO₂, Nb₂O₅ and so on.¹⁴⁷–¹⁴⁹ These ITMO fabrics with superior softness like a napkin and controllable defects break the traditional perception of brittle and inert oxide ceramics.
+
+Metal oxides with open interpenetrating tunnel network structures can be intercalated by cations with small atomic radius to form ITMO.¹⁵⁰ However, most studies focused on crystals, which increased the difficulty of studying the intercalation pathways. In this study, the flexible TiO₂ NF film had a porous structure formed by stacking random arranged NFs, which not only enhanced the continuity of electron conduction, but also improved the Li⁺-ion transfer speed due to the siphoning effect of nanopore structures. This topological NF structure provided a platform for observing the directional transfer pathways of Li⁺-ions and electrons, thus it is expected to control the intercalation reaction and prepare intercalation oxide compounds that meet the practical requirements. Studying the intercalation pathways and the control of the intercalation structures is of great significance to develop high-performance functional ITMO materials for fast-charge electrodes, electrochromic pattern designs, catalysis, and so on.
+
+# 4. Conclusion
+
+In summary, we have reported a visual topochemical synthesis strategy to study the intercalation chemical reaction pathways, processes, and structures from a new perspective that different with the traditional study on ITMO. The main findings of this study include the following three aspects. First, the intercalation reaction pathways are visualized in real-time, which has not been reported before. The topological nanofiber structure provides a platform for observing the directional transfer pathways of Li<sup>+</sup>-ions and electrons, thus it is expected to control the intercalation reaction and prepare intercalation oxides that meet the practical requirements. Second, the stability mechanism of the intercalation structures and the regulation mechanism of metal oxides’ electronic conductivity by intercalation structures are clarified. An intercalation-based electron conduction pathway is established in the intercalation reactions, and the stability of the intercalated structures is found to be related to the lattice oxygen vacancies, the concentration and originality of Ti<sup>3+</sup>-species. Third, the topochemical synthesis strategy is a universal method for the rapid synthesis of conductive metal oxide films on large scale. The control of metal oxides‘ conductivity is of great significance to develop high-performance ITMO materials for appealing applications in fast-charge electrodes, electrochromic pattern designs, catalysis and so on.
+
+# Methods
+
+Fabrication of flexible topological TiO₂ NF films and the design of different charge-driven models. Flexible TiO₂ NF films were prepared with an electrospinning method followed by a high temperature calcination. Specifically, a clear solution was first obtained by dissolving PEO (Mw = 600,000) powder in a mixed solvent of acetic acid and ethanol with a weight ratio of 3:4. Then, titanium isopropoxide (TIP) was added into the clear solution followed by stirring until obtaining a transparent and homogeneous sol for electrospinning. Next, the spinning sol was injected at a speed of 10 ml/h and stretched by an applied voltage of 15 kV during the electrospinning process at room temperature and a humidity environment of 45% ± 2%. The precursor electrospun NF films were collected on a rotating collector, which was 150 mm from the spinneret. The last step was sintering the precursor films in a muffle furnace at 600 ℃ for 2 h in air with a heating rate of 2 ℃/min and then naturally cooling the furnace by turning off the heating button. Five charge-driven models were designed to verify the proposed strategy. Specifically, for model 2, 3, 4, and 5, TiO₂ NF films and circular Li-sheets were clamped and placed on both sides of the H-type electrolytic cells that contained different solvents. The other ends of the two alligator clips were connected by a wire. The differences in these four models are listed as follows. The solvent was DMAc in model 2 and 4, and the solvent was a mixture of DMAc and LiClO₄ in model 3 and 5. In addition, there was an output power device with an applied voltage of 3.7 V in model 4 and 5, but a naked wire in model 2 and 3. On the other hand, in model 1, a naked Li-metal was directly connected with a TiO₂ NF film, and the latter was infiltrated with ~ 1 ml of DMAc. The experiments were recorded with a camera in real-time.
+
+Material characterization. The structures of TiO₂ and the intercalated LiₓTiO₂−δ NF films were characterized by field emission SEM (Hitachi S-4800) and HRTEM (JEM-2100F). The crystal and chemical structures were checked via Bruker XRD with Cu Kα radiation, XPS (PHI 5000C ESCA System) and Raman (LabRAM HR Evolution). UV–visible diffuse reflectance spectra were obtained by a spectrophotometer (Hitachi U-3900). The electronic conductivity was measured by an ST-2258C multifunction digital four-probe tester.
+
+Computational simulation. DFT calculations were performed by using the Vienna Ab-initio Simulation Package (VASP) to study the adsorption of Li⁺-ions on {101} surface of TiO₂ and their electronic structures. The exchange–correlation interactions were described by generalized gradient approximation (GGA) with the functional Perdew–Burke–Ernzerhof (PBE). The cut-off energies for plane waves were set to be 500 eV, and the residual force and energy on each atom during structure relaxation were converged to 0.005 eV Å⁻¹ and 10⁻⁵ eV, respectively. The adsorption energy (Eₐd) is defined as: Eₐd = Eₛᵤᵣf+Li + Eₗᵢ − Eₛᵤᵣf, Where Eₛᵤᵣf+Li is the total energy of TiO₂ (101) adsorbing lithium, Eₗᵢ is the total energy of Li, Eₛᵤᵣf is the total energy of TiO₂ (101).
+
+# References
+
+1. Griffith, K., Wiaderek, K., Cibin, G., Marbella, L., Grey, C. Niobium tungsten oxides for high-rate lithium-ion energy storage. *Nature* **559**, 556–563 (2018).
+
+2. Tao, R., Zhang, T., Tan, S., Jafta, C., Li, C., Liang, J., Sun, X., Wang, T., Fan, J., Lu, Z., et al. Insight into the fast-rechargeability of a novel Mo₁.₅W₁.₅Nb₁₄O₄₄ anode material for high-performance lithium-ion batteries. *Adv. Energy Mater.* **12**, 2200519 (2022).
+
+3. Jin, X., Han, Y., Zhang, Z., Chen, Y., Li, J., Yang, T., Wang, X., Li, W., Han, X., Wang, Z., et al. Mesoporous single-crystal lithium titanate enabling fast-charging Li-ion batteries. *Adv. Mater.* **34**, e2109356 (2022).
+
+4. He, H., Huang, D., Pang, W., Sun, D., Wang, Q., Tang, Y., Ji, X., Guo, Z., Wang. H. Plasma-induced amorphous shell and deep cation-site S doping endow TiO₂ with extraordinary sodium storage performance. *Adv. Mater.* **30**, 1801013 (2018).
+
+5. Li, Y., Zhang, L. Synthesis of highly substitutional nitrogen doped TiO₂ via oxygen vacancy mediated strategy for ultrafast-charging lithium ion storage. *Chem. Eng. J.* **431**, 134164 (2022).
+
+6. Fan, M., Lin, Z., Zhang, P., Ma, X., Wu, K., Liu, M., Xiong, X. Synergistic effect of nitrogen and sulfur dual‐doping endows TiO₂ with exceptional sodium storage performance. *Adv. Energy Mater.* **11**, 2003037 (2021).
+
+7. Hao, Z., Chen, Q., Dai, W., Ren, Y., Zhou, Y., Yang, J., Xie, S., Shen, Y., Wu, J., Chen, W., Xu, G. Oxygen‐deficient blue TiO₂ for ultrastable and fast lithium storage. *Adv. Energy Mater.* **10**, 1903107 (2020).
+
+8. Liu, H., Zhu, Z., Yan, Q., Yu, S., He, X., Chen, Y., Zhang, R., Ma, L., Liu, T., Li, M., et al. A disordered rock salt anode for fast-charging lithium-ion batteries. *Nature* **585**, 63–67 (2020).
+
+9. Dahlman C., Heo, S., Zhang, Y., Reimnitz, L., He, D., Tang, M., Milliron, D. Dynamics of lithium insertion in electrochromic titanium dioxide nanocrystal ensembles. *J. Am. Chem. Soc.* **143**, 8278 (2021).
+
+10. Nikitina, V., Vassiliev, S., Stevenson, K. Metal‐ion coupled electron transfer kinetics in intercalation‐based transition metal oxides. *Adv. Energy Mater.* **10**, 1903933 (2020).
+
+11. Wu, Y., Li, D., Wu, C., Hwang, H., Cui, Y. Electrostatic gating and intercalation in 2D materials. *Nat. Rev. Mater.* **8**, 41–53 (2022).
+
+12. Yu, Z., Duan, Y., Kong, Y., Zhang X., Feng, X., Chen, Y., Wang, H., Yu, X., Ma, T., Zheng, X., et al. General synthesis of tube-like nanostructured perovskite oxides with tunable transition metal–oxygen covalency for efficient water electrooxidation in neutral media. *J. Am. Chem. Soc.* **144**, 13163–13713 (2022).
+
+13. Qi, Y., Huang, J., Yan, L., Cao, Y., Xu, J., Bin, D., Liao, M., Xia, Y. Towards high-performance aqueous zinc-ion battery via cesium ion intercalated vanadium oxide nanorods. *Chem. Eng. J.* **442**, 136349 (2022).
+
+14. Hu, Z., Xiao, X., Jin, H., Li, T., Chen, M., Liang, Z., Guo, Z., Li, J., Wan, J., Huang, L., et al. Rapid mass production of two-dimensional metal oxides and hydroxides via the molten salts method. *Nat. Commun.* **8**, 15630 (2017).
+
+15. Barnes, P., Zuo, Y., Dixon, K. Hou, D., Lee, S., Ma, Z., Connell, J., Zhou, H., Deng, C., Smith, K., et al. Electrochemically induced amorphous-to-rock-salt phase transformation in niobium oxide electrode for Li-ion batteries. *Nat. Mater.* **21**, 795–803 (2022).
+
+16. Zhang, W., Seo, D., Wu, T., Topsakal, M., Zhu, Y., Lu, D., Ceder, G., Wang, F. Kinetic pathways of ionic transport in fast-charging lithium titanate. *Science* **367**, 1030 (2020).
+
+17. Park, H., Kwon, J., Song, T., Palk, U. Microstructural control of new intercalation layered titanoniobates with large and reversible d-spacing for easy Na⁺ ion uptake. *Sci. Adv.* **3**, e1700509 (2017).
+
+18. Kuai, X., Li, K., Chen, J., Wang, H., Yao, J., Chiang, C., Liu, T., Ye, H., Zhao, J., Lin, Y., et al. Interfacial engineered vanadium oxide nanoheterostructures synchronizing high-energy and long-term potassium-ion storage. *ACS Nano* **16**, 1502–1510 (2022).
+
+19. Griffith, K., Seymour, I., Hope, M., Butala, M., Lamontagne, L., Preefer, M., Koçer, C., Henkelman, G., Morris, A., Cliffe, M., Dutton, S., Grey, C. Ionic and electronic conduction in TiNb₂O₇. *J. Am. Chem. Soc.* **141**, 16706–16725 (2019).
+
+20. Hosono, H., Kitano, M. Advances in materials and applications of inorganic electrides. *Chem. Rev.* **121**, 3121–3185 (2021).
+
+21. Chen, W., Zhan, X., Yuan, R., Pidaparthy, S., Yong, A., An, H., Tang, Z., Yin, K., Patra, A., Jeong, H., et al. Formation and impact of nanoscopic oriented phase domains in electrochemical crystalline electrodes. *Nat. Mater.* **22**, 92–99 (2023).
+
+22. Yan, J., Zhang, Y., Zhao, Y., Song, J., Xia, S., Liu, S., Yu, J., Ding, B. Transformation of oxide ceramic textiles from insulation to conduction at room temperature. *Sci. Adv.* **6**, eaay8538 (2020).
+
+23. Wu, Y., Yang, W., Fan, Y., Song, Q., Xiao, S. TiO₂ metasurfaces: From visible planar photonics to photochemistry. *Sci. Adv.* **5**, eaax0939 (2019).
+
+24. Ou, G., Xu, Y., Wen, B., Lin, R., Ge, B., Tang, Y., Li, Y. Tuning defects in oxides at room temperature by lithium reduction. *Nat. Commun.* **9**, 1 (2018).
+
+25. Lim, J., Park, K., Lee, H., Kim, J., Kwak, K., Cho, M. Nanometric water channels in water-in-salt lithium ion battery electrolyte. *J. Am. Chem. Soc.* **140**, 15661–15667 (2018).
+
+26. Song, J., Sun, G., Yu, J., Si, Y., Ding, B. Construction of ternary Ag@ZnO/TiO₂ fibrous membranes with hierarchical nanostructures and mechanical flexibility for water purification. *Ceram. Int.* **46**, 468–475 (2020).
+
+27. Zhang, Z., Hedhili, N., Zhu, H., Wang, P.. Electrochemical reduction induced self-doping of Ti³⁺ for efficient water splitting performance on TiO₂ based photoelectrodes. *Phys. Chem. Chem. Phys.* **15**, 15637–15644 (2013).
+
+28. Ni, Q., Dong, R., Bai, Y., Wang, Z., Ren, H., Seam, S., Wu, F., Xu, H., Wu, C. Superior sodium-storage behavior of flexible anatase TiO₂ promoted by oxygen vacancies. *Energy Storage Mater.* **25**, 903–911 (2020).
+
+29. Wang, C., Yang, J., Li, T., Shen, Z., Guo, T., Zhang, H., Lu, Z. In situ tuning of defects and phase transition in titanium dioxide by lithiothermic reduction. *ACS Appl. Mater. Interfaces* **12**, 5750–5758 (2020).
+
+30. Abdullah, S., Sahdan, M., Nayan, N., Embong, Z., Hak, C., Adriyanto, F. Neutron beam interaction with rutile TiO₂ single crystal (1 1 1): Raman and XPS study on Ti³⁺-oxygen vacancy formation. *Mater. Lett.* **263**, 127143 (2020).
+
+31. Idriss, H. On the wrong assignment of the XPS O1s signal at 531–532 eV attributed to oxygen vacancies in photo- and electro-catalysts for water splitting and other materials applications. *Surf. Sci.* **712**, 121894 (2021).
+
+32. Li, J., Li, J., Ren, J., Hong, H., Liu, D., Liu, L., Wang, D. Electric-field-treated Ni/Co₃O₄ film as high-performance bifunctional electrocatalysts for efficient overall water splitting. *Nano-Micro Letters* **14**, 148 (2022).
+
+33. Wang, Wu, T., Lin, J., Sheng, P. A critical path approach for elucidating the temperature dependence of granular hopping conduction. *Front. Phys.* **13**, 137205 (2018).
+
+34. Zhang, X., Tang, F., Wang, M., Hu, H., Li, Y., Friend, R., Song, X. Femtosecond visualization of oxygen vacancies in metal oxides. *Sci. Adv.* **6**, eaax9427 (2020).
+
+35. Lazzeri, M., Vittadini, A., Selloni, A. Structure and energetics of stoichiometric TiO₂ anatase surfaces. *Phys. Rev. B* **63**, 155409 (2001).
+
+36. Cao, Y., Zhou, P., Tu, Y., Liu, Z., Dong, B. W., Azad, A., Ma, W. Modification of TiO₂ Nanoparticles with organodiboron molecules inducing stable surface Ti³⁺ complex. *iScience* **20**, 195–204 (2019).
+
+37. Li, G., Dimitrijevic, M., Chen, L., Nichols, M., Rajh, T., Gray, A. The important role of tetrahedral Ti⁴⁺ sites in the phase transformation and photocatalytic activity of TiO₂ nanocomposites. *J. Am. Chem. Soc.* **130**, 5402–5403 (2008).
+
+38. Li, H., Wang, S., Wang, M., Gao, Y., Tang, J., Zhao, S., Chi, H,. Zhang, P., Qu, J., Fan, F., Li, C. Enhancement of plasmon-induced photoelectrocatalytic water oxidation over Au/TiO₂ with lithium intercalation. *Angew. Chem. Int. Ed.* **61**, e202204272 (2022).
+
+39. Wang, J., Hou, K., Wen, Y., Liu, H., Wang, H., Chakarawet, K., Gong, M., Yang, X. Interlayer structure manipulation of iron oxychloride by potassium cation intercalation to steer H₂O₂ activation pathway. *J. Am. Chem. Soc.* **144**, 4294 (2022).
+
+40. Li, Z., Zhou, C., Hua, J., Hong, X., Sun, C., Li, H., Xu, X., Mai, L. Engineering oxygen vacancies in a polysulfide-blocking layer with enhanced catalytic ability. *Adv. Mater.* **32**, e1907444 (2020).
+
+41. Nandasiri, M., Shutthanandan, V., Manandhar, S., Schwarz, A., Oxenford, L., Kennedy, J., Thevuthasan, S., Henderson, M. Instability of hydrogenated TiO₂. *J. Phys. Chem. Lett.* **6**, 4627–4632 (2015).
+
+42. Liu, C., Wu, P., Wu, J., Hou, J., Bai, H., Liu, Z. Effective protect of oxygen vacancies in carbon layer coated black TiO₂₋ₓ/CNNS hetero-junction photocatalyst. *Chem. Eng. J.* **359**, 58 (2019).
+
+43. Whittingha, Meekins, B., Kamat, P. Got TiO₂ Nanotubes? Lithium ion intercalation can boost their photoelectrochemical performance. *ACS Nano* **3**, 3437–3446 (2009).
+
+44. Tsui, K., Saito, M., Homma, T., Zangari, G. Trap-state passivation of titania nanotubes by electrochemical doping for enhanced photoelectrochemical performance. *J. Mater. Chem. A* **3**, 360–367 (2015).
+
+45. Wang, Z., Yang, C., Lin, T., Yin, H., Chen, P., Wan, D., Xu, F., Huang, F., Lin, J., Xie, X., Jiang, M. Visible-light photocatalytic, solar thermal and photoelectrochemical properties of aluminium-reduced black titania. *Science* **6**, 3007–3014 (2013).
+
+46. Yan, J., Yu, J., Ding, B. Mixed ionic and electronic conductor for Li-metal anode protection. *Adv. Mater.* **30**, 1705105 (2018).
+
+47. Yan, J., Zhang, Y., Zhao, Y., Song, J., Xia, S., Liu, S., Yu, J., Ding, B. Superior flexibility in oxide ceramic crystal nanofibers. *Adv. Mater.* **33**, 2105011 (2021).
+
+48. Yan, J., Han, Y., Xia, S., Wang, X., Zhang, Y., Yu, J., Ding, B. Polymer template synthesis of flexible BaTiO₃ crystal Nanofibers. *Adv. Funct. Mater.* **29**, 1907919 (2019).
+
+49. Lin, X., Xia, S., Zhang, L., Zhang, Y., Sun, S., Chen, Y., Chen, S., Ding, B., Yu, J., Yan, J. Fabrication of flexible mesoporous black Nb₂O₅ nanofiber films for visible‐light‐driven photocatalytic CO₂ reduction into CH₄. *Adv. Mater.* **34**, e2200756 (2022).
+
+50. Whittingha, M. S. *Intercalation chemistry*. Ch. 16 (Elsevier, Press 2012).
+
+51. Perdew, J., Ernzerhof, M., Burke, K. Rationale for mixing exact exchange with density functional approximations. *J. Chem. Phys.* **105**, 9982 (1996).
+
+# Supplementary Files
+
+- [SupplementaryInformation.docx](https://assets-eu.researchsquare.com/files/rs-2991823/v1/bab5d7e3b381946f8d849677.docx)
+- [MovieS1.mp4](https://assets-eu.researchsquare.com/files/rs-2991823/v1/dc2db454b7a0613662ea8e35.mp4)  
+  Supplementary Movie S1
+- [MovieS2.mp4](https://assets-eu.researchsquare.com/files/rs-2991823/v1/3320361cb672a91cab8ba5cb.mp4)  
+  Supplementary Movie S2
+- [MovieS3.mp4](https://assets-eu.researchsquare.com/files/rs-2991823/v1/13acfb748aecef1df1c0863d.mp4)  
+  Supplementary Movie S3
+- [MovieS4.mp4](https://assets-eu.researchsquare.com/files/rs-2991823/v1/baf168f153851dd70af74185.mp4)  
+  Supplementary Movie S1

0d52f34ef10080a11c8fa1ec39fd540697528a60223e57f19842ced1e8fac875/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Trial profile and clinical response in r/r MM patients treated with CAR-T cell infusion. (A) Patient enrollment. (B) Anti\u2013BCMA single-chain variable fragment (scFv), a hinge and transmembrane regions, and 4-1BB costimulatory moiety, and CD3\u03b6 T-ce  ll activation domain. (C) Blood and fecal sample collection. (D) Clinical response; CRS grade distribution in 43 r/r MM patients. (E) Numbers of BCMA CAR-T cell percentages in PB assessed by FACS in different therapy stages after CAR-T cell infusion and serum concentrations of IL-6 and IFN-\u03b3 in different therapy stages among the CR, EPR, and PR groups. (F) Body temperature and serum concentrations of IL-6 and IFN-\u03b3 in different therapy stages among CRS grade groups. (G) Representative MM patients with impressive antimyeloma response. Positron emission tomography-computed tomography scans before and five months after CAR-T cell treatment showing complete elimination of large number of MM bone metastases. Before receiving CAR-T cell infusion, 43.5% of bone marrow cells of the patient were plasma cells, but after 1.5 months of infusion, dramatic eradication of MM from the bone marrow was observed; and MM cells became undetectable by flow cytometry.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Changes of microbial composition during CAR-T therapy. (A) Simpson diversity indices of gut microbiome across CAR-T stages in all myeloma patients by Wilcoxon rank-sum test. (B) Pairwise Spearman correlation of OTU-level bacterial abundance across different timepoints. Rho value for each significant correlation is labelled inside box. (C) Stacked bar plot of mean phylum-level phylogenetic composition of bacterial taxa in myeloma patients across therapy stages. (D, E) Relative abundance of phyla Firmicutes and Bacteroidetes across therapy stages. Significance was assessed by Wilcoxon rank-sum test. (F) Longitudinal analysis by Qiime2 \u201cfeature-volatility\u201d plugin to identify taxonomic features associated with therapy stages. Important genus-level features are labelled.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Association of compositional differences in gut microbiome with responses to CAR-T therapy. (A) Shannon diversity indices of gut microbiome differed between CR and PR groups across CAR-T stages. Significances were assessed by Wilcoxon rank-sum test. (B) Principal coordinate analysis of fecal samples by response using weighted UniFrac distances. (C) Summary of number of PR or CR-enriched OTUs in different therapy stages. Difference between CR and PR groups was assessed by Wilcoxon rank-sum test. P value significant cutoff was 0.05. (D) Heatmap for abundance of OTUs with significant temporal differences between CR and PR groups identified by maSigPro. Rows denote bacterial OTUs grouped into three sets according to regression coefficients and sorted by mean abundance within each set. Individual samples are organized in columns, with colored bars representing response group and therapy stage. (E) Profiles of significant gene clusters correspond to (D). Solid lines denote median profile of abundance of OTUs within cluster for each experimental group through time. Fitted curve of each group is displayed as dotted line. (F) Phylogenetic composition of OTUs within each cluster in (D) at phylum and order levels. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Determination of correlated genera with clinical response to CAR-T therapy. (A) Differentially abundant genera in CR and PR patients were identified by LefSe and maSigPro. Bar plots denote linear discriminant analysis (LDA) scores computed for differentially abundant genera in CR (blue) and PR (red) groups using LefSe. P < 0.05 for Kruskal-Wallis H statistic; LDA score > 2. Bubble plot on left marked p values from temporal group difference analysis for each genus. Bubble size and color are proportional to log-transformed p value. (B) Mean bacterial abundance (log transformed) of CR, VGPR, and PR myeloma patents before and after CAR-T cell infusion. Significances tested with Wilcoxon rank-sum test; * p < 0.05. (C) Relative abundance (log transformed) of top discriminative signatures at baseline (FCa) timepoint identified by RF feature selection procedure. Genera with highest scores of mean decreases in Gini were selected. Importance scores in RF classification model and fold-change levels in log2 scale are noted below plot for each genus. Underlined genera are those identified at both baseline and post-chemotherapy stages. (D) Same as panel C for post-chemotherapy (FCb) timepoint. Only signatures enriched in CR patents are displayed. Those depleted in CR patents are displayed in Fig. S2C. (E) Receiver operating characteristic (ROC) curve of RF model using discriminatory genera as predictors for baseline timepoint. (F) Same as panel E for post-chemotherapy timepoint. (G) Kaplan-Meier (KM) plot of PFS curves by log-rank test for patients with high (dark blue), median (green). or low (red) abundance of Sutterella. Abundance of genus Sutterella was in terms of median abundance of all timepoints. (H) Differential KEGG pathways in CR and PR groups measured by Welch\u2019s t-test. Bar plot on left denotes mean proportion of each pathway in CR and PR groups. Dot plot on right depicts difference in mean proportion. Blue and red dots represent pathways enriched in CR and PR groups, respectively. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Compositional differences between subjects with different CRS grades. (A) Differentially abundant genera in severe (CRS = 3) and mild (CRS = 1) CRS groups identified by LefSe and maSigPro. Bar plots denote linear discriminant analysis (LDA) scores computed for differentially abundant genera in CRS grades 1 (green) and 3 (orange) groups by LefSe (p < 0.05 and LDA score > 2). Bubble plot on left marks p values from temporal group difference analysis for each genus calculated by maSigPro. Bubble size and color are proportion to log-transformed p value. (B) Mean bacterial abundance in MM patients with different CRS grades before and during occurrence of CRS. Significances were assessed with Wilcoxon rank-sum test. (C) Concentrations of immune cells and inflammatory markers in different CRS grades across therapy stages. Significances were assessed by Kruskal-Wallis test. (D) Network representing correlations between gut microbes (blue nodes) and immune cells and inflammatory markers (green nodes) at FDR < 0.2 and \u03c1 > 0.2. Red edges indicate positive correlations and blue edges negative correlations. Edge width is proportional to correlation coefficient (\u03c1) calculated by Spearman correlation test. Only genera identified as associated with clinical response and CRS grade were included in correlation analysis. (E) Correlation plots for Leuconostoc and correlated immune cells and inflammatory markers from network shown in (D). Color of dots represents CRS grades. (F) Same as (E) for Bifidobacterium. p < 0.1; * p < 0.05; ** p < 0.01. \n\n",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0d52f34ef10080a11c8fa1ec39fd540697528a60223e57f19842ced1e8fac875/preprint/preprint.md

@@ -0,0 +1,231 @@
+# Abstract
+
+Chimeric antigen receptor (CAR)-T cell therapy has emerged as a promising immunotherapeutic treatment for hematologic malignancies. By comparing the diversity and composition of the gut microbiome during different stages of CAR-T therapy, significant changes were detected, not only in patients with relapsed/refractory multiple myeloma (MM; n = 43), but also in those with acute lymphocytic leukemia (ALL; n = 23) and non-Hodgkin lymphoma (NHL; n = 12). Analysis of treatment responses revealed significant temporal differences in diversity and abundance of *Bifidobacterium*, *Prevotella*, *Sutterella*, and *Collinsella* between MM patients in complete remission (n = 24) and those in partial remission (n = 11). Furthermore, we found that patients with severe cytokine release syndrome (CRS) exhibited higher abundance of *Bifidobacterium*, *Leuconostoc*, *Stenotrophomonas*, and *Staphylococcus*. This study has important implications for understanding the biological role of the microbiome in the CAR-T treatment of patients with hematologic malignancies (ChiCTR1800017404).
+
+Oncology  
+Translational Medicine  
+Chimeric Antigen Receptor T Cell Therapy  
+Multiple Myeloma  
+Acute Lymphocytic Leukemia  
+Non-Hodgkin Lymphoma  
+Cytokine Release Syndrome
+
+# Introduction
+
+B-cell-derived hematologic malignancies, including acute lymphoblastic leukemia (B-ALL), non-Hodgkin lymphoma (B-NHL), and multiple myeloma (MM), carry a high probability of relapse after conventional chemotherapy<sup>1</sup>. With novel therapeutic strategies incorporating monoclonal antibodies, bispecific T-cell engager (BiTE) antibodies, and hematopoietic stem cell transplantation (HSCT), treatment outcomes have greatly improved<sup>2,3,4</sup>. However, some patients progress to relapsed/refractory (r/r) status, with a poor prognosis<sup>5</sup>. The 5-year overall survival (OS) rate generally is < 10% with a median OS of 3–6 months for patients with r/r B-ALL<sup>6,7</sup>. The complete response (CR) rate is 7% with a median OS of 6.2 months for r/r diffuse large B-cell lymphoma (DLBCL)<sup>8</sup>. For r/r MM patients, the 1-year OS is about 40%<sup>9</sup>. There is an urgent need to explore novel treatment strategies for these malignancies.
+
+Chimeric antigen receptor (CAR) T-cell therapy (approved by the U.S. Food and Drug Administration) recently emerged as promising for r/r B-ALL, DLBCL, and mantle cell lymphoma (MCL)<sup>10,11,12</sup>. In multiple myeloma, investigations targeting the B-cell maturation antigen (BCMA) yielded encouraging outcomes with reversible toxic effects such as cytokine release syndrome (CRS) and pancytopenia<sup>13,14,15,16,17</sup>. However, the efficacy and toxicity have been inconsistent. No biomarker has been identified that predicts outcome and associated toxicities after CAR-T in patients.
+
+Several studies have reported that the differences in diversity and composition of the gut microbiome might influence cancer immunotherapy response<sup>18,19,20,21</sup>. After analyzing fecal samples from 43 melanoma patients treated with anti-programmed cell death 1 protein (PD-1) immunotherapy, significantly higher alpha diversity and abundance of Clostridiales/Ruminococcaceae were found in responders, whereas Bacteroidales were significantly enriched in non-responders<sup>19</sup>. In hematologic malignancies, intestinal bacteria also modulate the risk of graft-versus-host disease (GVHD) and infection after allogeneic hematopoietic stem cell transplantation (allo-HSCT). Greater bacterial diversity and abundance of the genus *Blautia* were associated with reduced GVHD-related death and improved OS<sup>22,23</sup>. However, no study has shown a potential role for the intestinal microbiota in the efficacy and toxicity of CAR-T therapy for B-cell malignancies.
+
+The primary aims of this study were to understand the intestinal microbiome changes in patients with r/r B–cell-derived hematologic malignancies undergoing CAR-T cell treatment and to investigate associations of the microbiota with clinical responses and CRS severity. Finally, the potential of the gut microbiome to predict treatment outcomes and CRS severity was explored.
+
+# Results
+
+A total of 92 patients with r/r B-cell-derived hematologic malignancies were screened. Ten patients were not eligible for inclusion. Another four patients were excluded because of lack of sufficient 16S sequencing depth. Thus, MM (n = 43), B-ALL (n = 23), and B-NHL (n = 12) patients were included (Fig. 1 A).
+
+The median age of the MM patients was 59 (range 39–75) years, and 55.8% were male (Table 1). The median number of prior lines of therapy was 4 (range 2–8), with all receiving proteasome inhibitor therapy and 95.3% immunomodulatory agents. At enrollment, 39.5% had received autologous stem cell transplantation, and 55.8% had extramedullary disease(s).
+
+**Table 1**  
+Baseline characteristics of multiple myeloma patients included in final fecal microbiome analyses cohorts.
+
+|  | Total N = 43(%) |
+|---|---|
+| Age |  |
+| Median | 59 |
+| Range | 39–75 |
+| Gender |  |
+| Male | 24 (55.8) |
+| Female | 19 (44.2) |
+| Number of prior lines of therapy |  |
+| Median | 4 |
+| Range | 2–8 |
+| CAR-T cell dose(×10⁶/kg) |  |
+| Median | 4.4 |
+| Range | 1.2–6.9 |
+| Autologous stem cell transplantation |  |
+| No | 26 (60.5) |
+| Yes | 17 (39.5) |
+| Extramedullary disease |  |
+| No | 19 (44.2) |
+| Yes | 24 (55.8) |
+| Prior PI therapy |  |
+| No | 0 |
+| Yes | 43 (100) |
+| Prior IMiD therapy |  |
+| No | 2 (4.7) |
+| Yes | 41 (95.3) |
+
+PI, Proteasome inhibitors (Bortezomib/Carfilzomib/Ixazomib).  
+IMiD, immunomodulatory agent (Lenalidomide/Thalidomid/Pomalidomide).
+
+Three months after infusion of a median dose of 4.4 × 10⁶/kg (range 1.2–6.9 × 10⁶/kg) of BCMA CAR-T cells, 55.8%, 14%, and 25.5% of patients had a complete remission (CR), very good partial response (VGPR), or partial response (PR), respectively. All 43 MM patients showed CRS, grade 1 in 8 patients (18.6%), grade 2 in 16 (37.2%), and grade 3 in 19 (44.2%). No higher grade was observed (Fig. 1 D). Two patients died: one from sepsis caused by *Pseudomonas aeruginosa* and the other from intracranial hemorrhage (Fig. 1 D). Both the BCMA CAR-T/CD3⁺ T-cell percentages in peripheral blood (PB) and serum concentrations of interleukin (IL)-10 and interferon (IFN)-γ increased during CRS and differed significantly in the CR and PR groups (Fig. 1 E). Patients’ temperature and C-reactive protein (CRP), ferritin, and lactic dehydrogenase (LDH) concentrations were elevated, and IL-6 and IFN-γ concentrations were significantly different in grade 3 vs grade 1 CRS (Fig. 1 F and Supplementary Fig. 1A–C). The serum immunoglobulins (IgG, IgA) and immunoglobulin κ and λ light chain concentrations decreased dramatically after CAR-T (Supplementary Fig. 1D–F). Figure 1 G shows the differences of positron emission tomography–computed tomography (PET-CT) scans and plasma cells detected by Wright’s stain of a bone marrow smear (43.5% vs. 0), as well as flow cytometry (68.9% vs. 0) of bone marrow cells before and after CAR-T infusion for a representative subject.
+
+## Changes in the intestinal microbiome during CAR-T cell therapy
+
+To detect changes in the gut microbiota during CAR-T, we collected fecal samples from each patient at five times (FCa, FCb, CRSa, CRSb, and CRSc; Fig. 1 C), where FCa denotes the baseline when patients were first enrolled; FCb after chemotherapy; CRSa after CAR T-cell infusion but before the onset of CRS; and CRSb and CRSc the peak and during the recovery phase of CRS, respectively.
+
+We first evaluated the diversity of the gut microbiota in all subjects during CAR-T cell therapy. There was a significant decrease in diversity (measured by the Simpson index) during and after CRS (at CRSb and CRSc) compared with baseline (Fig. 2 A). This decrease was observed in the microbiome of patients receiving CAR-T therapy for r/r ALL (Supplementary Fig. 4A) or r/r NHL (Supplementary Fig. 4B). Refer to Supplementary Table 1 for details on the characteristics of r/r B-ALL and B-NHL patients. To further assess the similarity of composition between different therapy stages, we performed pairwise Spearman correlation analysis of operational taxonomic unit (OTU) level bacterial abundance (Fig. 2 B) and found that stronger correlations emerged during the early stages with a ρ value of 0.71, 0.73, and 0.68, respectively, at FCa, FCb, and CRSa. Correlations between late stages (CRSb and CRSc) and early stages were weaker, suggesting that changes in microbiome composition might be related to CRS.
+
+We next explored community structure and temporal shift of bacterial abundance at multiple taxonomic levels during CAR-T therapy. In myeloma, bacterial communities were dominated by Firmicutes and Bacteroidetes at the phylum level (Fig. 2 C) and characterized by significant enrichment of Firmicutes and depletion of Bacteroidetes at the last two timepoints (Fig. 2 D, E and Supplementary Fig. 4C). By applying the longitudinal analysis in the Qiime2 microbiome analysis platform, we detected changes in the gut microbial communities at taxonomic levels from phylum to genus (Fig. 2 F and Supplementary Table 2). We further employed a negative binominal (NB) regression model-based time-course analysis to identify genera with significant temporal changes (Supplementary Table 3). Five genera were detected by both Qiime2 and maSigPro procedures, which included increases in *Enterococcus*, *Lactobacillus*, and *Actinomyces* and decreases in *Bifidobacterium* and *Lachnospira* (Supplementary Fig. 4D). Most changes were aggravated during the late stages. Moreover, by checking changes in the five genera in ALL and NHL patients, we observed consistent shift trends in NHL (four genera; Supplementary Fig. 4E) and ALL (two genera; Supplementary Fig. 4F), respectively.
+
+## Association between microbial communities and clinical response to CAR-T therapy
+
+We next determined whether microbial compositions or changes were associated with the response to CAR-T. Because we wanted to identify maximum differences and only six subjects presented in the VGPR group, we performed comparisons only between the CR and PR groups.
+
+Notable differences in microbial alpha and within-sample diversity were observed in patients with CR and PR (Fig. 3 A, B). Although no differences were detected at baseline, PR patients descended more dramatically in alpha diversity and had significantly lower Shannon indices than CR patients after CAR-T infusion (Fig. 3 A). As the degree of differences between CR and PR groups changed across therapeutic stages, we characterized the periods with greater differences by summarizing the amount of CR/PR-enriched OTU at each timepoint. The most pronounced differences occurred at CRSb (Fig. 3 C).
+
+To explore longitudinal differences between CR and PR across all therapeutic stages, we identified OTU features with differential dynamic profiles by applying negative binominal regression-based time-course differential analysis with the maSIgPro package. In total, 125 OTUs were found to have differential time-course patterns between CR and PR patients (Fig. 3 D and Supplementary Table 4). The significant OTUs were further grouped into three clusters according to profiles of their abundance. Most of these OTUs were in clusters 1 and 2 (Fig. 3 E). Cluster 1, characterized by enrichment in the CR group, was comprised mainly of OTUs, which belong to the phyla Firmicutes and Bacteroidetes and the orders Clostridiales and Bacteroidales. Cluster 2 was comprised of OTUs from a broader taxonomy, which included the orders Clostridiales, Bacteroidales, Lactobacillales, and Actinomycetales (Fig. 3 F).
+
+We identified 30 genera with differential time-course patterns in patients with CR and PR after CAR-T (Supplementary Table 5). To explore these differences further, we divided the therapeutic period into before and after CAR-T infusion and performed genus-level class comparisons using linear discriminant analysis (LDA) of effect size (LEfSe)²⁴. We detected 34 genera with differences in abundance in the CR and PR groups (Fig. 4 A). Eighteen genera were detected by both procedures (Supplementary Fig. 5A). Consistent with the results from OTU-level pattern analysis, most of the significant genera such as *Faecalibacterium*, *Roseburia*, and *Ruminococcus* were enriched in CR patients after CAR-T. The genera *Bifidobacterium*, *Prevotella*, *Sutterella*, *Oscillospira*, *Paraprevotella*, and *Collinsella* had a higher abundance in CR versus PR patients both before and after CAR-T (Fig. 4 A and Supplementary Fig. 5B). We also took patients with VGPR into consideration and analyzed the above-mentioned genera before and after CAR-T infusion. The bacterial abundance in VGPR patients fell somewhere between CR and PR patients, but no statistical significance was evident for most of genera (Fig. 4 B and Supplementary Fig. 5D).
+
+To explore whether early bacterial abundance was indicative of therapeutic response, we used RF feature selection to identify key discriminatory genera for responses²⁵. By defining the stages before CAR-T infusion as early, we applied feature selection procedures individually at both baseline (FCa) and post-chemotherapy (FCb) and identified gut microbiome signatures comprising 8 and 14 discriminatory genera separately for baseline and post-chemotherapy (Fig. 4 C, D and Supplementary Fig. 5C). The area under the receiver operating characteristic curve (ROC) of the two RF models using these discriminatory features was 0.73 and 0.85, respectively (Fig. 4 E, F). *Prevotella*, *Collinsella*, *Bifidobacterium*, and *Sutterella* were enriched in CR versus PR both before and after CAR-T infusion and were identified by RF analysis as significant at baseline and post-chemotherapy. This indicates potential associations between these genera and the response to CAR-T.
+
+We also checked the abundance of these genera in r/r NHL and ALL patients. In NHL, *Faecalibacterium*, *Bifidobacterium*, and *Ruminococcus* were significantly (or almost significantly) enriched in CR versus PR and in patients not having a remission (NR), consistent with our results in myeloma (Supplementary Fig. 5E). However, for ALL, we observed enrichment of *Bifidobacterium*, *Roseburia*, and *Collinsella* in NR (Supplementary Fig. 5F), which differed from the results for MM and NHL but might be determined by the small NR sample.
+
+To further demonstrate the association between these taxa and outcome, we assessed progression-free survival (PFS) following CAR-T therapy. By stratifying patients by tertile of bacterial abundance, we observed that for *Sutterella*, patients in the highest-abundance tertile had significantly prolonged PFS (Fig. 4 G). Even after stratification by timepoints, this association remained significant (Supplementary Fig. 6A). However, for genus *Faecalibacterium*, which was reported to be significantly associated with PFS and anti-PD-1 therapy¹⁹, we did not observe an association (Supplementary Fig. 6B, C).
+
+We performed pathway analysis using Phylogenetic Investigation of Communities by Reconstruction of Unobserved State (PICRUSt) and identified significant changes in amino acid metabolism (Fig. 4 H), important for immune function²⁶. For example, CR patients had higher lysine biosynthesis, whereas PR patients had higher lysine degradation. Glutathione metabolism, which can have different effects on functional immunity²⁷, was increased in PR patients. Peptidoglycans biosynthesis was increased in CR versus PR patients. Bacteria-derived peptidoglycans are an important pathogen-associated molecular pattern (PAMP) that can activate inflammatory signaling pathways and stimulate immune responses²⁸.
+
+## Associations between gut microbiome and CRS
+
+Manifestations of severe CRS, namely high fever and greater amounts of cytokines, typically develop within several days after CAR-T cell infusion and may cause death if untreated²⁹. We scaled CRS from level 1 to 5³⁰. To analyze associations between bacterial communities associated with CRS, we compared patients with severe (level 3) versus mild (level 1) CRS and severe and moderate CRS (level 2). We found 146 OTUs with different time patterns in the severe and mild groups (Supplementary Fig. 7 and Supplementary Table 6), and 99 OTUs with different patterns in the severe and moderate CRS groups (Supplementary Fig. 8 and Supplementary Table 7). The profiles of the OTU clusters for the comparisons were similar, with OTUs in clusters 1 and 3 having a higher abundance during late therapy in patients with severe versus mild CRS (Supplementary Fig. 7B and Supplementary Fig. 8B). By analyzing associations between CRS grade and taxa at the genus level, we identified signatures discriminating severe from mild CRS, including decreases in amount of *Bifidobacterium* and *Leuconostoc* in patients with severe CRS (Fig. 5A and Supplementary Table 8). *Bifidobacterium* was increased in patients with worse CRS, not only during the window of CRS, but also at early stages (Fig. 5A, B). *Leuconostoc* was significantly enriched during the window in patients with high CRS grade (Fig. 5A, B). In addition, the abundance of *Stenotrophomonas* and *Staphylococcus* differed severe vs moderate CRS during the window (Supplementary Fig. 8D and Supplementary Table 9).
+
+Comparisons of KEGG pathways across CRS groups showed that the gut microbiome of patients with severe CRS had high metabolism or biosynthesis related to inflammatory compounds, including several pathways associated with amino acid synthesis and metabolism, purine metabolism, lipoic acid metabolism, and biosynthesis of lipopolysaccharide and peptidoglycan (Supplementary Fig. 9 and Supplementary Fig. 10).
+
+Primary inflammatory markers of CRS are cytokines, such as IL-6, IL-2, IL-10, interferon gamma (IFN-γ), and tumor necrosis factor-α (TNF-α). Various cytokines are elevated in the serum of patients experiencing CRS after CAR-T cell infusion³¹. By assessing serum cytokine concentrations and immune cell numbers during CAR-T, we observed significantly increased amounts of serum inflammatory cytokines (IL-6, CRP, IFN-γ, D-dimer, ferritin) but low numbers of immune cells (monocytes, lymphocytes, neutrophils, leukocytes) in severe CRS (Fig. 5C). We also compared serum cytokine concentrations and immune cell numbers in CR and PR, observing significant differences for many of them (see Supplementary Fig. 11A).
+
+To explore further associations between the gut microbiome and CRS during CAR-T therapy, we determined whether serum cytokine concentrations and numbers of PB immune cells correlated with the abundance of gut microorganisms (Fig. 5D). The abundance of the genus *Leuconostoc*, previously linked to CRS grade, correlated positively with ferritin and D-dimer concentrations. The abundance *Bifidobacterium* correlated significantly negatively with PB monocytes (Fig. 5E). We also found a correlation between inflammatory markers and bacteria associated with the clinical response and PFS. For example, *Sutterella* correlated negatively with serum concentrations of CRP and D-dimer (Supplementary Fig. 11B). *Prevotella* correlated negatively with the number of multiple PB immune cells but positively with the serum D-dimer concentration (Supplementary Fig. 11B). *Faecalibacterium* correlated negatively with the serum concentrations of D-dimer and IFN-γ (Supplementary Fig. 11B).
+
+# Discussion
+
+Although several studies have revealed the critical role of the gut microbiome in treatment responses and survival after administration of another important immunotherapy — immune checkpoint inhibitor (e.g., PD-1, PD-L1) therapy<sup>20</sup>, no study has reported on the association between the gut microbiome and CAR-T therapy. In this study, we describe the changes of the gut microbiome during CAR-T therapy and associations with treatment responses and CRS severity in CAR-T-treated patients with B-cell malignancies.
+
+Some of the bacterial genera with differences in abundance in CR versus PR patients have been reported to be involved in the regulation of the immune response, including to immunotherapy. *Faecalibacterium*, reported to enhance antitumor immune responses and survival after anti-PD-1 therapy in melanoma<sup>19, 32</sup>, was in this study associated with CR. Multiple species within the genera *Bifidobacterium* and *Collinsella* increased in responders to anti-PD-1 therapy for melanoma<sup>33</sup>, resulting in depleted peripherally derived colonic regulatory T cells, increased Batf3-lineage dendritic cells (DCs), and augmented T-helper 1 cell (Th1) responses and thus better immune-mediated tumor control<sup>34</sup>. Here, we observed an increased abundance of these two bacteria in CR patients, suggesting a similar response-associated effect of these taxa on the immune system across cancer types and therapeutic strategies.
+
+Nevertheless, some taxa might have effects that are specific for cancer or therapy types. For example, high abundance of genus *Sutterella* was associated with both CR and prolonged survival after CAR-T therapy. However, previous studies reported higher numbers of *Sutterella* in non-responders versus responders in non-small-cell lung cancer (NSCLC) treated with nivolumab<sup>35</sup>. Besides, in this study, we observed contradictory results for the genus *Bifidobacterium*, *Roseburia*, and *Collinsella* in three types of hematologic malignancy (<b>Supplementary Fig. 2F</b>). This indicates a potentially distinct involvement or function of some bacteria in different cancer types and treatments. But these findings require confirmation in studies with larger cohorts.
+
+Gut microbial communities contribute to inter-individual variation in cytokine responses<sup>36</sup>. We propose that gut microbes are related to the intensity of CRS during CAR-T therapy. *Bifidobacterium*, *Leuconostoc*, *Stenotrophomonas*, and *Staphylococcus* were enriched in myeloma patients with severe CRS. Additional studies also demonstrated an association between these microbes and cytokine production. Previous research showed that *Bifidobacterium* correlated with the production of multiple cytokines (e.g., IFN-γ) in a stimulus-specific pattern<sup>36</sup>. The opportunistic pathogen *Stenotrophomonas maltophilia* can stimulate the expression of proinflammatory cytokine and chemokine genes in *in vitro* and *in vivo*<sup>37, 38</sup>. Moreover, superantigens, a family of potent exotoxins produced by *Staphylococcus*, were could eliciting T-cell-driven CRS during treatment with CAR T-cells, T-cell agonistic antibodies, immune check point inhibitors, haploidentical HSCT, and other therapies<sup>39</sup>.
+
+The mechanisms through which gut microbes modulate host immunity are largely unknown. Gut microbial communities modulate host defenses mainly through the release of intermediary metabolites rather than by direct interaction between specific microorganisms and immune cells<sup>36</sup>. Multiple bioactive gastrointestinal metabolites produced by gut microbes, such as amino acids, short-chain fatty acids (SCFAs; e.g., butyrate), and bile acids, exert immunomodulatory functions through immune cell metabolic reprogramming or transcriptional and epigenetic modulation of immune-related genes<sup>26</sup>. Lipopolysaccharide (LPS) from some pathogens is a well-known endotoxin that can stimulate the release of a variety of cytokines/chemokines<sup>40, 41</sup>. Peptidoglycans in bacterial cell walls are a conserved PAMP that trigger innate inflammatory responses throughout the body<sup>42</sup>.
+
+In addition to myeloma, CAR-T therapy has been applied other blood cancers and solid tumors. The link between the gut microbiome and different cancer types needs to be studied systematically. Our research describes associations between changes in the gut microbiome of CAR-T patients and clinical responses and survival. This will open an avenue for investigating the interaction of the gut microbiome and CAR-T cells and lead to novel ways to improve the therapeutic efficacy of CAR-T therapy by targeting the gut microbiome.
+
+As one of the most prominent treatment strategies for hematologic malignancies, CAR-T cell therapy has recent received great attention. Here for the first time, we found that the dynamic changes in the gut microbiome correlated significantly with therapeutic response and CRS during CAR-T treatment of hematologic malignancies (B-ALL, B-NHL, and MM). These findings will aid the development of novel biomarkers for predicting treatment outcome and CRS severity, thereby optimizing the management of these patients while reducing potential toxicities.
+
+# Methods
+
+## Study design and protocol
+
+The study was approved by the Institutional Review Board of the First Affiliated Hospital, School of Medicine, Zhejiang University and was registered in the Chinese Clinical Trial Registry (ChiCTR1800017404). All patients provided written informed consent for participation in accordance with the guidelines of the Declaration of Helsinki and signed agreement for collection and analysis of microbiome samples.
+
+Patient inclusion criteria were: (1) age < 75 years; (2) relapsed or refractory BCMA–positive MM before CAR-T cell treatment; and (3) expected survival > 12 weeks and adequate performance status and organ function to tolerate treatment. Exclusion criteria were: (1) pregnancy or lactation; (2) having received systemic (except inhaled) steroids in the previous 2 weeks or gene therapies; (3) having medical conditions such as severe mental illness, clinically significant cardiovascular disease, severe renal or hepatic dysfunction, or active infection; and (4) any conditions that might increase treatment risks. Patient information and the methods related to two types of cancer (ALL and NHL) are presented in the Supplementary Materials.
+
+Peripheral blood mononuclear cells (PBMCs) were obtained from each patient by leukapheresis for CAR-T cell preparation. The purified CD3<sup>+</sup> T cells were transduced with lentiviral vector to express BCMA CAR (<b>Fig. 1 B</b>). Then the engineered T cells were expanded <em>ex vivo</em> under interleukin-2 stimulation. All patients received lymphodepletion with fludarabine (30 mg/m<sup>2</sup> of body surface area daily on days − 4, -3, and − 2) and cyclophosphamide (500 mg/m<sup>2</sup> daily on days − 3 and − 2) followed by an infusion of BCMA CAR-T cells on day 0. The primary response outcome, defined by the guidelines from the International Myeloma Working Group (IMWG) as a complete response (CR), very good partial response (VGPR), or partial response (PR) in the third month after CAR-T treatment<sup><span citationid="CR43" class="CitationRef">43</span>, <span citationid="CR44" class="CitationRef">44</span></sup>. CRS was graded by the Lee criteria<sup><span citationid="CR30" class="CitationRef">30</span></sup>.
+
+## Microbiome sample collection and restoration
+
+Gut microbiome samples were collected at five timepoints (Fig. <span class="InternalRef" refid="Fig1">1</span> C). All fecal samples were collected with the GUHE Flora Storage kit (Zhejiang Hangzhou Equipment Preparation 20190682, GUHE Laboratories, Hangzhou, China), which maintains microbial DNA stability at room temperature for as long as one month. All samples were frozen at -80℃ prior to DNA extraction. The stages of FCa, FCb, and CRSa were defined as early stages and CRSb and CRSc as late stages. The CRS grade 1 was defined as Mild, CRS grade ≤ 2 as Moderate, and CRS grade ≥ 3 as Severe.
+
+## Assessment of serum cytokine concentrations
+
+All blood samples were stored at 4°C until centrifugation at 5000 rpm for 6 min. The supernatant liquids were quantified with the BD Cytometric Bead Array Human Th1/Th2/Th17 Cytokine Kit and its corresponding software (BD Biosciences) according to the manufacturer's instructions.
+
+## Assessment of CAR-T cell expansion and persistence
+
+Serial PB samples were collected in BD Vacutainer K<sub>2</sub> EDTA tubes (BD Biosciences) before and after CAR-T cell infusion. The expansion of CAR-T cells in vivo was determined by detecting the CAR-T ratio continuously in PB as described<sup><span citationid="CR45" class="CitationRef">45</span>, <span citationid="CR46" class="CitationRef">46</span></sup>. BCMA CAR-T expression was assessed using biotin-SP-conjugated F(ab')2 fragment goat anti-mouse IgG, F(ab')2 fragment-specific antibody, and the secondary staining reagent streptavidin-FITC (BioLegend, 405202) or streptavidin-PE (BioLegend,405204).
+
+## DNA Extraction
+
+Total bacterial genomic DNA samples were extracted using the MO BIO PowerSoil DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA). The quantity and quality of extracted DNA was assessed using both the NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel electrophoresis.
+
+## Bacterial 16S rRNA gene sequencing
+
+The V4 region of the 16S rRNA gene was amplified with bacterial universal primers: 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACH VGGGTWTCTAAT-3′). The primers used for amplification contain adapters for the HiSeq platform and single-end barcodes allowing pooling and demultiplexing sequences of PCR products. Amplified sequences were purified with AMPure XP beads (Agencourt, Inc, Beverly, Manchester, MA, USA) and AxyPrep DNA Gel Extraction Kit (Axygen, Inc, Union City, CA). Qualified PCR products were sequenced with the HiSeq platform (Illumina, Inc, San Diego, CA, USA) using the 2 × 150-bp paired-end sequencing protocol.
+
+## Amplicon data processing
+
+Sequenced reads were demultiplexed according to barcodes. Paired-end reads were merged with the <em>fastq_mergepairs</em> command from VSEARCH v. 2.4.4<sup>47</sup>. The minimum length of overlap between paired-end reads was set to 5. Merged reads were then imported into Qiime2 (v. 2020.2)<sup><span citationid="CR48" class="CitationRef">48</span></sup>. Jointed reads were processed by the <em>qiime quality-filter q-score-joined</em> command to filter sequences with low-quality scores. Sequences were denoised with the <em>Deblur</em> workflow<sup><span citationid="CR49" class="CitationRef">49</span></sup>. Amplicon sequence variants (ASVs) were summarized with the <em>feature-table summarize</em> command. To calculate phylogenetic diversity, a rooted phylogenetic tree was constructed using the <em>align-to-tree-mafft-fasttree</em> pipeline from the <em>q2-phylogeny</em> plugin within Qiime2. The pipeline performed a multiple sequence alignment of the ASV sequences and then masked the alignment to remove positions that are highly variable. The masked alignment was used to generate a phylogenetic tree by <em>FastTree</em> program<sup><span citationid="CR50" class="CitationRef">50</span></sup>. Alpha and beta diversity matrices were generated through the <em>q2-diversity</em> plugin using the above-mentioned ASV feature table and rooted phylogenetic tree. <em>De novo</em> clustering of ASVs was performed with the <em>cluster-features-de-novo</em> command within <em>vsearch</em> plugin<sup><span citationid="CR47" class="CitationRef">47</span></sup>. Input features were collapsed at 97% identity, resulting in new OTU features that are clusters of the ASV features. Representative OTU sequences were then annotated with pre-trained Naive Bayes classifier trained on the Greengenes 13_8 99% OTU database using the <em>feature-classifier</em> plugin<sup><span citationid="CR51" class="CitationRef">51</span></sup>. The sequences used for training were trimmed to include only the V4 region. Taxonomic composition was summarized with the <em>collapse</em> method from the <em>taxa</em> plugin within Qiime2.
+
+## Functional prediction
+
+We used the OTU feature table generated from Qiime2 to predict microbial community function with PICRUSt2<sup>52</sup>. The PICRUSt2 algorithm performed functional prediction based on marker gene sequencing profiles and searched for the most closely related organisms with annotated genomes to infer gene contents per OTU. Gene family abundance per sample was summarized and grouped into KEGG orthologs (KOs). To facilitate the interpretation of functional results, KOs were further summarized into KEGG pathways on the basis of structured pathway mappings. For differential pathway analysis, we applied the two-sided Welch’s <em>t</em>-test to identify discriminative KEGG pathways concerning clinical responses (PR versus CR) and CRS level (level 1 versus level 3).
+
+## Bioinformatics and statistical analysis
+
+Comparisons of alpha diversity and taxonomic abundances between two groups were conducted with the Wilcoxon rank-sum test, while comparisons among three or more groups were conducted using the Kruskal-Wallis rank-sum test. For beta diversity analysis, a PCoA plot was generated with weighted Unifrac distances. To test the significance of between-sample diversity alternation, permutational analysis of variance (PERMANOVA) was performed with the <em>adonis</em> function within the R package <em>vegan</em>.
+
+The <em>feature-volatility</em> plugin<sup><span citationid="CR53" class="CitationRef">53</span></sup> within Qiime2 was applied to implement longitudinal analysis to identify features that are associated with therapy stages. In this pipeline, supervised learning regressor was used to identify important features and assess their ability to predict therapy states. Unclassified taxonomic features, features absent in more than 90% of all samples, and features with low abundance (< 0.01%) were all excluded from the analysis. Net average change scores and importance scores, which denote the correlation between input features and therapy stages, were exported and visualized in a volcano plot. Only features with net average change scores more than 0.2% and importance scores within the first tertile of distribution were retained.
+
+For time-course differential analysis, the R package <em>maSigPro</em><sup><span citationid="CR54" class="CitationRef">54</span>, <span citationid="CR55" class="CitationRef">55</span></sup> was used to find taxonomic features with significant temporal changes and significant differences between experimental groups (e.g., clinical response and CRS grade groups). Specifically, the <em>maSigPro</em> algorithm defined a generalized regressive model by dummy variables followed by two regression steps: the first one selects features with non-flat profiles by the least-squared technique and the second step creates best regression models for each feature by using stepwise regression to identify features with different profiles between experimental groups. We used as input, the normalized relative abundance (scaled to 100 million) and excluded features that did not occur in more than 90% of all samples. We employed a negative binominal regressive model for the microbial counts data and ran <em>maSigPro</em> on therapy stages with a degree of 4. All features with a significant group difference were exported. The significant features were further clustered together using the <em>hclust</em> function method according to the patterns of their relative abundance. For each cluster, a median profile and fitted curve of all included features were summarized to visualize the profile pattern.
+
+The LAD effect size (LEfSe) algorithm<sup><span citationid="CR24" class="CitationRef">24</span></sup> was employed to identify differentially abundant features between groups (e.g., between clinical response and CRS grade). The method first detected features with significant differential abundance using the non-parametric factorial Kruskal-Wallis rank-sum test with pre-defined α of 0.05. Significant features were then used to build a Linear Discriminant Analysis (LDA) model for estimating the effect size of each differentially abundant feature. The LDA score threshold for discriminative features was set to 2.0.
+
+To identify early predictive biomarkers with respect to clinical response (PR vs. CR), we implemented a random forest (RF) feature selection procedure within the R package <em>caret</em>. The recursive feature elimination (RFE) algorithm with 5-fold cross validation was applied for feature selection. An optimized number of feature sets was determined by performance of 5-fold cross validation. To depict the receiver operating characteristic (ROC) curve and calculate the area under the curve (AUC), the <em>pROC</em> package was utilized.
+
+For progression-free survival (PFS) analysis, subjects were classified as high, medium, or low based on tertiles of the distribution of specific taxa abundance (e.g., genus <em>Sutterella</em>). Time to progression was defined as the interval (in days) from the date of CAR T-cell infusion to the date of disease progression. Survival curves were estimated using the Kaplan-Meier product-limit method and compared using the log-rank test within the R package <em>survminer</em>.
+
+We applied Spearman’s rank-order correlation to test the association between bacterial abundance and concentration of immune cells and inflammatory factors. Only genus-level features deemed to be associated with clinical response and CRS grades were included in this analysis. Associations with an absolute value of correlation coefficient higher than 0.2 and FDR less than 0.2 were depicted using Cytoscape<sup><span citationid="CR56" class="CitationRef">56</span></sup>.
+
+# References
+
+1. Stephenson R, Singh A. Drug discovery and therapeutic delivery for the treatment of B and T cell tumors. *Adv Drug Deliv Rev* **114**, 285–300 (2017).
+2. Habermann TM, *et al.* Rituximab-CHOP versus CHOP alone or with maintenance rituximab in older patients with diffuse large B-cell lymphoma. *J Clin Oncol* **24**, 3121–3127 (2006).
+3. Goebeler ME, *et al.* Bispecific T-Cell Engager (BiTE) Antibody Construct Blinatumomab for the Treatment of Patients With Relapsed/Refractory Non-Hodgkin Lymphoma: Final Results From a Phase I Study. *J Clin Oncol* **34**, 1104–1111 (2016).
+4. Lokhorst HM, *et al.* Targeting CD38 with Daratumumab Monotherapy in Multiple Myeloma. *N Engl J Med* **373**, 1207–1219 (2015).
+5. Sonneveld P. Management of multiple myeloma in the relapsed/refractory patient. *Hematology Am Soc Hematol Educ Program* **2017**, 508–517 (2017).
+6. Gokbuget N, *et al.* Outcome of relapsed adult lymphoblastic leukemia depends on response to salvage chemotherapy, prognostic factors, and performance of stem cell transplantation. *Blood* **120**, 2032–2041 (2012).
+7. Kantarjian HM, *et al.* Defining the course and prognosis of adults with acute lymphocytic leukemia in first salvage after induction failure or short first remission duration. *Cancer* **116**, 5568–5574 (2010).
+8. Crump M, *et al.* Outcomes in refractory diffuse large B-cell lymphoma: results from the international SCHOLAR-1 study. *Blood* **130**, 1800–1808 (2017).
+9. Kumar SK, *et al.* Risk of progression and survival in multiple myeloma relapsing after therapy with IMiDs and bortezomib: a multicenter international myeloma working group study. *Leukemia* **26**, 149–157 (2012).
+10. Maude SL, *et al.* Chimeric antigen receptor T cells for sustained remissions in leukemia. *N Engl J Med* **371**, 1507–1517 (2014).
+11. Maude SL, *et al.* Tisagenlecleucel in Children and Young Adults with B-Cell Lymphoblastic Leukemia. *N Engl J Med* **378**, 439–448 (2018).
+12. Neelapu SS, *et al.* Axicabtagene Ciloleucel CAR T-Cell Therapy in Refractory Large B-Cell Lymphoma. *N Engl J Med* **377**, 2531–2544 (2017).
+13. Zhao WH, *et al.* A phase 1, open-label study of LCAR-B38M, a chimeric antigen receptor T cell therapy directed against B cell maturation antigen, in patients with relapsed or refractory multiple myeloma. *J Hematol Oncol* **11**, 141 (2018).
+14. Raje N, *et al.* Anti-BCMA CAR T-Cell Therapy bb2121 in Relapsed or Refractory Multiple Myeloma. *N Engl J Med* **380**, 1726–1737 (2019).
+15. Brudno JN, *et al.* T Cells Genetically Modified to Express an Anti-B-Cell Maturation Antigen Chimeric Antigen Receptor Cause Remissions of Poor-Prognosis Relapsed Multiple Myeloma. *J Clin Oncol* **36**, 2267-+ (2018).
+16. Xu J, *et al.* Exploratory trial of a biepitopic CAR T-targeting B cell maturation antigen in relapsed/refractory multiple myeloma. *Proc Natl Acad Sci U S A* **116**, 9543–9551 (2019).
+17. Ali SA, *et al.* T cells expressing an anti-B-cell maturation antigen chimeric antigen receptor cause remissions of multiple myeloma. *Blood* **128**, 1688–1700 (2016).
+18. Zitvogel L, Ma Y, Raoult D, Kroemer G, Gajewski TF. The microbiome in cancer immunotherapy: Diagnostic tools and therapeutic strategies. *Science* **359**, 1366–1370 (2018).
+19. Gopalakrishnan V, *et al.* Gut microbiome modulates response to anti-PD-1 immunotherapy in melanoma patients. *Science* **359**, 97–103 (2018).
+20. Yi M, *et al.* Gut microbiome modulates efficacy of immune checkpoint inhibitors. *J Hematol Oncol* **11**, 47 (2018).
+21. Vetizou M, *et al.* Anticancer immunotherapy by CTLA-4 blockade relies on the gut microbiota. *Science* **350**, 1079–1084 (2015).
+22. Staffas A, Burgos da Silva M, van den Brink MR. The intestinal microbiota in allogeneic hematopoietic cell transplant and graft-versus-host disease. *Blood* **129**, 927–933 (2017).
+23. Jenq RR, *et al.* Intestinal Blautia Is Associated with Reduced Death from Graft-versus-Host Disease. *Biol Blood Marrow Tr* **21**, 1373–1383 (2015).
+24. Segata N, *et al.* Metagenomic biomarker discovery and explanation. *Genome Biol* **12**, R60 (2011).
+25. Oh TG, *et al.* A Universal Gut-Microbiome-Derived Signature Predicts Cirrhosis. *Cell Metab* **32**, 878–888 e876 (2020).
+26. Blacher E, Levy M, Tatirovsky E, Elinav E. Microbiome-Modulated Metabolites at the Interface of Host Immunity. *J Immunol* **198**, 572–580 (2017).
+27. Perricone C, De Carolis C, Perricone R. Glutathione: a key player in autoimmunity. *Autoimmun Rev* **8**, 697–701 (2009).
+28. McDonald C, Inohara N, Nunez G. Peptidoglycan signaling in innate immunity and inflammatory disease. *J Biol Chem* **280**, 20177–20180 (2005).
+29. Norelli M, *et al.* Monocyte-derived IL-1 and IL-6 are differentially required for cytokine-release syndrome and neurotoxicity due to CAR T cells. *Nature Medicine* **24**, 739–748 (2018).
+30. Lee DW, *et al.* Current concepts in the diagnosis and management of cytokine release syndrome. *Blood* **124**, 188–195 (2014).
+31. Brudno JN, Kochenderfer JN. Toxicities of chimeric antigen receptor T cells: recognition and management. *Blood* **127**, 3321–3330 (2016).
+32. Peters BA, *et al.* Relating the gut metagenome and metatranscriptome to immunotherapy responses in melanoma patients. *Genome Med* **11**, 61 (2019).
+33. Matson V, *et al.* The commensal microbiome is associated with anti–PD-1 efficacy in metastatic melanoma patients. *Science* **359**, 104–108 (2018).
+34. Geva-Zatorsky N, *et al.* Mining the Human Gut Microbiota for Immunomodulatory Organisms. *Cell* **168**, 928–943 e911 (2017).
+35. Botticelli A, *et al.* Changes of microbiome profile during nivolumab treatment in NSCLC patients. In: *Asco* (2018).
+36. Schirmer M, *et al.* Linking the Human Gut Microbiome to Inflammatory Cytokine Production Capacity. *Cell* **167**, 1897 (2016).
+37. Kim YJ, *et al.* Stenotrophomonas maltophilia outer membrane vesicles elicit a potent inflammatory response in vitro and in vivo. *Pathog Dis* **74**, (2016).
+38. Pompilio A, Crocetta V, Di Bonaventura G. Stenotrophomonas maltophilia mutant lacking flagella remains virulent in DBA/2N mice but is less efficient in stimulating TNF-alpha expression. *FEMS Microbiol Lett* **365**, (2018).
+39. Kale SD, *et al.* "Small" Intestinal Immunopathology Plays a "Big" Role in Lethal Cytokine Release Syndrome, and Its Modulation by Interferon-gamma, IL-17A, and a Janus Kinase Inhibitor. *Front Immunol* **11**, 1311 (2020).
+40. Matera G, *et al.* Procalcitonin neutralizes bacterial LPS and reduces LPS-induced cytokine release in human peripheral blood mononuclear cells. *BMC Microbiol* **12**, 68 (2012).
+41. Wong MH, Chapin OC, Johnson MD. LPS-stimulated cytokine production in type I cells is modulated by the renin-angiotensin system. *Am J Respir Cell Mol Biol* **46**, 641–650 (2012).
+42. Dziarski R. Recognition of bacterial peptidoglycan by the innate immune system. *Cell Mol Life Sci* **60**, 1793–1804 (2003).
+43. Durie BG, *et al.* International uniform response criteria for multiple myeloma. *Leukemia* **20**, 1467–1473 (2006).
+44. Rajkumar SV, *et al.* Consensus recommendations for the uniform reporting of clinical trials: report of the International Myeloma Workshop Consensus Panel 1. *Blood* **117**, 4691–4695 (2011).
+45. Hu Y, *et al.* Potent Anti-leukemia Activities of Chimeric Antigen Receptor-Modified T Cells against CD19 in Chinese Patients with Relapsed/Refractory Acute Lymphocytic Leukemia. *Clin Cancer Res* **23**, 3297–3306 (2017).
+46. Wang Y, *et al.* A retrospective comparison of CD19 single and CD19/CD22 bispecific targeted chimeric antigen receptor T cell therapy in patients with relapsed/refractory acute lymphoblastic leukemia. *Blood Cancer J* **10**, 105 (2020).
+47. Rognes T, Flouri T, Nichols B, Quince C, Mahe F. VSEARCH: a versatile open source tool for metagenomics. *PeerJ* **4**, e2584 (2016).
+48. Bolyen E, *et al.* Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. *Nat Biotechnol* **37**, 852–857 (2019).
+49. Amir A, *et al.* Deblur Rapidly Resolves Single-Nucleotide Community Sequence Patterns. *mSystems* **2**, (2017).
+50. Price MN, Dehal PS, Arkin AP. FastTree 2–approximately maximum-likelihood trees for large alignments. *PLoS One* **5**, e9490 (2010).
+51. Bokulich NA, *et al.* Optimizing taxonomic classification of marker-gene amplicon sequences with QIIME 2's q2-feature-classifier plugin. *Microbiome* **6**, 90 (2018).
+52. Douglas GM, *et al.* PICRUSt2: An improved and extensible approach for metagenome inference. *bioRxiv*, 672295 (2019).
+53. Bokulich NA, *et al.* q2-longitudinal: Longitudinal and Paired-Sample Analyses of Microbiome Data. *mSystems* **3**, (2018).
+54. Nueda MJ, Tarazona S, Conesa A. Next maSigPro: updating maSigPro bioconductor package for RNA-seq time series. *Bioinformatics* **30**, 2598–2602 (2014).
+55. Conesa A, Nueda MJ, Ferrer A, Talón M. maSigPro: a method to identify significantly differential expression profiles in time-course microarray experiments. *Bioinformatics* **22**, 1096–1102 (2006).
+56. Shannon P, *et al.* Cytoscape: a software environment for integrated models of biomolecular interaction networks. *Genome Res* **13**, 2498–2504 (2003).
+
+# Supplementary Files
+
+- [naturecommunicationSupplementaryMaterial.docx](https://assets-eu.researchsquare.com/files/rs-725566/v1/e86526213f3a8dfca5fab464.docx)

0d84d302744e3647ca6b8589491799090bf8c5dc68e3598a4832926b3bb6e3f0/preprint/images_list.json

@@ -0,0 +1,50 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Blood levels of SFN and known biomarkers for the detection of acute DILD. SOMAscan signals of SFN (a) and serum levels of SFN (b, c), KL-6 (d, e) and SP-D (f, g) measured by ELISA (b-c) or clinical chemistry kits (d-g) in healthy volunteers (HV) and DILD patients in the Discovery (b, d, f) and Validation cohorts (c, e, g) are shown. The boxes indicate interquartile ranges (75% and 25%) and medians. DAD: diffuse alveolar damage; OP: organizing pneumonia; NSIP: nonspecific interstitial pneumonia; Recov: all recovery patients. **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant. h. The area under the curve (AUC) was derived from receiver operating characteristic (ROC) curves, and the 95% confidence intervals (95% CIs) were determined for the biomarkers in comparative analyses. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Distribution of SFN and known biomarkers in healthy volunteers and patients with various lung diseases. Results for the Combined cohort are shown. Serum levels of KL-6 (a), SP-D (b), and SFN (e) were measured by ELISA. The boxes indicate interquartile ranges (75% and 25%) and medians. DAD: diffuse alveolar damage; OP: organizing pneumonia; NSIP: nonspecific interstitial pneumonia; Recov: all DILD patients in recovery; Tolerant: tolerant control; Lung Ca: lung cancer; IIPs: idiopathic interstitial pneumonia; CTD: lung disease associated with connective tissue disease; COPD: chronic obstructive pulmonary disease; NTM: nontuberculous mycobacteria; BA: bronchial asthma; infectious: bacterial and mycotic pneumonia. *p < 0.05, **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "ROC curves for SFN and known biomarkers. Data of the Combined cohort are shown. ROC curves for SFN and KAL as well as the existing markers (SP-D and KL-6) were drawn to discriminate the acute phase of DAD patients from (a) all patients in recovery (discrimination of the disease activity), (b) acute-phase patients with other DILD types (DAD diagnostic performance), (c) the DILD-tolerant control patients (DAD onset), and (d) patients with infectious lung disease. Values in parentheses indicate the area under the curve (AUC) values.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Serum levels of SFN and known biomarkers in idiopathic interstitial pneumonias. The distributions of SFN (a), SP-D (b), and KL-6 (c) are shown in box-plot graphs. The boxes indicate interquartile ranges (75% and 25%) and medians. The broken lines indicate the respective cutoff values. AE-IIPs: Acute exacerbation of idiopathic interstitial pneumonias (DAD-type disease); IPF: idiopathic pulmonary fibrosis; NSIP: nonspecific interstitial pneumonia; Other: other types of idiopathic interstitial pneumonia. *p < 0.05; **p < 0.01; ns, not significant.\n\n",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "SFN expression in lung tissue, and associations of serum SFN levels with those in BALF and with pulmonary functions. a Representative immunohistochemical staining for SFN. Clear expression was observed in the basal cells of the bronchioles (a1, 200x) and alveolar type II cells (a2, 200x) in DAD specimens. Positive staining was not observed in the cells of alveolitis with bleeding in lung cancer patients (a3, 100x) and normal-looking lung tissue (a4, 100x). b Correlation between serum SFN and BALF SFN. The data are for patients with DILD or IIPs (n=14) from whom both serum and BALF samples were taken. BALF (ELF) SFN levels were estimated using the urea method. c Correlation between serum SFN levels and the ratio of arterial partial pressure of oxygen to fractional inspired oxygen (PaO2/FiO2) in the acute-phase DILD patients (n=40). The values of the correlation coefficients rs and p were given by Spearman\u2019s correlation analysis.\n\u2003\n",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Extracellular release of SFN in vitro. a Dose-dependent SFN expression by JNJ26854165 treatment. Various concentrations of JNJ26854165 were added to the A549 cells. After 48 h, the cell proliferation and the levels of SFN and LDH released into the conditioned medium (CM) were measured. b Comparison of the SFN levels in the conditioned media (CM) and cell lysate. The A549 cells were treated with bleomycin (BLM), nutlin-3 (NUT), JNJ26854165 (JNJ) and H2O2 for 48 h, and then the SFN levels were measured by ELISA. The amounts of SFN relative to the total amounts of proteins in cell lysates are shown. The CM SFN levels stimulated by BLM and NUT are enlarged and inlet at the top left of the graph. c The SFN mRNA levels were measured by qRT-PCR. ACTB (\u03b2-actin) was used for normalization. d Comparison of the expression levels of p53 and p21 proteins. Representative results of Western blot are shown. e, f LDH, K18 and ccK18 levels in CM. Three to five independent experiments were performed for each measurement. Standard errors are shown as error bars. A vehicle control was used for each drug (DMSO for BLM, NUT, and JNJ, and PBS for H2O2). Statistically significant differences (p<0.05) by t-test in relation to the vehicle control are indicated with an asterisk (*). ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0d84d302744e3647ca6b8589491799090bf8c5dc68e3598a4832926b3bb6e3f0/preprint/preprint.md

@@ -0,0 +1,276 @@
+# Abstract
+
+Among the various histopathological patterns of drug-induced interstitial lung disease (DILD), diffuse alveolar damage (DAD) is associated with poor prognosis. However, there is no reliable biomarker for its accurate diagnosis. Here, we show stratifin/14-3-3σ (SFN) as a biomarker candidate found in a proteomic analysis. The study included two independent cohorts and controls (n = 432 samples). SFN was specifically elevated in DILD patients with DAD, and was superior to the known biomarkers, KL-6 and SP-D, in discrimination of DILD patients with DAD from patients with other DILD patterns or other lung diseases, including bacterial pneumonia. SFN was also increased in serum from patients with idiopathic DAD, and in lung tissues and bronchoalveolar lavage fluid of patients with DAD. In vitro analysis using the A549 cell line suggested that extracellular release of SFN occurred via p53 activation. We conclude that serum SFN is a promising biomarker for DAD diagnosis.
+
+Applied Biochemistry  
+Pulmonology  
+Drug-induced Interstitial Lung Disease  
+Proteomic Analysis  
+Bacterial Pneumonia  
+Bronchoalveolar Lavage Fluid  
+p53 Activation
+
+# Introduction
+
+Pharmaceutical products can cause lung injury as an adverse drug reaction; drug-induced interstitial lung disease (DILD) is a common example. In fact, studies have found that epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs) such as gefitinib and erlotinib frequently cause DILD<sup>1–3</sup>. Although DILD is a problem in Europe and the US (with an incidence of 0.3% for gefitinib), the incidence rate of DILD is higher in Japan (2–3% for gefitinib)<sup><em>4–7</em></sup>. The high incidence rate in Japan is considered a serious issue not only in the clinical setting, but also in the context of global drug development.
+
+There are several types of DILD, categorized based on histological patterns. The most serious type is diffuse alveolar damage (DAD). Of note, DAD is frequently observed in patients who have developed DILD due to the administration of tyrosine kinase inhibitors (TKIs), including EGFR-TKIs<sup><em>8</em></sup>, and is also a common characteristic among the majority of patients with acute lung injury, such as cases of acute exacerbation (AE) of idiopathic pulmonary fibrosis (IPF) and acute respiratory distress syndrome (ARDS). Importantly, patients with DAD do not sufficiently respond to treatment, and their prognosis is poor; additionally, even if they recover from the disease, they continue to suffer from fibrosis. Therefore, for the diagnosis of DILD, early identification of patients with DAD is necessary in order to select an appropriate treatment and avoid severe outcomes or DILD-related death.
+
+Biopsy for pathological evaluation is needed for the definitive diagnosis of DILD, but it is clinically and ethically difficult to obtain samples from patients with DAD because of their severe symptoms. Currently, high-resolution computed tomography (HRCT) chest scans are the most effective non-invasive method to diagnose the histological pattern of DILD<sup><em>9</em></sup>. However, it is difficult for general physicians other than respiratory specialists to correctly diagnose DAD by HRCT, because this method requires specialized knowledge and training. In addition, HRCT is highly expensive and poses a significant financial burden on health-care systems. Moreover, exposure to radiation during HRCT scanning is associated with an increase in the long-term risk of complications. Therefore, it is necessary to develop a biomarker that can be used specifically for the diagnosis of DAD.
+
+Currently, the following biomarkers are used in the clinical detection of interstitial pneumonia: surface protein (SP)-A, SP-D, and Krebs von den Lungen-6 (KL-6); all of these are high-molecular-weight glycoproteins that are expressed by type II pneumocytes<sup>10,11</sup>. KL-6 has been reported to be useful for the detection of EGFR-TKI-induced DILD in patients with non-small-cell lung carcinoma<sup>12</sup>. However, although these biomarkers can detect interstitial pneumonia in general, they are not used specifically for the diagnosis of DAD. Moreover, there is no correlation between these biomarkers and the severity of lung injury. In addition to DILD, several blood biomarkers have been proposed for the detection of acute lung injury, including acute interstitial pneumonia and ARDS. Such biomarkers include the molecular chaperone HSP47<sup>13</sup>, the inflammatory cytokines IL-1β, IL-8, and IL-6; and the inflammatory mediators HMGB1 and LBP<sup>14</sup>. However, none of these biomarkers are currently used in the clinical setting.
+
+In this study, we performed an aptamer-based proteome analysis using blood samples from DILD patients in order to identify a new diagnostic marker of DAD, and we validated the clinical usefulness of the thus-identified marker using two independent cohorts.
+
+# Results
+
+## Screening and validation of new biomarkers
+
+We performed a sample collection at the acute and recovery phases from consecutively recruited DILD-onset patients exhibiting the DAD pattern, the DAD pattern mixed with other dominant DILD patterns (DAD-mixed patterns), the organizing pneumonia (OP) pattern, the nonspecific interstitial pneumonia (NSIP) pattern, or the hypersensitivity pneumonitis (HP) pattern. To search for biomarkers specific to patients in the DAD group (the DAD and DAD-mixed patterns), we performed a proteomics study by SOMAscan assay in a “Discovery cohort” of patients with DILD who were enrolled in the early phase of this study, and validated the results using a “Validation cohort” consisting of additional enrolled DILD patients. Tolerant controls (who were administered similar drug(s) but exhibited no DILD onset) and various lung disease patients were also analyzed for biomarker evaluation. The clinical characteristics of these cohorts are shown in Table 1 (for age, sex) and Supplementary Table 1 (for laboratory test). The underlying diseases and the suspected causal drugs of DILD in the registered DILD patients were diverse in both Discovery and Validation cohorts (Supplementary Tables 2 and 3).
+
+**Table 1**  
+Sample cohorts used in this study: numbers and characteristics.
+
+| Group | Pattern | No. of cases (female) | Age (range) | No. of samples (acute/recovery) |
+|-------|---------|------------------------|-------------|----------------------------------|
+| Discovery cohort [95] | HV | 24 (12) | 61 (55–65) | 24 |
+| | DILD | DAD group | 10 (2) | 71 (56–86) | 17 (10/7) |
+| | | DAD* | 6 (1) | 72 (61–86) | 11 (6/5) |
+| | | DAD-mixed† | 4 (1) | 70 (56–82) | 6 (4/2) |
+| | | non-DAD group | 30 (13) | 61.5 (32–85) | 54 (30/24) |
+| | | OP | 13 (4) | 69 (32–85) | 22 (13/9) |
+| | | NSIP | 15 (6) | 63 (46–81) | 28 (15/13) |
+| | | Other‡ | 2 (1) | 61.5 (50–73) | 4 (2/2) |
+| Validation cohort [120] | HV | 53 (33) | 34 (25–64) | 53 |
+| | DILD | DAD group | 16 (3) | 69 (54–84) | 23 (16/7) |
+| | | DAD* | 11 (3) | 70.5 (60–80) | 13 (11/2) |
+| | | DAD-mixed† | 5 (0) | 67 (54–84) | 10 (5/5) |
+| | | non-DAD group | 28 (15) | 72.5 (52–79) | 44 (28/16) |
+| | | OP | 17 (11) | 72 (52–79) | 28 (17/11) |
+| | | NSIP | 7 (2) | 73 (63–77) | 10 (7/3) |
+| | | Other‡ | 4 (2) | 71.5 (56–75) | 6 (4/2) |
+| | Tolerant controls [31] | Tolerant controls | 31 (12) | 69 (33–83) | 31 |
+| | Disease controls [186] | Lung cancer | 58 (17) | 72 (44–81) | 58 |
+| | | Infectious | 19 (3) | 75 (36–81) | 19 |
+| | | NTM | 14 (9) | 65.5 (51–81) | 14 |
+| | | IIPs | 43 (8) | 72 (41–83) | 43 |
+| | | CTD | 25 (16) | 67 (50–83) | 25 |
+| | | COPD | 15 (2) | 67 (51–80) | 15 |
+| | | BA | 12 (9) | 60 (43–87) | 12 |
+
+* DAD: patients diagnosed with a DAD-only pattern or DAD-dominant pattern (examples of diagnosis: DAD > NSIP, DAD > OP).  
+† DAD-mixed: patients diagnosed with DILD subtypes other than a DAD-dominant pattern, but in whom co-presence of the DAD pattern was observed by HRCT (examples of diagnosis: OP > DAD, HP > DAD, DAD = HP).  
+‡ Other: cases including eosinophilic pneumonia (EP) and hypersensitivity pneumonitis (HP). OP: organizing pneumonia; NSIP: nonspecific interstitial pneumonia.  
+HV, healthy volunteer; IIP, idiopathic interstitial pneumonia; CTD, lung disease associated with connective tissue disease; COPD, chronic obstructive pulmonary disease; NTM, nontuberculous mycobacteria; BA, bronchial asthma; infection, bacterial and mycotic pneumonia. Data of the DAD and non-DAD groups are shown in bold.
+
+First, we performed a SOMAscan assay-based proteomic analysis using plasma samples collected from subjects in the Discovery cohort. Using the quantitative data on 1,310 proteins, we searched for proteins whose expression was particularly changed in the DAD group, i.e., proteins that were increased (fold change (FC) > 2) or decreased (FC < 0.5) in acute-phase patients with DAD versus the control group (healthy volunteers and patients in recovery). A total of 55 proteins met these criteria (listed in Supplementary Table 4). Next, we focused our analyses on the following 8 proteins, due to their high effect size (g ≥ 1.9) in the group of patients with acute DAD: the upregulated macrophage-capping protein (CAPG), C-C motif chemokine 18 (PARC), stratifin/14-3-3σ (SFN), interleukin-1 receptor antagonist (IL-1Ra) and secreted phospholipase A2 (sPLA2), and the down-regulated carbonic anhydrase 6 (CA6), kallistatin (KAL), and apolipoprotein AI (Apo-AI) (Supplementary Table 5). Among these candidates, SFN caught our attention because its levels were markedly increased in the DAD group (FC: 2.3-fold, g = 1.9), and did not change in patients with the OP and NSIP patterns (FC: 1.0 and 1.2, respectively; g values = 0.1 and 0.6, respectively; Supplementary Table 5). Therefore, we next developed an in-house ELISA for SFN to validate the SOMAscan results. As shown in Supplementary Table 6, the results indicated that the analytical performance of SFN ELISA was sufficient to warrant further evaluation of this protein. In a parallel analysis using commercial kits, we measured the levels of sPLA2, PARC, IL-1Ra, Apo-AI, KAL, and KL-6 and SP-D, which are known to be representative biomarkers of interstitial pneumonia, to evaluate their diagnostic performance as DAD-specific biomarkers. Both the results of the assay by ELISA and the results using the commercial kits aligned with the SOMAscan data, except for IL-1Ra (data not shown; no replication of SOMAscan data by ELISA). As expected, the levels of SFN (Fig. 1) and all other candidates (Supplementary Fig. 1) were changed in acute DILD patients compared to healthy volunteers, and returned to levels close to those in healthy volunteers in the recovery phases. Importantly, unlike the levels of KL-6 and SP-D, which were elevated not only in the DAD group but also in the non-DAD group (Fig. 1d and f), the SFN levels were significantly higher in the DAD group than the non-DAD group (p < 0.0001, Fig. 1b).
+
+Next, we validated these findings by measuring these biomarker candidates in DILD patients and healthy volunteers in the Validation cohort. Importantly, the concentrations of each biomarker candidate in DILD patients in the Validation cohort were similar to those obtained in the Discovery cohort (Fig. 1 and Table 2 show the results for SFN, KL-6 and SPD; Supplementary Fig. 1 and Supplementary Table 7 show results for the other candidates). Based on the receiver operating characteristic (ROC) curve analysis, the area under curve (AUC) values obtained with the Discovery cohort were reproduced in the Validation cohort (Fig. 1h and Supplementary Fig. 1m). The AUC values observed in the Validation cohort indicated that all the candidates, except for PARC and Apo-AI, had an extremely good performance (AUC > 0.98) for the discrimination between healthy volunteers and the DAD group. However, in terms of the determination of disease activity in patients with DAD (DAD group patients in the acute phase versus all DILD patients in the recovery phase), SFN (AUC [95%CI]: 0.93 [0.84-1.0]) was superior to KL-6 (0.65 [0.48–0.82], SP-D (0.84 [0.71–0.97]) (Fig. 1h), KAL (0.90 [0.81-1.0]) and other candidates (around 0.80) (Supplementary Fig. 1m). In addition, SFN (0.85 [0.73–0.97]) showed the highest DAD-diagnostic performance (discrimination between the DAD group and the non-DAD groups) among these candidates.
+
+**Table 2**  
+Comparison of the serum levels of SFN, KL-6, and SP-D in the Discovery and Validation cohorts.
+
+| Measurement | Min. Dilution | LLoQ (unit) | HV | DILD acute | DAD | DAD-mixed | OP | NSIP | Other | DILD recovery | All |
+|-------------|---------------|-------------|----|------------|-----|-----------|----|------|-------|---------------|-----|
+| SFN | 2-Fold | 0.1 (ng/mL) | 0.3 (0.1–0.6) | 10.0 (3.9–48.3) | 8.5 (2.1–34.4) | 0.9 (0.3–2.8) | 2.3 (0.7–7.5) | 3.4 (1.8–5.1) | 0.8 (0.1–87.0) |
+| KL-6 | 51-Fold | 10.0 (U/mL) | 198.6 (114.2–408.0) | 1128 (512.0–2018) | 1136 (316.2–5366) | 762.2 (482.5–1019) | 538.6 (164.2–2408) | 904.2 (229.0–6826) | 536.0 (112.7–18959) |
+| SP-D | 11-Fold | 17.2 (ng/mL) | 51.0 (20.2–164.2) | 278.1 (140.1–1072) | 320.9 (82.7–828.4) | 281.6 (220.7–635.1) | 128.1 (38.2–765.4) | 187.7 (8.6–1463) | 105.4 (8.6–436.7) |
+
+The matrix used, the minimum dilution factors, the lower limits of quantification (LLoQ), and the concentration of each biomarker in healthy volunteers and DILD patients (median and range), as well as the number of samples measured (n) in the Discovery and Validation cohorts are indicated. DAD, diffuse alveolar damage; OP, organizing pneumonia; NSIP, nonspecific interstitial pneumonia.
+
+## Comparison of disease specificity in the Combined sample cohort
+
+Figure 2 shows the distribution of the levels of SFN, and SP-D, and KL-6 in the Combined sample cohort (with combined data from the Discovery and Validation cohorts, and controls). The corresponding values are shown in Supplementary Table 8. The levels of SP-D and KL-6 were elevated not only in the acute phase in all types of DILD patients but also in patients with idiopathic interstitial pneumonias (IIPs) and lung diseases associated with connective tissue diseases (CTDs) (Fig. 2a, b). In contrast, the SFN levels were markedly elevated in a specific fashion in patients with the DAD and DAD-mixed patterns when compared those in the DILD patients with the non-DAD pattern, the tolerant controls or the 7 disease controls (p < 0.0001), although infrequent increases of SFN levels were observed in the non-DAD patients of the DILD group, patients with lung carcinoma and IIPs (Fig. 2c). A high level of SFN was not observed in patients with CTDs or infectious lung pneumonia, which are important diseases to discriminate from DILD (p < 0.0001, Fig. 2c). Meanwhile, the levels of sPLA2, PARC and KAL were changed in patients with infectious pneumonia, and the level of Apo-AI was dispersed within each patient group (Supplementary Fig. 2; the corresponding values are shown in supplementary Table 8). Thus, for all of these protein candidates, the serum SFN level showed relatively specific association with DAD.
+
+## Comparison of the biomarker performance in the Combined cohort
+
+Next, following on the above-described results, we performed a ROC analysis to precisely compare the biomarker performance of SFN to that of the known biomarkers in the Combined cohort (Fig. 3). In terms of the determination of disease activity (acute versus recovery patients) in the DAD group, the performance of SFN (AUC [95%CI] 0.93 [0.88–0.99]) was higher than the performances of SP-D (0.86 [0.78–0.95]) and KL-6 (0.68 [0.56–0.80]) (Fig. 3a). Moreover, in terms of the diagnostic performance for discriminating the DAD group from the non-DAD group, SFN (0.90 [0.83–0.97]) was superior to SP-D and KL-6 (AUC ≤ 0.61, Fig. 3b). Additionally, the discriminating performance of SFN (0.92 [0.83–1.0]) between the DAD group and tolerant control group (no DILD onset) was almost equivalent to that of SP-D (0.95 [0.89–1.0]), Fig. 3c.
+
+Importantly, the differential diagnosis of DILD depends on the exclusion of infectious pneumonia. Therefore, we also determined the diagnostic performance in this context. Again, SFN was the biomarker showing the strongest performance for discriminating between the acute phase of the DAD group and patients with fungal/bacterial pneumonia (0.98 [0.95–1.0]); the performance of SFN was superior to those of SP-D (0.87 [0.76–0.98]) and KL-6 (0.86 [0.74–0.98]) (Fig. 3d).
+
+Next, we deepened our analysis and focused on the positivity rates; Table 3 shows a comparison of the positivity rates obtained with SFN, SP-D, and KL-6 in various lung diseases. For KL-6 and SP-D, the normal reference levels used in the clinical settings were set as cut-off values (500 U/mL and 110 ng/mL, respectively). For SFN, we applied the Youden Index-based cutoff value of 3.6 ng/mL for discriminating between the DAD group and the tolerant control group. Note that although the Combined cohort data (Fig. 3c) were used to calculate the cutoff value, the Discovery cohort and Validation cohort separately yielded values close to 3.6 ng/mL, indicating the robustness of the cutoff value. Remarkably, none of the healthy volunteers tested positive for SFN, while a high SFN-positivity rate (92%) was observed in the acute-phase patients of the DAD group; meanwhile, the SFN-positivity rates were low in patients with the non-DAD pattern of DILD (10–33%), suggesting that serum SFN-based diagnosis of DAD is possible by setting an appropriate cut-off value for SFN. In addition, the positivity rates of SFN in the DILD patients in recovery, tolerant control patients (mostly lung cancer patients), and lung cancer patients (9%, 10%, and 19%, respectively) were lower than those for KL-6 (50%, 29%, 29%, respectively) and SP-D (44%, 29%, and 26%, respectively). Finally, in patients with infectious pneumonia (n = 19), while 4 and 6 patients were positive for KL-6 and SP-D, respectively, only 1 patient (pneumocystis pneumonia) was positive for SFN. Thus, SFN was found to have a good performance as a DAD-specific biomarker.
+
+**Table 3**  
+Comparison of the positivity rates for SFN, KL-6, and SP-D in patients with DILD and three types of controls.
+
+| Patient group | Positive | SFN (> 3.6 ng/mL) | KL-6 (> 500 U/mL) | SP-D (> 110 ng/mL) |
+|---------------|----------|-------------------|-------------------|-------------------|
+| HV | | 0 (0/77) | 0 (0/77) | 4 (3/77) |
+| DILD acute | DAD group | 92 (24/26) | 81 (21/26) | 92 (24/26) |
+| | DAD | 94 (16/17) | 94 (16/17) | 94 (16/17) |
+| | DAD-mixed | 89 (8/9) | 56 (5/9) | 89 (8/9) |
+| | non-DAD group | 17 (10/58) | 67 (39/58) | 74 (43/58) |
+| | OP | 10 (3/30) | 70 (21/30) | 70 (21/30) |
+| | NSIP | 23 (5/22) | 73 (16/22) | 86 (19/22) |
+| | Other | 33 (2/6) | 33 (2/6) | 50 (3/6) |
+| DILD recovery | All | 9 (5/54) | 50 (27/54) | 44 (24/54) |
+| Tolerant control | | 10 (3/31) | 29 (9/31) | 29 (9/31) |
+| Lung cancer | | 19 (11/58) | 29 (17/58) | 26 (15/58) |
+| Infectious | | 5 (1/19) | 21 (4/19) | 32 (6/19) |
+| NTM | | 7 (1/14) | 21 (3/14) | 50 (7/14) |
+| IIPs | | 16 (7/43) | 86 (37/43) | 88 (38/43) |
+| CTD | | 8 (2/25) | 76 (19/25) | 52 (13/25) |
+| COPD | | 13 (2/15) | 40 (6/15) | 40 (6/15) |
+| BA | | 0 (0/12) | 0 (0/12) | 0 (0/12) |
+
+The reference values used in the clinical settings (500 U/mL and 110 ng/mL) were used as the cut-off for KL-6 and SP-D. The Youden’s index-based cutoff of 3.6 ng/mL for SFN was obtained by comparing SNF levels in the DAD group and the tolerant control group. Data of the DAD and non-DAD groups are shown in bold. DAD: diffuse alveolar damage; OP: organizing pneumonia; NSIP: nonspecific interstitial pneumonia; IIP: idiopathic interstitial pneumonia; CTD: lung disease associated with connective tissue disease; COPD: chronic obstructive pulmonary disease; NTM: nontuberculous mycobacteria; BA: bronchial asthma; infectious: bacterial and mycotic pneumonia.
+
+## SFN levels were not associated with the underlying diseases or causative medications
+
+As previously mentioned, the DILD patients enrolled in this study had various underlying diseases and had been administered several drugs suspected to be the cause of their DILD. The underlying diseases of the DILD patients were classified into three disease groups: lung cancer, other tissue cancer (including pancreatic, esophageal, and breast cancers), and non-cancerous diseases (including heart failure and rheumatoid arthritis) (Supplementary Table 2). The suspected drugs were categorized into the following five types in accordance with their mechanism of action (Supplementary Table 3): DNA-damaging agents (DDAs, including platinum drugs, gemcitabine, irinotecan, bleomycin, and 5-fluorouracil); taxanes (paclitaxel and docetaxel); tyrosine kinase inhibitors (including EGFR-TKIs such as erlotinib and osimertinib, as well as VEGF inhibitors such as axitinib and bevacizumab); immune checkpoint inhibitors (ICIs, including nivolumab, and pembrolizumab); and other drugs (mTOR inhibitors and non-oncology drugs such as antibiotics, anti-rheumatoid drugs, antiarrhythmic agents, and Chinese herbal medicines). In the DAD group, there were no significant differences in the levels of SFN among patients with lung cancer, other tissue cancers, and non-cancerous diseases, and no significant differences among patients treated with the different classes of suspected drugs (Supplementary Fig. 3a and d); in fact, the SFN-positivity rates (> 3.6 ng/mL) were almost equivalent among the groups (Supplementary Tables 10 and 11). In addition, in the non-DAD group and tolerant control patients, no clear relationships were observed between the levels of SFN and underlying diseases or causative drugs (Supplementary Fig. 3b, c, e and f). Collectively, these findings suggest that the increase in the serum levels of SFN in patients with DAD was not affected by the patients’ underlying diseases or the administration of any drugs.
+
+## Changes in the biomarkers in AE-IIPs
+
+Among the 43 patients with IIPs analyzed (Fig. 2), AE-IIPs were observed in 6 patients (5 with AE-IPF and 1 with acute interstitial pneumonia). In general, DAD is also a histopathologic hallmark of AE-IIPs. Figure 4 shows the distribution of the blood levels of SFN and the two known biomarkers according to the categorization of patients with IIPs. While there was a tendency for the levels of KL-6 and SP-D to be higher in the patients with AE-IIPs, the levels of these biomarkers were also frequently higher (SP-D > 110 ng/mL, KL-6 > 500 U/mL) in patients with IPF and NSIP (Fig. 4b and c). In contrast, the levels of SFN were significantly higher in patients with AE-IIPs compared to patients with the other disease types, who were mostly negative for SFN (< 3.6 ng/mL, Fig. 4a). Therefore, our data suggest that SFN is strongly associated with DAD, and may also be a useful marker for AE-IIPs.
+
+## Origin of SFN
+
+To clarify whether the increase of serum SFN levels in patients with DAD was a result of the release of SFN from the damaged lung tissues, we investigated the expression of SFN in lung tissues and bronchoalveolar lavage fluid (BALF) from patients with DAD. We immunohistochemically analyzed 19 lung tissue specimens, including 9 DAD autopsy specimens (from patients with DILD and with AE-IIPs such as ARDS) and 10 control tissues (5 autopsy and 5 surgical specimens from patients with lung cancer). Among the 9 DAD cases, only 1 was negative (−) for SFN expression, 2 had occasional slightly positive cells (±), 2 had some positive lesions (+), and 4 cases had several positive lesions (++). In contrast, among 10 control specimens, 6 were −, 2 were ±, 1 was + and 1 was ++. Cytoplasmic staining was observed mostly in the basal cells of the bronchioles (Fig. 5a-1) and some alveolar type II cells (Fig. 5a-2). Positive staining was observed in the cells that tended to show type II pneumocyte hyperplasia or squamous cell metaplasia, the pathological features of advanced DAD. The strength of expression was associated with the grade of lesions observed. In contrast, SFN expression was not observed in the cells of alveolitis with bleeding in the lung cancer case (Fig. 5a-3) or in the normal-looking alveolar cells (Fig. 5a-4) of controls. In the non-tumor specimens from the lung cancer patients, a few positive cells were observed in the basal cells of the bronchioles where focal interstitial inflammation was observed.
+
+In addition, significant correlation was observed between the levels of SFN in serum and BALF, which were taken from patients with DILD or IIPs (n = 14, Spearman's rank correlation coefficient [rs] = 0.670, p = 0.0053) (Fig. 5b). Moreover, there was also a significant correlation between the serum levels of SFN and the PaO2/FiO2 ratio, an indicator of blood oxygenation, in DILD patients (n = 40, rs = − 0.439, p = 0.0023) (Fig. 5c). These results suggest that SFN may be upregulated in damaged cells at the alveoli and bronchioles, and then released into the blood. This finding is supported by the significant correlations observed between the serum SFN levels and the SpO2/FiO2 ratio (rs = − 0.434, p < 0.0001), as well as between the serum levels of SFN and the serum levels of C-reactive protein (CRP) (rs = 0.520, p < 0.0001) and lactate dehydrogenase (LDH) (rs = 0.437, p < 0.0001), which are indicators of inflammation and tissue injury, respectively (Supplementary Fig. 4).
+
+## Upregulation of SFN in vitro
+
+To elucidate whether lung epithelial cells have the potential to release SFN, the expression of SFN was analyzed in vitro using the A549 cell line (derived from type II pneumocytes with an intact p53 gene). Because SFN is known to be a transcription target of p53, we examined the expression of SFN in A549 cells treated with a p53-activating reagent (JNJ26854165, which activates p53 via the inhibition of the interaction between p53 and MDM2). Importantly, when A549 cells were treated with various concentrations of JNJ26854165, both the SFN levels and the LDH levels were increased in the conditioned medium, although the decrease in the cell proliferation was found to be dose-dependent (Fig. 6a). In fact, for the A549 cells treated with 10 µg/mL JNJ26854165, while there were slight changes in the amount of intracellularly expressed SFN, the amount of SFN in the conditioned medium was increased by over 500-fold compared to that in the untreated control cells (Fig. 6b). Of note, these results were much more prominent than those obtained with nutlin-3, another p53 activator, and bleomycin and hydrogen peroxide, both associated with ARDS and bleomycin-induced pneumonia, which exhibited modest increases of over 5-fold (Fig. 6b). The changes in the amount of SFN in the conditioned medium aligned with the results of the mRNA expression of SFN (Fig. 6c). We also observed an increase in the expression of the transcription factor p53 and its transcription target p21, indicating the activation of p53 (Fig. 6d). Simultaneously, the levels of extracellularly released LDH and keratin18 (K18) were increased, and in particular, the caspase-cleaved form of K18 (ccK18) was markedly elevated with JNJ26854165, indicating apoptosis induction (Fig. 6e and f). The changes in the extracellular release level and mRNA level of SFN were similar to that of ccK18. Taken together, these results suggest that SFN may be produced in alveolar epithelial cells in a p53-dependent manner (at least in A549 cells), and immediately released extracellularly via apoptosis.
+
+# Discussion
+
+This study sought to identify a blood-based biomarker of DILD and was the first study to evaluate the usefulness of SFN as a DAD marker. In fact, until now, the relationship between SFN and DAD-type interstitial pneumonia has been unclear, and, to the best of our knowledge, no study has reported the detailed behavior of SFN in the blood. In a validation study using our established SFN ELISA, we observed that the serum levels of SFN were specifically increased in patients with DAD. Thus, our findings on SFN were reproduced in samples from an independent cohort. Moreover, the biomarker performance of SFN for discriminating DAD was superior to those of known biomarkers such as KL-6 and SP-D, which are currently used in clinical settings. Additionally, our data suggests that SFN may be useful not only for the diagnosis of drug-induced DAD but also for the diagnosis of idiopathic DAD. We also showed that the levels of SFN were increased not only in the serum, but also in autopsy specimens and the BALF from patients with DAD. Serum SFN levels were significantly correlated with respiratory parameters. In addition, upregulation and extracellular release of SFN were observed in response to the activation of p53 in pulmonary epithelial-derived cells (A549). These findings suggest that serum SFN levels might reflect the degree of lung injury in patients with DAD; thus, this biomarker would be useful for the diagnosis of DILD patients with DAD as well as for the monitoring of their response to treatment.
+
+SFN (14-3-3σ) belongs to the 14-3-3 family, together with six other isoforms, namely β, γ, ε, ζ, η, and τ. This family is a group of highly evolutionarily conserved proteins with a molecular mass of 25–30 kDa, expressed in all eukaryotes and involved in the modulation of various cellular processes via the binding to phosphorylated proteins<sup>22</sup>. While most 14-3-3 isoforms are expressed in various normal tissues in a ubiquitous manner, SFN is expressed specifically in stratified epithelia<sup>23</sup>. In fact, immunohistochemistry staining images of SFN included in the Human Protein Atlas database (<span class="ExternalRef"><span class="RefSource">https://www.proteinatlas.org</span></span>) show that SFN is expressed permanently and abundantly in the squamous epithelium of the skin and esophageal tissues; on the other hand, SFN is only very weakly expressed in the normal alveolar epithelium. However, our immunohistochemical analysis using DAD autopsy specimens revealed that SFN was expressed mostly in the basal cells of the bronchioles and some alveolar type II cells that tended to show squamous metaplasia, which is generally observed in the mid to late phase of advanced DAD. Although SFN has been reported to be associated with lung adenocarcinoma<sup>24,25</sup>, no such association was observed in the serum samples of our lung cancer patients (Fig.<span class="InternalRef" refid="Fig2">2</span> and data not shown). Instead, we would like to emphasize that the marked increase of serum SFN levels occurred specifically in the patients with DAD.
+
+The pathological changes of DAD proceed consistently through discrete but overlapping phases: the early exudative phase, the mid proliferative (organizing) phase, and the late fibrotic phase. They reflect the global mechanisms of wound repair, and are thought to be involved in cell cycle regulation and apoptosis<sup>26</sup>. Denudation and apoptosis of alveolar epithelia may be important early features of acute lung injury. Guinee et al. reported that the expression of the transcription factor p53 as well as the expression of its transcription targets WAF1 and BAX were increased in type II pneumocytes in the lung of a patient with DAD<sup>27,28</sup>. In addition, Bardales et al. showed that decreases in cell proliferation and increases in apoptosis were frequently observed in the lung tissues of patients with DAD, with both changes tending to be stronger in more severe cases, while apoptosis was undetectable in non-DAD patients<sup>29</sup>.
+
+It is also worth noting that SFN is a direct transcriptional target of p53<sup>30</sup>. In the current study, we showed that A549 cells, which were derived from type II alveolar epithelial cells, were able to release SFN<em>in vitro</em> via apoptosis in the context of p53 activation (Fig.<span class="InternalRef" refid="Fig6">6</span>). These findings suggest that SFN may be upregulated by p53 activation when lung epithelial cells are damaged in patients with acute lung injury such as DAD, and extracellularly released via apoptosis, resulting in increased blood SFN levels. In particular, release of SFN into the alveolar epithelium, which is the main field for gas exchange from the lungs to the blood, may contribute significantly to the increase in circulating SFN levels.
+
+There have been many studies on the intracellular function and biological activity of SFN. The expression of SFN has been shown to be induced in a p53-dependent manner in response to DNA damage, and the progression of the G2/M phase of the cell cycle is arrested<sup>30</sup>. In addition, SFN is a marker of the differentiation of epidermal keratinocytes. An analysis based on skin biopsy samples revealed that the expression of SFN in keratinocytes in the squamous epithelium was stronger in areas close to the epidermis; moreover, an experiment using cultured cells found that the upregulation of SFN was closely related to a reduction in cell proliferation and the progression of cellular senescence<sup>23</sup>.
+
+Moreover, in addition to intracellular SFN, the role of extracellular released-SFN has also been reported. Previous studies have demonstrated that the expression of SFN increases when keratinocytes are exposed to ultraviolet rays, and that keratinocyte-released SFN affects dermal fibroblasts via paracrine actions, facilitating the expression of matrix metalloproteinase (MMP)-1<sup>31–33</sup>. In addition, in keratinocytes, the extracellular expression of SFN protein is markedly increased compared to the intracellular expression<sup>33</sup>, consistent with our data showing that the extracellular SFN levels following p53 activation were much higher than the intracellular levels (Fig.<span class="InternalRef" refid="Fig6">6</span> b). The above-mentioned studies also suggested that keratinocyte-released SFN may be associated with skin wound healing and suppression of skin fibrosis<sup>31–33</sup>. Based on our results, we believe that the lung epithelia-released SFN may also be associated with pulmonary damage and remodeling. However, the mechanistic link between the increase in SFN in blood and the pathophysiology of DAD requires further study.
+
+DNA-damaging agents (DDAs) such as cisplatin, carboplatin, and bleomycin may potentially activate p53 in cells. However, in this study we observed no significant difference in the serum SFN levels between DILD patients who were administered DDAs and those who were administered other types of drugs (Supplementary Fig. 3). Importantly, these findings show that the measurement of serum SFN can detect DILD patients with DAD irrespective of the therapeutic agents used, including DDAs. On the other hand, high SFN levels were rarely observed in patients with lung cancer (or other types of cancer) (Table<span class="InternalRef" refid="Tab3">3</span> and Supplementary Table 10). Therefore, when using serum SFN for monitoring the onset of DAD in oncology patients undergoing chemotherapy, it may also be necessary to measure the SFN levels prior to drug administration.
+
+Several limitations of this study bear mention. First, although we were able to replicate the current findings, the number of patients enrolled was small. Second, overall, as only the serum levels were examined, the findings of this study do not provide sufficient evidence to indicate a direct relationship between the pathology of DAD and SFN. Indeed, immunohistochemistry data was obtained using autopsy samples from patients with DAD (not from patients assessed for serum SFN levels). Third, high levels of SFN were observed in certain DILD patients with the NSIP or OP pattern, as well as in some lung cancer patients. Therefore, to clarify the reason behind the high levels of SFN in patients without DAD, it is necessary to analyze a larger number of samples. Lastly, this was a retrospective study. Although we demonstrated that SFN increases in patients with DAD, irrespective of either disease types or suspected causative drug types, we cannot completely rule out a potential bias in the selection of patients. Currently, we are collecting serum samples chronologically and prospectively from patients prescribed EGFR-TKI and immune checkpoint inhibitors for future analysis. The analysis of these samples should shed light on the continuous changes in the levels of SFN that accompany the onset of DAD.
+
+In summary, in this study, we identified a protein biomarker for the diagnosis of DAD. SFN showed DAD-specific diagnostic performance, unlike known biomarkers such as KL-6 and SP-D. Since DILD and AE-IIP patients with DAD are likely to have a poor prognosis, it is necessary to make a diagnosis for such patients at the earliest period. Our data suggest that measuring the serum levels of SFN helps to distinguish DAD from other types of DILD, such as OP and NSIP, facilitating the diagnosis of DAD. Therefore, adopting this method may help in the determination of treatment plans for patients with DAD (in addition to the current diagnostic approach based on HRCT findings). Moreover, because SFN is also effective in discriminating the acute phase from the recovery phase, this protein may be useful for monitoring the response to treatment.
+
+# Methods
+
+## Study population and clinical samples
+
+Clinical samples were collected from the four hub hospitals: Shinshu University, Nippon Medical School, Chiba University, and Hiroshima University. The National Institute of Health Sciences, the Kihara Foundation, Astellas Pharma Inc., and Daiichi Sankyo Company, Ltd. also participated in this study. Approval was obtained from each research ethics review committee and samples were collected in accordance with the approved protocols. Serum and plasma samples were collected from patients who were suspected to have developed drug-induced interstitial pneumonia during the acute (around the most severe phase) and recovery phases of DILD, after informed consent was obtained. DILD was diagnosed according to the Japanese diagnostic criteria by the respiratory specialists, as follows: 1) history of ingestion of a drug that is known to induce lung injury, 2) appearance of clinical manifestations after drug administration, 3) improvement of clinical manifestations after drug discontinuation, 4) exclusion of other causes of the clinical manifestations, and 5) exacerbation of clinical manifestations after resuming drug administration (challenge testing). Of note, criterion 5) was not used in this study due to ethical issues. Patient recovery was judged by respiratory specialists of each hospital at least two weeks following the onset of DILD, based on recovery of clinical manifestations, improvement of lung imaging findings (e.g., HRCT), and improvement of oxygenation (e.g., SpO₂). Blood samples were collected from patients with any of the following conditions: lung cancer, non-tuberculosis mycobacteriosis, idiopathic interstitial pneumonia (IIP), lung disease associated with connective tissue disease, chronic obstructive pulmonary disease, bronchial asthma, and mycotic/bacterial pneumonia. Blood samples were also collected from tolerant control patients, most of whom were lung cancer patients undergoing chemotherapy without the development of DILD over the course of at least three months. The following tubes were used for blood sampling: Vacutainer (Becton Dickinson, Franklin Lakes, NJ, USA) vacuum blood sampling tubes (6 mL) containing a blood coagulation accelerant for serum collection, and Vacutainer vacuum blood sampling tubes (7 mL) containing EDTA-2Na for plasma collection. The tubes containing the blood were centrifuged for 10 min at 1300 × g, and serum and plasma were divided into small amounts and placed into polypropylene tubes with a screw cap. The samples were then stored in a deep freezer (−80°C) until they were used.
+
+Samples from healthy volunteers were collected at Yaesu Sakura-Dori Clinic after approval was obtained from the research ethics review committees of the participating organizations. The above-described protocols were followed to obtain serum and plasma samples. The inclusion criteria for healthy volunteers were as follows: (1) Japanese (self-declaration that all family members up to grandparents were Japanese) and diagnosed as healthy by their physician, (2) fasted for at least 14 h before blood collection (only drinking water was permitted), (3) no consumption of any drugs for at least 1 week, and (4) a normal body mass index (18.5 ≤ BMI < 25). The exclusion criterion was (5) female individuals menstruating.
+
+The “Discovery cohort” consisted of samples from DILD patients in the acute and recovery phases collected from April 2015 to November 2016 and samples from age-matched healthy volunteers. The Discovery cohort was used for biomarker discovery. Thereafter, an independent “Validation cohort” was created, consisting of samples from DILD patients collected from December 2016 to May 2019 and samples from the healthy volunteers that were not included in the Discovery cohort. Importantly, samples from patients with related lung diseases (lung cancer, IIP, lung disease associated with connective tissue disease, chronic obstructive pulmonary disease, nontuberculous mycobacteria, bronchial asthma, and bacterial and mycotic pneumonia) were also included in the Validation cohort for biomarker validation. In some cases, BALF samples collected at Chiba University were also used for analysis. DAD autopsy specimens from patients with interstitial pneumonia, and non-tumor autopsy or surgical specimens from patients with lung cancer, which were obtained at Shinshu University, were also used for immunohistochemical analysis.
+
+## DILD disease classification
+
+Respiratory specialists at each hospital who were extensively experienced in DILD diagnosis classified the histopathological subtypes of DILD (e.g., DAD, OP, and NSIP). The DAD pattern was diagnosed based on the presence of diffuse ground-glass opacities and/or infiltrated shadows in the bilateral lung fields on HRCT. Typical DAD patterns and DAD-dominant patterns (e.g., DAD > HP, DAD > OP) were classified as the DAD pattern. Cases with co-presence of the DAD and non-DAD patterns, but not DAD-dominant (e.g., OP > DAD, HP > DAD, DAD = HP) were classified as the DAD-mixed pattern. Patients with the DAD and DAD-mixed patterns were studied as the DAD-group. In contrast, patients with the OP pattern, NSIP pattern, and HP pattern were categorized as the non-DAD group. Specialists from the four hub hospitals held a meeting and reviewed the patterns based on clinical findings and HRCT images to reach a consensus on their diagnosis.
+
+## SOMAscan assay
+
+The SOMAscan system (SomaLogic, Boulder, CO, USA) was used to perform proteomic analyses. Briefly, frozen plasma samples were sent to SomaLogic (Boulder, CO, USA) through the NEC Corp. (Tokyo, Japan). A total of 1,310 proteins were obtained through the SOMAscan assay. For each protein probe, the relative fluorescence units (RFUs) were logarithmically transformed. Thereafter, a control group was set in which data from all recovered patients and healthy volunteers in the Discovery cohort were integrated. A search was performed for proteins whose abundance was significantly different between the DAD group (acute phase, with the DAD or the DAD-mixed patterns) and the control group, based on changes in mean values (fold change: FC), and effect size (Hedge’s, *g*).
+
+## Measurement of SFN by in-house ELISA
+
+Stratifin/14-3-3 (SFN) was measured using an in-house ELISA kit, which was developed using two commercially available anti-SFN mouse monoclonal antibodies, and recombinant human SFN protein produced in *E. coli* (ATGen, now NKMAX, Seongnam, Korea) as a standard. Briefly, a mouse anti-SFN monoclonal antibody (mAb) used as the primary capture antibody (clone CS112-2A8; Merck, Kenilworth, NJ, USA) was added to Nunc MaxiSorp 96-well microtiter plates (Thermo Fischer Scientific, Waltham, MA, USA) to prepare a solid phase. The plates were washed with a commercial washing buffer (Quantikine Wash Buffer; R&D Systems, Minneapolis, MN, USA) three times at 25°C, and then incubated with the SuperBlock reagent (Thermo Fischer Scientific). The plates were then washed three times at 25°C, and 50 µL samples (serum samples or standard SFN) diluted with General Serum Diluent (Immunochemistry Technologies, Bloomington, MN, USA) were added, followed by 50 µL of the assay buffer (SuperBlock reagent containing 2M KCL and 50 µg/mL of the HBR-1 heterophilic antibody blocking reagent [Scantibodies Laboratory, Santee, CA, USA]), and the plates were stirred for 2 h. Following three washes, the SuperBlock reagent containing biotin-labeled secondary detection antibody (mouse anti-SFN mAb, 3c3; Sigma-Aldrich, St. Louis, MO, USA) was added and incubated for 1 h. After washing six times, plates were reacted with 10% SuperBlock/PBS solution containing Streptavidin-Poly HRP40 (Stereospecific Detection Technologies, Baesweiler, Germany). SFN was then quantitatively measured using the QuantaRed Enhanced Chemifluorescent HRP Substrate (Thermo Fischer Scientific).
+
+This ELISA assay was validated with reference to the Japanese Guidelines on Bioanalytical Method Validation (Ligand Binding Assay) in Pharmaceutical Development (https://www.pmda.go.jp/files/000206208.pdf). We confirmed that the analytical parameters were sufficient for our current purposes within the criteria of the guidelines (Supplementary Table 2). Moreover, the ELISA method was used to measure SFN concentrations in the cell extracts, conditioned medium, and BALF. SFN concentrations in BALF were normalized with each concentration of urea in BALF and serum according the method of Rennard et al.³⁴, and then represented as the concentration in epithelial lining fluid (ELF). The urea concentration was determined using a urea nitrogen colorimetric detection kit (Arbor Assays, Ann Arbor, MI, USA).
+
+## Commercially available kits
+
+The serum levels of SP-D and KL-6 were measured using in vitro diagnostic kits, SP-D kit YAMASA EIA II (Yamasa, Chiba, Japan) and the E test TOSOH II (Tosoh, Tokyo), respectively. Serum kallistatin (KAL) was measured using DuoSet ELISA, and plasma C-C motif chemokine 18 (PARC) and plasma interleukin-1 receptor antagonist (IL-1Ra) were using Quantikine ELISA kits from R&D Systems (Minneapolis, MN, USA). Secreted phospholipase A2 (sPLA2) and apolipoprotein A-I (apo A-I) in plasma were measured using ELISA kits from Cayman Chemical (Ann Arbor, MI, USA) and Abcam (Cambridge, UK), respectively. All ELISA kits were used in accordance with the manufacturers’ instructions.
+
+## Immunohistochemistry of SFN
+
+To investigate the expression and localization of SFN in the human lung, immunohistochemical analysis was performed with DAD autopsy specimens from patients with DILD or IIPs, and non-tumor autopsy or surgical specimens from patients with lung cancer. The sections were deparaffinized, hydrated, and then immersed in 3% H₂O₂/methanol solution for 10 min at 25 ºC for inactivation of endogenous peroxidase activity. After blocking nonspecific reactions with 10% normal goat serum, the sections were incubated with a primary antibody targeting SFN (diluted 1:1000; rabbit anti-SFN polyclonal antibody [HPA011105], Atlas Antibodies, Bromma, Sweden) overnight at 4 ºC. Visualization of antibody binding was performed using a VectaStain Elite ABC Kit (Vector Laboratories, Burlingame, CA, USA) and 3,3′-diaminobenzidine. All sections were counterstained with hematoxylin. The strength of staining was scored under a light microscope using the following criteria: −, almost all negative cells; ±, occasional slightly positive cells; +, some apparently positive lesions; ++, several positive lesions for SFN.
+
+## In vitro experiments
+
+The A549 cancer cell line derived from human type II pneumocytes was obtained from the JCRB Cell Bank and cultured by the recommended methods. Briefly, the cells were plated in 6-well plates at a density of 1.0 × 10⁵ cells/well. Twenty-four hours later, bleomycin (Cayman Chemical, Ann Arbor, MI, USA), JNJ26854165 (Cayman Chemical), Nutlin-3a (Selleck Chemicals, Houston, TX, USA), or hydrogen peroxide solution (Nacalai Tesque, Kyoto, Japan) was added to the cells. Forty-eight hours after the addition of the agents, the conditioned medium was collected. Cell extracts were prepared using RIPA buffer (Wako, Richmond, VA, USA). SFN in conditioned medium and cell extracts was measured using the in-house ELISA method. The Viability/Cytotoxicity Multiplex Assay Kit (Dojindo Rockville MD, USA) was used to measure the cell proliferation rate and LDH activity. For Western blotting of p53 and p21, the mAbs mouse anti-p53 (DO-1; Invitrogen, Carlsbad, CA, USA) and rabbit anti-p21 Waf1/Cip1 (12D1; Cell Signaling Technology, Leiden, the Netherlands) were used. The K18 and ccK18 ELISA kits were used for detecting apoptosis.
+
+## qRT-PCR analysis
+
+SFN mRNA was measured by real-time PCR. Total RNA was extracted from the cells using RNeasy spin columns (Qiagen, Hilden, Germany). The extracted total RNA was then quantitatively measured using a One Step SYBR PrimeScript RT-PCR Kit (Takara, Kyoto, Japan) and the ABI PRISM 7700 PCR thermal cycler. The β-actin gene (*ACTB*) was used as the internal standard. The PCR primers with the following sequences were used: SFN-F, 5′-TGACGACAAGAAGCGCATCAT; SFN-R, 5′-GTAGTGGAAGACGGAAAAGTTCA; ACTB-F 5′-CACCATTGGCAATGAGCGGTTC; ACTB-R, 5′-AGGTCTTTGCGGATGTCCACGT.
+
+## Statistical analysis
+
+The difference in the degree of changes in the amounts of proteins, as per the SOMAscan analysis, was evaluated via the calculation of Hedge’s g values. Statistically significant differences between two groups of candidate marker proteins were calculated by the Mann-Whitney U test. The diagnostic performance of each candidate was evaluated based on the AUC values of ROC curves. The cut-off value for SFN was set based on Youden's index in comparison with the DAD group and the tolerant control. Spearman's nonparametric analysis was used to evaluate the correlation between variables. All data were analyzed using the following software: GraphPad PRISM (version 8.4.3; GraphPad Software, San Diego, CA, USA) and Microsoft Excel.
+
+# Data Availability
+
+The source data underlying all figures and supplementary figures are provided as a Source Data file.
+
+# References
+
+1. Mok, T. S. et al. Gefitinib or carboplatin-paclitaxel in pulmonary adenocarcinoma. *N Engl J Med* **361**, 947–957, doi:10.1056/NEJMoa0810699 (2009).
+
+2. Zhou, C. et al. Erlotinib versus chemotherapy as first-line treatment for patients with advanced EGFR mutation-positive non-small-cell lung cancer (OPTIMAL, CTONG-0802): a multicentre, open-label, randomised, phase 3 study. *Lancet Oncol* **12**, 735–742, doi:10.1016/s1470-2045(11)70184-x (2011).
+
+3. Qi, W. X., Sun, Y. J., Shen, Z. & Yao, Y. Risk of interstitial lung disease associated with EGFR-TKIs in advanced non-small-cell lung cancer: a meta-analysis of 24 phase III clinical trials. *J Chemother* **27**, 40–51, doi:10.1179/1973947814y.0000000189 (2015).
+
+4. Cohen, M. H., Williams, G. A., Sridhara, R., Chen, G. & Pazdur, R. FDA drug approval summary: gefitinib (ZD1839) (Iressa) tablets. *Oncologist* **8**, 303–306, doi:10.1634/theoncologist.8-4-303 (2003).
+
+5. Cohen, M. H., Johnson, J. R., Chen, Y. F., Sridhara, R. & Pazdur, R. FDA drug approval summary: erlotinib (Tarceva) tablets. *Oncologist* **10**, 461–466, doi:10.1634/theoncologist.10-7-461 (2005).
+
+6. Ando, M. et al. Predictive factors for interstitial lung disease, antitumor response, and survival in non-small-cell lung cancer patients treated with gefitinib. *J Clin Oncol* **24**, 2549–2556, doi:10.1200/jco.2005.04.9866 (2006).
+
+7. Endo, M., Johkoh, T., Kimura, K. & Yamamoto, N. Imaging of gefitinib-related interstitial lung disease: multi-institutional analysis by the West Japan Thoracic Oncology Group. *Lung Cancer* **52**, 135–140, doi:10.1016/j.lungcan.2006.02.002 (2006).
+
+8. Lind, J. S., Smit, E. F., Grünberg, K., Senan, S. & Lagerwaard, F. J. Fatal interstitial lung disease after erlotinib for non-small cell lung cancer. *J Thorac Oncol* **3**, 1050–1053, doi:10.1097/JTO.0b013e318183a9f5 (2008).
+
+9. Silva, C. I. & Müller, N. L. Drug-induced lung diseases: most common reaction patterns and corresponding high-resolution CT manifestations. *Semin Ultrasound CT MR* **27**, 111–116, doi:10.1053/j.sult.2006.01.001 (2006).
+
+10. Ohnishi, H. et al. Comparative study of KL-6, surfactant protein-A, surfactant protein-D, and monocyte chemoattractant protein-1 as serum markers for interstitial lung diseases. *Am J Respir Crit Care Med* **165**, 378–381, doi:10.1164/ajrccm.165.3.2107134 (2002).
+
+11. Kinder, B. W. et al. Serum surfactant protein-A is a strong predictor of early mortality in idiopathic pulmonary fibrosis. *Chest* **135**, 1557–1563, doi:10.1378/chest.08-2209 (2009).
+
+12. Kawase, S. et al. Change in serum KL-6 level from baseline is useful for predicting life-threatening EGFR-TKIs induced interstitial lung disease. *Respir Res* **12**, 97, doi:10.1186/1465-9921-12-97 (2011).
+
+13. Kakugawa, T. et al. Serum heat shock protein 47 levels in patients with drug-induced lung disease. *Respir Res* **14**, 133, doi:10.1186/1465-9921-14-133 (2013).
+
+14. Mokra, D. & Kosutova, P. Biomarkers in acute lung injury. *Respir Physiol Neurobiol* **209**, 52–58, doi:10.1016/j.resp.2014.10.006 (2015).
+
+15. Kitsiouli, E., Nakos, G. & Lekka, M. E. Phospholipase A2 subclasses in acute respiratory distress syndrome. *Biochim Biophys Acta* **1792**, 941–953, doi:10.1016/j.bbadis.2009.06.007 (2009).
+
+16. Papiris, S. A. et al. High levels of IL-6 and IL-8 characterize early-on idiopathic pulmonary fibrosis acute exacerbations. *Cytokine* **102**, 168–172, doi:10.1016/j.cyto.2017.08.019 (2018).
+
+17. Kaarteenaho, R. & Kinnula, V. L. Diffuse alveolar damage: a common phenomenon in progressive interstitial lung disorders. *Pulm Med* 2011, 531302, doi:10.1155/2011/531302 (2011).
+
+18. Hermeking, H. et al. 14-3-3sigma is a p53-regulated inhibitor of G2/M progression. *Mol Cell* **1**, 3–11, doi:10.1016/s1097-2765(00)80002-7 (1997).
+
+19. Xu, X. et al. A2BAR activation attenuates acute lung injury by inhibiting alveolar epithelial cell apoptosis both in vivo and in vitro. *Am J Physiol Cell Physiol* **315**, C558-c570, doi:10.1152/ajpcell.00294.2017 (2018).
+
+20. Yan, X. et al. Nrf2/Keap1/ARE Signaling Mediated an Antioxidative Protection of Human Placental Mesenchymal Stem Cells of Fetal Origin in Alveolar Epithelial Cells. *Oxid Med Cell Longev* 2019, 2654910, doi:10.1155/2019/2654910 (2019).
+
+21. Wallach-Dayan, S. B. et al. Bleomycin initiates apoptosis of lung epithelial cells by ROS but not by Fas/FasL pathway. *Am J Physiol Lung Cell Mol Physiol* **290**, L790-l796, doi:10.1152/ajplung.00300.2004 (2006).
+
+22. Chan, T. A., Hermeking, H., Lengauer, C., Kinzler, K. W. & Vogelstein, B. 14-3-3Sigma is required to prevent mitotic catastrophe after DNA damage. *Nature* **401**, 616–620, doi:10.1038/44188 (1999).
+
+23. Dellambra, E. et al. Downregulation of 14-3-3sigma prevents clonal evolution and leads to immortalization of primary human keratinocytes. *J Cell Biol* **149**, 1117–1130, doi:10.1083/jcb.149.5.1117 (2000).
+
+24. Shiba-Ishii, A. et al. High expression of stratifin is a universal abnormality during the course of malignant progression of early-stage lung adenocarcinoma. *Int J Cancer* **129**, 2445–2453, doi:10.1002/ijc.25907 (2011).
+
+25. Shiba-Ishii, A. & Noguchi, M. Aberrant stratifin overexpression is regulated by tumor-associated CpG demethylation in lung adenocarcinoma. *Am J Pathol* **180**, 1653–1662, doi:10.1016/j.ajpath.2011.12.014 (2012).
+
+26. Martin, T. R., Nakamura, M. & Matute-Bello, G. The role of apoptosis in acute lung injury. *Crit Care Med* **31**, S184–188, doi:10.1097/01.CCM.0000057841.33876.B1 (2003).
+
+27. Guinee, D., Jr. et al. The potential role of BAX and BCL-2 expression in diffuse alveolar damage. *Am J Pathol* **151**, 999–1007 (1997).
+
+28. Guinee, D., Jr. et al. Association of p53 and WAF1 expression with apoptosis in diffuse alveolar damage. *Am J Pathol* **149**, 531–538 (1996).
+
+29. Bardales, R. H., Xie, S. S., Schaefer, R. F. & Hsu, S. M. Apoptosis is a major pathway responsible for the resolution of type II pneumocytes in acute lung injury. *Am J Pathol* **149**, 845–852 (1996).
+
+30. Taylor, W. R. & Stark, G. R. Regulation of the G2/M transition by p53. *Oncogene* **20**, 1803–1815, doi:10.1038/sj.onc.1204252 (2001).
+
+31. Ghaffari, A. et al. Fibroblast extracellular matrix gene expression in response to keratinocyte-releasable stratifin. *J Cell Biochem* **98**, 383–393, doi:10.1002/jcb.20782 (2006).
+
+32. Medina, A., Ghaffari, A., Kilani, R. T. & Ghahary, A. The role of stratifin in fibroblast-keratinocyte interaction. *Mol Cell Biochem* **305**, 255–264, doi:10.1007/s11010-007-9538-y (2007).
+
+33. Seok, J. K. & Boo, Y. C. p-Coumaric Acid Attenuates UVB-Induced Release of Stratifin from Keratinocytes and Indirectly Regulates Matrix Metalloproteinase 1 Release from Fibroblasts. *Korean J Physiol Pharmacol* **19**, 241–247, doi:10.4196/kjpp.2015.19.3.241 (2015).
+
+34. Rennard, S. I. et al. Estimation of volume of epithelial lining fluid recovered by lavage using urea as marker of dilution. *J Appl Physiol (1985)* **60**, 532–538, doi:10.1152/jappl.1986.60.2.532 (1986).
+
+# Supplementary Files
+
+- [supplementarymaterials.pdf](https://assets-eu.researchsquare.com/files/rs-690487/v1/48b8363ed9745e5e90d6c2c1.pdf)

0e3cab1ca0d8fe7f1b4b2f06d84ffeccaca02757d5ebd74dc71cc42ac91f2a1a/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "(a) Synthesis and shape of TCA_NH4. (b) NH4 charge-assisted hydrogen-bonded networks within the unit cell. (c) Simplified topological network of TCA_NH4 with TCA as a 3-connected nodes\u00a0 and octahedral cage. (d) SSBU of (NH4)4(COOH)8(H2O)2 with concentrated polynuclear clusters and charge-assisted hydrogen bonds. Nonhydrogen bonding hydrogen atoms are omitted for clarity.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "(a) Encapsulation and decapsulation of trihalomethane (CHX3, X=Cl, Br, I) within octahedral cage. (b) C-H\u2026N hydrogen bonds between trihalomethane and four noncontiguous TCA. (c) C-H...I hydrogen bonds for encapsulated CHI3. (d) NMR spectra of TCA_NH4@CHCl3, TCA_NH4@CHBr3, and TCA_NH4@CHI3 in DMSO-d6. Nonhydrogen bonding hydrogen atoms are omitted for clarity.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "(a) Color and configuration of encapsulated trihalomethane with defined distances between X1 and its neighboring atoms, the horizontal X1-X3 were decreased. (b) Distances of neighboring halogen atoms in encapsulated TCA_NH4.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "(a) Normalized emission spectra of TCA_NH4, TCA_NH4@CHCl3, TCA_NH4@CHBr3, and TCA_NH4@CHI3 excited at 360 nm. (b) Decay profile of TCA_NH4, TCA_NH4@CHCl3, TCA_NH4@CHBr3, and TCA_NH4@CHI3.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "(a) Photochromic reaction and thermal recovery of TCA_NH4@CHI3. (b) Solid-state absorption spectrum of TCA_NH4@CHI3 irradiated with a Xe lamp (400 nm, 200 mW/cm2). (c) EPR spectrum of TCA_NH4@CHI3 before and after irradiation under a UV-vis light source (200 mW/cm2). (d) Solid-state absorption spectrum of TCA_NH4@CHI3 after heating for an hour under dark environment. (e) UV-vis absorption spectrum of solid CHI3 dissolved in ethanol (6 g/L) under irradiation with a Xe lamp for 5 minutes. (f) In-situ emission spectrum of TCA_NH4@CHI3 before and after irradiation with a Xe lamp for 60 seconds.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0e3cab1ca0d8fe7f1b4b2f06d84ffeccaca02757d5ebd74dc71cc42ac91f2a1a/preprint/preprint.md

@@ -0,0 +1,156 @@
+# Abstract
+
+Developing supramolecular porous crystalline frameworks with tailor-made architectures from advanced secondary building units (SBUs) remains a pivotal challenge in reticular chemistry. Particularly for hydrogen-bonded organic frameworks (HOFs), construction of geometrical cavities through secondary units has been rarely achieved. Herein, a body-centered cubic HOF (TCA_NH₄) with octahedral cages was constructed by a C₃-symmetric building block and NH₄⁺ node-assembled cluster (NH₄)₄(COOH)₈(H₂O)₂ that served as supramolecular secondary building units (SSBUs), akin to the polynuclear SBUs in reticular chemistry. Specifically, the octahedral cages could encapsulate four homogenous haloforms including CHCl₃, CHBr₃, and CHI₃ with truncated octahedron configuration. Crystallographic evidence revealed the cages served as spatially-confined nanoreactors, enabling fast, broadband photochromic effect associated with the reversible photo/thermal transformation between encapsulated CHI₃ and I₂. Overall, this work provides a new strategy by shaping SSBUs to expand the framework topology of HOFs, and a new prototype of hydrogen-bonded nanoreactors to accommodate reversible photochromic reactions.  
+
+[Physical sciences/Chemistry/Supramolecular chemistry/Molecular capsules](/browse?subjectArea=Physical%20sciences%2FChemistry%2FSupramolecular%20chemistry%2FMolecular%20capsules)  
+[Physical sciences/Materials science/Soft materials/Self-assembly](/browse?subjectArea=Physical%20sciences%2FMaterials%20science%2FSoft%20materials%2FSelf-assembly)  
+[Physical sciences/Chemistry/Supramolecular chemistry/Crystal engineering](/browse?subjectArea=Physical%20sciences%2FChemistry%2FSupramolecular%20chemistry%2FCrystal%20engineering)
+
+# Introduction
+
+Construction of supramolecular porous structures with regular polyhedral architectures capturing guest molecules<sup>1-2</sup> offer opportunities for catalytic reaction,<sup>3,4</sup> molecular recognition,<sup>5</sup> and gas separation.<sup>6,7</sup> Most of the well-developed frameworks so far show open cavities with one or two-dimensional channels,<sup>8</sup> lacking specific selectivity and dynamic sites for targeted objects.<sup>9,10</sup> Oppositely, synthesis of enclosed cavities with advanced symmetric three-dimensional cage architecture<sup>11,12</sup> or well-defined polyhedron<sup>13</sup> could implement efficient recognition and encapsulation of guest molecules.<sup>14,15</sup> At the same time, the isolated cavity could provide a potential platform and confined microenvironment for further physicochemical reactions.<sup>11,16</sup>
+
+For building sophisticated structures, nature gives representative examples to take advantage of secondary structures of proteins such as α-helix and β-sheet to assemble them into advanced living entities. There is no doubt that SBUs play important roles as nodes and clusters<sup>17</sup> in designing sophisticated substances.<sup>18</sup> The architectural and mechanical stability of metal-organic frameworks (MOFs) imparted by corresponding SBUs have given rise to unique framework chemistry. However, this concept has been rarely discussed in the design of HOFs with regular porosity, as HOFs are assembled by rigid building molecules through intermolecular hydrogen bonding.<sup>19-24</sup> Pristine hydrogen bonding interactions were too weak to stabilize rigid and directional polynuclear clusters.<sup>25,26</sup> The rational synthesis of porous HOFs with preorganized and highly symmetrical networks has been a long-term challenge.<sup>27</sup> Therefore, conventional hydrogen bonding motifs usually result in one-dimensional channels<sup>28</sup> instead of hierarchical cavities,<sup>29</sup> exhibiting poor adaptability to accommodate specific guest molecules.<sup>30,31</sup> The stacking manner of the building blocks defines the topology and porosity of HOFs.<sup>32-36</sup> Nevertheless, the weak nature of hydrogen bonds provides the possibility for the hybrid synthesis of hydrogen-bonded networks.<sup>37,38</sup> As a consequence, the employment of polyhedral clusters<sup>39,40</sup>, diverse auxiliary interactions<sup>41-43</sup> such as charge-assisted hydrogen bonds<sup>44-47</sup> and multiple components<sup>48-50</sup> expands the opportunity to construct hydrogen-bonded networks based on more advanced, robust building units. At this point, NH<sub>4</sub><sup>+</sup>, with a tetrahedral geometric structure stabilized by four equivalent N-H bonds and appropriate size (<em>r</em><sub>ionic</sub> ≈ 1.5 Å), behaves in some ways like metal cations in crystal engineering. Cationic interaction of NH<sub>4</sub><sup>+</sup> could reliably contribute to electrostatic interactions through charge-assisted hydrogen bonds and potentially regulate the assembly of hydrogen-bonded networks. By donating ionic regulator in the assembly with organic building molecules, we envision that NH<sub>4</sub><sup>+</sup> offers tremendous opportunities for constructing advanced supramolecular architectures.<sup>51</sup>
+
+In this work, NH<sub>4</sub><sup>+</sup> node-assembled polynuclear clusters, akin to the SBUs in reticular chemistry, are exploited in HOF construction for the first time, enabling the formation of a novel body-centered cubic HOF (<strong>TCA_NH<sub>4</sub></strong>) with distinctive molecular encapsulation, reversible and broadband photochromism, especially sensitive under 400 nm light irradiation. The polynuclear clusters (NH<sub>4</sub>)<sub>4</sub>(COOH)<sub>8</sub>(H<sub>2</sub>O)<sub>2</sub>, sustained by charge-assisted hydrogen bonds, are not as rigid and directional as the classic SBUs stabilized by metal-ligand coordination bonds, which are thus named as supramolecular secondary building units (SSBUs). The unit cell of <strong>TCA_NH<sub>4</sub></strong> contained an enclosed octahedral cage that could accommodate four haloforms such as CHCl<sub>3</sub>, CHBr<sub>3</sub>, and CHI<sub>3</sub>. All of the encapsulated haloforms exhibited truncated octahedron configurations that suitably matched the enclosed octahedral cage. The distance of adjacent halogens within the encapsulated cavity decreased from Cl-Cl (4.513 Å), Br-Br (4.894 Å) to I-I (3.991 Å) along with decayed fluorescence intensity and shortened lifetime for the corresponding HOFs. Specifically, the host HOF with CHI<sub>3</sub> encapsulated in the confined octahedral cages, i.e., <strong>TCA_NH<sub>4</sub>@CHI<sub>3</sub></strong>, showed fast photochromism (60 s) accompanied by fluorescence quenching outcome. Single crystal−single crystal transformation (SCSCT) analysis and electron paramagnetic resonance (EPR) revealed that CHI<sub>3</sub> was photolyzed to metastable I<sub>2</sub>. More interestingly, such a photochromic reaction, occurring within the isolated octahedral cage of the host HOF, was reversible, and the color of <strong>TCA_NH<sub>4</sub>@CHI<sub>3</sub></strong> could be recovered through the heating-induced regeneration of CHI<sub>3</sub> from the metastable I<sub>2</sub>.
+
+# Results
+
+A cubic single crystal of TCA_NH₄ was obtained by evaporating the mixed solution of C₃-symmetric 4,4',4''-nitrilotribenzoic acid (H₃TCA)⁵²,⁵³ and NH₃·H₂O at room temperature (Supporting information, Figure S1). Blocky TCA_NH₄ was also formed by CH₂Cl₂ diffusion in the mixed solution of H₃TCA and NH₄Cl for weeks (Figure S2). SCXRD results showed H₃TCA crystallized in a body-centered cubic system and I23 space group (Figure 1a, Table S1) with a cell length of 20.9737 Å. NH₄⁺ served as nodes and bridged two neighboring H₃TCA molecules through neutral (2.509 Å) and charge-assisted N-H(+)…O(−) (1.830 Å) hydrogen bonds (Figure 1b, Table S2). Water molecules strengthened the polynuclear clusters through O-H...O (2.069 Å) and N-H...O (2.478 Å) hydrogen bonds. Topologically, H₃TCA molecules can be considered as 3-connected nodes, it can be found NH₄⁺ exactly arranged as the round to connect eight H₃TCA molecules within the adjacent unit cell. Consequently, an octahedral cage with a size of about 12 Å was encompassed by NH₄⁺ and H₃TCA molecules (Figure 1c, Figure S3) and every unit cell contained an isolated cage (Figure S4). As expected, NH₄⁺ acted as the node and every four of them formed supramolecular clusters to connect neighbouring cells (Figure 1d, Figure S5).
+
+Specifically, the above-obtained supramolecular polynuclear clusters (NH₄)₄(COOH)₈(H₂O)₂, with well-defined composition and structure sustained by hybrid hydrogen bonds, served as SSBUs in building the cubic network of TCA_NH₄ with octahedral cavities. As shown in Figure S5, every NH₄⁺ bridged two carboxy groups, and two H₂O molecules further linked carboxy groups and NH₄⁺ from two sides of the SSBUs. The supramolecular clusters could be further considered as two oxygen (two H₂O) and four nitrogen (four NH₄⁺) models. The simplified SSBUs interlaced orthogonally and connected with each other by O-H…O hydrogen bonds (Figure S5, Figure S6). On the whole, the NH₄⁺ node-assembled clusters not only connected each cell through six surfaces of cubic but also derived into SSBUs to construct advanced frameworks with enclosed octahedral cages.
+
+The enclosed cage or cavity⁵⁴ appeared scarce since most HOFs possessed open channels⁵⁵,⁵⁶ and pores⁵⁷,⁵⁸ to transport guest molecules.⁵⁹–⁶² The pore structure of TCA_NH₄ appeared an octahedral shape with solvent accessible volume of 14.9 %. Interestingly, we found the void of TCA_NH₄ could accommodate homogeneous haloforms including CHCl₃, CHBr₃, and CHI₃. The encapsulation was carried out in solvothermal synthesis at 70°C in the presence of corresponding trihalomethane molecules. Cubic single crystals of TCA_NH₄@CHCl₃, TCA_NH₄@CHBr₃, and TCA_NH₄@CHI₃ were obtained within about 6 hours (Supporting information, Figure S1). The encapsulation could also be achieved by CH₂Cl₂ diffusion in mixed solutions of H₃TCA and NH₄Cl for weeks. As shown in Figure 2a, SCXRD results found four homogeneous CHX₃ (X=Cl, Br, I) were assembled in the octahedral void within the unit cell. Trihalomethane formed C-H…N hydrogen bonds with four nonadjacent H₃TCA molecules (Figure 2b). The angle of the three hydrogen bonds was all 180° but the distances slightly elongated with the increase of halogen radius (Table S3). It was found that four trihalomethanes were precisely accommodated through truncated octahedron configuration (Figure 2a, Figure S7), and each trihalomethane was parallel to one plane that contained one H₃TCA molecule. Additionally, CHI₃ formed C-H…I hydrogen bonds (2.817 Å) to enhance the encapsulation (Figure 2c, Table S4). SCXRD results of encapsulated crystals indicated that the encapsulation of trihalomethanes didn’t change the crystal structure and the lattice constant and the volume of the unit cell was almost kept constant (Table S5). But due to the inefficient and incompact encapsulation of CHX₃, the diffracted intensity of halogen atoms especially for iodine was slightly disordered, which implied that the octahedral cages were partly unoccupied by CHX₃ molecules. Still, the encapsulation of four trihalomethanes with truncated octahedron shape compatibly matched the octahedral cavity of the hydrogen-bonded framework. Powder X-ray diffraction (PXRD) analysis proved that after encapsulation of trihalomethane, the crystal structure of TCA_NH₄ remained stable (Figure S8). Thermogravimetric (TG) results suggested that TCA_NH₄, TCA_NH₄@CHCl₃, TCA_NH₄@CHBr₃, and TCA_NH₄@CHI₃ had similar thermal degradation behaviors (Figure S9) due to the encapsulation efficiency were too low and the weight percent of CHX₃ were also too low (less than 40 mg for encapsulation). It can be found that TCA_NH₄ held moderate stability in organic solvent and water (Figure S10). The thermal and solvent stability of crystals indicated that SSBUs built by charged-assisted⁶³ supramolecular clusters stabilized TCA_NH₄ and avoided further phase change⁶⁴ and collapse of the framework. Taking advantage of the solution-processible benefit of HOFs,⁶⁵ the encapsulated trihalomethane could be decapsulated by mild hydrolysis of hydrogen-bonded networks. After heating and dissolving in dimethyl sulfoxide (DMSO), nuclear magnetic resonance (NMR) results revealed that the encapsulated TCA_NH₄@CHCl₃, TCA_NH₄@CHBr₃, and TCA_NH₄@CHI₃ were observed to decompose and trihalomethane were observed to release (Figure 2d). The encapsulation efficiency was roughly estimated by intensity of NMR spectra and all encapsulated samples were less than 10%, showing an incompact capture of trihalomethane.
+
+With a closer examination, although the trihalomethane molecules were all encapsulated in truncated octahedrons, their configuration appeared slightly different. As shown in Figure 3a, the distances of the nearest halogen atoms were classified as X1-X1, X1-X2, and X1-X3 (X =Cl, Br, I, Figure S11). Each classification included two directions. SCXRD results revealed that the distances of X1-X1 and X1-X2 were almost maintained after assembling CHX₃ (Figure 3b). However, one distance of X1-X3 decreased from Cl1-Cl3 (4.513 Å) to Br1-Br3 (4.894 Å), and then to I1-I3 (3.99 Å). The shortened distances of adjacent halogen atoms provided the possibility for bonding reactions, especially for iodine. The color of TCA_NH₄ was changed after encapsulating haloforms, particularly for TCA_NH₄@CHBr₃ and TCA_NH₄@CHI₃. The color of crystals was resulted from the guest trihalomethane. In addition, the fluorescence emission spectrum of TCA_NH₄ showed no shift after assembling with CHCl₃ and CHBr₃ (Figure 4a), but the emission spectrum red-shifted from 430 nm to 480 nm for TCA_NH₄@CHI₃ attributed to the absorbed energy transfer from H₃TCA to CHI₃. Solid-state fluorescence reflection spectrum indicated that the fluorescence efficiency of TCA_NH₄ decreased (Figure S12) after encapsulating trihalomethane, because of heavy atom effect-induced fluorescence quenching. In the same way, the excitation spectrum of TCA_NH₄ also decreased after assembling haloforms, and the maximum excitation wavelength slightly blue-shifted from TCA_NH₄ at 410 nm to TCA_NH₄@CHI₃ at 380 nm (Figure S13) because of higher excitation energy for hydrogen-bonded networks of H₃TCA and trihalomethane. The lifetime of TCA_NH₄ was 2.09 ns (Figure 4b, Table S6) and it slightly decreased to 2.05 ns and 1.69 ns after assembling CHCl₃ and CHBr₃, respectively, following the quenching mechanism. But the lifetime increased to 2.43 ns after assembling CHI₃ maybe due to the energy exchange between excited species of H₃TCA and CHI₃ contributing to the delayed decay of the excited state. Solid-state absorption spectrum proved the structure of TCA_NH₄ did not vary obviously when compared with TCA_NH₄@CHCl₃, TCA_NH₄@CHBr₃, and TCA_NH₄@CHI₃ (Figure S14). The absorption peak appeared at 400 nm, indicating the crystals were more sensitive to ultraviolet light, especially for 400 nm of light. The above results indicated that the octahedral cavity within TCA_NH₄ precisely accommodated four trihalomethanes with matched truncated octahedron shape. The color of assembled crystals changed by varied guest molecules. The fluorescence efficiency decreased after encapsulating haloforms. Particularly, the distances between adjacent halogen atoms were shortened especially for TCA_NH₄@CHI₃, which promoted energy transfer from excited H₃TCA to CHI₃ and prolonged lifetime and provided the possibility for bonding formation.
+
+The encapsulation of photoactive CHI₃ inspired us to examine the photo-responsive behaviors of TCA_NH₄@CHI₃. Interestingly, TCA_NH₄@CHI₃ showed a 60-second fast photochromic phenomenon under irradiation of a UV-vis xenon lamp at 400 nm with a light power density of about 200 mW/cm² (Figure 5a). SCSCT analysis revealed the distance of I1-I3 was further shortened to 2.822 Å of metastable I₂ approximately equivalent to I₂ (2.6 Å), and yellow TCA_NH₄@CHI₃ quickly transformed to brown TCA_NH₄@I₂. The solid-state absorption spectrum recorded the photolysis of iodoform (Figure 5b). The absorption peak of CHI₃ at 418 nm decreased and the absorption peak of I₂ at 450 nm appeared after irradiation with UV-vis light for 5 minutes. Electron paramagnetic resonance (EPR) of TCA_NH₄@CHI₃ was conducted to examine the photolysis of iodoform. Disordered signals of TCA_NH₄@CHI₃ turned to specific radical signals (Figure 5c) after in-situ illumination by a UV-vis light source with a light power density of about 200 mW/cm². The g value at about 2.005 indicated the free electron resulted from the homolytic cleavage of the C-I bond⁶⁶ (Figure S15). Accordingly, TCA_NH₄@CHCl₃ was also irradiated by a UV-vis light source to examine free radical signals. As shown in Figure S16, disordered signals indicated that no free radicals were generated or chemical bonds broken after irradiation. The SCXRD and EPR results proved that the photochromic reaction of TCA_NH₄@CHI₃ stemmed from the photolysis of iodoform and the formation of metastable I₂. We consider the free radical generated within the enclosed cavity could further react with metastable I₂, on account of the isolated octahedral cavity providing a spatially-confined platform as nanoreactor that could stabilize free radicals and metastable I₂. As a result, after the photochromic reaction, brown TCA_NH₄@I₂ was immediately immersed in liquid paraffin to equably heat under a dark environment for an hour. As expected, the brown crystal returned to yellow and solid-state absorption spectrum results confirmed the recovery of iodoform (Figure 5a, Figure 5d). Photochromic reaction and thermal recovery were repeated to investigate the reversible cycle of photo-responsive performance of TCA_NH₄@CHI₃. After repeating 5 rounds of irradiating and heating processes, TCA_NH₄@CHI₃ still maintained photochromic activity and the yellow crystal was still renewed (Figure S17, Figure S18). The results indicated that fast photochromic reaction and thermal recovery of encapsulated CHI₃ within an octahedral cage maintained a steady process to contribute as a new type of photochromic material.
+
+For a photochromic process, usually, the dependence of light wavelength with a limited range restricts general applications in environment probing.⁶⁷–⁶⁹ HOFs have shown great potential in luminescent fields.⁷⁰ Surprisingly, TCA_NH₄@CHI₃ not only turned color at 400 nm of light but also held a broad photosensitive scope ranging from X-ray, ultraviolet light, sunlight, and even low-energy light-emitting diode (8 mW/cm²). The varied light source and light intensity made TCA_NH₄@CHI₃ with different photosensitive efficiency ranging from 60 seconds to 5 minutes (Table S7). The broad photosensitive scope and fast photochromic efficiency of TCA_NH₄@CHI₃ originated from the combination of multiple effects. Firstly, as shown by SCXRD in Figure 3b, the shortened distance between adjacent I1-I3 in the enclosed octahedral cavity provided the possibility for a bond-formation reaction. Secondly, the decreased fluorescence efficiency of TCA_NH₄ after assembling trihalomethane promoted the transfer of the absorbed energy from H₃TCA to CHI₃, so that CHI₃ was energetic for potential physicochemical reactions. Thirdly, CHI₃ is intrinsically photosensitive. As shown in Figure 5e, the pure CHI₃ powder was dissolved in ethanol (6 g/L) and irradiated under a xenon lamp with a light power density of about 200 mW/cm². The color of CHI₃ quickly turned brown within minutes (Inserted graph) and the UV-vis absorption spectrum proved the photolysis of CHI₃ and generation of I₂. The characteristic peaks of CHI₃ at 267 nm, 301 nm, and 339 nm decreased and characteristic peaks of I₂ at 293 nm and 349 nm appeared. Solid powder of CHI₃ also showed photosensitive decomposition except that the color of the powder showed no visible change under 10 minutes irradiation with Xe lamp (Figure S19). Nevertheless, after irradiation with Xe lamp, both dissolved CHI₃ and solid powder of CHI₃ could not recover whether heating or standing due to the lack of an enclosed void for the restriction and stabilization of decomposed products and free radicals. In addition, the fluorescence emission spectrum of TCA_NH₄@CHI₃ was markedly quenched to 5.59% after in-situ irradiation with Xe lamp for 60 seconds because of fluorescence quenching caused by metastable I₂ (Figure 5f), and emission spectrum blue-shifted from 480 nm to 450 nm (Figure S20) that was close to TCA_NH₄@CHCl₃ and TCA_NH₄@CHBr₃ due to the break of energy transfer between H₃TCA and CHI₃. The excitation spectrum was also quenched after the photochromic reaction (Figure S21) and the lifetime slightly decreased (Table S6, 2.43 ns vs 2.37 ns). The above results confirmed that the octahedral cavity within TCA_NH₄ offered unique nanoreactors for photochromic reaction and spatially-confined thermal recovery of CHI₃. The fast photochromic efficiency, broad photosensitivity, and unique fluorescence quenching of TCA_NH₄@CHI₃ showed great potential in sensing and environmental probing.
+
+# Discussion
+
+In summary, taking advantage of NH₄⁺ as nodes to bridge H₃TCA building blocks through charge-assisted hydrogen bonds, supramolecular polynuclear clusters (NH₄)₄(COOH)₈(H₂O)₂ were obtained and, for the first time, served as SSBUs to construct a distinctive body-centered cubic hydrogen-bonded organic framework. The assembled TCA_NH₄ contained enclosed octahedral cages that precisely accommodated four homogenous trihalomethanes including CHCl₃, CHBr₃, and CHI₃. The truncated octahedron form of encapsulated trihalomethane compatibly matched the octahedral cavity of TCA_NH₄ through strong C-H…N hydrogen bonds. Besides, the fluorescence emission efficiency of TCA_NH₄ decreased after assembling different haloforms, promoting the transfer of the absorbed light energy to haloforms. The distances between adjacent halogen atoms within the octahedral cavity were shortened especially for CHI₃. As a result, TCA_NH₄@CHI₃ showed fast photochromic efficiency, broad photosensitivity, and unique fluorescence quenching behavior. The brown TCA_NH₄@I₂ could be recovered to TCA_NH₄@CHI₃ after heating due to the enclosed octahedral cages that provided spatially-confined nanoreactors to stabilize the free radical and metastable byproducts. This study provides new insights into the framework chemistry of HOFs and particularly an adaptable strategy for HOF construction from supramolecular polynuclear clusters sustained by charge-assisted H-bonds. The established methodology would expand the framework topology of HOFs and accelerate the customized development of HOFs with predetermined architectures. Moreover, this study also provided a new prototype of sensitive photochromic materials based on CHI₃ encapsulation within the isolated cavities of HOF nanoreactor platforms.
+
+# Methods
+
+General Remark  
+4,4\',4\'\'-nitrilotribenzoic acid (H₃TCA) and other reagents were purchased from commercial approach with guarantee reagent. Powder X-ray diffraction data (PXRD) were performed by a Rigaku MiniFlex600 (40 kV, 15 mA) with a graphite-monochromatized Cu Kα radiation. Electron paramagnetic resonance (EPR) was conducted on a spectrometer (JEOL, JES FA-200). The data was collected with microwave field power of 0.7 mW. The modulation frequency was 100 kGHz and the microwave frequency was 9.7 GHz. Thermogravimetric analysis (TGA) was conducted on an SDT Q600 analyzer with a heating rate of 10 °C/min under N₂ (100 mL/min) atmosphere. ¹H NMR spectra was conducted in DMSO-d6 solution by JNM-ECZ400S (400 MHz) spectrometer.
+
+Synthesis of single crystals  
+TCA_NH₄: 120 mg H₃TCA was dissolved in 0.8 mL N, N-Dimethylformamide (DMF) in a vial. 20 μL NH₃H₂O was added and the mixed solution was slightly heated to get a clear solution. Then the solution was stood for evaporation and a cubic single crystal can be obtained after 3 weeks (Figure S1). Block single crystals also can be obtained by CH₂Cl₂ diffusion in a mixed solution of H₃TCA and NH₄Cl for 3 weeks.  
+TCA_NH₄@CHCl₃: 120 mg H₃TCA was dissolved in 1 mL DMF in a vial. 40 μL NH₃H₂O was added to get precipitation. 30 μL deionized water was added and the vial was slightly heated to get a clear solution. 30 μL CHCl₃ was added and the mixed solution was quietly evaporated at 70°C. Cubic single crystals can be obtained within 6 hours (Figure S1).  
+TCA_NH₄@CHBr₃: 120 mg H₃TCA was dissolved in 0.8 mL DMF in a vial. 30 μL NH₃H₂O was added to get precipitation. 40 μL deionized water was added and the vial was slightly heated to get a clear solution. 30 μL CHBr₃ was added and the mixed solution was quietly evaporated at 70°C. Cubic single crystals can be obtained within 6 hours (Figure S1).  
+TCA_NH₄@CHI₃: 200 mg H₃TCA was dissolved in 0.8 mL DMF in a vial. 40 μL NH₃H₂O was added to get precipitation. 80 μL deionized water was added and the vial was slightly heated to get a clear solution. 20 mg CHI₃ was added and the vial was heated to get a clear solution. Then the mixed solution was quietly evaporated at 66°C under a dark environment. Yellow single crystals can be obtained within 6 hours (Figure S1).  
+The encapsulation of CHCl₃, CHBr₃, and CHI₃ can also be obtained by CH₂Cl₂ diffusion in a mixed solution of H₃TCA, NH₄Cl, and trihalomethane for 3 weeks.
+
+Single-crystal X-ray diffraction analysis  
+Single crystal measurement was conducted on a Bruker APEX-II CCD with Cu Kα (λ = 1.54184 Å) X-ray sources. SADABS-2016/2(Bruker,2016/2) was used for absorption correction. The structure refinement was performed with Olex2 1.5 program. The structure was solved by ShelXT intrinsic phasing method and was refined by ShelXL least-squares techniques. All nonhydrogen atoms were refined with anisotropic displacement parameters. SQUEEZE function from PLATON program was taken to treat disordered solvent molecules in voids for TCA_NH₄. Data collection and refinement details are listed in Table S1 and supplementary crystallographic CIF files have been deposited on Cambridge Crystallographic Data Centre (CCDC) with the number 2309944 for TCA_NH₄, 2309942 for TCA_NH₄@CHCl₃, 2309941 for TCA_NH₄@CHBr₃, 2309943 for TCA_NH₄@CHI₃, and 2309940 for TCA_NH₄@I₂.
+
+Solid-state spectroscopy measurement  
+Solid-state absorption spectroscopy was measured on a CRAIC 20/30 microspectrophotometer. A Xe lamp (90 W) was used as UV-vis and fluorescence source for measurement. For the UV-vis absorption test, parameters were set as Time1=57ms: Objective=15X: Aperture=4×4 mm². For the fluorescence reflection test, 365 nm channel was used and parameters were set as Time1=1000ms: Objective=15X: Aperture=4×4 mm².
+
+The fluorescence emission and absorption measurement  
+The data was collected on a FLS1000 spectrometer with Visible/red-PMT detector and Xe lamp (300 W) as the light source for measurement. The excitation wavelength was 360 nm (bandwidth: 1 nm) and the emission wavelength was 470 nm (bandwidth: 0.4 nm). A nF920 nanosecond flashlamp (H₂ padded in chamber) was used for lifetime measurement. The time range was less than 500 millisecond and channels were fixed to 1024, and stop condition was set to 1000 counts. The decay lifetime was fitted according to multiexponential function models. The matching rate of fitting results were estimated by χ² (1.0-1.1, Table S6).
+
+Photochromic reaction and thermal recovery  
+The synthesized yellow TCA_NH₄@CHI₃ crystal was washed with dimethyl formamide (DMF) and acetone and then dried out. The yellow crystal was put into a vial and heated at 60°C under a dark environment for 30 minutes to activate the crystal. Then the crystal was put under a Xe lamp (200 mW/cm², 400 nm) to be exposed for 1 min or directly exposed under intense sunlight for 1-5 mins. The color of the crystal quickly turned to brown and then the crystal was immediately immersed in liquid paraffin to equably heat at 100°C for an hour under dark surroundings. The color of the crystal recovered to yellow. The crystal was taken out and washed with organic solvent to remove liquid paraffin and then dried out. The recovered crystal was preserved under dark surroundings.
+
+# References
+
+1. Katsoulidis AP et al (2019) Chemical control of structure and guest uptake by a conformationally mobile porous material. Nature 565:213–217
+2. Yun L, Zhao W (2019) Chloride capture using a C–H hydrogen-bonding cage. Science 365
+3. Han B et al (2019) Postsynthetic metalation of a robust hydrogen-bonded organic framework for heterogeneous catalysis. J Am Chem Soc 141:8737–8740
+4. Gong W et al (2019) Permanent porous hydrogen-bonded frameworks with two types of Bronsted acid sites for heterogeneous asymmetric catalysis. Nat Commun 10:600
+5. Deng J-H (2020) π-π stacking interactions: non-negligible forces for stabilizing porous supramolecular frameworks. Sci Adv
+6. Yang Y et al (2021) Ethylene/ethane separation in a stable hydrogen-bonded organic framework through a gating mechanism. Nat Chem 13:933–939
+7. Cai Y et al (2023) Pore modulation of hydrogen-bonded organic frameworks for efficient separation of propylene. Angew Chem Int Ed 62:e202308579
+8. Liu Y et al (2022) Hydrogen-bonded metal-nucleobase frameworks for efficient separation of xenon and krypton. Angew Chem Int Ed 61:e202117609
+9. Zeng H et al (2021) Orthogonal-array dynamic molecular sieving of propylene/propane mixtures. Nature 595:542–548
+10. Samanta J et al (2022) An ultra-dynamic anion-cluster-based organic framework. Chem 8:253–267
+11. Liu Y, Hu C, Comotti A, Ward MD (2011) Supramolecular archimedean cages assembled with 72 hydrogen bonds. Science 333
+12. Zhu Q et al (2022) Analogy powered by prediction and structural invariants: computationally led discovery of a mesoporous hydrogen-bonded organic cage crystal. J Am Chem Soc 144:9893–9901
+13. He Y et al (2008) Hierarchical self-assembly of DNA into symmetric supramolecular polyhedra. Nature 452:198–201
+14. Liang W et al (2019) Enzyme encapsulation in a porous hydrogen-bonded organic framework. J Am Chem Soc 141:14298–14305
+15. Chen G, Huang S, Ma X, He R, Ouyang G (2023) Encapsulating and stabilizing enzymes using hydrogen-bonded organic frameworks. Nat Protocols
+16. Hisaki I et al (2019) Acid responsive hydrogen-bonded organic frameworks. J Am Chem Soc 141:2111–2121
+17. Farid N et al (2008) Supermolecular building blocks (SBBs) for the design and synthesis of highly porous metal-organic frameworks. J Am Chem Soc 130:1833–1835
+18. Yang S (2015) Supramolecular lattices from tetrahedral nanobuilding blocks. Science 348
+19. Zhang Z, Ye Y, Xiang S, Chen B (2022) Exploring multifunctional hydrogen-bonded organic framework materials. Acc Chem Res 55:3752–3766
+20. Ermer O (1988) Fivefold-diamond structure of adamantane-1,3,5,7-tetracarboxylic acid. J Am Chem Soc 110:3474–3754
+21. Nadia Malek T, Maris M Simard, Wuest JD (2005) Molecular tectonics. selective exchange of cations in porous anionic hydrogen-bonded networks built from derivatives of tetraphenylborate. J Am Chem Soc 127:5910–5916
+22. Soleimani Abhari P et al (2023) Recent progress in gas separation platforms based on hydrogen-bonded organic frameworks (HOFs). Inorg Chem Front 10:6134–6159
+23. Lin R-B, Chen B (2022) Hydrogen-bonded organic frameworks: Chemistry and functions. Chem 8:2114–2135
+24. Song X et al (2022) Design rules of hydrogen-bonded organic frameworks with high chemical and thermal stabilities. J Am Chem Soc 144:10663–10687
+25. Lin ZJ, Mahammed SAR, Liu TF, Cao R (2022) Multifunctional porous hydrogen-bonded organic frameworks: current status and future perspectives. ACS Cent Sci 8:1589–1608
+26. Xi X-J et al (2023) Robust porous hydrogen-bonded organic frameworks: Synthesis and applications in gas adsorption and separation. Giant 16
+27. Qin WK et al (2022) Reticular synthesis of hydrogen-bonded organic frameworks and their derivatives via mechanochemistry. Angew Chem Int Ed 61:e202202089
+28. Chen X et al (2022) A proton conductive porous framework of an 18-crown-6-ether derivative networked by rigid hydrogen bonding modules. Angew Chem Int Ed 61:e202211686
+29. Shields CE et al (2023) Experimental confirmation of a predicted porous hydrogen-bonded organic framework. Angew Chem Int Ed 62:e202303167
+30. Pulido A et al (2017) Functional materials discovery using energy-structure-function maps. Nature 543:657–664
+31. Cui P et al (2020) An expandable hydrogen-bonded organic framework characterized by three-dimensional electron diffraction. J Am Chem Soc 142:12743–12750
+32. Liu Y et al (2023) Hybrid hydrogen-bonded organic frameworks: structures and functional applications. Chem -Eur J 29:e202202655
+33. Wang B, Lin RB, Zhang Z, Xiang S, Chen B (2020) Hydrogen-bonded organic frameworks as a tunable platform for functional materials. J Am Chem Soc 142:14399–14416
+34. Yang Y et al (2022) An ultramicroporous hydrogen-bonded organic framework exhibiting high C₂H₂/CO₂ separation. Angew Chem Int Ed 61:e202207579
+35. Chen L et al (2022) A flexible hydrogen-bonded organic framework constructed from a tetrabenzaldehyde with a carbazole N-H binding site for the highly selective recognition and separation of acetone. Angew Chem Int Ed 61:e202213959
+36. Li YL et al (2020) Record complexity in the polycatenation of three porous hydrogen-bonded organic frameworks with stepwise adsorption behaviors. J Am Chem Soc 142:7218–7224
+37. Hashimoto T, Oketani R, Nobuoka M, Seki S, Hisaki I (2023) Single crystalline, non-stoichiometric cocrystals of hydrogen-bonded organic frameworks. Angew Chem Int Ed 62:e202215836
+38. Bao Z et al (2018) Fine tuning and specific binding sites with a porous hydrogen-bonded metal-complex framework for gas selective separations. J Am Chem Soc 140:4596–4603
+39. Montis R et al (2021) Complex structures arising from the self-assembly of a simple organic salt. Nature 590:275–278
+40. Halliwell CA et al (2022) Hierarchical assembly of a micro- and macroporous hydrogen-bonded organic framework with tailored single-crystal size. Angew Chem Int Ed 61:e202208677
+41. Madueno R, Raisanen MT, Silien C, Buck M (2008) Functionalizing hydrogen-bonded surface networks with self-assembled monolayers. Nature 454:618–621
+42. Sun Y et al (2023) Bio-inspired synthetic hydrogen-bonded organic frameworks for efficient proton conduction. Adv Mater 35:e2208625
+43. Zhang AA et al (2022) Partial metalation of porphyrin moieties in hydrogen-bonded organic frameworks provides enhanced CO₂ photoreduction activity. Angew Chem Int Ed 61:e202203955
+44. Bai X-T et al (2023) Ultra-high proton conductivity iHOF based on guanidinium arylphosphonate for proton exchange membrane fuel cells. Chem Mater 35:3172–3180
+45. Ding X et al (2022) Binary solvent regulated architecture of ultra-microporous hydrogen-bonded organic frameworks with tunable polarization for highly-selective gas separation. Angew Chem Int Ed 61:e202116483
+46. Zhou Y et al (2023) Tuning pore polarization to boost ethane/ethylene separation performance in hydrogen-bonded organic frameworks. Angew Chem Int Ed 62:e202305041
+47. Li Y et al (2018) Guest exchange through facilitated transport in a seemingly impenetrable hydrogen-bonded framework. J Am Chem Soc 140:12915–12921
+48. Zhang J, Feng Y, Staples RJ, Zhang J, Shreeve JM (2021) Taming nitroformate through encapsulation with nitrogen-rich hydrogen-bonded organic frameworks. Nat Commun 12:2146
+49. Han Z et al (2023) High-efficiency lithium-ion transport in a porous coordination chain-based hydrogen-bonded framework. J Am Chem Soc 145:10149–10158
+50. Xing G et al (2019) A double helix of opposite charges to form channels with unique CO₂ selectivity and dynamics. Chem Sci 10:730–736
+51. Ding X et al (2023) Charge-assisted hydrogen-bonded organic frameworks with inorganic ammonium regulated switchable open polar sites. Small 19:e2207771
+52. Nandi S, Chakraborty D, Vaidhyanathan R (2016) A permanently porous single molecule H-bonded organic framework for selective CO₂ capture. Chem Commun 52:7249–7252
+53. Cai Y et al (2023) Robust microporous hydrogen – bonded organic framework for highly selective purification of methane from natural gas. Microporous Mesoporous Mater 352
+54. Marchetti D et al (2022) Selective and reversible solvent uptake in Tetra-4-(4-pyridyl) phenylmethane-based supramolecular organic frameworks. Chem -Eur J 28:e202202977
+55. Wang J-X, Gu X-W, Lin Y-X, Li B, Qian G (2021) A novel hydrogen-bonded organic framework with highly permanent porosity for boosting ethane/ethylene separation. ACS Mater Lett 3:497–503
+56. Vicent-Morales M et al (2022) Semiconductor porous hydrogen-bonded organic frameworks based on tetrathiafulvalene derivatives. J Am Chem Soc 144:9074–9082
+57. Liu X et al (2022) A robust redox-active hydrogen-bonded organic framework for rechargeable batteries. J Mater Chem A 10:1808–1814
+58. Guo C et al (2022) Hydrogen-bonded organic framework for high-performance lithium/sodium-iodine organic batteries. Angew Chem Int Ed 61:e202213276
+59. Chen G et al (2022) Hydrogen-bonded organic framework biomimetic entrapment allowing non-native biocatalytic activity in enzyme. Nat Commun 13:4816
+60. Huang Q et al (2022) Achieving bright mechanoluminescence in a hydrogen-bonded organic framework by polar molecular rotor incorporation. CCS Chem 4:1643–1653
+61. Wang B et al (2020) Microporous hydrogen-bonded organic framework for highly efficient turn-up fluorescent sensing of aniline. J Am Chem Soc 142:12478–12485
+62. Bassanetti I et al (2018) Flexible porous molecular materials responsive to CO₂, CH₄ and Xe stimuli. J Mater Chem A 6:14231–14239
+63. Ward MD (2005) Design of crystalline molecular networks with charge-assisted hydrogen bonds. Chem Commun, 5838–5842
+64. Slavney AH et al (2022) Liquid and glass phases of an alkylguanidinium sulfonate hydrogen-bonded organic framework. J Am Chem Soc 144:11064–11068
+65. Yang X et al (2023) Solution-processed hydrogen-bonded organic framework nanofilms for high-performance resistive memory devices. Adv Mater 35:e2305344
+66. Bosnidou AE, Muniz K (2019) Intermolecular radical Csp3-H amination under iodine catalysis. Angew Chem Int Ed 58:7485–7489
+67. Cai S, An Z, Huang W (2022) Recent advances in luminescent hydrogen-bonded organic frameworks: structures, photophysical properties, applications. Adv Funct Mater 32
+68. Shi Y et al (2022) Multiple yet switchable hydrogen-bonded organic frameworks with white-light emission. Nat Commun 13:1882
+69. Chen Q et al (2022) Tunable fluorescence in two-component hydrogen-bonded organic frameworks based on energy transfer. ACS Appl Mater Interfaces 14:24509–24517
+70. Tholen P et al (2022) Tuning structural and optical properties of porphyrin-based hydrogen-bonded organic frameworks by metal insertion. Small 18:e2204578
+
+# Supplementary Files
+
+- [ManuscriptSI.docx](https://assets-eu.researchsquare.com/files/rs-3670187/v1/a56b3c9f2acb434dbf969839.docx)
+- [CheckCIFTCANH4.pdf](https://assets-eu.researchsquare.com/files/rs-3670187/v1/b7dbaf47e5c13d097cc8464c.pdf)  
+  Checking file of crystallographic data
+- [CheckCIFTCANH4CHBr3.pdf](https://assets-eu.researchsquare.com/files/rs-3670187/v1/35a31ed6925407b700e5a35d.pdf)  
+  Checking file of crystallographic data
+- [CheckCIFTCANH4I2.pdf](https://assets-eu.researchsquare.com/files/rs-3670187/v1/858936427381eaa688ed511f.pdf)  
+  Checking file of crystallographic data
+- [CheckCIFTCANH4CHCl3.pdf](https://assets-eu.researchsquare.com/files/rs-3670187/v1/a81d2880bf5c242cea1d4d22.pdf)  
+  Checking file of crystallographic data
+- [CheckCIFTCANH4CHI3.pdf](https://assets-eu.researchsquare.com/files/rs-3670187/v1/fffb87bb4cefa2a3a507175d.pdf)  
+  Checking file of crystallographic data
+- [TCANH4CHBr3.cif](https://assets-eu.researchsquare.com/files/rs-3670187/v1/8c9c217c70c649876544e59a.cif)  
+  Crystallographic data
+- [TCANH4I2.cif](https://assets-eu.researchsquare.com/files/rs-3670187/v1/07ee67fed293c3c043f6be19.cif)  
+  Crystallographic data
+- [TCANH4CHCl3.cif](https://assets-eu.researchsquare.com/files/rs-3670187/v1/469b855025bcc361234b1560.cif)  
+  Crystallographic data
+- [TCANH4CHI3.cif](https://assets-eu.researchsquare.com/files/rs-3670187/v1/216f888b0aa3e8febd8b5b4d.cif)  
+  Crystallographic data
+- [TCANH4.cif](https://assets-eu.researchsquare.com/files/rs-3670187/v1/20208ce926e68bc49824a373.cif)  
+  Crystallographic data
+- [sourcedata.xlsx](https://assets-eu.researchsquare.com/files/rs-3670187/v1/7366e8c96e34ec0eec536e6f.xlsx)  
+  Source data for figures reproduction

0f122001ca5580c28dc179520fb16c08ec19f506fb2f6f5a1a1194f51bc98d0b/preprint/images_list.json

@@ -0,0 +1,34 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "ivoryOS architecture and interoperable SDL control through module serialization. The backend automatically creates an SDL snapshot and interactive webforms. The Python frontend requests the snapshot and generates SDL control Application Programming Interface (API) for proxying the HTTP requests. The GUI frontend utilizes the webforms for visualizing Python functions in workflow design and direct SDL control.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "ivoryOS interfaces using the above abstract SDL script as an example. aDirect control interface dynamically shows available device operations. Customization options are hide/unhide and reorder function cards. b Workflow design interface of the abstract SDL. Workflow building starts with device selection (step 1) and parameters filling (step 2) for a selected method. The action steps will show interactively on the design canvas. c Workflow database interface for loading saved workflow to the design canvas.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Workflow phases and execution options. a Workflow phases, including sequential execution of preparation, experiment and clean up. b Execution options for experiment phase. cscreenshot of a fixed iteration interface. d screenshot of configurable iterations interface through uploading a CSV file. e screenshot of configurable iterations interface using online form entry. f adaptive iterations interface with automatically generated parameters and objective configuration menu.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "SDL control structure used in Hein Group PurPOSE Platform.33 The structure demonstrates progression from individual hardware components to high-level user functions and composed experiments. Launching ivoryOS extension on any script allows seamless interaction across different control levels.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0f122001ca5580c28dc179520fb16c08ec19f506fb2f6f5a1a1194f51bc98d0b/preprint/preprint.md

@@ -0,0 +1,119 @@
+# Abstract
+
+Graphical User Interfaces (GUIs) are essential for providing a user-friendly experience to chemists employing Self-Driving Laboratories (SDLs). However, building an interoperable GUI is challenging due to the diverse layouts, frameworks and functionalities of SDLs, which often limit the use of existing GUIs on these SDLs. In this work, we introduce ivoryOS, a configuration-free software that automatically generates a snapshot of an SDL Python script, capturing all device instances, functionalities and argument information. This snapshot allows for the automatic creation of interactive forms that enable visual programming for control and workflow design across any SDLs. We demonstrate an example use case with the Hein Group Purification Platform Optimizing Solubility based Experimentation (PurPOSE) platform. This plug-and-play operating system for SDLs streamlines robot and lab hardware interaction, democratizing access to advanced SDLs for a broader range of scientists.
+
+[Physical sciences/Chemistry/Chemical synthesis/Automation](/browse?subjectArea=Physical%20sciences%2FChemistry%2FChemical%20synthesis%2FAutomation)
+
+[Physical sciences/Chemistry/Cheminformatics](/browse?subjectArea=Physical%20sciences%2FChemistry%2FCheminformatics)
+
+# Introduction
+
+Please have a look at courier new font provided for text in article.
+
+Self-Driving Laboratories (SDLs) integrate commercial or custom-designed hardware, such as liquid handlers, automated reactors, analytical instruments, robotic arms and etc., to automate and guide chemistry experimentation.¹–³ Utilizing machine learning algorithms, such as Bayesian Optimization, SDLs dynamically suggest subsequent experimental conditions based on prior data.¹,⁴,⁵ The application of SDLs in various research domains, including materials chemistry, drug discovery, and formulation,⁶–⁹ means that SDLs are inherently diverse in their layout and components. The use of modular equipment and flexible frameworks allows labor- and cost-efficient reconfiguration to different experimental requirements.¹⁰,¹¹
+
+The most widely used programming language in SDLs for orchestrating hardware control and adaptive experimentation is Python. This popularity is driven by several factors: (1) robust hardware protocol supports;¹²–¹⁴ (2) powerful machine learning tools;¹⁵–¹⁷ (3) comprehensive data analysis and visualization libraries; and (4) diverse and active online presence that provides strong support within the SDL communities.¹⁸,¹⁹ While other languages like C/C++, MATLAB/R, and JavaScript may offer native advantages in specific areas such as hardware control, data analysis, or interfacing, Python excels in its ability to integrate these diverse components, making it an ideal orchestration layer for SDLs.
+
+Despite Python being recognized for its ease of learning, operating a sophisticated SDL in Python can still pose significant challenges, especially for users without a programming background. While individual commercial hardware modules tend to incorporate Graphical User Interfaces (GUIs) to facilitate their use by non-experts, the development of a GUI for fully integrated SDLs incorporating several custom and commercial modules remains a challenge. The task is particularly difficult when the goal is to offer greater flexibility and usability across different configurations. Previous work, such as ChemIDE, enables visual programming using chemical description language (χDL), a hardware-independent language that allows chemists to script transferable chemical syntheses.²⁰–²² SDL developers can map their own hardware control scripts to χDL operations, enabling the execution of standardized χDL scripts across various systems.²³,²⁴ AlabOS,²⁵ developed for operating an autonomous lab for solid-state synthesis of inorganic powders,²⁶ proposed a reconfigurable workflow management framework. When adapting the use of AlabOS web interface, developers can tailor their SDL scripts into the AlabOS framework, thereby enabling managing complex SDL operations within an established operating system. ChemOS 2.0 is an orchestration platform that combines experimental planning with real-time execution in the chemistry domain.²⁷ It supports specific modular device communication through SiLA2 with simulation mode, making it adaptable across all stages of SDL development when some instruments are not accessible. However, the inclusion of new devices or other SDL control requires manual integration and SiLA2 adaptation. Overall, despite these advancements, adapting these GUIs to existing SDL systems still requires manual modifications on the current orchestration workflows, thereby limiting their applicability as a universal solution for diverse prototyping needs across different domains.
+
+In this work, we propose ivoryOS: an interoperable web interface for Python-based SDLs. This system provides a configuration-free solution for managing diverse SDL systems and accommodating continuous developing requirements. Specifically, a snapshot of the SDL control Python script is generated from all instances of Python class, capturing available functions and their parameter signatures (Fig. 1; backend). Using the extracted information, we then generate interactive web forms for all functions onto a customizable control interface for direct interaction with SDLs (Fig. 1; direct control). To support experimental design, ivoryOS also features a workflow design tool, an iteration manager and a database, allowing users to create, execute and manage customizable workflows (Fig. 1; workflows). We demonstrate the plug-and-play feature of ivoryOS for individual hardware or integrated SDLs, allowing for a standardized user-experience across different robotic platforms without constraints on framework.
+
+# Results
+
+## Web server architecture
+
+In this robot control architecture using Flask, a lightweight web application framework for Python, the source SDL control module is dynamically loaded into the system through global configuration upon launching the server, allowing the backend to access robot control methods (Fig. 1). Clients interact with the robot through HyperText Transfer Protocol (HTTP) requests that specify the instance through the Uniform Resource Locator (URL) and include a POST payload with task details (see Supplementary Information, S1.1 for function calling post request). Upon receiving the request, Flask routes it to the backend, where the loaded SDL control module is accessed to execute the corresponding function (example of method execution in backend in Supplementary Information, S1.2). Once the task is completed, the server sends a response back to the client, returning the result and a server status update. The web server allows remote SDL control from other devices in the network. Additionally, a socket connection is used to stream real-time logs to the client, allowing for continuous monitoring of the SDL’s activities during task execution (Fig. 1).
+
+## Python client
+
+To facilitate a more direct control for clients over SDL and abstract the complexity of HTTP communication, the ivoryOS can automatically generate Python Application Programming Interface (API) classes for all available SDL control methods by requesting the snapshot from the backend (Fig. 1 – Python API frontend; Supplementary Information, section 2). These generated classes allow developers to programmatically interface with the robot using Python, just like they would with a local SDL instance. Each generated Python class will act as a proxy that hides the underlying HTTP communication. When a method is called, the class sends the appropriate HTTP request to the server (via post) with the necessary function name and arguments. The recreated Python API mirrors the original functions from the SDL control script, allowing developers to continue utilizing their existing (source SDL control) Python code to interact with the robot.
+
+## SDL direct control interface
+
+The displayed web forms within the direct control interface show all available methods of an SDL or laboratory equipment (Fig. 2a). Users can navigate through available methods and interact with the SDL by inputting required parameters. The control interface can be customized and reorganized through drag-and-drop interactions to suit frequent usage and visibility preferences. Such flexibility enhances the usability and adaptability of the interface across various SDL frameworks or individual laboratory equipment APIs, permitting a plug-and-play GUI for prototyping hardware before the development of a dedicated GUI.
+
+## Workflow design
+
+To facilitate experimental design through visual programming, the interactive web forms (Fig. 1 - backend) are enhanced with a save field for storing function return values. Users can start by browsing systematic or logic operations or available SDL-specific instances (Fig. 2b; step 1). Upon entering function parameters in the web form and clicking “Add” (step 2 and 3), the defined workflow task is scripted into a structured JavaScript Object Notation (JSON) format (Supplementary Information, section 3). This task is then visually represented on the design canvas, where users can interactively reorder tasks, remove unnecessary steps, and adjust function parameters by clicking the corresponding entries (step 4). Once the design is complete, users can click the “Compile and Run” button to validate their design and proceed to the workflow execution page (step 5).
+
+To enhance experimental design intuitiveness, each SDL device instance integrates a text-to-code option that attempts to translate natural language task descriptions into SDL code (Fig. 2b; step 6). Our previous work successfully translated literature into machine-executable scripts using large language models like GPT-4. The translation is achieved through strategic prompting by giving sample code and instructions that maps chemistry tasks description to machine operations. In the context of flexible SDLs, where device functionalities are highly adaptable and lack strict operation mapping rules, we adapt this approach by appending the SDL snapshot instead of defined mapping rules. An example of the expected JSON format used for visual workflow programming is also attached to the prompt, ensuring successful display on the design canvas. Additionally, a post-processing step validates the scripted functions to further guarantee the alignment with the module capabilities. The full prompt using the example SDL is in Supplementary Information, section 4.
+
+The database interface, as shown in Fig. 2c, is implemented using Flask-SQLAlchemy to facilitate comprehensive management of experimental workflows within the system. This interface supports functionalities for saving, organizing, and loading workflows into the design interface from the database. Users can edit existing non-protected workflows, with changes being saved to the same or new workflows. Additionally, the saved SDL snapshot allows users to design and manage workflows without the need for direct connection to lab hardware.
+
+## Experimental phases and parameters
+
+Each workflow consists of three experimental phases: a preparation phase for preliminary steps, like purging solvent lines, a main experimental phase involving processes such mixing, heating, purification etc., and a cleanup phase to reset the platform post-experiment. Users can change between editing phases using the tabs on the top of the workflow canvas (Fig. 2b; step 7). While the preparation and cleanup phases execute once with only constant parameters, the main experimental phase is repeatable and supports dynamic parameters (Fig. 3a). The use of dynamic parameters, in contrast to a constant value, can change during execution, allowing different iterations without modifying the workflow. In the workflow design interface, users can define a configurable (dynamic) parameter by employing the “#” notation followed by the parameter name (e.g., “#amount” instead of “5” in Fig. 2b; step 2). This feature allows for flexible iterations with various execution options, such as parameter entries or an optimizer that adaptively suggests parameters (Fig. 3b, see Workflow execution section).
+
+## Workflow execution
+
+After a successful compilation of a workflow, three Python functions, representing the preparation, experimentation and clean up phases of the current design, will be generated (Supplementary Information, section 5). The script is available to download for usage without using ivoryOS web interface. Over the execution interface, the available execution options will be available depending on the parameter type and the presence of output values (Fig. 3b). The repeat option is designed for workflows with constant parameters, allowing either a single iteration or repeated iterations (Fig. 3c). For configurable iterations, defined parameter names and their respective data types are used to generate an empty configuration CSV file (Fig. 3d). Users can populate this CSV file with parameter values for each iteration and then upload it back into the system. For designs with fewer than six parameters, there is also online form option for quick iteration design (Fig. 3e). Additionally, for flexible workflows that also yield at least one numerical output, the Bayesian Optimization option with Ax platform becomes available, enabling adaptive exploration of the experimental space for optimized results (Fig. 3f). In this interface, all defined parameters and output values will be listed in the Parameters and Objectives sections respectively. Note that only numeric outputs can be used as optimization objectives; non-numeric outputs should have their optimization option set to “none”. For all workflows that generate output values, a CSV file will be created to record the data, which will be available for download.
+
+## PurPOSE platform integration
+
+Although the SDL development and frameworks exhibit variations across institutions and developers, Fig. 4 showcases a typical SDL development structure using the Purification Platform Optimizing Solubility based Experimentation (PurPOSE) platform developed in the Hein Group. The development process begins with core hardware components and evolving into multiple user functions that utilize this hardware synergistically. For instance, the user function “add solid” necessitates the coordinated operation of an automated balance, a capping mechanism, and a robotic arm for vessel translocation.
+
+As developers transition from controlling individual hardware components to designing higher-level user functions, custom layers can be added to manage various levels of robotic control. These layers may handle low-level tasks such as vessel transportation or coordinating the automated balance with the robotic arm. The experiments level (as shown in Fig. 4) is typically constructed using user functions as building blocks along with decision-making steps, resulting in ‘ready-to-execute’ scripts. A typical application of the ivoryOS is its interface with experiment conditions inputs during (high-throughput) executions, allowing users to browse available experiments and configure iteration parameters. To support workflow building from user functions simultaneously, developers can define which operations are accessible by importing instances from different control levels. By integrating ivoryOS into the PurPOSE, we provide access to both hardware handling functions and experimental operations, enabling low-level control, workflow design as well as experimental execution. A demonstration, along with a code example and a video, is available in Supplementary Information, section 6.
+
+# Discussion
+
+We have demonstrated an adaptable approach to controlling SDLs through the integration of ivoryOS with the PurPOSE platform. By generating dynamic snapshots of Python scripts, it ensures that the GUI are not rigidly tied to predefined configurations. This feature allows users to rapidly adjust to new experiments, instruments, or processes without needing to rewrite or manually configure scripts. Such adaptability is particularly valuable in SDLs, where platform development is continuous and evolves to accommodate dynamic research objectives. <sup><span citationid="CR2" class="CitationRef">2</span>, <span citationid="CR10" class="CitationRef">10</span>, <span citationid="CR11" class="CitationRef">11</span></sup> Additionally, launching the ivoryOS server requires only a single line of code and does not necessitate any modifications to the Python script, further guaranteeing its adaptability to workflows at various stages of development. The web server architecture also supports remote access to the SDL, facilitating seamless integration of current SDL system to various analytical instruments or other SDL platforms.
+
+A key advantage of providing interoperable and configuration-free GUI is the democratization of SDL technology, particularly by lowering the barrier to entry for users unfamiliar with Python or programming in general. Traditionally, SDLs and other robotic control systems have required users to possess a certain level of coding proficiency. However, with ivoryOS, users can now interact with the system through the intuitive text-to-code generator and user-friendly interfaces without needing to write scripts manually. This significantly lowers the learning curve, making advanced laboratory automation accessible to a broader scientific community. While programmers can continue using Python for development and day-to-day experimentation, ivoryOS serves as an extension, making the complicated SDL platform ready to share anytime, even during ongoing development. In this way, SDL platforms promote a more inclusive approach, empowering researchers across various scientific disciplines to leverage SDLs while focusing on solving domain-specific scientific problems.
+
+While the ivoryOS is primarily designed with SDLs in mind, its underlying principles and architecture are versatile and lend themselves to a wide range of use cases, including other robotic control systems and complex workflows management. The ability to manage and coordinate hardware components and control functions is equally valuable in industrial robotics, healthcare automation, and manufacturing. This adaptability suggests that the same software could be applied to orchestrate other hardware or software workflows, such as data science pipelines or automated laboratory procedures.
+
+Looking ahead, we plan to enhance ivoryOS by introducing concurrent workflow designs and scheduling, allowing parallel workflows that could significantly improve efficiency and flexibility. Another area of improvement is the support for a broader range of input types, including custom or collection data types. Additionally, we aim to incorporate modular interfaces, enabling users to tailor the interface to their specific needs—whether focusing solely on control, omitting the database, or selecting other customized configurations.
+
+# Methods
+
+## SDL serialization
+
+From the main SDL Python script (referred to as the “deck”), ivoryOS begins by systemically iterating over variables in the current scope, finding instances that are of custom Python class types. Attributes that start with an uppercase letter or an underscore are also excluded because they are the modules names or private variables (Supplementary Information, Fig. S1). This ensures that only instances of custom methods relevant to the SDL functions are captured (Fig. 5a). Next, ivoryOS delves into each class’s methods, selectively extracting those that are public and non-property functions (Fig. 5b). The retrieved parameter signatures and associated docstrings for each method are then organized into a structured dictionary, with each entry corresponding to an SDL component (e.g. deck.balance; Fig. 5c). The resulting snapshot captures the functionality and expected argument types of SDL devices for further function calling or workflow designing. The serialization of the SDL is done by creating a .pkl file, facilitating offline access to the experimental design without the need for direct hardware connection. A complete snapshot of this abstract SDL is in Supplementary Information, section 7.
+
+## Web form generation
+
+For robot interaction through a GUI, we generate a unique web form for each method in the serialized SDL snapshot, with input fields tailored to parameters types and default values (Fig. 5d). For primitive data types (int, float, str and bool), the input fields natively support direct input and type validation. However, due to the limitation of form input field, data type validation is not supported for user-defined data types or collection data types (list, tuple, dict and set). This process enables a basic visualization of SDL methods, allowing parameter inputting through a web interface. These forms are rendered in the direct control and workflow design interfaces (Fig. 1).
+
+# Conclusions
+
+In conclusion, we develop ivoryOS for creating a plug-and-play operating system for Python-based SDLs. On an existing SDL script, ivoryOS can automatically capture a snapshot of current components and functionalities, effectively accommodating the hardware diversity and necessity of continuous development of SDLs. The featured control interface enables users to customize function order and visibility for seamless interfacing of various hardware or SDLs. The integration of workflow design, execution and database interfaces empowers researchers to design and conduct high-throughput or adaptive experiments. Overall, the standardized design and control interface promotes low-code programming and interoperability between various SDLs, and enhances broader accessibility to SDL technologies across the scientific community.
+
+# References
+
+1. Tom G et al (2024) Self-Driving Laboratories for Chemistry and Materials Science. Chem Rev acs. chemrev.4c00055
+2. Christensen M et al (2021) Automation isn’t automatic. Chem Sci 12:15473–15490
+3. Liu J, Sato Y, Yang F, Kukor AJ, Hein JE (2022) An Adaptive Auto-Synthesizer using Online PAT Feedback to Flexibly Perform a Multistep Reaction. *Chemistry—Methods* 2, e202200009
+4. Hickman RJ, Aldeghi M, Häse F, Aspuru-Guzik A (2022) Bayesian optimization with known experimental and design constraints for chemistry applications. Digit Discovery 1:732–744
+5. Shields BJ et al (2021) Bayesian reaction optimization as a tool for chemical synthesis. Nature 590:89–96
+6. Christensen M et al (2021) Data-science driven autonomous process optimization. Commun Chem 4:112
+7. Vaddi K, Chiang HT, Pozzo LD (2022) Autonomous retrosynthesis of gold nanoparticles via spectral shape matching. *Digit Discovery* 1:502–510
+8. Mehr M, Caramelli SH, D., Cronin L (2023) Digitizing chemical discovery with a Bayesian explorer for interpreting reactivity data. *Proc. Natl. Acad. Sci. U.S.A.* 120, e2220045120
+9. Nambiar AMK et al (2022) Bayesian Optimization of Computer-Proposed Multistep Synthetic Routes on an Automated Robotic Flow Platform. ACS Cent Sci 8:825–836
+10. MacLeod BP, Parlane FGL, Brown AK, Hein JE, Berlinguette CP (2022) Flexible automation accelerates materials discovery. Nat Mater 21:722–726
+11. El-khawaldeh R, Hein JE (2024) Balancing act: when to flex and when to stay fixed. Trends Chem 6:1–4
+12. pySerial 3 0 documentation. [http://pythonhosted.org/pyserial/](http://pythonhosted.org/pyserial/)
+13. Eclipse paho-mqtt [https://eclipse.dev/paho/files/paho.mqtt.python/html/](https://eclipse.dev/paho/files/paho.mqtt.python/html/)
+14. Python opcua-asyncio Documentation [https://opcua-asyncio.readthedocs.io/en/latest/](https://opcua-asyncio.readthedocs.io/en/latest/)
+15. TensorFlow [https://www.tensorflow.org/](https://www.tensorflow.org/)
+16. PyTorch [https://pytorch.org/](https://pytorch.org/)
+17. BoTorch [https://botorch.org/](https://botorch.org/)
+18. Accelerated Discovery [https://accelerated-discovery.org/](https://accelerated-discovery.org/)
+19. Lab Automation Forums [https://labautomation.io/](https://labautomation.io/)
+20. Mehr SHM, Craven M, Leonov AI, Keenan G (2020) Cronin, L. A universal system for digitization and automatic execution of the chemical synthesis literature. Science 370:101–108
+21. Hammer AJS, Leonov AI, Bell NL, Cronin L (2021) Chemputation and the Standardization of Chemical Informatics. JACS Au 1:1572–1587
+22. ChemIDE [https://croningroup.gitlab.io/chemputer/xdlapp/](https://croningroup.gitlab.io/chemputer/xdlapp/)
+23. Rauschen R, Guy M, Hein JE, Cronin L (2024) Universal chemical programming language for robotic synthesis repeatability. Nat Synth 3:488–496
+24. Strieth-Kalthoff F et al (2024) Delocalized, asynchronous, closed-loop discovery of organic laser emitters. Science 384:eadk9227
+25. Fei Y (2024) *AlabOS: A Python-Based Reconfigurable Workflow Management Framework for Autonomous Laboratories*. [http://arxiv.org/abs/2405.13930](http://arxiv.org/abs/2405.13930)
+26. Szymanski NJ et al (2023) An autonomous laboratory for the accelerated synthesis of novel materials. Nature 624:86–91
+27. Sim M et al (2024) ChemOS 2.0: An orchestration architecture for chemical self-driving laboratories. Matter 7:2959–2977
+28. Hein Device API [https://gitlab.com/heingroup/device-api](https://gitlab.com/heingroup/device-api)
+29. Devices [https://gitlab.com/aspuru-guzik-group/self-driving-lab/devices](https://gitlab.com/aspuru-guzik-group/self-driving-lab/devices)
+30. pylabware [https://github.com/croningp/pylabware](https://github.com/croningp/pylabware)
+31. Zhang W et al (2024) Leveraging GPT-4 to transform chemistry from paper to practice. Digit Discovery. 10.1039.D4DD00248B
+32. Adaptive Experimentation Platform [https://ax.dev/](https://ax.dev/)
+33. PurPOSE [https://gitlab.com/heingroup/purpose](https://gitlab.com/heingroup/purpose)
+
+# Supplementary Files
+
+- [ivoryosv8SI.docx](https://assets-eu.researchsquare.com/files/rs-5307798/v1/5f8135ea7fc9e154251fe4f9.docx)

1265f8b06778239434bb4b6e3d1588d191708414dc2de58cfdb2da9350869262/preprint/images_list.json

@@ -0,0 +1,34 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpg",
+        "caption": "PRISMA Flow chart illustrating our search strategy and article screening and selection.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpg",
+        "caption": "Illustration of the global breakdown and distribution of the institutions in the 47 countries of 834 co-authors of the 226 articles included in the GPOC systematic review and meta-analysis.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpg",
+        "caption": "Forest plot for the key generation time meta-analysis, with a heterogeneity chi squared of 2429.59 (degrees of freedom 45), p=0.0005, I-squared (variation in effect size attributable to heterogeneity) 98.1%. For all 18 forests plots of the meta-analysis, see the Supplementary Meta-Analysis Forest Plots portable document format file.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpg",
+        "caption": "Pictorial representation of the Results from the risk of bias analysis indicating low to moderate risk of bias presented in the studies.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

1265f8b06778239434bb4b6e3d1588d191708414dc2de58cfdb2da9350869262/preprint/preprint.md

@@ -0,0 +1,191 @@
+# Abstract
+
+Current geopolitical tensions together with the global pandemic have provided important lessons for the need to independently re-evaluate our healthcare needs, guide and promote patient self-awareness and patient-centred care and to consider how cross-border medical information needs have become connected. The pandemic and war have also led to various humanitarian and healthcare crises for which there’s a need to re-evaluate and develop technologies to better manage Personal Health Records (PHRs) for displaced refugees with chronic diseases crossing borders. The recent trend of mobile platform-based, and electronic health record for e-health technologies enabled cloud-based PHR management as a paradigm for patient centred care. However, these platforms are yet to gain use-ubiquity globally. Here we performed a *Prospective Register of Systematic Reviews* (PROSPERO) registered and *Preferred Reporting Items Systematic and Meta-Analyses* (PRISMA)-guided systematic review and meta-analysis of the Personal Health Record looking at outcomes such as data security, efficiency, privacy, cost-based measures to generate a benchmark for future studies in this area. A meta-analysis of twelve axes for a future *Global Patient co-Owned Cloud* (GPOC) highlight the potential in health economics, performance, cryptography and speed of the basic techniques that are currently available, that would facilitate the construction of a GPOC. Whilst the field is early in its development, we highlight barriers, limitations and solutions through a proposed global consensus to ensure appropriate value delivery, safety and ethical governance for global digital personal health record adoption that can fundamentally beneficially transform the future of healthcare.
+
+Health sciences/Health care/Health policy  
+Health sciences/Health care/Medical ethics  
+Health sciences/Health care/Health care economics  
+meta-analysis  
+global patient co-owned cloud  
+personal health records  
+electronic health records  
+economic impact modelling  
+GPOC
+
+# MAIN
+
+Cloud-based infrastructures have gained prominence in healthcare for supporting patient management. The effectiveness of patient care is dependent on upholding integrity, privacy, security and efficiency of the stored data and its rapid retrieval for patient management by the clinician and healthcare provider. <sup>1, 2</sup> Aside from clinical reports of poorly designed centralised EHR systems being responsible for clinician burnouts, in 2017 the WannaCry ransomware, propagated through the National Security Agencies stolen EternalBlue targeted less secure centralised health architectures with records being deleted or held for ransom. <sup>3</sup> This compromised effective treatment due to care disruption achieved through centralised PHR (ISO/TR 14292:2012) systems.
+
+Arguments have therefore been made against centralised architectures, which only facilitate unilateral organisational control without honouring the concept of the patient’s human right to the ease of access to their information. As such various cloud-based models have been researched and discussed from around the globe. <sup>2</sup> Notable examples include personal health cloud-based systems built on top of distributed blockchain technologies for privacy, asserting superior security and encryption, access control. Newer approaches to PHR blockchain architectures adopt methods to even circumvent the permanency of the blockchain by developing specific rights to enable deletion of patient information when deemed necessary. <sup>1, 2</sup>
+
+Another important issue is cross-border travel by refugees displaced during conflict and natural disasters. This usually limits effective healthcare as clinicians in host nations have no access to the patient’s health record from their home nations who may be managing complex medical problems. For the management of chronic medical conditions such as diabetes and endocrine diseases, cancers, cardiorespiratory and gastrointestinal diseases. A secure cloud-based personal health records that can support patient care and migration become of interest.
+
+Securing these cloud-based PHR infrastructures from unauthorised access also requires the highest level of system-based encryption requiring advanced cryptographic primitives and a topic of multiple research foci. These encryption primitives require security researchers and cryptanalysts to constantly design tests that enable collision detection of cryptographic key infrastructures for at-risk systems to allow the development of newer systems.
+
+However, with the rapid evolution of technology, newer systems also have regulatory and ethico-legal challenges. These pertain to legal considerations of data custodianship, ownership and responsibility that become necessary areas for emerging jurisprudence for the organisation, clinician and for the co-ownership model of the patient as well. Very few have considered the co-ownership concept for patient healthcare data. In our meta-analysis we study the impact of these newer technologies on co-ownership across multiple borders and use this literature as the basis for studying outcomes in PHR management and design for a global patient co-owned cloud.
+
+# METHODS
+
+## Search Strategy
+The PRISMA-guided multi-platform database review was registered on PROSPERO (CRD42022342597) and supported by Karolinska Institutet and Imperial College London librarians. Thematic keyword searches were performed on Ovid Medline, PubMed, Cochrane Library, EMBASE, Web of Science core collection, CINAHL, SCOPUS and Engineering village (Inspec and Knovel) using the search presented in the supplementary methods. The overarching themes were global cloud-based, decentralised, patient co-ownership, personal/electronic health record systems. Keywords encompassed/included *data co-ownership, patient rights, artificial intelligence, ethics, data infrastructure, economics, regulatory*, *patient outcomes*, and *auditing*. We present the full list of databases and our search strategy in the supplementary methods.
+
+## Screening
+Articles were imported into EndNote (Programmer - The EndNote Team, Year − 2013 Title - EndNote Place Published – Philadelphia, PA Publisher – Clarivate Version: EndNote 20 Type: 64 bit) referencing software, deduplicated and exported to Rayyan<sup><span citationid="CR4" class="CitationRef">4</span></sup> (Harvard, USA) for article screening by NL and JD with HA resolving any conflicts.
+
+### Inclusion and Exclusion Criteria
+Relevant primary articles addressing global patient co-ownership electronic health systems (EHRs), Personal Health Records (PHR) and includes data-co-ownership, patient rights, ethics, economics of EHR patient systems, personal care record, patient outcomes from threatened security were identified. This included randomised controlled clinical trials performed on EHRs. Articles were included if they discussed cloud-based personal health records that had patient and healthcare provider co-ownership. Abstracts, reviews, conference proceedings, articles that do not reference PHR/EHR systems and with unclear outcomes were excluded. Specific exclusions included lack of reference to patient co-ownership with and without cloud-based infrastructures.
+
+It initially records the number of articles found and then makes the selection process transparent by reporting on decisions made at various stages of the systematic review. Numbers of articles are recorded at the different stages.
+
+## Meta-Analysis
+A meta-analysis was performed for PHR domains investigating efficiency, security, and cost-based parameters. This was based on access policies, runtimes, encryption and decryption times, key generation times, distributed network related data transfer cost (gas cost) and other time-based activities. Log-transformation was applied when necessary. We also performed a ratio of means standards effect size estimation on encryption calculated using the following formulae (Mean of intervention – Mean of control)/Mean of control. Analysis was performed using STATA (StataCorp 2013, Statistical Software, Release 13 College Station, TX StataCorp LP) for random effects modelling due to result heterogeneity. Significance was set at a p < 0.05. Authors were contacted for completion of data if unclear.
+
+## Validity and Bias
+Risk of bias was assessed using the seven domains of the ROBINS-I and the RobVis tool. Figure<span class="InternalRef" refid="Fig4">4</span> visualises the various risk-of-bias domains for each included study.
+
+# RESULTS
+
+The PRISMA flow diagram in Fig. 1 summarises the screening process. Search results retrieved 16,045 references with 6683 duplicates removed and 9,362 references screened. Thirty-four were selected for final inclusion in the review and 12 were included in the meta-analysis. Figure 2 shows the geographical distribution of the institutions included in the GPOC systematic review and meta-analysis. All forests plots of the meta-analysis are presented in the Supplementary Meta-Analysis Forest Plots portable document format file.
+
+## Efficiency-based parameters
+
+*Runtimes* defines the amount of time it takes for a program or piece of code to run (ms). In 117 sub studies on runtimes, a pooled effect size estimate of 12874 milliseconds (CI 12867.35–12880.80, I² 100%; *p* = 0.0005). A log transformed meta-analysis of the 117 sub studies on runtimes also showed an effect size estimate of 1.98 milliseconds (CI 1.97–1.98, I² 100%; *p* = 0.0005).
+
+*Key generation times* was defined as the time required for the process of generating cryptographic keys (ms). In 46 sub studies on key generation time, a pooled effect size estimate of 143.12 milliseconds (CI 121.01–165.22, I² 98.1%; *p* = 0.0005). A log transformed meta-analysis of the 46 sub studies on key generation time also showed an effect size estimate of 4.49 milliseconds (CI 4.52–4.47, I² 99.9%; *p* = 0.0005). Figure 3 illustrates the forest plot for the key generation time meta-analysis.
+
+## Other time-based activities:
+
+In 26 sub studies on time analysis such as key management and increased keyword query search time for PHR server transfer, a pooled effect size estimate of 3951.64 milliseconds (CI 3948.637–3954.642, I² 100%; *p* = 0.0005). A log transformed meta-analysis of the 26 sub studies on usage policy also showed an effect size estimate of 2.56 milliseconds (CI 2.55–2.56, I² 100%; *p* = 0.0005).
+
+## Security-based parameters
+
+*Access policies define the* protection of cloud data access and devices. These are set up to block access to all unauthorised uploads. In 34 sub studies on usage policy, a pooled effect size estimate of 30076 security-based policy of granularity of data access and response (CI 30072.82–30079.15, I² 100%; *p* = 0.0005) was identified. A log transformed meta-analysis of the 34 sub studies on usage policy also showed an effect size estimate of 3.98 policies (CI 3.97–3.98, I² 100%; *p* = 0.0005).
+
+*Encryption* ensures the conversion of information secretly to hide its original contents and was defined as the total encrypted data (bytes) divided by the encryption time (ms). In 86 sub studies on encryption, a pooled effect size estimate of 80.76 milliseconds (CI 80.76–80.76, I² 100%; *p* = 0.0005). A log transformed meta-analysis of the 86 sub studies on encryption also showed an effect size estimate of 1.86 milliseconds (CI 1.86–1.86, I² 100%; *p* = 0.0005).
+
+*Decryption* reverses the coded information to its original content and was defined as the total decrypted data (bytes) divided by the decryption time (ms). In 73 sub studies on decryption, a pooled effect size estimate of 59.50 milliseconds (CI 59.50–59.51, I² 100%; *p* = 0.0005). A log transformed meta-analysis of the 73 sub studies on decryption also showed an effect size estimate of 1.70 milliseconds (CI 1.70–1.70, I² 100%; *p* = 0.0005).
+
+## Ratio of means
+
+In 20 sub studies on ratio of means of encryption, a pooled effect size estimate of 0.16 milliseconds (CI 0.11–0.21, I² 100%; *p* = 0.0005). A log transformed meta-analysis of the 20 sub studies on ratio of means of encryption also exhibited an effect size estimate of 0.162 milliseconds (CI 0.110–0.214, I² 100%; *p* = 0.0005).
+
+## Cost-based parameters
+
+*Data transfer cost (gas cost)* was defined as gas, which is the price per unit of computation that is performed on the Ethereum network. In 8 sub studies on gas analysis, a pooled effect size estimate of 70192.81 Ethereum (CI 70113.20–70272.43, I² 100%; *p* = 0.0005). A log transformed meta-analysis of the 8 sub studies on gas analysis also showed an effect size estimate of 1.71 Ethereum (CI 1.63–1.79, I² 99.9%; *p* = 0.0005).
+
+## Risk of Bias (ROB)
+
+Figure 4 illustrates risk of bias of the 12 meta-analysed studies across seven bias domains, with 31% moderate and 69% of low risk. The studies presented moderate risks of bias: 8% due to confounding, 75% due to selection of participants, 25% in classification of interventions, 42% due to deviations from intended interventions, 25% due to missing data, 17% in measurement of outcomes, and 25% in selection of the reported result.
+
+# DISCUSSION
+
+The utility of PHRs is yet to gain global adoption with considerations for security, provenance, efficiency, and costs being a topic of debate and research focus. Although the world has become heavily reliant on data—now deemed the biggest commodity powering machine learning, such datasets for personal/electronic health records remain fragmented. This fragmented structure affects the ability to ensure effective transfer and delivery of cross-border healthcare in the 21st century model of healthcare. There are also issues surrounding healthcare equity, data security, ethical and legal challenges that are directly impacted by the way the patient's data is handled by the clinician and also the patient themselves. (CROSSREF FRAMEWORKS) Data oligopoly has arisen because of companies being provided patients data without their explicit consent. <sup>5</sup> Additional concerns also arise with omics data for precision medicine. Patients are therefore conscious about the effects of misuse of their data that could impact on their health independence, missed opportunity to be co-responsible for their own data, non-compensation for the use and misuse of their information etc. <sup>6</sup> The new PHR model of information management is yet to receive a global literature appraisal and this work is the first that is able to comprehensively address this issue.
+
+In our meta-analysis we addressed efficiency-based parameters that could impact the retrieval of data from PHR and noted a significantly fast run-times for PHR’s across 117 sub studies. These studies demonstrate that the speed of accessing information can impact decision-making, particularly for busy patients looking to manage their data, or the busy clinician deriving insights to optimise the patient's care. Communication of information with a healthcare professional in a collaborative model benefit from accelerated delivery of alerts to the patient and clinician for areas such as appointment scheduling, results notification etc.
+
+Akter et al studied the efficiency and performance of cloud based PHRs and demonstrated through areas of analysis of capabilities that included chunking, bundling, deduplication, delta-encoding, and data compression <sup>7</sup>. Performance indicators included control data overhead quantification of an average packet transmission rate of 93% benchmarked against other cloud storage services. <sup>7</sup> Others have shown similar results, e.g., Bhargavi et al performed a time efficiency comparison of values for different attributes with different files (n = 10–50, and key generation times 401.2-997.5 ms), <sup>8</sup> Preetha and colleagues presented computation times (|S| = 100) for six models (WTCM, WTCF, SADS, VAKF, LSTM, MLPPT-MHS), with key generation times ranging 823.75-4,093 ms. <sup>9, 10</sup> Saravanan and team presented the computation of delegation in key verification by comparing two models (proposed ECP-ABE vs existing CP-ABE) with key generation times varying from 1 vs 1.2 ms to 3.2 vs 4.7 ms. <sup>11, 12</sup> Further Sukte et al compared six models (Blowfish, RSA, ASE, El-Gamal, ECC, Modified El-Gamal, Modified ECC) with key generation times from 1–14 ms. <sup>13</sup> Others argue that centralised cloud providers by organisations affect the ease of movement of various PHR datasets. <sup>14</sup>
+
+Burns and colleagues also discuss efficiency of sharing cancer care data that has been generated affected by barriers highlighting the difficulty that these create when data must be shared to optimise the patient treatment and further diagnosis. <sup>15</sup> They further argue that not sharing cancer data because of patient privacy results in a learning-gap that impacts patient care. Under certain circumstances to make management efficient, patients will be willing to share their cancer treatment experiences as a substrate for research, despite potential risks to privacy. However, the counter argument is whether a regulatory framework must enable the privacy to be respected as we have discussed elsewhere (CROSSREF FRAMEWORKS).
+
+Efficiency in ensuring the security of the PHR record was presented in the form of the type of encryption used and how cryptographic keys are generated to support activity across the PHR, safeguarding from unauthorised access and breaches. PHRs were reported to be significantly more efficient than other methods of record keeping. These breaches can compromise healthcare information integrity and lead to ransomware attacks, distributed denial of service attacks, which can further exacerbate health-related issues through unnecessary stress.
+
+Similarly, other time-based analysis including file transfer times had a significantly better efficiency-based measure recorded for electronic and personal health records.
+
+Securing PHR’s is of paramount importance to prevent unauthorised access requiring access policies that optimise the speed and efficiency of shared information. Objectively quantifying this is difficult. Moreover, patients may not have the technical knowledge about security risks that could lead to unauthorised access and ransomware attacks leading to data loss. In our meta-analysis of access policies, a heterogenous pooled effect size estimate of 30076 security related policies that impacts granularity of data access and response was identified. The significant effect size estimate of 3.98 policies (p = 0.0005) in log transformed meta-analysis was also identified. This has not been reported in the literature before. However, several authors have discussed innovations such as patient-controlled health access brokerage services with necessity for implementing security logs and unique methods of intrusion detection. <sup>16, 41, 46</sup>
+
+Encryption forms the backbone of all modern electronic systems security systems that power digital technologies today. Privacy and security concerns were discussed by a majority of all included articles, 148 articles (65.4%), and was the commonest term elaborated by all covered facets. The type of encryption used is crucial to safeguard transfer and sharing of information from the patient health record to clinicians or the healthcare provider, as well as encrypting compressed information on the record etc. In 86 substudies on encryption, a pooled effect size estimate of encryption speed of 80.76 milliseconds was seen. The response ratio was performed on 20 substudies looking at mean encryption times which demonstrated an effect size estimate of 0.16 milliseconds. The literature review of the meta-analysed articles presented several studies on encryption (Abaid et al presented 800–1200 ms, <sup>16</sup> Chennam et al 8654–10025 ms, <sup>17</sup> Florence et al 29–98 ms, <sup>18</sup> Kocabas et al 80-5040 ms, <sup>19</sup> Preetha 9,919-280.43 ms, <sup>8</sup> Saravannan et al 8-11.7 ms, and Sukte et al compared six schemes (Blowfish, RSA, ASE, El-Gamal, ECC, Modified El-Gamal, and Modified ECC) with ranges 0.000058–0.027614 ms, <sup>13</sup> and where the meta-analysis gave a pooled effect size estimate of 80.76 milliseconds, an effect size estimate of 1.86 milliseconds and p = 0.0005).
+
+By comparison, decryption time, that is necessary for retrieving information by a patient or clinician, had a pooled effect size estimate of 59.50 milliseconds. These are clinical benchmarks for PHR’s that subsequent studies building on this infrastructure could improve upon. In the review several studies presented decryption, e.g., Chennam et al 4236–7546 ms, <sup>1716</sup> Florence et al 16–74 ms, <sup>18</sup> Kocabas et al 30-2290 ms, <sup>19</sup> Sangeetha et al 4–12 ms, <sup>20</sup> Preetha et al 90.11-71,167 ms <sup>8</sup>. Moreover, Sukte et al compared six schemes (Blowfish, RSA, ASE, El-Gamal, ECC, Modified El-Gamal, and Modified ECC) with ranges 0.000086–0.00054 ms, <sup>13</sup> The meta-analysis gave a pooled effect size estimate of 59.50 milliseconds, an effect size estimate of 1.70 milliseconds and p = 0.0005) performances for their proposed algorithms and security solutions. <sup>21–25</sup> As mentioned, there are several types of encryptions - Raisaro et al presented a collective homomorphic encryption, which allows users to analyse on encrypted datasets. <sup>26</sup>
+
+Earlier proof of work blockchain technologies enabled an organisation to quantify the precise amount of costs of executing software and mathematical operations that facilitate digital tokenization and activity. This expressed as gas on the Ethereum virtual machine smart contract. Such innovations provide advantages, because those that operate on the Ethereum virtual machine do so by a measurable gas cost for executing programs that support the functioning of the PHR using smart executable contracts. Based on inherent technical constraints imposed upon the design standard for smart contracts, these could be tailored to one specific action without affecting other necessary components of the PHR. This makes useability costs highly measurable and auditable. Our gas analysis demonstrated a pooled effect size estimate of 70193 milliseconds with a log transformed effect size estimate of 1.71 milliseconds. An ideal PHR should allow accurate estimation of costs for information transfer, data mining and interdisciplinary access for decision support to compensate users in a co-ownership model.
+
+In middle eastern countries such as Bahrain, Al-Aswad and colleagues apply blockchain technologies in a patient-centric model for PHR data management allowing for smarter interconnectivity between healthcare and the Internet of Things that power smart cities. <sup>27</sup> The aim is to streamline the provision of higher quality privacy powered healthcare services using zero-knowledge proofs. The intended consequence is a fusion of a zero-knowledge proof for encryption whilst ensuring patient consent is acquired for data insight discovery to maintain privacy and anonymity.
+
+Chen and colleagues offer a patient-centred argument for a PHR model and discuss a unique information access control scheme using Lagrange interpolation polynomials for secure multi-user permissible information access. <sup>41</sup>
+
+Alshammari et al and others discuss machine learning analysis of cloud-based datasets and IoT. <sup>28</sup> Regulatory and ethico-legal challenges for responsible data custodianship and the assumption of how responsibility is apportioned opens the need for new legal frameworks, policies as well as organisational, clinician and patient jurisprudence for a newer co-ownership model.
+
+> “A clear lesson of this whole arrangement is that attempts to deliver public healthcare services should not be launched without disclosing the details, documentation, and approvals—the legal bedrock—of the partnerships that underlie them. This lesson applies no less to companies offering algorithmic tools on big datasets than it does to pharmaceutical and biotech companies” <sup>29</sup>
+
+> Effective runtimes of a GPOC reduces the effects of data retrieval lag. Currently no meta-analysis exists for runtimes. In 117 substudies on runtimes, a pooled effect size estimate of 12874 milliseconds (CI 12867.35–12880.80, I<sup>2</sup> 100%; p = 0.0005). A log transformed meta-analysis of the 117 substudies on runtimes also showed an effect size estimate of 1.98 milliseconds (CI 1.97–1.98, I<sup>2</sup> 100%; p = 0.0005).
+
+> Security efficiency relates to key generation times and remains an area that has not got a clinical benchmark. In 46 substudies on key generation time, a pooled effect size estimate of 143.12 milliseconds (CI 121.01–165.22, I<sup>2</sup> 98.1%; p = 0.0005). A log transformed meta-analysis of the 46 substudies on key generation time also showed an effect size estimate of 4.49 milliseconds (CI 4.52–4.47, I<sup>2</sup> 99.9%; p = 0.0005).
+
+> Other time-based activities included server transfer times yet to be meta-analytically benchmarked. In 26 substudies on time analysis such as key management and increased keyword query search time for PHR server transfer, a pooled effect size estimate of 3951.64 milliseconds (CI 3948.64.82–3954.64, I<sup>2</sup> 100%; p = 0.0005). A log transformed meta-analysis of the 26 substudies on usage policy also showed an effect size estimate of 2.56 milliseconds (CI 2.55–2.56, I<sup>2</sup> 100%; p = 0.0005).
+
+Multiple problems exist that need to be addressed but we identify four main areas summarised below.
+
+1. No effective global healthcare data, interaction, or communication platform. One of the major issues is that there is no effective global healthcare data interaction, or communication platform. In response to the emergence of the global health catastrophe of COVID-19, researchers tried to design a global pandemic monitoring platform. <sup>30</sup> Others conclude that the present centralised systems cannot adapt to the vast volumes of globalised PHRs. <sup>6</sup> An optimal and complete use of PHRs could become prophylactic and have a major impact on global health. <sup>31</sup> Another team concludes that COVID-19 a global PHRs platform, would play a pivotal role in combatting the pandemic. <sup>32</sup>
+2. Siloed use of AI on health data, and no pipeline for future AI improvement. <sup>7</sup> Furthermore, the use of AI on health data is siloed, and there is no pipeline for future AI improvement. The fact that patients sharing their PHR contents is a game changer, as well as the use of AI on their data. <sup>33</sup> An AI-empowered cloud-based PHR system, which could possibly decrease healthcare errors, costs, and improve quality and effectiveness has been suggested. <sup>34</sup> Another viewpoint is that although PHRs facilitate healthcare, they are most often outsourced to third party cloud service providers, bringing severe security issues, and increasing the risk of malicious usage and leakages. <sup>35</sup>
+3. Current EHRs are expensive, non-interactive and studies have shown they are so badly designed that it causes health worker burnouts. <sup>8</sup> Others underline that with the rise of cloud based PHRs as the important health sharing platform, these must even more importantly allow the users to share in a UX/UI simplified manner. <sup>37</sup> Notably, that sharing can be both to family and to any professional of preference. A team presented the Bluefish algorithm to improve the security, flexibility, and transmissibility associated with third-party provided cloud solutions <sup>37</sup>. Furthermore, the present technical solutions appear not secure enough to sophisticated threats. <sup>38</sup> Thus, there are proposals of a re-encryption solution in response to white-box attacks. This to maintain efficiency even if there are multiple recipients. Turner et al. identified easy accessibility and straight provider access as key vulnerabilities. <sup>39</sup> They further specify the high expenses, low IT knowledge and safety awareness as three other main factors. In particular, they bring up design issues and other practicalities associated to elderly patient users and suggest a set of tools to cope with the problem. There are economic and access advantages for cloud based PHR platforms. <sup>40</sup> Even though cloud storage can cut costs and improve health data sharing, the security issues are still substantial <sup>41</sup>
+4. The EHRs are hindering effective use of health data, and hence the progress of AI in medicine with grave socioeconomic consequences. Present PHRs are hindering effective use of health data, and hence the progress of AI in medicine with grave socioeconomic consequences. Black et al applies globally relevant ethnic and social perspectives, delving into the patient journey and the PHR adoption in detail. <sup>42</sup> As a continuation to these, the needs of the disabled persons from ethical, social, and judicial perspectives, have been elaborated. <sup>43</sup> It is further emphasised that a global PHR must be all-encompassing, and not exclude anyone. Another team also showed how multi-source PHRs with both socioeconomic and genetic data will have a pivotal role in the realisation of true glob.al and individual-centred precision healthcare of the near future. <sup>44</sup>
+5. Lack of interaction and communication leads to one fifth of EHRs having serious errors. Current EHRs are costing time, money, and lives. <sup>45</sup> Though the lack of interaction and communication leads to one fifth of EHRs having serious errors. Current EHRs are costing time, money, and lives. Chen et al provides the insight that patients should self-manage their PHRs, which would then provide them with a “strong sense” of control and even of ownership. <sup>46</sup> They point out how this notion changes the sharing patterns, and further decreases the amount of PHR medical errors, i.e., a lowered amount of nosocomial adverse effects. Health expenses are rising with an older global population, and an intelligent cloud-based electronic health record (ICEHR) has been suggested byKhansa et al, in a way to diminish medical mistakes. <sup>34</sup> Other groups emphasise the concept of individual-focused PHR to maintain long term error-free healthcare qualities. <sup>42</sup> As one of many practical examples Uchimura’s team presented a mobile app self-administrative medical solution, to further avoid mistakes and increase correctness of the PHRs. <sup>47</sup>
+
+## LIMITATIONS AND FUTURE WORK
+
+As this is prima facie in this area, we observed marked heterogeneity and this is because there is yet to be a global standard for the conduct and reporting of EHR/PHR-related meta-analysis. Heterogeneity was partially controlled for using a random effects model that showed significant results across the domains of security, efficiency and cost-based parameters associated with EHR/PHR management for patients. Since there was significant heterogeneity identified throughout, results should be taken into context based on the fact that no effective standardisation for a meta-analysis within this area has ever been attempted. This would be the first study to benchmark this. Future studies should explore measures of ensuring that acquisition of data and its reporting obeys a standard that is yet to be defined.
+
+Future work will be to develop a consensus-driven approach that would standardise PHR data management for patients, supporting effective and secure access for clinicians and organisations. This will also enable researchers to adopt a standardised approach to data mining and build effective artificially intelligent systems that will support and optimise patient care.
+
+Moreover, a consensus-driven data standard for security - especially in the age of distributed ledger technologies allows patient involvement, influence, and engagement to be securely highlighted during the search for optimal treatment protocols that will impact their care. This will fit the current paradigm of patient centred care to facilitate co-ownership, as well as allow the ethical and legal co-responsibility models to evolve.
+
+# CONCLUSIONS
+
+We conclude that the meta-analysis of twelve axes for a future GPOC currently demonstrates marked heterogeneity. This is a consequence of the naissance of a terra nova field without incumbent standardisation or modus operandi. Although we have meta-analysed the cryptographic, cost, performance and speed of the basic techniques that are currently available, and which would facilitate and make the construction of a Global Patient co-Owned Cloud (GPOC) possible. We have highlighted the limitations with the lack of standard and foresee a consensus in the field to emerge within the pivotal field of privacy and security for cloud based Blockchain PHRs.
+
+# References
+
+1. Cao S, Wang J, Du X, Zhang X, Qin X, editors, "CEPS: A Cross-Blockchain based Electronic Health Records Privacy-Preserving Scheme," ICC 2020 - 2020 IEEE International Conference on Communications (ICC), 2020, pp. 1-6.
+2. Cao S, Zhang XS, Xu RX. Toward Secure Storage in Cloud-based eHealth Systems: A Blockchain-Assisted Approach," in IEEE Network, 2020; vol. 34, no. 2, pp. 64-70.
+3. Jones S, Neville S, Chaffin J. Hackers use tools stolen from NSA in worldwide cyber attack, Financial Times, 12<sup>th</sup> May 2017 [Retrieved 19<sup>th</sup> November 2022 from: https://www.ft.com/content/e96924f0-3722-11e7-99bd-13beb0903fa3]
+4. Mourad Ouzzani, Hossam Hammady, Zbys Fedorowicz, and Ahmed Elmagarmid. Rayyan — a web and mobile app for systematic reviews. Systematic Reviews (2016) 5:210
+5. Guddati VCAK. Ethical Issues in Patient Data Ownership. Interact J Med Res. 2021; v.10(2); Apr-Jun 2021.
+6. Karabekmez ME. Data Ethics in Digital Health and Genomics. New Bioeth. 2021 Dec;27(4):320-333.
+7. Akter M, Gani A, Rahman MO, Hassan MM, Almogren A, Ahmad S. Performance Analysis of Personal Cloud Storage Services for Mobile Multimedia Health Record Management. IEEE Access. 2018;6:52625-38.
+8. Bhargavi M, Bharath Siva Varma P. Privacy protection for e-health records over mobile cloudlet. International Journal of Recent Technology and Engineering. 2019;8(2):6014-9.
+9. Preetha AD, Kumar TSP. Securing IoT-based healthcare systems from counterfeit medicine penetration using Blockchain. Appl Nanosci 13, 1263–1275 (2023).
+10. Preetha AD, Kumar TSP, editors. MLPPT-MHS: Multi-Layered Privacy Preserving and Traceable Mobile Health System, Procedia Computer Science, Vol 165, 2019, Pages 598-614.
+11. Saravanan N, Umamakeswari A. Enhanced attribute based encryption technique for secured access in cloud storage for personal health records. Concurrency and Computation-Practice & Experience.2022; Vol 34:11, Wiley.
+12. Saravanan N, Umamakeswari A. Hap-Cp-Abe Based Encryption Technique With Hashed Access Policy Based Authentication Scheme For Privacy Preserving Of Phr. Microprocessors and Microsystems.80:10, 2021, 103540.
+13. Sukte C, Emmanuel M, Deshmukh RR. Modified Elliptic Curve Cryptography Model for Personal Health Record Sharing in Cloud with Trust Valuation. International Journal of Computer Science and Network Security. 2022;22(1):593-601.
+14. Al-Issa Y, Ottom MA, Tamrawi A. eHealth Cloud Security Challenges: A Survey. Journal of healthcare engineering. 2019;2019:7516035.
+15. Burns S, Collisson EA. Blockchain-authenticated sharing of cancer patient genomic and clinical outcomes data. Journal of Clinical Oncology. 2020;38(15).
+16. Abaid Z, Shaghaghi A, Gunawardena R, Seneviratne S, Seneviratne A, Jha S. Health access broker: Secure, patient-controlled management of personal health records in the cloud. 13th International Conference on Computational Intelligence in Security for Information Systems (CISIS). 2021, p. 111-21.
+17. Chennam K, Muddana L. An efficient two stage encryption for securing personal health records in cloud computing. International Journal of Services Operations and Informatics. 2018;9(4):277-96.
+18. Florence ML, Suresh D. Enhanced secure sharing of PHR's in cloud using user usage based attribute based encryption and signature with keyword search. Cluster Computing-the Journal of Networks Software Tools and Applications. 2017; 22:13119-30.
+19. Kocabas O, Soyata T. Towards privacy-preserving medical cloud computing using homomorphic encryption. 2015, p. 213-46.
+20. Sangeetha, D., Chakkaravarthy, S.S., Satapathy, S.C. et al. Multi keyword searchable attribute based encryption for efficient retrieval of health Records in Cloud. Multimed Tools Appl 81, 22065–22085 (2022).
+21. Liu Qin, Liu Xuhui, HU Baishuang, Zhang Shaobo. Fine-grained Access Control with User Revocation in Cloud-based Personal Health Record System[J]. Journal of Electronics & Information Technology, 2017, 39(5): 1206-1212.
+22. Liu X, Liu Q, Peng T, Wu J. HCBE: Achieving fine-grained access control in cloud-based PHR systems. 2015. p. 562-76.
+23. Liu XH, Liu Q, Peng T, Wu J. Dynamic access policy in cloud-based personal health record (PHR) systems. Information Sciences. 2017; 379:62-81.
+24. Meddah, N., Jebrane, A., Toumanari, A. (2018). Scalable Lightweight ABAC Scheme for Secure Sharing PHR in Cloud Computing. In: Ezziyyani, M., Bahaj, M., Khoukhi, F. (eds) Advanced Information Technology, Services and Systems. AIT2S 2017. Lecture Notes in Networks and Systems, vol 25. Springer, Cham.
+25. Niu S, Song M, Fang L, Wang C. Cloud Storage Data Sharing Based on Attribute Encryption in Smart Healthcare. Dianzi Yu Xinxi Xuebao/Journal of Electronics and Information Technology. 2022;44(1):107-17.
+26. Raisaro JL, Troncoso-Pastoriza JR, Misbach M, Sousa JS, Praderv, S, et al. MedCo: Enabling Secure and Privacy-Preserving Exploration of Distributed Clinical and Genomic Data. IEEE/ACM transactions on computational biology and bioinformatics. 2019;16(4):1328-41.
+27. Al-Aswad H, El-Medany WM, Balakrishna C, Ababneh N, Curran K. BZKP: Blockchain-based zero-knowledge proof model for enhancing healthcare security in Bahrain IoT smart cities and COVID-19 risk mitigation. Arab Journal of Basic and Applied Sciences. 2021;28(1):154-71.
+28. Alshammari H, Abd El-Ghany S, Shehab A. Big IoT Healthcare Data Analytics Framework Based on Fog and Cloud Computing. Journal of Information Processing Systems. 2020; 16(6):1238-49.
+29. Powles J, Hodson H. Google DeepMind and healthcare in an age of algorithms. Health Technol (Berl). 2017;7(4):351-367.
+30. Lee H-A, Kung H-H, Lee Y-J, Chao JCJ, Udayasankaran JG, Fan H-C, et al. Global Infectious Disease Surveillance and Case Tracking System for COVID-19: Development Study. JMIR medical informatics. 2020;8(12):e20567.
+31. Ramu G. A secure cloud framework to share EHRs using modified CP-ABE and the attribute bloom filter. Education and Information Technologies. 2018; 23(5):2213-33.
+32. Devi T, Ramachandran A, Deepa N.A Biometric Approach for Electronic Healthcare Database System using SAML - A Touchfree Technology," 2021 Second International Conference on Electronics and Sustainable Communication Systems<em>(ICESC)</em>, 2021, pp. 174-178.
+33. Kumar S, Wajeed MA, Kunabeva R, Dwivedi N, Singhal P, Jamal SS, et al. Novel Method for Safeguarding Personal Health Record in Cloud Connection Using Deep Learning Models. Computational intelligence and neuroscience. 2022;2022:3564436.
+34. Khansa L, Forcade J, Nambari G, Parasuraman S, Cox P. Proposing an intelligent cloud-based electronic health record system. International Journal of Business Data Communications and Networking. 2012;8(3):57-71.
+35. Pussewalage HSG, Oleshchuk VA, editors. A Patient-Centric Attribute Based Access Control Scheme for Secure Sharing of Personal Health Records Using Cloud Computing," 2016 IEEE 2nd International Conference on Collaboration and Internet Computing (CIC), 2016, pp. 46-53.
+36. Topol E. The Patient Will See You Now: The Future of Medicine Is in Your Hands. First edition. ed. 2016: Basic Books: New York, NY. ; 2016.
+37. Rinesh S, Baskaran K. A secure and efficient data sharing in cloud using multiple authority atturibute based biometric encryption. International Journal of Applied Engineering Research. 2015;10(20):19490-504.
+38. Shin JYJ, Man K, Zhou W. International and global issues - Differences in health systems, patient populations, and medical practice. 2021. p. 257-72.
+39. Turner AM, Osterhage K, Hartzler A, Joe J, Lin L, Kanagat N, Demiris G. Use of Patient Portals for Personal Health Information Management: The Older Adult Perspective. AMIA Annu Symp Proc. 2015 Nov 5;2015:1234-41.
+40. Almutiry O, Wills G, Alwabel A, Crowder R, Walters R, editors. Toward a framework for data quality in cloud-based health information system," International Conference on Information Society (i-Society 2013), 2013, pp. 153-157
+41. Chen Y, editor. The role of patients in transiting personal health information: a field study. Stud Health Technol Inform. 2010;160(Pt 1):3-7.
+42. Black AS, Sahama T. Chronicling the patient journey: Co-creating value with digital health ecosystems. In Maeder, A & Williams, T (Eds.) Proceedings of the Australasian Computer Science Week Multiconference. Association for Computing Machinery, United States of America, 2016, pp. 1-10.
+43. Knapfel S, Plattner B, Santo T, Tyndall S. Promotion of meaningful use of a personal health record in second life. Studies in health technology and informatics. 2014;201:413-7.
+44. Koufi V, Malamateniou F, Tsohou A, Vassilacopoulos G. A framework for privacy-preserving access to next-generation EHRs. Studies in health technology and informatics. 2014;205:740-4.
+45. Hecht J. The future of electronic health records. Nature. 2019;573, S114-S116
+46. Liu JH, Huang XY, Liu JK. Secure sharing of Personal Health Records in cloud computing: Ciphertext-Policy Attribute-Based Signcryption. Future Generation Computer Systems-the International Journal of Escience. 2015; 52:67-76.
+47. Uchimura Y, Fujita H. Development of medical and health information system using mobile devices. IEEJ Transactions on Sensors and Micromachines. 2012;132(11):381-6.
+
+# Supplementary Files
+
+- [DocumentationofsearchstrategiesNiklasLidstrXXmer1.docx](https://assets-eu.researchsquare.com/files/rs-3004559/v1/a03f8a0120d5fcac6dc3f50b.docx)
+- [GPOCARTICLENAVIGATIONINDEX.docx](https://assets-eu.researchsquare.com/files/rs-3004559/v1/3b0d1db03e0599a7d005fd3e.docx)
+- [GPOCIDENTITYImageAbstract.pdf](https://assets-eu.researchsquare.com/files/rs-3004559/v1/15525e972ac8c4a42f1f9ff4.pdf)
+- [PRISMA2020checklist.GPOC.docx](https://assets-eu.researchsquare.com/files/rs-3004559/v1/6210f8dff43c6e4221be057c.docx)

1401d26f7b25827b37c95ec379dd5bfa46a6f70c061d00b95cd9a260a4be417d/preprint/images_list.json

@@ -0,0 +1,34 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Ultraflexible, Self-powered Photoplethysmogram Sensor. a, Schematic diagram of the ultraflexible, self-powered photoplethysmograpm (PPG) sensor on human hands. Ultraflexible organic photovoltaic (OPV) module generates electrical power from sunlight and drives polymer light-emitting diode (PLED) and organic photodiode (OPD). b, Schematic diagram of the self-powered PPG sensor with PLED, OPD, and OPV module. c, A photograph of the ultraflexible OPV module under one-sun illumination. Scale bar indicates 1 cm. d, A top view photograph of the ultraflexible, self-powered PPG sensor. Scale bar indicates 1 cm. e, An electrical circuit of self-powered PPG sensor.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Air Operational Stable, Ultraflexible Polymer Light-emitting Diodes. a, Schematic diagram of cross-sectional view of the stack present in polymer light-emitting diode (PLED). b, Comparison of the lifetime and turn-on voltage of previously reported organic LED and PLED without encapsulation. c, Current density-luminance-voltage curves of PLED on glass and ultraflexible substrate. d, A photograph of the ultraflexible PLED during operation. Operation voltage of PLED was 6 V. Scale bar indicates 1 cm. e, Current efficiency characteristics of the PLED on glass and ultraflexible substrate. f, Air operational stability of the PLED under constant-current operation in ambient air. Current density to drive PLED was set to be 8 mA/cm2. The black line indicates the PLED on the glass substrate without any encapsulation; the red line indicates the PLED on the glass substrate with parylene encapsulation; and the blue line indicates the PLED on the ultraflexible substrate with parylene encapsulation. The ultraflexible PLEDs were measured in the freestanding state.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Optical Responses and Linearities of Organic Photodiode for Various Light Sources. a, Schematic diagram of the cross-sectional view of the stack of organic photodiode (OPD). b, Light intensity dependence of OPD with 1-sun illumination form solar simulator. Light intensity of solar simulator was varied from 0.01-sun to 1-sun with optical filters. The red dash line represents a linear fitted line of short circuit current and light intensity. The blue dash line represents a logarithmically fitted line of open circuit voltage and light intensity. c, External quantum efficiency spectra of OPD and Electroluminescence spectra of polymer light-emitting diode (PLED). d, J-V characteristics of OPD illuminated by PLED. Optical power of PLED was varied by changing driving voltage of PLED. Each of the devices are stacked together with a face-to-face configuration. e, Light intensity dependence of OPD with PLED illumination. The red dash line represents a linear fitted line of short circuit current and light intensity. The blue dash line represents a logarithmically fitted line of open circuit voltage and light intensity.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Blood Pulse Detection of Self-powered Photoplethysmogram sensors. a, Schematic diagram describing the operational mechanism of photoplethysmogram (PPG) by polymer light-emitting diode (PLED) and organic photodiode (OPD). b, Output voltage characteristics of OPD with PPG measurement. The PLED was powered by 10 series-connected organic photovoltaic (OPV) modules with one-sun illumination from a solar simulator. All devices are on glass substrate. c, A photograph of the ultraflexible, self-powered PPG sensor. Scale bar indicates 1 cm. d, Output voltage characteristics of OPD with PPG measurement. The PLED was powered by 10 series-connected OPV modules with one-sun illumination from a solar simulator. All devices are on ultraflexible substrate. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

1401d26f7b25827b37c95ec379dd5bfa46a6f70c061d00b95cd9a260a4be417d/preprint/preprint.md

@@ -0,0 +1,144 @@
+# Abstract
+
+Flexible organic optical devices have been used extensively in next-generation wearable electronics owing to advantages such as light-weight, thinness, and flexibility. Making such organic optical devices ultrathin enables long-term monitoring of health conditions owing to the increased conformability of ultrathin devices to human skin. Long-term biological signal monitoring also requires the integration of organic optical devices with an energy harvesting power sources which does not require recharging; to make devices self-powered. However, system-level integration of such thin organic optical sensors with power sources is challenging. An important obstacle to this type of integration is the insufficient operational stability of ultrathin organic light-emitting diodes under ambient air conditions. Here we develop an ultrathin self-powered organic optical system for photoplethysmogram (PPG) monitoring. This system consists of three types of ultrathin electronic devices: polymer light-emitting diodes (PLEDs), organic solar cells, and organic photodetectors. By adopting an inverted structure and a polyethylenimine ethoxylated layer doped with 8-hydroxyquinolinato-lithium as the electron-transport layer, PLEDs exhibit improved operational stability under ambient air conditions without external encapsulation. Ultraflexible PLEDs with no passivation retain 70% of the initial luminance lifetime of 11.3 h under ambient air. Integrated optical sensors exhibit a high linearity with the light intensity exponent of 0.98 by PLED light source. Such self-powered, ultraflexible PPG sensors can perform long-term stable monitoring of blood pulse signals on human hands.
+
+Electronic Materials and Devices  
+Electrical Engineering  
+photonic skin  
+continuous bio-signal detection  
+light-emitting diodes
+
+# Introduction
+
+Organic semiconducting devices have been used extensively in next-generation wearable electronics because of their light-weight, thinness, and flexibility [1–3]. Flexible displays consisting of organic light-emitting diodes (LEDs) have been installed in smartwatches and wristband-type applications, contributing to a reduction in power consumption [4]. In addition, an all-organic optoelectronic sensor has been developed for pulse oximetry, by integrating organic LEDs and organic photodetectors [5]. Imparting flexibility to such organic optical sensors enables long-term monitoring of health conditions with reduced discomfort when these sensors are attached directly on human skin.
+
+Recent advances in the field of flexible optical integrated devices have made these devices favourable for optical sensor applications, such as photoplethysmogram (PPG) [6]–[10], pulse oximetry [10]–[12]. High-resolution imaging devices have been used in advanced applications, such as vein authentication [13, 14], fingerprint imaging [13, 15], and X-ray imaging [16, 17]. Self-powered systems such as thermoelectric conversion devices [18] and organic solar cells [19] have been developed by integrating organic optical sensors and flexible energy-harvesting devices. Another approach involves combining organic LEDs with battery modules to develop a patch system for photobiomodulation that can be used for wound healing [20].
+
+In addition to integrated devices, devices with improved conformability to biological tissues are necessary for stable long-term monitoring of biological signals. Conformability can be achieved by combining two approaches: using soft materials having Young’s moduli similar to biological tissues, and reducing total device thickness [21]. Thanks to smaller Young’s moduli of organic materials compared to inorganic materials, and the compatibility of organic devices to low-temperature and solution processes, the total thickness can be reduced to 1–2 µm without sacrificing the device performance leading to the reduction of discomfort on the skin [22]–[25]. An ultrathin organic optical sensor has been developed with several microns thick [26]–[29].
+
+However, system-level integration of such thin organic optical sensors with power sources remains challenging. An significant obstacle to this type of integration is the insufficient operational stability of ultrathin organic LEDs under ambient air conditions [26]–[28], which hinders the long-term monitoring of biological signals.
+
+Here we propose an ultrathin self-powered organic optical system for plethysmogram monitoring. This system consists of three ultrathin electronic devices: polymer light-emitting diodes (PLEDs), organic solar cells, and organic photodetectors. By adopting an inverted structure and a polyethylenimine ethoxylated (PEIE) layer doped with 8-quinolinolato lithium (Liq) [30] as the electron transport layer (ETL), organic light-emitting diodes exhibit improved operation stability under ambient air conditions. Because of the intrinsic air operational stability of inverted PLED with PEIE:Liq electron injection layers, the fabricated ultraflexible PLED with no passivation retains 70% of its initial luminance lifetime of 11.3 h under ambient air, which is more than three times the luminance lifetime of conventional ultraflexible PLEDs. Integrated optical sensor exhibits a high linearity with the light intensity exponent of 0.98 by PLED light source. Such self-powered, ultraflexible PPG sensor shows a blood pulse monitoring during 7 s and can detect blood pulse rate of 77 beats per minute (bpm) on human hands.
+
+# Air-operation-stable Pleds
+
+Figure 1a and Supplementary Fig. 1 shows a schematic and photographs of ultraflexible, self-powered, PPG sensor operated on a human hand. The PPG sensors consisted of two types of ultraflexible devices: an ultraflexible PLED and ultraflexible organic photodiodes (OPD) (Supplementary Fig. 2). The ultraflexible PPG sensors and organic photovoltaic (OPV) modules were connected with flexible gold wiring, which was a 100-nm-thick gold electrode fabricated on a 12.5-µm-thick polyimide film (Fig. 1b). A top-view image of an OPV module and a PPG sensor are shown in Fig. 1c, and 1d, respectively. Figure 1e shows a schematic of the electrical circuit of a self-powered PPG device. The PLEDs were powered by ultraflexible OPV modules, and PPG data obtained from the OPD was transferred to the oscilloscope. The PLEDs were fabricated on a 1.5-µm-thick flexible substrate made with a 1-µm-thick parylene substrate and a 500-nm-thick planarisation layer of SU-8. The PLED comprised stacked layers of ITO/Zinc oxide (ZnO) /PEIE:Liq/Super Yellow (SY)/MoX/Al (Fig. 2a).
+
+A polymer of SY was used as the emission layer, providing highly efficient light emission [31]. To achieve air-stable operation of the PLED, an inverted structure was introduced with an electron injection layer of Liq-doped PEIE. While the inverted structure allows the use of an air-stable cathode in PLEDs and improves the air stability of PLEDs [32], doping PEIE with Liq can significantly improve the operational stability of PLEDs. To optimize the doping concentration of PEIE with Liq, the effect of the Liq wt% on the inverted PLED was examined (Supplementary Fig. 3). The J–L–V characteristics of the PLED show comparable electrical and optical outputs (> 10⁴ cd m⁻² at 10 V of applied voltage) for each doping concentration (Supplementary Fig. 3a and 3b). As shown in Supplementary Fig. 3c, 30 wt% of Liq was observed to be an optimal ratio in terms of current efficiency. For the air-operation stability, 10 wt% and 30 wt% of Liq show longer lifetime to be half of initial luminance (Supplementary Fig. 3d). Supplementary Fig. 3e summarises the variations in the efficiency and lifetime of the PLED with varying doping concentrations of Liq in the PEIE layer. From these results, we concluded that 30 wt% is the optimal Liq doping concentration to achieve an air-operation-stable inverted PLED. Therefore, the PEIE layer doped with 30 wt% Liq was used as the ETL for successive experiments. While the WF of PEIE:Liq shifts to the shallower level until the concentration of Liq up to 50 wt%, WF shifts back to the deeper level after the concentration of Liq over 50%. This WF shifts correspond to the electron injection of the PLED, which is improved by WF shifts to the shallower level. As a result of WF shift, 30 wt% showed the best result of efficiency [30]. The device stabilities under constant 8-V operation in ambient air were compared between various reported structures and ETLs (Supplementary Fig. 4). As a conventional structure, ITO/Poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS) /SY/NaF/Al was used. As inverted structures with different ETLs, ITO/ZnO/SY/MoX/Al, ITO/ZnO/Poly-ethylenimine (PEI) /SY/MoX/Al, and ITO/ZnO/PEIE:Liq/SY/MoX/Al were compared. While the inverted structure with PEI layers initially showed an efficient luminance of 5 cd A⁻¹, its air stability became lower than that of the structure without PEI layers. The conventional structure showed a relatively stable luminance for up to 200 min, after which its luminance decreased abruptly. The PEIE:Liq layer showed the best air-stability among all the examined PLED structures. A comparison of the current efficiencies of various inverted-structure PLEDs is shown in Supplementary Fig. 5. We chose F8BT as a conventional emission layer polymer and PEI as a conventional ETL polymer. When the current efficiency of the inverted-structure PLED with F8BT and PEI reached 3.4 cd A⁻¹, the current efficiencies of the inverted-structure PLED with SY and PEI and that with SY and PEIE reached 6.9 cd A⁻¹ and 11.7 cd A⁻¹, respectively. The operational lifetime and turn-on voltage of the inverted PLED with PEIE:Liq and SY were compared to those in previous studies (Fig. 2b) [28, 33–45]. The inverted PLED has an excellent half luminance lifetime of 41.5 h with 8-V operation in ambient air without encapsulation, while maintaining a low turn-on voltage (4.7 V). Turn-on voltage is determined as the voltage at which the luminance reaches to more than 1000 cd m⁻². Also, their angular distribution follows with Lambert's cosine law, which means our PLED has almost ideal diffusely reflecting surface (Supplementary Fig. 6).
+
+After the characteristics of the inverted PLED on the glass substrate were examined, the inverted PLED on the ultraflexible substrate was fabricated. When the PLED was fabricated on an ultraflexible substrate, it showed a turn-on voltage of 4.7 V, which is equal to that of the PLED on the glass substrate (4.7 V) (Fig. 2c). Figure 2d shows the ultraflexible inverted PLED during 6-V operation in air. As shown in Fig. 2e, the ultraflexible PLED exhibits a current efficiency of 14 cd A⁻¹ at a current density of 24 mA cm⁻²; this efficiency is comparable to or even higher than that of the PLED on the glass substrate. The EQE of PLED on ultraflexible substrate also shows comparable value to that on glass substrate (Supplementary Fig. 7). The mechanical durability of ultraflexible PLED was evaluated with the cyclic bending test. In the cyclic bending test, the ultraflexible PLED was bent cyclically to the bending radius of 5 mm. The results of the bending test is shown in Supplementary Fig. 8. Even after the device was bent cyclically with 50 times, the J-V curves did not show any degradation (Supplementary Fig. 8). Also, the luminance of PLED still remains 982 cd/m² which is comparable value as that of the initial value (918 cd/m²) after the 50 cyclic bending (Supplementary Fig. 8b). Finally, to evaluate the operational stability under ambient air, the operational lifetimes of the PLEDs on both substrates under ambient air were evaluated under a constant current of 8 mA cm⁻² (Fig. 2f). The inverted PLED on the glass substrate showed lifetime for 70% of initial luminance as 14.7 h. In the inverted PLED encapsulated with a 1-µm-thick parylene layer, the air-operation stability of the PLED improved to 33.4 h. When the PLED was fabricated on the ultraflexible substrate, the lifetime for 70% of its initial luminance decreased to 11.3 h. This is because of the low gas barrier property of ultraflexible substrates compared with that of glass [46]. Although the lifetime of the ultraflexible PLED was lower than that of the PLED fabricated on glass, it was still more than three times to the lifetime of a conventional ultraflexible PLED, which is 4 h [28]. To assess the reason of non-diode behaviour of our ultraflexible PLED (Fig. 2c), surface morphology of the PLED interface layer was studied by using atomic force microscope (AFM) (Supplementary Fig. 9). In the AFM images, the morphological difference clearly observed between bare ITO surface and ZnO interface layer (Supplementary Fig. 9a and b), which shows surface roughness increase in ZnO layer. However, addition of Liq into ZnO layer does not affect morphology dramatically (Supplementary Fig. 9b and c). These results proof us that the ZnO layer is a main contribution for increasing roughness, which would result in the non-diode behavior of our ultraflexible inverted PLED.
+
+# Ultraflexible Organic Photodetectors
+
+For an ultraflexible OPD, a blended film of PTzNTz-BOBO and PC<sub>71</sub>BM is used as the active layer. A stack of ITO/ZnO/PTzNTz-BOBO:PC<sub>71</sub>BM/MoOx/Ag was chosen as the OPD structure in this study (Fig. 3a). To use OPD as an application of PPG sensors, linearity factors determined by using the input light intensity (*L*), short circuit current (*J*<sub>SC</sub>), logarithmic *L*, and open-circuit voltage (*V*<sub>OC</sub>) are required [13]. Thus, the linearity of the OPD was examined for several light sources, such as simulated sunlight and light emitted from PLEDs. The dependence of the light intensity on *J*<sub>SC</sub> and *V*<sub>OC</sub> for the OPD, with simulated sunlight, is shown in Fig. 3a. The red dashed line and blue dashed line in Fig. 3b represent linearly fitted lines for *L* and *J*<sub>SC</sub> and for *L* and log(*V*<sub>OC</sub>), respectively. The *J*<sub>SC</sub> of OPD and the *L* of simulated sunlight were fitted with the equation , in which A is a constant value determined using the power-conversion efficiency (PCE) of the OPD. α is the light intensity exponent, which indicates the linearity and accuracy of the OPD’s response to the input light intensity. The α calculated from the experimental result was 0.95 for sunlight. The obtained value of α for sunlight was excellent because the value was quite close to the theoretical limit (1) of linearity of the OPD. Additionally, the external quantum efficiency (EQE) of the OPD at wavelengths of 300–900 nm was measured (Fig. 3c). The electroluminescence (EL) spectra of the ultraflexible inverted PLED are also shown in Fig. 3c. The EQE results represent broad light conversion of the OPD from 300 nm to 800 nm. The EL spectra of PLED had a peak wavelength of 590 nm and the width ranging from 500 nm to 800 nm. These results prove that the EQE spectra of the OPD and the EL spectra of the PLED are considerably overlapped over a range of wavelengths; thus, the fabricated OPD is capable of efficiently absorbing the light emitted from the ultraflexible PLED. The EL wavelength of our PLED has favourable properties to PPG because of their penetration depth. Since our body tissues have a greater absorption in yellow/green wavelength [47], the penetration depth of yellow/green light is shorter compared to that of infrared. Thus, the yellow/green light will suppress the artifact of the reflection and scattering from deeper tissue and leads advantageous light properties for reflection-mode PPG [48–51]. To ensure efficient and accurate PLED light detection, the linearity of the OPD with the input PLED light was examined, as shown in Fig. 3d and Fig. 3e. *L* and *J*<sub>SC</sub> were fitted well with the same equation as used in the sunlight experiment. The α with PLED light was calculated as 0.98 which proves that our OPD can accurately measure the light intensity of the PLED. The frequency response of the OPD was examined (Supplementary Fig. 10a). The rising time and falling time were calculated as 1 ms, respectively, which is fast enough to measure blood pulse with the frequency around 1 Hz (Supplementary Fig. 10b, c). The mechanical durability of ultraflexible OPD was evaluated with the cyclic bending test. Even after the device was bent cyclically with 150 times, the *J-V* curves did not show any degradation as a difference in the current density of light state and dark state (Supplementary Fig. 11a). The *J*<sub>SC</sub> difference between light state and dark state maintains the value of 2.15 mA/cm<sup>2</sup>, which remains 98% of the initial value before bending (Supplementary Fig. 11b). Also, our ultraflexible OPD shows excellent storage stability up to 100000 min under ambient air (Supplementary Fig. 12). The excellent air stability proofs a favourable characteristics of our ultraflexible OPD for self-powered devices. The OPD and PLED were used for the PPG sensor in further experiments.
+
+# Self-powered Ultraflexible Ppg Sensors
+
+A sufficient output power from the OPV is required to drive the PLED and make the PPG sensor self-powered. In order to increase the output power and voltage of OPV single cells, OPV modules containing several OPV single cells were designed and optimized. To optimize the OPV module design, the output current and voltage of the OPV module for varying numbers of series connections were calculated from the single-cell characteristics. The output current and voltage were calculated as  
+$$I_{OUT}=0.8\times \frac{I_{SC}}{n}$$  
+,  
+$$V_{OUT}=0.8\times V_{OC}\times n$$  
+, in which $I_{SC}$ is the short-circuit current of the single cell; $V_{OC}$ is the open-circuit voltage of the single cell; and $n$ is the number of series-connected single cells that constitute the module. $J_{SC}$ and $V_{OC}$ are multiplied by 0.8 to compromise typical value of fill-factor as 0.64 of OPV modules. Making the output current and voltage of the OPV modules equal to that of the PLED is crucial for efficient operation of self-powered PPG sensors. The results of the calculation indicated that OPV modules with 10 series-connected single cells have the highest overlap with the $I–V$ characteristics of the PLED (Supplementary Fig. 13). Based on this observation, OPV modules with series numbers of 6, 8, and 10 were fabricated (Supplementary Fig. 14a–c). In the OPV modules, $I_{OUT}$ decreased and $V_{OUT}$ increased proportionally as the increase in the number of series connections. As the maximum output powers of all the OPV modules became similar, the voltage required to obtain the maximum power point increased proportionally with an increase in the number of series connections (Supplementary Fig. 14d). The constant output power characteristics were examined by keeping the total area of the module, which is the area of individual cells multiplied by series number, constant (1.2 cm²) for every OPV module in this experiment. The operational point of the PLED powered by the OPV module was examined by overlaying the reversed $I–V$ curves of PLED with the $I–V$ curves of the OPV module (Supplementary Fig. 14e). From the results, the number of series connections required for the optimal OPV module design was calculated to be 10. This finding was similar to the result obtained from the previous calculation using single-cell properties; hence, we concluded that having 10 series connections is optimal for PLED operation. Hence, OPV modules with 10 series connections were fabricated on an ultraflexible substrate (Supplementary Fig. 15a and 15b). These ultraflexible OPV modules exhibited a PCE of 5.8%, which is comparable to the PCE of the module on the glass substrate (6.5%) (Supplementary Table 1). The operational stability of our OPV modules under air is also important. The ultraflexible OPV cells with same material showed 90% of initial PCE even after 3 h continuous 1-sun illumination with maximum power point (MPP) tracking test under ambient air, which indicates our OPV module should also has comparably stable characteristics under MPP tracking [52].
+
+Subsequently, PLED operation with a 10 series connected OPV module was experimentally examined. Simulated sunlight with a neutral density filter and with various optical densities was used as a variable-intensity input light source. A photograph depicting PLED operation with the OPV, and a schematic of the electrical circuit with the PLED and OPV are shown in Supplementary Fig. 16a and 16b, respectively. The light-intensity dependences of the OPV modules were measured and overlaid with the PLED $I-V$ curve (Supplementary Fig. 16c). From the results of Supplementary Fig. 16c, circuit current of PLED and OPV were expected to increase with the increase in the input light intensity. Supplementary Fig. 16d shows the experimentally measured dependence of the circuit current and PLED luminance on the light intensity. As expected, circuit current and luminance proportionally increased with the increase in the light intensity. The PLED was observed to emit bright light with an input light intensity of about 5000 µW/cm², which is 0.05 times the intensity of sunlight (Supplementary Fig. 16e). The PCE of OPV under indoor light shows 28.1% with 1000 lux of fluorescent lamp illumination [53]. This value is equivalent as power output of 78.2 µW/cm² whereas power output under 5000 µW/cm² simulated sunlight is 290 µW/cm². With these results, our self-powered PPG system should work under 1000 lux of indoor light with OPV modules of 4.5 cm² device area.
+
+Finally, the operation of the self-powered PPG sensor was demonstrated. The operation mechanism of a PPG sensor is shown in Fig. 4a. When light is emitted from the PLED and penetrates the finger, a part of the light is reflected at the blood vessels, while the remaining part is absorbed into the blood. The volume of blood vessels changes with every blood pulse, and the intensity of the reflected light is consistent with the volume pulsation of the blood vessel. The intensity of the reflected light can be detected by an OPD placed on the finger, and the blood pulse rate can be calculated from the OPD signals [5]. First, a self-powered PPG sensor consisting of an OPD, PLED, and OPV module on a glass substrate was fabricated and examined. The PLED was powered by an OPV module with 10 series connections, and the OPV module was illuminated with simulated sunlight of one-sun intensity. The PLED switched on when the OPV module was exposed to the sunlight, and it switched off when the OPV module was stored in the dark. Supplementary Fig. 17 shows the OPD signals when the PLED light was switched on and off by intermittently exposing the OPV module to the simulated sunlight. When the PLED light switched off, the OPD showed an almost constant voltage, as indicated by the black line in Supplementary Fig. 17. However, the OPD showed a clear periodic peak at 1.2 Hz, which is the frequency of the blood pulse, when the PLED was turned on. As shown in Fig. 4b, our PPG sensor performs very stable signal detection, even when the measurement duration is 20 s. The raw data of PPG signal without signal filtering also shows a large artifact in the low frequency (Supplementary Fig. 18). Also, there should be certain artifact from the body and finger movement. Therefore, there will be a certain challenge in the external circuit to process signal during the measurement and to monitor the signal continuously for the long-term measurement.
+
+After the self-powered PPG detection was confirmed to be operational on the glass substrate, self-powered PPG sensors on an ultraflexible substrate were fabricated. For this fabrication, previously reported lamination methods with 6 µm-thick adhesive tape¹ were used to combine the ultraflexible OPV module, OPD, and PLED (Supplementary Fig. 19). A photograph of the ultraflexible, self-powered PPG sensor is shown in Fig. 4c. Finally, the PPG signals from the self-powered ultraflexible PPG sensor were measured (Fig. 4d). Based on the results, blood pulse frequency was calculated to be 77 bpm. Although the periodic signal of the OPD was visible in the result, a large noise in the detected signal was also observed. Some of the reasons for the high noise in the ultraflexible PPG sensor were the large leak current of the ultraflexible OPD and the weak light intensity of the ultraflexible PLED due to the limited power supply from the ultraflexible OPV module.
+
+The reason of poor signal quality with ultraflexible PPG sensors might be the dark current increase of ultraflexible OPDs. While the previous research used the conventional structure OPD [28], this work used the inverted structure OPD with ZnO, which is air stable, but would have a higher rigidity and roughness [54], which would leads high dark current of OPD and poor signal quality of PPG. The active voltage regulator (AVR) circuit would be critical for stable power output in self-powered devices [55]. Since the AVR has an input voltage range from 7–15 V, the AVR have a feasibility to be combined with our 10 series-connected OPV module (8.2 V output) and regulate the voltage to 6.5 V for stable power output. Another critical requirement for self-powered device is the auxiliary battery, especially for their stable operation in dark situation. In Supplementary Fig. 20, the circuit with auxiliary rechargeable battery are explained. While the OPV module differs their current and voltage by the light intensity, the battery has stable current and voltage output and thus, the stable output light of PLED will be achieved. 180 µm-thick battery [56] and 40 µm-thick supercapacitor [57] would be used to achieve self-powered PPG sensors with auxiliary battery. Lastly, signal sampling is a critical problem for state-of-the-art flexible, self-powered electronics. In both an ultraflexible self-powered organic electrochemical transistor [58] and a flexible self-powered PPG sensor [56], the sampling systems were rigid ICs and separately connected as external circuits. Though our OPV module is capable to operate signal monitoring ICs because their power comsumtion is is ~ 0.2µW (TWILITE, Mono Wireless), certain effort is necessary to make these ICs ultraflexible and integrate to the self-powered systems.
+
+# Conclusion
+
+In this study, a self-powered, ultraflexible PPG sensor employing air-operation-stable, ultraflexible PLEDs was fabricated. To the best of our knowledge, this is the first study that demonstrates the use of self-powered optoelectrical sensors in ultraflexible organic devices. This self-power technology used in this study will pave the wave of ultraflexible wearable optoelectronic devices which take an important role for future ubiquitous healthcare society.
+
+# Methods
+
+## Materials
+The PTzNTz-BOBO samples were synthesised according to a previously reported procedure [59]. PC₇₁BM was purchased from Solenne BV Corporation. PEIE, Liq, Zinc acetate dihydrate, 2-methoxyethanol, and 2-aminoethanol were purchased from Wako Chemical. The yellow light-emitting polymer, SY, was purchased from Sigma-Aldrich. The glass substrate with a patterned ITO electrode was purchased from EHC Ltd.. The sheet resistance of ITO was <10 Ω/sq.
+
+## Ultraflexible Film Substrate Fabrication
+The ultra-thin film substrates were composed of double-layered parylene (diX-SR, Daisan Kasei Co., Ltd.) and epoxy (SU-8 3005, MicroChem). First, a 1-μm-thick parylene layer was deposited via chemical vapour deposition (CVD) on a glass substrate whose surface was coated with a fluorinated polymer (Novec 1700, 3M). A 500-nm-thick epoxy layer was then spin-coated (5000 rpm for 60 s) on a 1-µm-thick parylene layer as the planarisation layer. The film was then annealed at 95 °C for 3 min after ultraviolet exposure. were annealed under a nitrogen atmosphere at 180 °C for 30 min. All ultra-thin devices were electrically connected with thin Au wiring on a 12.5-μm-thick polyimide substrate using anisotropic conductive film (3M, ECATT 9703) tapes as contact parts.
+
+## Inverted PLED Fabrication
+A 100-nm-thick ITO layer was subsequently formed as the transparent electrode by DC sputtering. The ITO electrodes were patterned via photolithography using a ZPN (Nihon Zeon) negative resist and an ITO-07N (Kanto Chemical) ITO etcher. A 20-nm-thick ZnO layer was used as the electron-transport layer. The substrates were spin-coated with a ZnO precursor (5000 rpm for 30 s) solution prepared by dissolving zinc acetate dehydrate (0.5 g) and ethanolamine (0.16 mL) in 5 mL 2-methoxyethanol [60]. The substrates were then baked in air at 180 °C for 30 min. To form PEIE:Liq layer, the mixture of 1.5 wt% solution of PEIE and Liq with 2-methoxyethanol was deposited on a substrate via spin-coating (4000 rpm 60 s). After annealing the substrate at 100 °C for 1 min, the device was rinsed with ethanol and annealed at 100 °C for 1 min again. The SY layer was deposited via spin-coating using a toluene solution of 6 mg/ml. Subsequently, the MoOₓ (10 nm) and Al (80 nm) layers were deposited via vacuum evaporation. Finally, a 1-μm-thick parylene layer was deposited via CVD to form a passivation layer.
+
+## OPV and OPD Fabrication
+A ZnO layer was deposited using the same methods as that used for the PLED. PTzNTZ-BOBO and PC₇₁BM were mixed as a 1:1.5 of weight ratio. Then, chlorobenzene was added to a mixture with 5 g L⁻¹ PTzNTZ-BOBO concentration. The solution was stirred at 100 °C for 1 h, and 1 vol% of 1,8-Diiodooctane was added based on the amount of chlorobenzene after the temperature of the solution decreased back to 20-30 °C in under N₂ atmosphere. The solution was spin-coated on the device in a glove box at 600 rpm for 20 s. The active layer thickness is 200 nm. MoOₓ (7.5 nm) and Ag (100 nm) were sequentially deposited via thermal evaporation, under a pressure of less than 1 × 10⁻³ Pa, as a top interlayer and an anode, respectively. Finally, a 1-μm-thick parylene layer was deposited via CVD to form a passivation layer.
+
+## Device characterisation
+The OPV modules with a total cell area of 1.2 cm² and OPD with an active area of 0.04 cm² were characterised under one-sun illumination using a solar simulator (AM 1.5 global spectrum with 1000 W m⁻² intensity calibrated using a silicon reference diode, Bunkokeiki). The J–V characteristics were recorded using a Keithley 2400 source metre for every 10 mV in an ambient laboratory atmosphere, without humidity or temperature control (with an approximate temperature of 20 °C and humidity of 30% relative humidity (RH)). Accordingly, external gold wirings were used for the electrical contacts of the OPVs on the ultrathin substrates. The 100-nm-thick gold wirings were deposited under a vacuum through a shadow mask onto 12.5-μm-thick polyimide films. One side of the wirings were connected to the electrodes on the freestanding foils using an electrically conductive adhesive transfer tape (3M™, ECATT 9703). The other sides were connected to the source metre using alligator clips. Light power of the PLED was calibrated with optical power meter (1936-R, Newport).
+
+To evaluate the PLED characteristics, a light distribution measurement system (C9920-11, Hamamatsu Photonics) and an external quantum efficiency measurement system (C9920-12, Hamamatsu Photonics) were used. For the air-stability tests, the devices were operated under ambient air (~20 °C, 30% RH). The measurements were also performed in an ambient atmosphere.
+
+To perform the PLED operation experiment with the OPV as a power source, a solar simulator and an optical filter were used to modulate the light intensity for simulated sunlight input. A schematic of the experimental setup is shown in Supplementary Fig. 21. Anode of PLED and OPV module are electrically connected. Cathode of the PLED and the OPV module are connected each other through digital multi meter.
+
+## Data Availability
+The data that support the plots present in this paper and the other findings of this study are available from the corresponding author upon reasonable request.
+
+# References
+
+1. Kim, J. et al. "Wearable smart sensor systems integrated on soft contact lenses for wireless ocular diagnostics." *Nat. Commun.* **8**, 14997 (2017).
+2. Wang, S. et al. "Skin electronics from scalable fabrication of an intrinsically stretchable transistor array." *Nature* **555**, 83–88 (2018).
+3. Kondo, M. et al. "Imperceptible magnetic sensor matrix system integrated with organic driver and amplifier circuits." *Sci. Adv.* **6**, eaay6094 (2020).
+4. Sanford, J. L., and Eugene S. S. "Direct View Active Matrix VGA OLED-on‐Crystalline‐Silicon Display." *Soc. Information Display Dig.* **32**, 376–379 (2001).
+5. Lochner, C. M. et al. "All-organic optoelectronic sensor for pulse oximetry." *Nat. Commun.* **5**, 5745 (2014).
+6. Lee, Y. et al. "Sticker-type hybrid photoplethysmogram monitoring system integrating CMOS IC with organic optical sensors." *IEEE J. Emerg. Top. Circ. Syst.* **7**, 50–59 (2017).
+7. Bilgaiyan, A. et al. "Enhancing small molecule organic photodetector performance for reflectance mode photoplethysmography sensor applications." *ACS Appl. Electron. Mater.* (2020). DOI: 10.1021/acsaelm.0c00082.
+8. Xu, H. et al. Flexible Organic/Inorganic Hybrid Near-Infrared Photoplethysmogram Sensor for Cardiovascular Monitoring. *Adv. Mater.* **29**, 1–6 (2017).
+9. Elsamnah, F. et al. Reflectance-based organic pulse meter sensor for wireless monitoring of photoplethysmogram signal. *Biosensors* **9**, 1–13 (2019).
+10. Simone, G. et al. “High-Accuracy Photoplethysmography Array Using Near-Infrared Organic Photodiodes with Ultralow Dark Current.” *Adv. Opt. Mater.* **8**, 1901989 (2020).
+11. Khan, Y. et al. "A flexible organic reflectance oximeter array." *Proc. Natl. Acad. Sci. USA* **115**, E11015–E11024 (2018).
+12. Lee, H. et al. "Toward all-day wearable health monitoring: An ultralow-power, reflective organic pulse oximetry sensing patch." *Sci. Adv.* **4**, eaas9530 (2018).
+13. Yokota, T. et al. "A conformable imager for biometric authentication and vital sign measurement." *Nat. Electron.* **3**, 113–121 (2020).
+14. Tordera, D. et al. "Vein detection with near-infrared organic photodetectors for biometric authentication." *J. Soc. Inf. Display* **28**, 381–391 (2020).
+15. Lee, B. et al. “Ultraflexible and transparent electroluminescent skin for real-time and super-resolution imaging of pressure distribution.” *Nat. Commun.* **11**, 663 (2020).
+16. Gelinck, G. H. et al. "X-ray imager using solution processed organic transistor arrays and bulk heterojunction photodiodes on thin, flexible plastic substrate." *Org. Electron.* **14**, 2602–2609 (2013).
+17. Basirico, L. et al. "Direct X-ray photoconversion in flexible organic thin film devices operated below 1 V." *Nat. Commun.* **7**, 13063 (2016).
+18. Zhang, M., and Yeow, J. T. "A flexible, scalable, and self-powered mid-infrared detector based on transparent PEDOT: PSS/graphene composite." *Carbon* **156**, 339–345 (2020).
+19. Lu, H. et al. "A self-powered and stable all‐perovskite photodetector–solar cell nanosystem." *Adv. Func. Mater.* **26**, 1296–1302 (2016).
+20. Jeon, Y. et al. "A wearable photobiomodulation patch using a flexible red-wavelength OLED and its in vitro differential cell proliferation effects." *Adv. Mater. Technol.* **3**, 1700391 (2018).
+21. Yamagishi, K., Takeoka, S., and Fujie, T. “Printed nanofilms mechanically conforming to living bodies.” *Biomater. Sci.* **7**, 520–531 (2019).
+22. Kim, D. H. et al. "Dissolvable films of silk fibroin for ultrathin conformal bio-integrated electronics." *Nat. Mater.* **9**, 511–517 (2010).
+23. Kim, D. H. et al. "Epidermal electronics." *Science* **333**, 838–843 (2011).
+24. Kaltenbrunner, M. et al. "An ultra-lightweight design for imperceptible plastic electronics." *Nature* **499**, 458–463 (2013).
+25. Liu, Z. et al. "Ultraflexible in-plane micro‐supercapacitors by direct printing of solution‐processable electrochemically exfoliated graphene." *Adv. Mater.* **28**, 2217–2222 (2016).
+26. Kaltenbrunner, M. et al. "Ultrathin and lightweight organic solar cells with high flexibility." *Nat. Commun.* **3**, 770 (2012).
+27. White, M. S., et al. "Ultrathin, highly flexible and stretchable PLEDs." *Nat. Photonics* **7**, 811–816 (2013).
+28. Yokota, T. et al. "Ultraflexible organic photonic skin." *Sci. Adv.* **2**, e1501856 (2016).
+29. Yin, D. et al. "Efficient and mechanically robust stretchable organic light-emitting devices by a laser-programmable buckling process." *Nat. Commun.* **7**, 11573 (2016).
+30. Chiba, T. et al. “Addition of lithium 8-quinolate into polyethylenimine electron-injection layer in OLEDs: Not only reducing driving voltage but also improving device lifetime.” *ACS Appl. Mater. Interfaces* **9**, 18113–18119 (2017).
+31. Höfle, S., Schienle, A., Bruns, M., Lemmer, U., and Colsmann, A. “Enhanced electron injection into inverted polymer light-emitting diodes by combined solution-processed Zinc Oxide/Polyethylenimine interlayers.” *Adv. Mater.* **26**, 2750–2754 (2014).
+32. Spreitzer, H. et al. “Soluble phenyl-substituted PPVs–New materials for highly efficient polymer LEDs.” *Adv. Mater.* **10**, 1340–1343 (1998).
+33. Bae, S. J. et al. “Enhancement of barrier properties using ultrathin hybrid passivation layer for organic light emitting diodes.” *Jpn. J. Appl. Phys.* **45**, 5970–5973 (2006).
+34. Park, S. H. K. et al. “Ultrathin film encapsulation of an OLED by ALD.” *Electrochem. Solid-State Lett.* **8**, H21–H23 (2005).
+35. Kim, G. H., Oh, J., Yang, Y. S., Do, L. M., and Suh, K. S. “Lamination process encapsulation for longevity of plastic-based organic light-emitting devices.” *Thin Solid Films* **467**, 1–3 (2004).
+36. Jeong, F. G. et al. “A mechanically enhanced hybrid nano-stratified barrier with a defect suppression mechanism for highly reliable flexible OLEDs.” *Nanoscale* **9**, 6370–6379 (2017).
+37. Seo, H. K. et al. “Laminated graphene films for flexible transparent thin film encapsulation.” *ACS Appl. Mater. Interfaces* **8**, 14725–14731 (2016).
+38. Li, H.-Y., Liu, Y.-F., Duan, Y., Yang, Y.-Q. & Lu, Y.-N. “Method for Aluminum Oxide Thin Films Prepared through Low Temperature Atomic Layer Deposition for Encapsulating Organic Electroluminescent Devices.” *Materials.* **8**, 600–610 (2015).
+39. Yang, Y. Q. et al. “Realization of thin film encapsulation by atomic layer deposition of Al2O3 at low temperature.” *J. Phys. Chem. C* **117**, 20308–20312 (2013).
+40. Duan, Y. et al. “High-performance barrier using a dual-layer inorganic/organic hybrid thin-film encapsulation for organic light-emitting diodes.” *Org. Electron.* **15**, 1936–1941 (2014).
+41. Grover, R., Srivastava, R., Kamalasanan, M. N. & Mehta, D. S. “Multilayer thin film encapsulation for organic light emitting diodes.” *RSC Adv.* **4**, 10808–10814 (2014).
+42. Wu, Z., Wang, L., Chang, C. & Qiu, Y. “A hybrid encapsulation of organic light-emitting devices.” *J. Phys. D. Appl. Phys.* **38**, 981–984 (2005).
+43. Kim, G. H., Oh, J., Yang, Y. S., Do, L. M. & Suh, K. S. “Encapsulation of organic light-emitting devices by means of photopolymerized polyacrylate films.” *Polymer.* **45**, 1879–1883 (2004).
+44. Kho, S., Cho, D. & Jung, D. “Passivation of organic light-emitting diodes by the plasma polymerized para-xylene thin film.” *Japanese J. Appl. Physics, Part 2 Lett.* **41** (2002).
+45. Yamashita, K., Mori, T. & Mizutani, T. “Encapsulation of organic light-emitting diode using thermal chemical-vapour-deposition polymer film.” *J. Phys. D. Appl. Phys.* **34**, 740–743 (2001).
+46. Dennler, G., Lungenschmied, C., Neugebauer, H., Sariciftci, N. S., and Labouret. “A. flexible, conjugated polymer-fullerene-based bulk-heterojunction solar cells: Basics, encapsulation, and integration.” *J. Mater. Res.* **20**, 3224–3233 (2005).
+47. Song, K. et al. Subdermal Flexible Solar Cell Arrays for Powering Medical Electronic Implants. *Adv. Healthc. Mater.* 5, 1572–1580 (2016).
+48. Maeda, Y. et al. Comparison of reflected green light and infrared photoplethysmography. *Proc. 30th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS’08 - "Personalized Healthc. through Technol.* 2270–2272 (2008).
+49. Maeda, Y., Sekine, M. & Tamura, T. The advantages of wearable green reflected photoplethysmography. *J. Med. Syst.* 35, 829–834 (2011).
+50. Lee, J. et al. Comparison between red, green and blue light reflection photoplethysmography for heart rate monitoring during motion. *Proc. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS* 1724–1727 (2013) doi: 10.1109/EMBC.2013.6609852.
+51. Maeda, Y., Sekine, M. & Tamura, T. Relationship between measurement site and motion artifacts in wearable reflected photoplethysmography. *J. Med. Syst.* 35, 969–976 (2011).
+52. Kimura, H. et al. High Operation Stability of Ultraflexible Organic Solar Cells with Ultraviolet-Filtering Substrates. *Adv. Mater.* 31, 1808033 (2019).
+53. Lee, H. K. H. et al. "Organic photovoltaic cells – promising indoor light harvesters for self-sustainable electronics." *J. Mater. Chem. A* 6, 5618–5626 (2018).
+54. Krebs, F. C., Gevorgyan, S. A. & Alstrup, J. A roll-to-roll process to flexible polymer solar cells: model studies, manufacture and operational stability studies. *J. Mater. Chem.* 19, 5442 (2009).
+55. Fuketa, H. et al. 16.4 Energy-autonomous fever alarm armband integrating fully flexible solar cells, piezoelectric speaker, temperature detector, and 12V organic complementary FET circuits. in *2015 IEEE International Solid-State Circuits Conference - (ISSCC) Digest of Technical Papers* vol. 58 1–3 (IEEE, 2015).
+56. Ostfeld, A. E., Gaikwad, A. M., Khan, Y. & Arias, A. C. High-performance flexible energy storage and harvesting system for wearable electronics. *Sci. Rep.* 6, 26122 (2016).
+57. Liu, R. et al. "An Efficient Ultra-Flexible Photo-Charging System Integrating Organic Photovoltaics and Supercapacitors." *Adv. Energy Mater.* 10, 1–8 (2020).
+58. Park, S. et al. Self-powered ultra-flexible electronics via nano-grating-patterned organic photovoltaics. *Nature* 561, 516–521 (2018).
+59. Saito, M. et al. “Highly efficient and stable solar cells based on Thiazolothiazole and Naphthobisthiadiazole copolymers.” *Sci. Rep.* **5**, 14202 (2015).
+60. Vohra, V. et al. “Efficient inverted polymer solar cells employing favourable molecular orientation.” *Nat. Photonics* **9**, 403–408 (2015).
+
+# Supplementary Files
+
+- [SIJinnno.docx](https://assets-eu.researchsquare.com/files/rs-117283/v1/6a32ce7968f49b86fa4409fd.docx)

17d228c8d89c44bbe230367eab1751d3f9314fc331665758932a22c1041549fc/preprint/images_list.json

@@ -0,0 +1,58 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Structure of the inactive FZD7 dimer mediated by a lipidic interface. (A) Cryo-EM density map of the FZD7 dimer with C2 symmetry, colored by monomer, with an intermolecular layer of lipids sandwiched between the monomers, shown in red. (B) Atomic arrangement of the FZD7 dimer shown with a close-up side view of the compact interface presenting with endogenous 1-palmitoyl-2-oleoyl-phosphatidylcholine (POPC) and two major cholesterol hemisuccinate (CHS) molecules, which altogether mediate and stabilize the interface between the monomers. (E) Overall organization of the extracellular region of FZD7 with ECL1, ECL2, ECL3, and the hinge region which altogether form a peripheral lid over the transmembrane bundle. (D) A top view of the heptahelical bundle depicts the folded beta-stranded structure of ECL2 with K5337.41 and Y5347.42 rendering blockage of the receptor core. (E) Map of the surface electrostatic potential of the internal cavity highlighting the bottleneck in modulating entry of water molecules.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Structural comparison of conserved motifs in the inactive and active FZD7 conformations. \u00a0\nThe intracellular rearrangements of conserved residue (A) P4816.43, (B) extended molecular switch residues, F4746.36 - W5476.36, (E) Y4786.40 and (F) molecular switch residues W5477.55 - R4706.32 are observed upon agonist binding. (C) A network of extended polar contacts in ICL3 is made between the two conformations of FZD7. (D) Comparison of the nucleotide-free G\u03b1s-bound FZD7 (pink, PDB: 7EVW) and inactive FZD7 (purple) highlighting structural changes of conserved residues and regions of interest. (G) G protein binding rearranges residues in ICL1 involving D2781.57, F282, R281, R280, and M279. Molecular dynamic profiles depict the dihedral angles of rotamers (H) W3543.43 and (I) W5477.55 between inactive and active conformations.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Structural insight into the activation mechanisms across GPCR classes.\u00a0 \u00a0\nThe dynamic nature of TM6 in GPCRs between the inactive and active conformations are shown in (A) Class A, b2AR (PDB: 6PS3, 3SN6), (B) Class B, GCGR (PDB: 5XEZ and 6WPW) and (C) Class F, FZD7 (PDB: 9EPO, 7EVW). (D) The molecular switch in Class F, R6.32-W7.55 acts as a hinge limiter restricting the swing out of TM6.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Water mediated-interactions and conserved residues of FZD7.\n(A) Overall organization of the internal water pocket of inactive FZD7. (B) The region of the cavity referred to as the bottleneck is comprised of the following residues: L415ECL1, D405ECL1, Y4896.51, Y5347.42, and K5337.41. (C) Phylogenetic analysis of FZD7 depicting highly conserved and interacting residue pairs involved in the internal water pocket.\u00a0 (D,E,F) The tight base of the internal water cavity facilitates a reorientation of residues W3543.43, Y4786.40, Y2962.51, and V5407.48 in the (D) inactive conformation and (E) active conformation. (F) A top view of the transmembrane domains between the two structures highlighting residues of interest. \u00a0",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Molecular dynamics of the water network in the inactive conformation of FZD7.\n(A) The occupancy and location of water molecules attributed to the internal water pocket with green depicting high occupancy and red depicting low occupancy during molecular dynamics simulations of the inactive FZD7. Volumetric map for water molecules calculated for 20% occupancy (blue density) with inactive (B) and (C) active FZD7. (D) Key residues in the bottleneck region for the first conformation (transient bottleneck closing) with a volumetric map of the bottleneck residues computed with a threshold of 20% occupancy (blue density). (E) The second conformation represents a reorganization of ECL2 (disruption) with a volumetric map of the bottleneck residues that was computed with a threshold of 20% occupancy (brown density). (F) Distance calculations between L415ECL2/CD2 and K5337.41/NZ depict the two transient conformations shown in (D, turquoise) and (E, orange) over the course of 6000 frames of the MD simulations.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Workflow for the investigation of a conserved cholesterol binding site across Class F GPCRs.\n(A) A phylogenetic analysis was performed across Class F GPCRs where highly conserved residues involved in cholesterol binding were identified. Fully conserved residues are colored in dark red and residues with similar conservation are colored in yellow. (B) Overview of the cryo-EM pipeline from data collection to model building and structural characterization of the cholesterol binding site. (C) Based on the structural analysis, site-directed mutagenesis of cholesterol-interacting residues was performed and then used for analysis of functional downstream readouts of transducer coupling and signaling.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_7.png",
+        "caption": "Role of cholesterol in transducer coupling and signaling.\n(A) Organization of the inactive FZD7 monomer with the cholesterol (CHS; green) binding site. (B) Analysis of the cholesterol binding site reveals various interacting residues of which three were mutated for functional studies; F3453.34, H3824.46, and W3864.50. (C) G\u03b1s translocation assay showing BRET2 ratios of \u0394FZD1-10 HEK293T cells transiently transfected with rGFP-CAAX and GS-67-RlucII, with either negative control (pcDNA), wild-type FZD7, or the indicated FZD7 mutants normalized to conditions with pcDNA. Data are presented as means \u00b1 SEM of normalized BRET2 ratios from three independent experiments. (**p < 0.001, one-way ANOVA followed by Tukey\u2019s multiple comparison). (D) A NanoBiT-based DEP recruitment BRET assay was performed with FZD7-SmBiT, its respective cholesterol binding site mutants, and \u03b22AR-SmBiT as a control in HEK293A cells. For functional reconstitution of Nluc, membrane-anchored FLAG-LgBiT-CAAX was cotransfected. Data are presented in triplicates of three independent experiments with baseline-corrected BRET values based on conditions with 0% DEP-Venus.\u00a0 (E) \u0394FZD1-10 HEK293T cells were transfected with Renilla (Rluc), the Super 8x TOPFlash reporter, and wild-type FZD7, or the indicated mutants. Cells were stimulated with 300 ng/mL of recombinant WNT-3A overnight. Data are presented as Fluc/Rluc ratios, normalized to the respective vehicle control, from three independent experiments performed in triplicates analyzed using one-way ANOVA (Dunnett post hoc test) (**p < 0.01, *p<0.05).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

17eab058ec6b01f33dab22a05914fcc299f37652a28c56c8fa8876a919caccf3/preprint/images_list.json

@@ -0,0 +1,50 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Development of the VEDIC system for high efficiency intracellular protein delivery by EVs. a, Graphic abstract of the development of VEDIC and VFIC systems for high efficiency intracellular protein delivery in vitro and in vivo. Intein in tripartite fusion protein (EV-sorting Domain-Intein-Cargo) performs C-terminal cleavage during the process of EV-biogenesis, resulting in enriched free cargo protein inside EVs. Together with a fusogenic protein, VSV-G, these engineered EVs achieve high-efficiency intracellular delivery of cargo protein (Cre and super repressor of NF-\u0138B) or protein complex (Cas9/sgRNA RNPs) both in reporter cells and in mice models. b, Fluorescence reporter construct expressed in the reporter cells generated to measure Cre delivery. c, Constructs used for developing the VEDIC system. d, Schematic of intein cleavage and intraluminal cargo release during EV biogenesis, MVB: multivesicular body. e, Percentage of GFP positive reporter cells after adding EVs for 2 days, as evaluated by flow cytometry. f, Fusogen screen in T47D-TL cells after a two-day incubation period with EVs. g,Percentage of GFP positive reporter cells after exposure to EVs derived from VSV-G co-transfected cells. h, EV dose-and time-dependent recombination in HeLa-TL reporter cells mediated by VEDIC EVs. i, EV dose-and time-dependent recombination in B16F10-TL reporter cells mediated by VEDIC EVs. j, Cre and VSV-G protein was detected in T47D-TL reporter cells by Western blot (WB) analysis, 48 hours (h) after addition of engineered EVs loaded with Cre in 24-well plates. Two-way ANOVA multiple comparisons test was used for analysis of (g) and (i); One-way ANOVA multiple comparisons test was used for analysis of (f). Data are shown as mean+SD, ** p < 0.01; *** p < 0.001.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Development of the VFIC system to further improve intracellular protein delivery by EVs. a, VSV-G+ and CD63+ EV concentrations as determined by single-vesicle flow cytometry after transfection with VSV-G-mNG alone or VSV-G-mNG and CD63 together. b, Constructs generated for developing the VFIC system. The last construct was generated where Cre was replaced with the bacteriophage protein MS2, as a negative control for Cre delivery 32. c, EV dose-dependent recombination in B16F10-TL cells mediated by VSV-G-Foldon-Intein-Cre and VSV-G-Intein-Cre EVs as evaluated by flow cytometry. d, Representative images showing GFP positive HeLa-TL cells 24, 48 and 72 h after exposure to VFIC EVs at different doses. Scale bar, 100 \u00b5m. e-h, Recombination in hard-to-transfect reporter cells (MSC-TL, THP-1-TL, Raw264.7-TL and K562-TL) mediated by VFIC EVs after 48 h. Two-way ANOVA multiple comparisons test was used for analysis of (c) and (e-h). Data are shown as mean+SD, * p < 0.05; ** p < 0.01; *** p < 0.001; ns: non-significant.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "The underlying mechanisms responsible for intein cleavage and endosomal escape of VSV-G. a, Schematic illustration of the cleavage mechanism of different engineered intein variants. b, Protein expression of different engineered mutant intein constructs in whole cell lysates (WCL) and isolated EVs derived from HEK293T cells evaluated by WB analysis. Lysates from 5\u00d7105 EV-producing cells and 1\u00d71010 engineered EVs were used for the assay. TSG101, syntenin-1 and \u03b2-actin were used as EV markers and Calnexin was used as a cellular organelle marker (endoplasmic reticulum) that should be absent in EV samples. c,d, Properties of the two VSV-G mutants: VSV-G P127D loses the capacity to mediate fusion between the EV-and endosomal membranes and VSV-G K47Q is unable to bind to LDLR on the cell surface. e, Confocal immunofluorescence demonstrating the subcellular distribution of mNG in the presence or absence of wild type VSV-G engineered EVs in Huh7 cells. Scale bar, 20 \u00b5m, representative images. f, Subcellular distribution of mNG in different groups after adding the indicated engineered EVs determined by confocal immunofluorescence. Scale bar, 20 \u00b5m, representative images. g, WB evaluation of protein levels of Cre and VSV-G in HeLa-TL reporter cells after addition of engineered EVs with wild type, P127D or K47Q VSV-G in 24-well plates.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Robust gene editing by Cas9/sgRNA RNP and meganuclease targeting PCSK9 using the VFIC and VEDIC systems. a, Constructs generated for Cas9/sgRNA RNPs delivery. b, Schematic illustration about how the Cas9/sgRNA RNPs were encapsulated into engineered EVs. c, Schematic of reporter cells used to functionally assess Cas9/sgRNA RNPs delivery by engineered EVs. d,Percentage of eGFP positive cells after addition of engineered EVs, as measured by flow cytometry 48, 72 and 96 h after EV addition. e, Immunofluorescence demonstrated gene-editing in recipient cells after treatment with different doses of engineered EVs after 4 days. Scale bar, 100 \u00b5m, representative images. f, Constructs generated for EV-mediated delivery of meganuclease targeting PCSK9. g, WB analysis of PCSK9 and VSV-G protein in Huh7 cells after treatment with different doses of EVs in 24-well plates. Two-way ANOVA multiple comparisons test was used for analysis of (d). Data are shown as mean+SD, * p < 0.05; ** p < 0.01; *** p < 0.001.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Cre recombination in R26-LSL-tdTomato reporter mice by VEDIC and VFIC engineered EVs after local (ICV or osmotic pump ICV) and systemic (IP) injections. a,Workflow for the intracerebroventricular (ICV) injection of engineered EVs to deliver Cre in the brains of R26-LSL-tdTomato reporter mice. b, TdTomato expression in different regions of brain after ICV injection of engineered EVs, as determined by immunofluorescence. Scale bar, 50 \u00b5m for cerebellum and cortex, and 200 \u00b5m for hippocampus. c, Workflow for the osmotic pump ICV injection of engineered EVs to transfer Cre to the brain tissues of R26-LSL-tdTomato reporter mice. d, Percentage of tdTomato+ cells in the brain tissues after osmotic pump ICV injection of engineered EVs for 5 days. n=4 mice for PBS group whereas n=5 mice for other groups. e, Schematic workflow for the intraperitoneal (IP) injection of engineered EVs into R26-LSL-tdTomato reporter mice. f, TdTomato expression in liver and spleen after IP injection of engineered EVs for one week. Scale bar, 50 \u00b5m. g, Co-staining of tdTomato with the T cell marker CD3 in spleen as detected by immunofluorescence one week after IP injection of engineered EVs. Scale bar, 50 \u00b5m. h, Co-staining of tdTomato with the B cell marker B220 in spleen one week after IP injection of engineered EVs. Scale bar, 50 \u00b5m.\ni, Co-staining of tdTomato and the macrophage marker F4/80 in spleen one week after IP injection of engineered EVs. Scale bar, 50 \u00b5m. n=3 mice per group, representative images for (b) and (f-i). Two-way ANOVA multiple comparisons test was used for analysis of (d). Data are shown as mean+SD, * p < 0.05; ** p < 0.01; **** p < 0.0001.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Treatment of LPS-induced systemic inflammation using VEDIC and VFIC-EV mediated delivery of a super-repressor of NF-\u0138B. a, b, Design of the constructs and reporter cells utilized for delivery and assessment of a super-repressor of NF-\u0138B by engineered EVs. c, Schematic illustration how the EV-delivered super-repressor of NF-\u0138B inhibits NF-\u0138B activity. d,e, Luciferase activity from HEK-NF-\u0138B reporter cells, 24 or 48 h after treatment with engineered EVs respectively (TNF-\u03b1 stimulation for 6 h before harvesting cells), in 24-well plates. f, Schematic illustration of the workflow for the treatment of LPS-induced systemic inflammation by engineered EVs in mice (5\u00d71011 EVs/mouse per dose). g,h, Percentage of body weight loss and group survival in mice after LPS and engineered EV injections. n=10 mice per group. i, Representative histology images (hematoxylin-eosin stain) of liver to show the aggregation of inflammatory cells (upper panel, yellow arrows indicate the aggregated inflammatory cells in portal areas) and the hydropic degeneration of hepatocytes (lower panel, red arrows indicate the hydropic degeneration of hepatocytes) after LPS induction. Scale bar, 50 \u00b5m. Two-way ANOVA multiple comparisons test was used for analysis of (d, e, and g); Log-rank (Mantel-Cox) test was used for the analysis of survival curve (h). Data are shown as mean+SD for (d, e, and g). * p < 0.05; ** p < 0.01.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

17eab058ec6b01f33dab22a05914fcc299f37652a28c56c8fa8876a919caccf3/preprint/preprint.md

@@ -0,0 +1,287 @@
+# Abstract
+
+Intracellular delivery of protein and RNA therapeutics represents a major challenge. Here, we developed highly potent engineered extracellular vesicles (EVs) by incorporating essential bio-inspired attributes required for effective delivery. These comprise engineered mini-intein proteins with self-cleavage activity for active cargo loading and release, and fusogenic VSV-G protein to activate productive endosomal escape. Combining these components allowed high efficiency recombination and genome editing *in vitro* following EV-mediated delivery of Cre recombinase and Cas9/sgRNA RNP cargoes, respectively. *In vivo*, single dose EV-mediated Cre delivery to the brains of Cre-LoxP R26-LSL-tdTomato reporter mice resulted in greater than 40% and 30% recombined cells in hippocampus and cortex respectively. In addition, we demonstrate therapeutic potential of this platform by showing inhibition of LPS-induced systemic inflammation via delivery of a super-repressor of NF-ĸB activity. Our data establish these engineered EVs as a novel platform for effective delivery of multimodal therapeutic cargoes, including for efficient genome editing.
+
+[Biological sciences/Biotechnology/Protein delivery](/browse?subjectArea=Biological%20sciences%2FBiotechnology%2FProtein%20delivery)  
+[Biological sciences/Biological techniques/Nanobiotechnology](/browse?subjectArea=Biological%20sciences%2FBiological%20techniques%2FNanobiotechnology)
+
+# Full Text
+
+Protein-based therapeutics have unique potential for the treatment of diverse diseases. Whereas there has been great success developing therapeutic proteins against extracellular targets, of which a broad range have been approved for clinical application<sup>1,2</sup>, intracellular delivery of proteins remains challenging due to the inherent impermeability of the plasma membrane<sup>3,4</sup>. Thus, numerous strategies have been developed to facilitate intracellular protein delivery. For instance, the iTOP system exploits NaCl-mediated hyperosmolarity and transduction compounds to achieve high efficiency delivery of proteins into primary cells, but this has limited potential for *in vivo* applications<sup>5</sup>. Although Cell-penetrating peptides (CPPs) have also shown promise in some applications, the limitations related to endosomal entrapment and toxicity have been reported<sup>6–8</sup>. Finally, various nanocarriers, such as lipid nanoparticles and polymers, are frequently utilized for intracellular delivery of proteins, but with similar drawbacks to CPPs<sup>9–13</sup>.
+
+Another limitation of these strategies is that their synthetic properties induce various side effects when they are applied *in vivo*. Thus, one strategy to overcome these issues is to harness natural delivery vehicles – extracellular vesicles (EVs). EVs are lipid bilayer enclosed particles that are secreted and taken up by all cell types to mediate intercellular trafficking of biologically active molecules<sup>14–16</sup>. However, there are presently two main challenges that must be solved to achieve efficient intracellular delivery of proteins by EVs: i) enrichment of therapeutic proteins inside EVs in a soluble (or non-tethered), active form; and ii) endosomal escape of therapeutic proteins into the cytosol of the target cell.
+
+The enrichment of specific proteins inside of EVs has previously been achieved by fusing the target proteins to the cytoplasmic domain of EV-sorting proteins, such as CD63<sup>17</sup>. However, as these proteins remain bound to the EV membrane, this approach is not suitable for cytosolic delivery of soluble proteins. To address this limitation, several technologies introducing cleavable linker peptides between the target protein and EV-sorting domains, and the EXPLORs system have been developed<sup>18,19</sup>,<sup>20</sup>. Although several of the above technologies have achieved intracellular protein delivery by modifying EVs, these solutions were either dependent on extracellular conditions or required multiple components to be co-expressed. In the present study, we innovatively exploited the engineered self-cleaving mini-intein (intein) derived from *Mycobacterium tuberculosis (Mtu) recA*<sup>21</sup><em>,</em> to connect the EV-sorting protein with the cargo protein, facilitating liberation of soluble cargo protein from the EV-sorting protein inside the EV lumen<em>.</em> This intein comprises the first 110 and last 58 amino acids of the 441 amino acid *Mtu recA* intein, with four additional mutations (C1A, D24G, V67L, and D422G) introduced to enable C-terminal cleavage in a pH-sensitive manner at 37°C<sup>21</sup>.
+
+Since endocytosis is the most common mode of biomacromolecular uptake by cells, endosomal entrapment constitutes the primary barrier to the functional intracellular delivery of both therapeutic proteins and EVs<sup>22,23</sup>. Interestingly, fusogenic proteins derived from viruses have been found to mediate endosomal escape and facilitate the release of EV cargo into the cell cytosol<sup>24,25</sup>. In this study, we made use of the fusogenic protein, vesicular stomatitis virus G glycoprotein (VSV-G), as both an efficient endosomal escape activator and EV-sorting protein. Here, we have developed two systems that solve the aforementioned problems above in tandem to achieve high efficiency intracellular protein delivery harnessing engineered EVs (Fig. 1a). Altogether, this work demonstrates the great potential for EV-based therapeutic protein delivery.
+
+## Development of the VEDIC system for high efficiency intracellular protein delivery by EVs
+
+Traffic Light (TL) fluorescent Cre reporter cells were exploited to assess the potential of EVs for intracellular delivery of functional proteins (Fig.<span class="InternalRef" refid="Fig1">1</span>b)<sup>26</sup>. LoxP recombination by Cre results in the excision of the red fluorescent protein (RFP) DsRed which subsequently leads to permanent expression of green fluorescent protein (GFP) (Fig.<span class="InternalRef" refid="Fig1">1</span>b). EVs derived from Cre alone, CD63-Cre (Cre fused to CD63 to enhance EV-enrichment), CD63-Intein-Cre (self-cleaving protein intein was introduced between CD63 and Cre to liberate Cre from CD63 inside the EV lumen)<sup>21</sup> or Intein-Cre (no EV-sorting domain) overexpressing cells could not achieve any recombination in reporter cells analyzed by flow cytometry (Fig.<span class="InternalRef" refid="Fig1">1</span>c-e and Supplementary Fig. 1a,c,d). Next, a comprehensive screen of 40 human- and 2 virus-derived fusogenic proteins to identify efficient candidates was performed since fusogenic proteins were reported to enhance EV cargo endosomal escape<sup>27</sup>. While none of the human-derived fusogenic proteins induced Cre-mediated recombination, the fusogenic viral protein VSV-G significantly boosted Cre delivery after co-transfection with CD63-Intein-Cre, such that 66% and 98% of HeLa-TL and T47D-TL cells, respectively, expressed GFP two days after treatment with EVs (Fig.<span class="InternalRef" refid="Fig1">1</span>f and Supplementary Fig. 1b). Subsequently, we co-expressed VSV-G in the EV producing cells alongside our previously engineered constructs (Fig.<span class="InternalRef" refid="Fig1">1</span>c) and incubated isolated EVs with reporter cells for 48 hours (h). Of the conditions tested, only VSV-G plus CD63-Intein-Cre EVs achieved significant activity in reporter cells (Fig.<span class="InternalRef" refid="Fig1">1</span>g), indicating that the EV-sorting domain (CD63), self-cleaving protein (intein) and endosomal escape booster (VSV-G) were indispensable for the intracellular delivery of Cre by engineered EVs. We term this approach the VEDIC (VSV-G plus EV-Sorting Domain-Intein-Cargo) system.
+
+Next, different doses of the VEDIC EVs were added to reporter cells. We observed a clear pattern of dose-dependent recombination in B16F10-TL, HeLa-TL and T47D-TL cells (Fig.<span class="InternalRef" refid="Fig1">1</span>h,i, Supplementary Fig. 1e,f and Supplementary Fig. 2a). For VEDIC EVs, a time-lapse video obtained from live imaging of HeLa-TL recipient cells showed increasing recombination over time (Supplementary Video 1), while no GFP signal was observed in the absence of VSV-G (Supplementary Video 2). Furthermore, the protein level of Cre and VSV-G in T47D-TL and B16F10-TL cells was evaluated and showed a correlation with activation of the GFP signal in these cell lines (Fig.<span class="InternalRef" refid="Fig1">1</span>j and Supplementary Fig. 2c). Moreover, VEDIC EVs were applied to multiple hard-to-transfect cell lines, and we observed significant recombination 2 days after EV addition in all cases (Supplementary Fig. 1g-j and Supplementary Fig. 2b). Finally, CD63 was replaced with other known EV-sorting domains (CD81, CD9 and PTGFRN) and we found that addition of EVs isolated from cells transfected with these additional constructs resulted in significant recombination when combined with VSV-G (Supplementary Fig. 2d).
+
+In addition to adding engineered EVs to reporter cells, the EV-mediated recombination efficiency in co-culture was assessed using direct co-culture (co-culture), IBIDI co-culture µ-slide (IBIDI) and Transwell (Transwell) assays (Supplementary Fig. 3)<sup>29</sup>. For direct co-culture assays, EV-producing and reporter cells were cultured at different ratios and GFP levels were evaluated one day after incubation (Supplementary Fig. 3a). Here, VEDIC expressing cells enabled significant recombination in various reporter cells (Supplementary Fig. 3b-e). For IBIDI co-culture assays, reporter cells were seeded in the inner reservoir as shown in Supplementary Fig. 3f, and the feeder cells were seeded in the 8 surrounding reservoirs. After 4 days of incubation, co-culture with the VEDIC donor cells showed significant levels of GFP activation in the reporter cells (Supplementary Fig. 3g,j). For Transwell assays, the reporter cells were seeded in the lower compartment and the EV-producing cells in the upper chamber (Supplementary Fig. 3f). In line with the previous results, significant increases in the number of GFP positive cells was only observed after co-culture with the VEDIC donor cells, after 4 days of incubation (Supplementary Fig. 3h,i,k).
+
+## Development of the VFIC system to further improve intracellular protein delivery by EVs
+
+To further improve the VEDIC system, which requires transfection of several plasmids, we aimed to combine the essential components for intracellular EV cargo delivery into a single construct. To determine whether VSV-G itself could be employed as an efficient EV loading protein, VSV-G-mNG was transfected into HEK293T cells and the isolated vesicles were analyzed by single vesicle flow cytometry after the EVs were stained with a CD63-APC antibody. The total number of mNG + vesicles was far greater than CD63 + vesicles (Fig.<span class="InternalRef" refid="Fig2">2</span>a). Upon VSV-G-mNG co-transfection with CD63 plasmid in HEK293T cells, the total number of isolated mNG + vesicles was similar to that of CD63 + vesicles (Fig.<span class="InternalRef" refid="Fig2">2</span>b). These data imply that VSV-G could function as an EV-sorting domain, similar to CD63. Hence, a VSV-G-Intein-Cre fusion protein was constructed, with and without a foldon component that has been demonstrated to enhance VSV-G trimer formation and function<sup>30,31</sup>, and expression of constructs was confirmed (Supplementary Fig. 4d). Upon adding these EVs to reporter cells, a dose-dependent recombination in reporter cells was observed, and nearly 100% recombination in B16F10-TL and T47D-TL cells were detected in the high-dose treated groups (Fig.<span class="InternalRef" refid="Fig2">2</span>c,d and Supplementary Fig. 4a,b). The VSV-G-Foldon-Intein-Cre (VFIC) performed better than the VSV-G-Intein-Cre construct in low EV-dose experiments (Fig.<span class="InternalRef" refid="Fig2">2</span>c and Supplementary Fig. 4a), demonstrating the ability of the foldon domain to enhance VSV-G loading and/or function. Time-lapse fluorescence microscopy video analysis revealed similar recombination dynamics using VFIC EVs as was previously observed using the VEDIC system in HeLa-TL cells (Supplementary Video 3). VFIC EV-mediated recombination was also observed in hard-to-transfect cells (Fig.<span class="InternalRef" refid="Fig2">2</span>e-h and Supplementary Fig. 4c). We subsequently performed the direct co-culture, IBIDI, and Transwell assays using the VFIC system, and got significant recombination in recipient cells by transfer of Cre protein from donor cells (Supplementary Fig. 3k and Supplementary Fig. 4e-h).
+
+## The pH-sensitive intein performs C-terminal cleavage during EV-biogenesis
+
+To confirm that the intein we used in this study performed C-terminal cleavage in a pH-dependent manner, we introduced 2 mutants (Fig.<span class="InternalRef" refid="Fig3">3</span>a). The H439Q variant has a lower pH-sensitivity, which should lead to a lower cleavage rate during EV-biogenesis, since the pH in MVBs is around 6 (Fig.<span class="InternalRef" refid="Fig3">3</span>a, middle panel)<sup>33</sup>. In contrast, the N440A variant is supposed to abolish the C-terminal cleavage which should lead to minimal cleavage both in EV-producing cells and the isolated EVs (Fig.<span class="InternalRef" refid="Fig3">3</span>a, lower panel)<sup>34</sup>. As expected, the decrease of C-terminal cleavage to form released Cre cargo protein was corroborated by western blot for the VEDIC system (Fig.<span class="InternalRef" refid="Fig3">3</span>b). Accordingly, the functional assays (adding isolated EVs to recipient cells or direct co-culture with VEDIC producer cells and reporter cells) showed significantly decreased recombination in reporter cells when the 2 mutants were included in the VEDIC system (Supplementary Fig. 5a-e and Supplementary Fig. 6a). These results were corroborated for the VFIC system (Supplementary Fig. 5f-j and Supplementary Fig. 6b,c). Altogether, these findings support the notion that the intein used for the VEDIC and VFIC technologies enables robust C-terminal cleavage to form soluble cargo proteins in the EV-lumen during EV-biogenesis in a pH-dependent manner (Fig.<span class="InternalRef" refid="Fig1">1</span>d).
+
+## VSV-G boosts endosomal escape following receptor-mediated endocytosis in recipient cells
+
+To ascertain the role of VSV-G in endosomal escape and endocytosis, we introduced mutations that have been described to disrupt fusogenic (P127D) or LDL-R receptor binding capacity (K47Q), respectively (Fig.<span class="InternalRef" refid="Fig3">3</span>c,d)<sup>35–37</sup>. In order to assess EV uptake and trafficking, Cre was replaced by mNG in the VEDIC system, which was co-transfected with VSV-G to produce EVs that were added to Huh7 cells<sup>38</sup>. Confocal microscopy was used to evaluate the uptake of the vesicles 48 h after incubation and showed a punctate distribution of green fluorescent signal in recipient cells in the absence of VSV-G, indicating endosomal entrapment (Fig.<span class="InternalRef" refid="Fig3">3</span>e). However, when VSV-G was included, the mNG signal diffused into the cytosol (Fig.<span class="InternalRef" refid="Fig3">3</span>e), suggesting endosomal escape. Of note, when P127D was used, mNG instead showed a punctate distribution (Fig.<span class="InternalRef" refid="Fig3">3</span>f), confirming that endosomal escape was mediated by the fusogenic activity of VSV-G. In line with our hypothesis, mNG was furthermore observed in punctate form when a directly fused CD63-mNG, without intein to liberate cargo protein, was co-transfected with VSV-G (Fig.<span class="InternalRef" refid="Fig3">3</span>f).
+
+Next, the two mutant VSV-G constructs were used for Cre protein delivery. Decreased Cre protein delivery with the same doses of EVs in HeLa-TL recipient cells was observed using western blot analysis for both mutants compared to the wide-type VSV-G group, given that free Cre protein level was equal in these EV groups (Fig.<span class="InternalRef" refid="Fig3">3</span>g and Supplementary Fig. 7g)<sup>37</sup>. Utilization of the two VSV-G mutants in the VEDIC system showed significantly decreased Cre-mediated recombination in reporter cells with K47Q and complete loss of reporter activity with P127D after EV treatment, direct co-culture, IBIDI and Transwell assays in different reporter cells (Supplementary Fig. 7a-f). Furthermore, we reproduced these results using the VFIC system (Supplementary Fig. 8).
+
+### Robust gene editing by Cas9/sgRNA RNPs and meganuclease targeting PCSK9 using the VEDIC and VFIC systems
+
+Since Cre and mNG proteins were successfully delivered, we next tested whether more relevant therapeutic cargoes could be delivered. To this end, Cre was replaced with Cas9 to encapsulate Cas9-ribonucleoprotein (Cas9-RNP) into engineered EVs (Fig.<span class="InternalRef" refid="Fig4">4</span>a,b and Supplementary Fig. 9b). As a read-out for Cas9-RNP delivery, we employed CRISPR reporter (Stoplight, SL) cells that constitutively express mCherry followed by a short linker region that includes a sgRNA target region and a stop codon, followed by two eGFP open reading frames (+1nt and +2nt out of frame, respectively). Upon successful Cas9-RNP delivery, non-homology end joining (NHEJ)-mediated +1nt and +2nt frameshifts in the targeting linker region will result in bypassing of the stop codon and the permanent expression of eGFP. (Fig.<span class="InternalRef" refid="Fig4">4</span>c)<sup>39</sup>. Significant dose- and time-dependent genome editing was observed in cells treated with VEDIC and VFIC EVs (Fig.<span class="InternalRef" refid="Fig4">4</span>d,e and Supplementary Fig. 9a). For VFIC EVs, close to 80% gene editing efficiency was achieved, which is most likely the maximum achievable for this reporter construct (Fig.<span class="InternalRef" refid="Fig4">4</span>d)<sup>39</sup>.
+
+Next, VEDIC and VFIC constructs for the delivery of a previously described meganuclease targeting *PCSK9* were generated (Fig.<span class="InternalRef" refid="Fig4">4</span>f)<sup>40</sup>. Upon EV exposure to cells, a significant decrease of PCSK9 protein level was observed in a dose-dependent manner (Fig.<span class="InternalRef" refid="Fig4">4</span>g).
+
+### Cre-mediated recombination in melanoma-xenograft and R26-LSL-tdTomato reporter mice by VEDIC and VFIC systems after local injection
+
+Based on the promising results achieved *in vitro*, the *in vivo* applicability of VEDIC and VFIC-engineered EVs was assessed. To evaluate VEDIC and VFIC-mediated Cre delivery *in vivo*, an intratumoral (IT) injection of engineered EVs was conducted in C57BL/6 mice bearing subcutaneous B16F10-TL melanoma xenografts (Supplementary Fig. 10a). 4 days after injection, tumors were harvested for immunohistochemistry analysis, which showed significant GFP signals following treatment with VEDIC and VFIC EVs (Supplementary Fig. 10b). Next, R26-LSL-tdTomato reporter mice, whereby successful Cre delivery excises stop cassettes that are present between the CAG promoter and an ORF of the red fluorescent protein, tdTomato, were exploited for *in vivo* study. These mice were injected with VEDIC and VFIC EVs through intracerebroventricular (ICV) injection for one week to assess Cre delivery in the brain (Fig.<span class="InternalRef" refid="Fig5">5</span>a)<sup>26</sup>. As shown in Fig.<span class="InternalRef" refid="Fig5">5</span>b, tdTomato expression was detected in the cerebellum, cortex, and hippocampus of VEDIC- and VFIC EV treated mice. Moreover, we additionally observed trace amounts of tdTomato expression in the olfactory bulb and thalamus of VEDIC and VFIC treated animals (Supplementary Fig. 11a). To further identify which cell types internalized the engineered EVs, the slides were co-stained for both tdTomato and specific cell-marker genes. We found good colocalization of tdTomato with GFAP (astrocyte marker) and IBA1 (microglia marker), but only marginal colocalization of NeuN (neuronal marker) in the corpus callosum and hippocampus in the engineered EV-treated animals (Supplementary Fig. 11b-f). Based on the above results, we concluded that the engineered EVs mainly delivered their cargo to astrocytes and microglia in the brain.
+
+To enhance functional CNS delivery, we next repeated the ICV experiments using osmotic minipumps, allowing for chronic dosing of engineered EVs over 24 h (Fig.<span class="InternalRef" refid="Fig5">5</span>c). Strikingly, upon flow cytometry analysis of single cells we observed more than 40% and 30% of cells in hippocampus and cortex respectively were edited by Cre delivered via VEDIC system (Fig.<span class="InternalRef" refid="Fig5">5</span>d). In addition, close to 10% cells in cerebellum were also edited by VEDIC EVs.
+
+## Cre recombination in R26-LSL-tdTomato reporter mice following systemic VEDIC and VFIC EV-mediated Cre delivery
+
+Next, one pilot *ex vivo* study was performed by adding engineered EVs to liver primary cells harvested from R26-LSL-tdTomato reporter mice, which showed significant recombination of the cells by VEDIC and VFIC EVs (Supplementary Fig. 12a). Next, engineered EVs were administered via intraperitoneal (IP) injection into R26-LSL-tdTomato reporter mice. The liver, spleen and heart were harvested for analysis by immunofluorescence one week after injection (Fig.<span class="InternalRef" refid="Fig5">5</span>e). A substantial number of cells in the liver and spleen were found to be tdTomato positive following injection of both VEDIC and VFIC EVs, but not after injection of CD63-Intein-Cre EVs (Fig.<span class="InternalRef" refid="Fig5">5</span>f). In contrast, we did not observe any significant tdTomato expression in the EV hard-to-reach organ, heart (Supplementary Fig. 13b).
+
+After co-staining of tdTomato together with cell specific markers, a high degree of functional delivery to leukocyte (CD45), especially to the T cell population (CD3) and macrophages (F4/80), was detected (Fig.<span class="InternalRef" refid="Fig6">6</span>g,i and Supplementary Fig. 12c,d,e,g). In contrast, B cells (B220) showed low recombination events *in vivo* (Fig.<span class="InternalRef" refid="Fig5">5</span>h and Supplementary Fig. 12f).
+
+## Treatment of Lipopolysacharide (LPS)-induced systemic inflammation by VEDIC and VFIC-mediated delivery of super-repressor of NF-ĸB
+
+To demonstrate the applicability of our systems for the treatment of disease, we applied the VEDIC and VFIC EVs to treat lipopolysaccharide (LPS)-induced systemic inflammation by delivering a previously reported super-repressor of NF-ĸB activity (SR) (Supplementary Fig. 13a)<sup>41</sup>. To accomplish this, CD63-Intein-SR, VSV-G-Intein-SR and VSV-G-Foldon-Intein-SR constructs were generated, and HEK-Blue-NF-ĸB luciferase reporter cells were used as a read-out for functional *in vitro* assessment of the system (Fig.<span class="InternalRef" refid="Fig6">6</span>a,b and Supplementary Fig. 13d). In this reporter cell line, IĸB is degraded upon inflammatory stimuli such as LPS or TNF-α stimulation, which allows NF-ĸB to translocate to the nucleus and drive downstream luciferase reporter gene expression, since luciferase is driven by a minimal NF-ĸB promoter (Fig.<span class="InternalRef" refid="Fig6">6</span>c)<sup>42</sup>. First, we confirmed that the NF-ĸB SR, which is constitutionally active and maintains the NF-ĸB in the cytoplasm, inhibited the HEK-Blue-NF-ĸB luciferase reporter activation (Supplementary Fig. 13b,c)<sup>43</sup>. SR delivered by VSV-G-Foldon-Intein-SR, VSV-G-Intein-SR, and VSV-G + CD63-Intein-SR EVs successfully inhibited TNF-α-mediated NF-ĸB signaling activation evidenced by decreased reporter gene expression (Fig.<span class="InternalRef" refid="Fig6">6</span>d,e). This provided the rationale for testing these EVs in a murine LPS-induced model of systemic inflammation (Fig.<span class="InternalRef" refid="Fig6">6</span>f). Thus, engineered EVs were injected 4 h before and 6 h after injecting LPS to allow for, and enhance, the binding of EV-delivered SR to NF-ĸB, respectively. Subsequently, the weight and mortality of the mice were measured at 24 and 48 h after LPS treatment respectively. Compared to the PBS and CD63-Intein-SR injected groups, the body weight and survival of VSV-G + CD63-Intein-SR and VSV-G-Foldon-Intein-SR treated animals was significantly improved after 48 h (Fig.<span class="InternalRef" refid="Fig6">6</span>g,h). Histology of the liver revealed a decrease of inflammatory cells at the portal areas as well as a significant alleviation of the hydropic degeneration of hepatocytes by treatment with both VEDIC and VFIC engineered EVs (Fig.<span class="InternalRef" refid="Fig6">6</span>i). These results demonstrate the adaptability and therapeutic potential of the VEDIC and VFIC systems for the treatment of disease.
+
+# Conclusions
+
+By fine-tuning our engineering strategies, we have identified solutions to the major bottlenecks, the enrichment of free-form active cargo into EVs and their endosomal escape in recipient cells, to EV-mediated intracellular protein delivery. The resulting VEDIC and VFIC systems achieved an unprecedented level of efficiency for EV-mediated intracellular delivery of functional proteins *in vitro* and *in vivo*. Owing to the versatility of the VEDIC and VFIC systems, we anticipate that many other therapeutic proteins or protein complexes of interest could be efficiently delivered, apart from Cre, super-repressor of NF-ĸB and Cas9-RNPs explored in this work. As such, these approaches may hold great potential for the further therapeutic development.
+
+Importantly, more than 40% and 30% cells in hippocampus and cortex respectively were edited by Cre delivered via VEDIC system after osmotic pump ICV injection, the efficiency of which is even comparable to the traditional AAV-mediated protein delivery<sup>44,45</sup>. Encouraged by this, we anticipate the development of therapeutics for central nervous system (CNS) genetic diseases, such as Huntington's disease and spinal muscular atrophy<sup>46,47</sup>, through delivery of gene editing tools (CRISPR/Cas9 or base editors) to the brain using engineered EVs.
+
+Furthermore, we applied engineered EVs for the treatment of LPS-induced systemic inflammation, demonstrating that therapeutic levels of intracellular protein delivery were achieved by our systems *in vivo*. These observations demonstrate the therapeutic potential of these approaches, which shows exciting promise for the potential development of treatments for a wide array of pathologies, such as lysosomal storage diseases (LSDs) and enzymatic deficiencies<sup>48,49</sup>. To summarize, the VEDIC and VFIC systems developed in this study allow for robust intracellular functional delivery of therapeutic proteins, both *in vitro* and *in vivo*. In addition, the high genome editing efficiency achieved by Cas9-RNP delivery implies that this strategy may lead to potential applications in the treatment of genetic diseases.
+
+# References
+
+1. Messersmith, W. A. & Ahnen, D. J. Targeting EGFR in Colorectal Cancer. *N. Engl. J. Med.* **359**, 1834–1836 (2022).
+
+2. Reck, M. et al. Safety and patient-reported outcomes of atezolizumab plus chemotherapy with or without bevacizumab versus bevacizumab plus chemotherapy in non-small-cell lung cancer. *Journal of Clinical Oncology* **38**, 2530–2542 (2020).
+
+3. Stewart, M. P. et al. In vitro and ex vivo strategies for intracellular delivery. *Nature* **538**, 183–192 (2016).
+
+4. Goswami, R., Jeon, T., Nagaraj, H., Zhai, S. & Rotello, V. M. Accessing Intracellular Targets through Nanocarrier-Mediated Cytosolic Protein Delivery. *Trends Pharmacol Sci* **41**, 743–754 (2020).
+
+5. D’Astolfo, D. S. et al. Efficient intracellular delivery of native proteins. *Cell* **161**, 674–690 (2015).
+
+6. Fu, J., Yu, C., Li, L. & Yao, S. Q. Intracellular Delivery of Functional Proteins and Native Drugs by Cell-Penetrating Poly(disulfide)s. *J Am Chem Soc* **137**, 12153–12160 (2015).
+
+7. Steinauer, A. et al. HOPS-dependent endosomal fusion required for efficient cytosolic delivery of therapeutic peptides and small proteins. *Proc Natl Acad Sci U S A* **116**, 512–521 (2019).
+
+8. Wadia, J. S., Stan, R. V. & Dowdy, S. F. Transducible TAT-HA fusogenic peptide enhances escape of TAT-fusion proteins after lipid raft macropinocytosis. *Nat Med* **10**, 310–315 (2004).
+
+9. Boehnke, N. et al. Massively parallel pooled screening reveals genomic determinants of nanoparticle delivery. *Science (1979)* **377**, (2022).
+
+10. Chakravarty, P., Qian, W., El-Sayed, M. A. & Prausnitz, M. R. Delivery of molecules into cells using carbon nanoparticles activated by femtosecond laser pulses. *Nat Nanotechnol* **5**, 607–611 (2010).
+
+11. Liu, B., Ianosi-Irimie, M. & Thayumanavan, S. Reversible Click Chemistry for Ultrafast and Quantitative Formation of Protein-Polymer Nanoassembly and Intracellular Protein Delivery. *ACS Nano* **13**, 9408–9420 (2019).
+
+12. Yan, M. et al. A novel intracellular protein delivery platform based on single-protein nanocapsules. *Nat Nanotechnol* **5**, 48–53 (2010).
+
+13. Zuris, J. A. et al. Cationic lipid-mediated delivery of proteins enables efficient protein-based genome editing in vitro and in vivo. *Nat Biotechnol* **33**, 73–80 (2015).
+
+14. Kamerkar, S. et al. Exosomes facilitate therapeutic targeting of oncogenic KRAS in pancreatic cancer. *Nature* **546**, 498–503 (2017).
+
+15. Herrmann, I. K., Wood, M. J. A. & Fuhrmann, G. Extracellular vesicles as a next-generation drug delivery platform. *Nat Nanotechnol* **16**, 748–759 (2021).
+
+16. Mehanny, M., Lehr, C. M. & Fuhrmann, G. Extracellular vesicles as antigen carriers for novel vaccination avenues. *Adv Drug Deliv Rev* **173**, 164–180 (2021).
+
+17. Liang, X. et al. Extracellular vesicles engineered to bind albumin demonstrate extended circulation time and lymph node accumulation in mouse models. *J Extracell Vesicles* **11**, (2022).
+
+18. Banskota, S. et al. Engineered virus-like particles for efficient in vivo delivery of therapeutic proteins. *Cell* **185**, 250-265.e16 (2022).
+
+19. Somiya, M. & Kuroda, S. Reporter gene assay for membrane fusion of extracellular vesicles. *J Extracell Vesicles* **10**, (2021).
+
+20. Yim, N. et al. Exosome engineering for efficient intracellular delivery of soluble proteins using optically reversible protein-protein interaction module. *Nat Commun* **7**, 1–9 (2016).
+
+21. Wood, D. W., Wu, W., Belfort, G., Derbyshire, V. & Belfort, M. A genetic system yields self-cleaving inteins for bioseparations. *Nat Biotechnol* **17**, 889–892 (1999).
+
+22. Mulcahy, L. A., Pink, R. C. & Carter, D. R. F. Routes and mechanisms of extracellular vesicle uptake. *J Extracell Vesicles* **3**, 1–14 (2014).
+
+23. Gilleron, J. et al. Image-based analysis of lipid nanoparticle-mediated siRNA delivery, intracellular trafficking and endosomal escape. *Nat Biotechnol* **31**, 638–646 (2013).
+
+24. Zhang, X. et al. Programmable Extracellular Vesicles for Macromolecule Delivery and Genome Modifications. *Dev Cell* **55**, 784-801.e9 (2020).
+
+25. Albanese, M. et al. *MicroRNAs are minor constituents of extracellular vesicles that are rarely delivered to target cells*. *PLoS Genetics* vol. 17 (2021).
+
+26. Zomer, A., Steenbeek, S. C., Maynard, C. & Van Rheenen, J. Studying extracellular vesicle transfer by a Cre-loxP method. *Nat Protoc* **11**, 87–101 (2016).
+
+27. Mangeot, P. E. et al. Protein transfer into human cells by vsv-g-induced nanovesicles. *Molecular Therapy* **19**, 1656–1666 (2011).
+
+28. Görgens, A. et al. Optimisation of imaging flow cytometry for the analysis of single extracellular vesicles by using fluorescence-tagged vesicles as biological reference material. *J Extracell Vesicles* **00**, (2019).
+
+29. Zomer, A. et al. In vivo imaging reveals extracellular vesicle-mediated phenocopying of metastatic behavior. *Cell* **161**, 1046–1057 (2015).
+
+30. Vogel, A. B. et al. BNT162b vaccines protect rhesus macaques from SARS-CoV-2. *Nature* **592**, 283–289 (2021).
+
+31. Papanikolopoulou, K., Forge, V., Goeltz, P. & Mitraki, A. Formation of Highly Stable Chimeric Trimers by Fusion of an Adenovirus Fiber Shaft Fragment with the Foldon Domain of Bacteriophage T4 Fibritin. *Journal of Biological Chemistry* **279**, 8991–8998 (2004).
+
+32. Callanan, J. et al. Expansion of known ssRNA phage genomes: From tens to over a thousand. *Sci Adv* **6**, (2020).
+
+33. Chong, S. et al. Utilizing the C-terminal cleavage activity of a protein splicing element to purify recombinant proteins in a single chromatographic step. **26**, 5109–5115 (1998).
+
+34. Roey, P. Van et al. Crystallographic and Mutational Studies of Mycobacterium tuberculosis recA Mini-inteins Suggest a Pivotal Role for a Highly Conserved Aspartate Residue. *J Mol Bio.* **1**, 162–173 (2007).
+
+35. Votteler, J. et al. Designed proteins induce the formation of nanocage-containing extracellular vesicles. *Nature* **540**, 292–295 (2016).
+
+36. Fredericksen, B. L. & Whitt, M. A. Vesicular stomatitis virus glycoprotein mutations that affect membrane fusion activity and abolish virus infectivity. *J Virol* **69**, 1435–1443 (1995).
+
+37. Nikolic, J. et al. Structural basis for the recognition of LDL-receptor family members by VSV glycoprotein. *Nat Commun* **9**, 1–12 (2018).
+
+38. Wan, T. et al. Exosome-mediated delivery of Cas9 ribonucleoprotein complexes for tissue-specific gene therapy of liver diseases. *Sci Adv.* **2017**, 1–14 (2022).
+
+39. de Jong, O. G. et al. A CRISPR-Cas9-based reporter system for single-cell detection of extracellular vesicle-mediated functional transfer of RNA. *Nat Commun* **11**, 1–13 (2020).
+
+40. Wang, L. et al. Meganuclease targeting of PCSK9 in macaque liver leads to stable reduction in serum cholesterol. *Nat Biotechnol* **36**, 717–725 (2018).
+
+41. Schreiber, J., Efron, P. A., Park, J. E., Moldawer, L. L. & Barbul, A. Adenoviral gene transfer of an NF-κB super-repressor increases collagen deposition in rodent cutaneous wound healing. *Surgery* **138**, 940–946 (2005).
+
+42. Liu, T., Zhang, L., Joo, D. & Sun, S. C. NF-κB signaling in inflammation. *Signal Transduct Target Ther* **2**, (2017).
+
+43. Kim, S. et al. Exosome-based delivery of super-repressor IκBα ameliorates kidney ischemia-reperfusion injury. *Kidney Int* **100**, 570–584 (2021).
+
+44. Lang, J. F., Toulmin, S. A., Brida, K. L., Eisenlohr, L. C. & Davidson, B. L. Standard screening methods underreport. *Nat Commun* 1–10 doi:10.1038/s41467-019-11321-7.
+
+45. Kaspar, B. K. et al. Adeno-associated virus effectively mediates conditional gene modification in the brain. *Proc. Natl. Acad. Sci. U. S. A.* **99**, 2320-2325 (2002).
+
+46. Tu, Z. et al. Cas9-mediated replacement of expanded CAG repeats in a pig model of Huntington ’ s disease. *Nat. Biomed. Eng.* **7**, 629-646 (2023).
+
+47. To, L. & Editor, T. H. E. Base editing-mediated splicing correction therapy for spinal muscular atrophy. *Cell Res.* **30**, 548-550 (2020).
+
+48. Platt, F. M., d’Azzo, A., Davidson, B. L., Neufeld, E. F. & Tifft, C. J. Lysosomal storage diseases. *Nat Rev Dis Primers* **4**, (2018).
+
+49. Burton, B. K., Kar, S. & Kirkpatrick, P. Sapropterin. *Nat Rev Drug Discov* **7**, 199–200 (2008).
+
+# Methods
+
+Materials: see the Supplementary Table 1 key reagents resources.
+
+## Cell lines
+
+HEK293T cells used to produce the functional EVs in this study were maintained in Dulbecco’s Modified Eagle Medium (DMEM) (high glucose) supplemented with 10% fetal bovine serum (FBS) (Gibco, USA) and 1% Antibiotic-Antimycotic (Anti-anti) (Gibco, USA). Cells were cultured at 37℃ in a humidified air atmosphere containing 5% CO₂. The reporter cell lines (Hela-TL, T47D-TL, B16F10-TL, Raw264.7-TL, HEK-Blue-NF-ĸB and HEK293T-SL) were cultured using the same medium and the same conditions as HEK293T cells. THP-1-TL, K562-TL and MSC-TL cells were cultivated in Roswell Park Memorial Institute (RPMI) 1640 medium (high glucose) supplemented with 10% FBS and 1% Anti-anti.
+
+## Mice model (Intra-tumor injection)
+
+C57BL/6 mice (5 weeks of age, 20 g body weight) were acclimated to their new surroundings at least one week before the experiment. B16F10-TL cells resuspended in PBS were inoculated subcutaneously into the mice at a density of 0.5 million cells per mouse. Ten days after inoculation when obvious tumors were formed, engineered EVs were injected directly into the tumors. The injected volume was 50 µL per mouse with 7.5×10¹⁰ EVs. Four days after intra-tumor injection of EVs, the mice were sacrificed, and tumors were harvested and fixed in PFA. Tumor tissues were stained immunohistochemically for GFP expression while tissues in lysis buffer were homogenised using a tissue lyser machine.
+
+## Mice model (LPS-induced inflammation)
+
+C57BL/6 mice (5 weeks of age, 20 g body weight) were acclimated to their new surroundings at least one week before the experiment. Animals were injected (IP) with engineered EVs 4 h before IP injection of LPS (Sigma, USA) at the dose of 7.5 mg/Kg. Six hours after LPS induction, engineered EVs were IP injected once more to boost the intracellular delivery of the protein cargos by EVs. The survival rate and body weight of the mice with LPS induction were recorded for 2 days. 48 hours after LPS induction, the mice were euthanized and sacrificed, and main organs, such as liver, were harvested and fixed with PFA. H&E (haematoxylin and eosin) staining was performed to check the extent of damage to the organs induced by LPS. The damage of the tissues was evaluated by a professional pathologist.
+
+## Mice model (ICV delivery of EVs via osmotic pump)
+
+For a constant intracerebroventricular delivery of the EV preparations, Alzet® osmotic pumps (2001D) were used. These pumps were filled with 200 µl of the EV preparation at a concentration of 4x10¹² EVs/ml. The osmotic pumps were prepared and implanted as described by Sanchez-Mendoza and colleagues. In short, 15 weeks old R26-LSL-tdTomato reporter mice (B6.Cg-Gt(ROSA)26Sor<sup>tm9(CAG-tdTomato)Hze</sup>/J (Rosa26.tdTomato)) reporter mice were anesthetized with isoflurane and mounted on a stereotactic frame. A constant body temperature of 37 °C was maintained using a heating pad. Next, a small incision in the skin was made from the base of the neck to up in between the eyes. The injection needle of the pump system was placed in the ventricles based on coordinates measured relative to the bregma intersection (anteroposterior 0.07 cm, mediolateral 0.1 cm, dorsoventral 0.2 cm). The cannula was fixed to the skull with Loctite 454. A subcutaneous pocket for the osmotic pump was made by using a blunted scissor to slide underneath the skin at the base of the neck. Next the pumps were inserted under the skin at the base of the neck and pushed to the back as far as was possible without resistance. The osmotic pump was connected to the cannula via a vinyl catheter (0.71 mm outer diameter). An immediate constant delivery of the EV preparations started upon implantation of the pumps at a flow rate of 8 µl/hour. After surgery the incision was sutured and 48 hours after the implantation, the pumps were removed, and bone wax was used to close the skull again.
+
+## Construct generation
+
+All the transgenes, except the ones purchased from Addgene and the human-derived fusogens, were ordered from IDT (Integrated DNA Technologies, USA). The transgenes were first cloned into the pLEX vector backbone using EcoRI and XhoI sites. The constructs used in this study were then generated from the ordered fragments through restriction enzyme digestion and subsequent self-ligation. VSV-G-Intein-Cre and VSVG-Foldon-Cre were generated from the VSV-G-Foldon-Intein-Cre construct using Kpn2I and MluI, respectively. For VSV-G-Cre, VSV-G-Intein-Cre was digested with MluI. CD63-Intein-RS was generated by digesting VSV-G-Foldon-Intein-RS and CD63-Intein-Cre with BamHI and Xhol, and then inserting RS into the resulting CD63-Intein vector. Similarly, to synthesize CD63-Intein-Cas9 and VSV-G-Foldon-Intein-Cas9 constructs, CD63-Intein-Cre and VSV-G-Foldon-Intein-Cre were digested with BamHI and Xhol, and Cas9 inserted into CD63-Intein and VSV-G-Foldon-Intein, respectively. The CD63-Intein-M1 and VSV-G-Foldon-Intein-M1 constructs were generated by replacing Cre with M1 (M1 PCSK9 meganuclease) using BamHI and Xhol digestion. The 40 human-derived fusogens were ordered from Twist (Twist Bioscience) and the vector used was pTwist CMV BetaGlobin.
+
+## Plasmid transfection
+
+HEK293T cells were seeded into 15-cm dishes, with the numbers of dishes decided according to the amount of EVs to be used in indicated experiments. Polyethylenimine (PEI, Polysciences) was utilized for the transfection of plasmid/s according to the protocol provided by the manufacturer. The ratio of PEI to plasmid was 2:1 in this study. For single plasmid transfection, 30 µg plasmid was used for each plate. For co-transfection of 2 plasmids, 20 µg of each plasmid was used while for 3 plasmids, 15 µg of each plasmid was used.
+
+## EV production
+
+EVs were produced by transient transfection of the transgenes using polyethylenimine. HEK293T cells were seeded into 15-cm dishes at a density of 5 million cells per dish using complete DMEM medium. After 2 days, the cells were transfected with the transgenes and the medium changed to Opti-MEM (Gibco, USA) with 1% Anti-anti 6 h post-transfection. After 48 h, the conditioned medium (CM) was collected and centrifuged (700 x g for 5 min followed by 2,000 x g for 10 min). The supernatant was then filtered through a 0.22 µm filter system.
+
+## EV isolation
+
+Tangential flow filtration (TFF, MicroKross, 20 cm², Spectrum labs) was used to isolate EVs from the filtered CM. Particles greater than the 300 kDa cutoff of the TFF were retained in the system and concentrated. These particles were further concentrated using Amicon Ultra-15 100 kDa (Millipore) spin filters, which were centrifuged at 4,000 x g for 30 min to several hours at 4°C, depending on the amount EVs in the samples. Lastly, the concentrated EVs were collected in maxirecovery 1.5 ml Eppendorf tubes (Axygene, USA) and quantified using Nanoparticle Tracking Analysis (NTA).
+
+## Nanoparticle Tracking Analysis (NTA)
+
+EV samples were diluted with freshly 0.22 µm-filtered PBS before checking the particle sizes and concentrations using the NanoSight NS500 instrument. Five videos of more than 30 second durations each were taken at the camera level of 15 in light scatter mode. All the samples were analysed with the same setting using the NTA 2.3 software.
+
+## Traditional flow cytometry
+
+After the different traffic-light reporter cells were added into EVs at different time points, or after the reporter cells were co-cultured with EV-producing cells for 24 h, GFP expression was quantified using the MACSQuant Analyzer 10 flow cytometer (Miltenyi Biotec, Germany). Briefly, the cells in 96-well plates were washed with PBS once and trypsinized for 5 min at 37°C. The trypsin was then neutralized using cell medium supplemented with 10% FBS. After adding DAPI to check the cell viability, the cells were sampled by the MACSQuant analyzer using the settings of one specific reporter cell line for all the measurements. The FlowJo software (version 10.6.2) was used to calculate the percentage of GFP positive cells.
+
+## Single-vesicle flow cytometry
+
+HEK293T cells were either transfected with VSV-G-mNeonGreen construct only or co-transfected with VSV-G-mNeonGreen and CD63. Six hours post-transfection, the medium was changed to Opti-MEM medium (Gibco, USA) with 1% Anti-anti. After 2 days, the medium was harvested and centrifuged at 700 x g for 5 min, followed by 2,000 x g for 10 min and subsequently filtered through a 0.22 µm filter system. In a v-bottom 96-well plate, 25 µl of each sample was incubated with APC-labelled CD63 antibody (Miltenyi Biotec, Germany; 1 nM per well) overnight under dark conditions. The samples were then diluted 1000 times and transferred into an R-bottom 96-well plate. Amnis® CellStream instrument (Luminex, US) was utilized to evaluate the engineered EVs at single-vesicle level. The collected data was then analysed using FlowJo software (version 10.6.2).
+
+## EV-addition assays in reporter cells
+
+The reporter cells used in this study were seeded into 96-well plates at the following densities: 1×10⁴ (Hela-TL), 2×10⁴ (T47D-TL), 1.5×10⁴ (B16F10-TL), 5×10⁴ (Raw264.7-TL), 5×10⁴ (THP-1-TL), 5×10⁴ (K562-TL), 1×10⁴ (MSC-TL), 8×10³ (HEK-SL) and 2×10⁴ (HEK-Blue-NF-ĸB) cells per well. The following day, different doses of EVs were added directly into each of the reporter cells, except for HEK-Blue-NF-ĸB. The doses of the EVs used and the time for incubation are indicated in each Fig.. GFP positive cells were confirmed either by fluorescent microscopy or by MACSQuant flow cytometry. For HEK-Blue-NF-ĸB cells, the RS EVs were added directly into the wells for 48 hours and the stimulation (TNF-α, 10 ng/ml) added after another 6 hours. Luciferase signals from the cell lysate were evaluated using the GloMax® 96 Microplate Luminometer machine (Promega, USA).
+
+## Virus production
+
+The transgenes were subcloned into our lentiviral vectors (Transfer plasmid, 22.5 µg/T175 flask) which were co-transfected with pCD/NL-BH (Helper plasmid, 22.5 µg/T175 flask) and pcoPE01 (Envelope plasmid, 3.5 µg/T175 flask) into HEK293T cells and incubated overnight. The next morning, cell medium was changed to complete DMEM medium (with 10% FBS and 1% Anti-anti) supplemented with sodium butyrate (Sigma-Aldrich). After 6 to 8 hours, the sodium butyrate containing DMEM medium was changed back to complete DMEM medium without additional chemicals. Nalgene® Oak Ridge Centrifuge Tubes (Thermo Scientific) were used for harvesting viruses 22 to 24 hours after incubation. Briefly, the virus-particle-containing medium was collected and filtered using a 0.45 µm syringe filter (VWR), and then centrifuged at 25,000 x g for 90 min at 4°C. The supernatant was aspirated, and freshly prepared medium (IMDM with 20%FBS) was used to resuspend the virus pellets. The viruses were added directly into the target cells or stored at -80°C for long-term use.
+
+## Stable reporter cell generation
+
+HeLa, T47D, B16F10, Raw264.7, THP-1, K562, and MSC cells were seeded into 6-well plates and the viruses were added into the cells the next day. Titration of viruses was done using 3 doses: 2 µl, 10 µl, and 50 µl per well. After one day incubation of the viruses with target cells, the medium was changed back to normal complete medium (DMEM + 10% FBS + 1% Anti-anti for B16F10 and Raw264.7 cells; RPMI-1640 + 10% FBS + 1% Anti-anti for THP-1, K562, and MSC cells). Two days after virus transduction, the cells were trypsinized and resuspended in fresh medium. Resistance selection was performed by adding puromycin (2 µg/ml for B16F10 and MSC cells, 4 µg/ml for THP-1, K562, HeLa, and T47D cells, and 6 µg/ml for Raw264.7 cells). Untransduced cells died from the puromycin whereas successfully transduced cells survived and continued to grow. The cells were passaged under puromycin selection for approximately one week before the cells were utilized for downstream experiments.
+
+## Direct co-culture of EV-producing cells with reporter cells
+
+HEK293T cells were seeded into a 6-well plate at a density of 0.5 million cells per well. The next day when the cells reached 60-70% confluence, corresponding constructs were transfected into the wells using Lipofectamine2000 (Invitrogen, USA) according to the manufacturer’s protocol. Six hours after transfection, the medium was changed to fresh medium (DMEM + 10%FBS + 1% Anti-anti) to reduce the toxicity of the Lipofectamine 2000. Twenty-four hours after plasmid transfection, the cells were trypsinized, counted, and mixed with the corresponding reporter cells at ratios of 1:1 or 1:5, or other ratios as indicated in the Fig.s (ratio=EV-producing cells: reporter cells) in a 96-well plate. After co-culturing for 24 h, the cells were trypsinized and measured using the MACSQuant flow cytometer to check the percentage of GFP positive cells.
+
+## IBIDI co-culture µ-slide assay
+
+HEK293T cells were seeded into a 6-well plate at a density of 5×10⁵ cells per well. The day after, indicated constructs were transfected into cells using Lipofectamine 2000 (Invitrogen, USA) according to the manufacturer’s protocol. To avoid the toxicity of the Lipofectamine2000 on the HEK293T cells, the medium was changed to fresh complete medium (DMEM + 10%FBS + 1% Anti-anti) after 6 h. The following day, the transfected cells were trypsinized and counted. The transfected cells (feeder cells or EV-producing cells) were seeded into the surrounding reservoirs of the ibidi µ-Slide while the recipient cells (traffic-light reporter cells) were added to the central reservoir, following cell numbers indicated in the Fig.s. The volume of medium used for each reservoir was 40 µl. Once the cells were attached to the bottom, another 400 µl of complete medium (DMEM + 10%FBS + 1% Anti-anti) was added into the slide slowly and carefully to immerse the walls between the central reservoir and the surrounding reservoirs such that cell-cell communication could be mediated by engineered EVs. Four days later, the GFP positive cells were measured using either a fluorescent microscope or the MACSQuant flow cytometer.
+
+## Transwell co-culture assay
+
+Similar to the IBIDI assay, HEK293T cells were seeded into a 6-well plate for 24 h and then transfected with the indicated constructs. The medium was changed to fresh complete medium after 6 h of transfection. One day after transfection, the cells were trypsinized and counted, and the EV-producing cells were added to the top chamber of the transwell system (pore size=0.4 µm) while the reporter cells were seeded at the bottom. After 4 days of cell-cell communication by engineered EVs, a fluorescent microscope or MACSQuant flow cytometer was used to check for GFP positive cells.
+
+## Dynamic live imaging assay
+
+Huh7 cells were plated 1 day before the experiment in a polymer-bottom cell culture plate (Ibidi, cat. No. 82426), with 5×10⁴ cells per well. 5×10¹⁰ EVs were added to the cells 3 hours prior imaging and Hoechst dye for nucleus staining was added just before the live cell imaging.
+
+Confocal images were acquired on a Nikon C2 + confocal microscope equipped with an oil-immersion 60x objective with numerical aperture 1.4 (Nikon Instruments, Amsterdam, The Netherlands). The sample was excited and detected with appropriate excitation laser lines and emission filters and the fluorophores were imaged sequentially. The images were taken every hour over the course of 72 hours. The corresponding videos were generated by using the Nikon NIS-Elements Imaging Software.
+
+## Confocal microscopy
+
+Huh7 cells were seeded in polymer-bottom cell culture plates (Ibidi, cat. no. 82426) one day before the experiment, similar to the dynamic live imaging assay. Indicated number of engineered EVs were added to the cells one day after seeding. After adding EVs for 48 hours, Hoechst dye was added before confocal microscopy imaging. The confocal images were taken the same way as described in dynamic live imaging assay. Image processing was performed by using Fiji software.
+
+## Fluorescent microscopy
+
+After addition of EVs, co-culture, IBIDI, and Transwell assays, the GFP positive cells were visualized under a fluorescent microscope. We chose the area for taking pictures randomly and set up the same parameters for the groups using one experiment. All the images were then processed with the same parameters in the machine or using the Fiji software.
+
+## Western blot analysis
+
+Whole cell protein was isolated using RIPA buffer supplemented with a protease inhibitor cocktail, mixed with sample buffer (4×), and heated at 70℃ for 10 min. For EV samples, 1×10¹⁰ EVs were mixed with sample buffer (4×) and heated at 70℃ for 10 min. Samples were then loaded onto a NuPAGE™ 4-12% Bis-Tris Protein Gel (Thermo Scientific) and ran at 120 V for 2 h in NuPAGE™ MES SDS running buffer (Thermo Scientific). Proteins were transferred from the gel to the membrane using iBlot™ 2 Transfer Stacks (Thermo Scientific). The membrane was blocked with Intercept™ blocking buffer (LI-COR Biosciences) for 1 h at room temperature in a shaker after which it was incubated with primary antibodies overnight at 4℃. The membrane was washed with TBS-T buffer three times for 5 min each and incubated with corresponding secondary antibodies for 1 h at room temperature in a shaker. After washing with TBS-T buffer three times and with PBS once, the membrane was scanned using the Odyssey infrared imaging system (LI-COR).
+
+## IHC staining for melanoma tissues
+
+Tissue sections were fixed at 65℃ for 1 hour before the slides were subjected to deparaffinization and rehydration as follows: Xylene for 20 min, 100% ethanol for 3 min twice, 95% ethanol for 3 min, 70% ethanol for 3 min, and then 50% ethanol for 3 min. Afterwards, the slides were rinsed in running cold tap water for 5 min followed by antigen retrieval using citrate buffer, pH 6.0 (Sigma). After antigen retrieval, the slides were washed with PBS three times for 5 min each and immersed in blocking buffer for 30 min at 37℃. The slides were then incubated with primary anti-GFP antibody (Abcam, ab290, 1:200 dilution) overnight after blocking. The following day, after washing the slides with PBS three times for 5 min each, the slides were incubated with secondary antibody Goat Anti Rabbit IgG H&L (Alexa Fluor® 488) (Abcam, ab150077, 1:500 dilution) for 30 min at 37℃, followed by washing with PBS three times for 5 min each. The slides were mounted using ProLong™ Diamond Antifade Mountant with DAPI (Thermo Scientific) and sealed with nail polish. Images were taken using a confocal microscope (Nikon, Japan).
+
+## IHC staining for tissues from Cre-LoxP R26-LSL-tdTomato reporter mice
+
+Tissue sections (5 µm) were made from organs of ICV and IP injected R26-LSL-tdTomato reporter mice and then were deparaffinized in xylene and ethanol, boiled in citrate buffer for 20 min, and blocked with 5% goat serum in PBS-T (PBS containing 0.3% Triton X-100) solution for 1 h at room temperature. The sections were then stained with primary antibodies in blocking buffer at 4°C overnight. After washing with PBS, sections were stained with appropriate fluorophore-conjugated secondary antibodies in PBS or PBS containing 0.1% Triton X-100 for 1 to 2 h before washing and mounting. A Zeiss LSM780 confocal microscope or Zeiss Axioscan Z.1 was used for imaging.
+
+## Statistics
+
+Statistical tests for the biological replicates used in this study are reported in each Fig. legend. GraphPad software was utilized for the statistical analysis and the data presented as + SD. Two-tailed T-test was used for the comparisons of two individual groups. One-way ANOVA or Two-way ANOVA multiple comparisons test was used for the analysis of the multiple groups. Log-rank (Mantel-Cox) test was used for the survival comparisons. Statistical significance was set up as * p < 0.05, ** p < 0.01; *** p < 0.001; **** p < 0.0001; ns: non-significant.
+
+## Data availability
+
+All the data are available in the manuscript or in the supplementary information, and all the raw data from this study are available from the corresponding authors upon request. Materials are available upon signing the material transfer agreement (MTA) submitted to S.E.-A. and Evox Therapeutics Limited, Oxford, United Kingdom.
+
+# Supplementary Files
+
+- [SupplementaryInformation.docx](https://assets-eu.researchsquare.com/files/rs-3329019/v1/bc2cd6077f20aa762bb936ea.docx)
+
+- [SupplementaryVideo1.VEDICEVgroup..mov](https://assets-eu.researchsquare.com/files/rs-3329019/v1/4b5c37dc139f1aa990f15953.mov)  
+  Dynamic GFP expression in HeLa-TL cells by VEDIC EVs.
+
+- [SupplementaryVideo2.ControlEVgroup..mov](https://assets-eu.researchsquare.com/files/rs-3329019/v1/2043eef19773ea4036a33fb3.mov)  
+  Dynamic GFP expression in HeLa-TL cells by control EVs.
+
+- [SupplementaryVideo3.VFICEVgroup..mov](https://assets-eu.researchsquare.com/files/rs-3329019/v1/4f8ae89c84d2a4df28b1ab97.mov)  
+  Dynamic GFP expression in HeLa-TL cells by VFIC EVs.

1860c1110668d418e560d07a0d1e24d403746fd4d1d4179bfdca1a17997bdf06/preprint/images_list.json

@@ -0,0 +1,34 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpeg",
+        "caption": "Geology of Ritchey crater. A: Ritchey overview, CTX image mosaic (see Methods). B: Eastern crater rim area of Ritchey, showing the smooth and dark-toned sheet unit. HiRISE grayscale\u00a0 and enhanced color images (ESP_011846_1515, ESP_012914_1515) on CTX basemap. Cyan area: regions of interest (ROI) used to extract average CRISM spectra in Fig. 3. Inset 1 at bottom left shows the increasingly brecciated texture toward the margin of the sheet unit. C: Close-up view of the sheet unit margin. Note the brecciated slope/escarpment of the sheet unit (see Fig. 2C for 3-dimensional view). White arrows: fragmented breccia underlying the sheet unit. Yellow arrows: NE-SW ridges in the light-toned basement unit. Inset 1 shows decameter-long vertical structures (black arrows) across the escarpment of the sheet unit resemble degassing pipes33 or hydrothermal channels34 in melt-rich breccia on Earth. Inset 2 shows a carbonate-bearing vein (red arrows) related to the fault zone, see Fig. 3 ROI d for spectrum. D: Comparison of the brecciated escarpment of the sheet unit (see Fig. 2A for 3D view) and a smooth escarpment of bedrock. The breccia and altered bedrock units are stratigraphically confined between the sheet unit to the NW and dark-toned underlying bedrock unit in the SE. E: cross section of the impactite stratigraphy, see E-E\u2019 in Fig 1B.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpeg",
+        "caption": "Comparison of the impactite stratigraphy between Ritchey (A and C) and terrestrial craters (B and D). A and C: 3D view of the sheet unit overlying the fragmented breccia unit at the eastern rim of the Ritchey crater, HiRISE image on HRSC DEM (architected from Fig. 1C/D). B and D: the impactite stratigraphy at Aum\u00fchle outcrop, Ries crater (B, Germany) and Mistastin lake impact structure (D, Canada). Photo D by Cassandra Marion in 2021 at Cot\u00e9 Creek ( https://craterexplorer.ca/mistastin-impact-crater/).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpeg",
+        "caption": "Details of the sheet unit, ridged unit and ejecta, and spectra for ROIs. A: The continuous cauliflower-like lobates/wrinkles and flow features (black arrows) of the sheet unit at the northern rim of the Ritchey, HiRISE image (ESP_013125_1520_RED). See Fig S1B and S1D for more lobate textures for fluidized evidence of sheet unit. B: Close-up view of the box area at Fig 3A from HiRISE IRB (ESP_013125_1520_IRB) color image. The yellow to copper gold material indicates the sheet unit, whereas the basement rocks are white and at lower topographies. The black arrows indicate the pitted area on the sheet unit, similar to the pitted materials described by35, 36. C: Ridged unit at the western rim, HiRISE image PSP_007363_1510_RED. Green area indicates the ROI for spectral analysis. Boxed area shows the brecciated fabrics at the slope of a ridge. D: altered basement rocks that contain carbonates at the eastern inner rim (PSP_009090_1515_RED). White arrows: ridged area; black arrow: putative mound structure. E: continuous ejecta that overlies the ridged unit (ESP_011846_1515_RED). White arrow: ridge, black arrow: blocks and breccia embedded at the slope/escarpment of the ridge. Spectra a and g resemble basaltic glass37. F: Occurrences of serpentine (cyan ROIs) nearby the ridged unit at the southern crater rim (ESP_017371_1505_RED). White arrows: ridged surfaces. G: CRISM ratio spectra from numbered ROI annotations in Figures. Black spectra are from standard minerals in the NASA RELAB database38 (http://www.planetary.brown.edu./relab/, see Methods).\u00a0",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpeg",
+        "caption": "Schematic model (based on50) of the early post-impact phase with active groundwater hydrothermal percolation (A and B) and alteration processes at microscale. A: Early post-impact phase with a hot central uplift and melt sheet covering the impact crater. The temperature of the central uplift is high enough to only allow steaming, instead of hot water circulation and spring discharge. Aqueous percolation started from the crater wall via the basement fractures once the temperature cooled down to below the boiling point. B: Cross-section of melt-sheet induced hydrothermal alteration. The heat of the melt sheet was transmitted downward to melt the cryosphere and alter the water-saturated basement rocks. C: Low-temperature, alkaline fluid altering an olivine-rich rock similar to Ritchey basement. The hydrothermal vein on Earth may preserve potential biosignature at a microscale65 in 1: microstromatolite; 2: carbonate minerals (such as dolomite); 3: mineralized fungi.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

1860c1110668d418e560d07a0d1e24d403746fd4d1d4179bfdca1a17997bdf06/preprint/preprint.md

@@ -0,0 +1,149 @@
+# Abstract
+
+Impact-induced hydrothermal systems have the potential to support long-lived aqueous environments throughout the history of Mars, but their nature and distribution are not well-understood. We reported impact-induced alteration within a well-preserved impactite stratigraphy across the inner rim of Ritchey crater. The impactite stratigraphy is characterized by a smooth and uniform sheet unit overlying fragmented breccia, consistent with emplacement of impact melt rocks on ballistic ejecta deposits, as observed in the typical double layer impactites of complex craters on Earth. CRISM hyperspectral data revealed a series of alteration minerals including serpentine, chlorite, Mg carbonate in fractured bedrock, veins, and fragmented breccia in erosional windows underneath the unaltered sheet unit. These alteration minerals neither correlate with post-impact fluvio-lacustrine facies, nor with the pre-impact target bedrock. Instead, the formation of these alteration minerals is most plausibly triggered by the emplacement of hot impact melt, facilitating percolation of groundwater reacting with the extensively fractured and permeable bedrocks at the crater rims. The ubiquitous distribution of the alteration minerals on the inner rim indicates habitable environments triggered by impact cratering and potential preservation of biosignatures in the precipitated minerals.
+
+Earth and environmental sciences/Planetary science/Mineralogy  
+Earth and environmental sciences/Planetary science/Geochemistry  
+Earth and environmental sciences/Planetary science/Asteroids, comets and Kuiper belt  
+Earth and environmental sciences/Planetary science/Petrology
+
+# Introduction
+
+In hypervelocity impact events, substantial kinetic energy of the impactor undergoes conversion into internal energy, where shock-induced high pressures and temperatures generate impact melt<sup>1, 2</sup>. On water-bearing or icy planets<sup>3, 4</sup>, and even on asteroids<sup>5</sup>, the ensuing propagation of heat and water after meteoric impact culminates in the development of impact-induced hydrothermal systems, which have the potential to support diverse forms of life<sup>6</sup>.
+
+Numerical simulations have also predicted impact-generated hydrothermal systems on Mars driven by heat from impact melt, triggering basin-wide groundwater percolation<sup>7, 8, 9</sup> and subsequent chemical alteration<sup>10, 11</sup>. These sites potentially fulfill important requirements for habitability, i.e., liquid water, nutrients and energy<sup>3, 12, 13</sup>. Moreover, potential biosignatures on Mars, if present, were likely to be preserved in chemically precipitated minerals (e.g., carbonates and silica) from various settings, particularly in association with hydrothermal activity<sup>3, 14, 15</sup>. However, the extent, mechanism and distribution of such hydrothermal systems in impact craters on Mars are unclear. Most previous research has searched for signatures of hydrothermal alteration within central peaks, where a series of alteration minerals associated with vein-<sup>16, 17</sup> or mound-like features<sup>18, 19</sup> have been identified in orbital datasets. Less commonly, alteration minerals at crater rims were detected in many craters<sup>20, 21, 22</sup> including Jezero crater<sup>23</sup>, but no clear relationship with the impact or impact-related hydrothermal alteration was established. As the Mars 2020 Perseverance rover is approaching the crater rim of Jezero, it is crucial to scrutinize the origin of alteration minerals for potential links to post-impact hydrothermal systems. One challenge of making this link is that ancient Martian craters like Jezero are often heavily eroded, so interpreting the alteration and impactite stratigraphy is difficult. A sufficiently large, young, and well-preserved impact crater would reveal more about the nature of impact-induced hydrothermal alterations within crater rims.
+
+Ritchey crater is located ~ 200 km south of Valles Marineris and is a complex crater 78 km in diameter, with exceptionally well-preserved central uplift, terraces, rim, and ejecta. Crater counting on the Ritchey ejecta yields a retention age of 3.46 Ga (Early Hesperian<sup>21</sup>). At the crater center, a smooth, coherent layer and a dark, rough unit were speculated to be a preserved impact melt deposit<sup>21</sup>. Although the crater floor is covered by aeolian deposits, the inner-rim exhibits well-exposed and intact bedrock. Fluvial channels and related fan deposits emanating from the wall indicate at least some post-impact aqueous activity occurred<sup>21</sup>.
+
+The well-preserved state of Ritchey crater offers a unique opportunity to study impactite-alteration stratigraphy on the rim for large, complex craters on Mars. Here we use Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) hyperspectral visible/near-infrared<sup>24</sup> and High Resolution Imaging Science Experiment (HiRISE<sup>25</sup>) images to identify alteration minerals, impactite stratigraphy, and post-impact geological processes on the rim of Ritchey, in order to investigate their relationship to impact-induced hydrothermal activity. Through these observations, we seek to provide an important framework to understand alteration processes and astrobiological potential of martian crater rims.
+
+# 1 A well-preserved impactite stratigraphy
+
+On the inner rim of Ritchey crater, the most common stratigraphic sequence presents as light-toned bedrock overlain by a fragmented breccia unit (Fig. 1 C and D), and capped by a darker-toned sheet-like unit (Fig. 1 C and D). Outcrops of light-toned massive bedrock at the base of the sequence exhibit a ridged texture, with smooth escarpments along their slope (Fig. 1 D). The fragmented breccia unit contains large boulders (up to ~70 meters) (Fig. 2 A). On the top, the sheet unit forms a ten- to decameters-thick coherent layer that can be traced laterally across several km (Fig. 1 B). With a homogeneous and smooth surface, the sheet unit is increasingly brecciated towards the margin, embedded with smaller blocks (usually <10 meters) than the underlying fragmented breccia unit (Fig. 1 C, D, 2 A, and 2 C). Some light-toned, elongated, and vertical pipe-like structures are observed inside the sheet unit (Fig. 1 C).
+
+We hypothesize that the light-toned bedrock, brecciated outcrops, and smooth sheet units correspond to altered bedrock, continuous ejecta, and melt-rich breccia, respectively. The same stratigraphic sequence of melt-rich breccia (e.g., suevite) conformingly overlying continuous ejecta is an important feature (Fig. 3 A and C) in double-layered impactites typically produced by complex craters on Earth 26, 27. This unique impactite stratigraphy has been verified from well-preserved terrestrial impact craters, such as Ries (Fig. 3 B), Mistastin Lake (Fig. 3 D) 28 and many others 27. Similar sheet-like melt-rich breccia units have been identified at Hargraves crater 29 and Negril crater 30 on Mars and in the Orientale Basin on the Moon 31. A lobate apron of the sheet unit (northern rimwall, Fig. 3 A and B) is consistent with convergent returning flow 32 and/or movement along the steep wall area during its emplacement and cooling.
+
+At its marginal slope/escarpment, the sheet unit sometimes grades to a rough-surfaced pitted unit (Fig. 2 A and 2 C) or a ridged unit (Fig. 1 C, 3 C and 3 F) that overlies the light-toned basement rocks. The pitted unit is a less regular and coherent, patchy material (Fig. 3 B), comprising closely-spaced, decameter- to hundred-meter sized pits that are reminiscent of “soap froth” networks in other Martian craters 35, 36. Blocks and brecciated fabrics are embedded at the interior walls of the pits (Fig S2G). The rims of the pits are not surrounded by ejecta, so unlikely to be impact craters. Between the pitted unit and the basement, a ridged unit is present, showing elongated, near-linear ridge features (Fig. 3 C, 3 F) and brecciated fabrics at the side slopes (Fig. 2 C).
+
+Pitted units in other Martian craters are hypothesized to result from degassing of volatile-rich components after the emplacement of the melt sheet 35, 36. At Ritchey, the presence of elongated vertical features (Fig. 1 C) within the sheet unit bears a resemblance to degassing pipes observed in the Ries suevite 33 and hydrothermal veins in the Chicxulub suevite 34. Such additional morphological evidence supports the interpretation of the sheet unit as the impact melt-rich breccia.
+
+While dark-toned smooth mantling units could be produced by a variety of processes, the morphology and distribution of the Ritchey sheet unit are most consistent with an impact melt sheet. Although lava flows are abundant in the Thaumasia Planum region, the complete, ramparted rim of Ritchey (Fig. 1 A) has protected the crater from infill by externally-sourced flows. In addition, the sheet unit possesses a consistently brecciated slope/escarpment, whereas the slope of the nearby mafic basement is indurated, intact and smooth (Fig. 1 D). Hence, it is unlikely that the sheet unit covering all azimuths of the crater rim is a volcanic flow deposit.
+
+### 2 Spectral properties of the impactites
+
+CRISM spectra of typical basement rocks in the Ritchey crater rim show strong pyroxene absorption bands (0.9–1.0 µm, 1.9–2.0 µm) most consistent with low Ca-pyroxene (LCP), with no associated alteration mineral detections<sup>21</sup>. The sheet unit also shows a broad absorption centered near 0.9 and 1.9 µm consistent with LCP, but with a shoulder near 1.2 µm potentially due to mixing with olivine or glass<sup>37</sup> (Fig. <span class="InternalRef" refid="Fig3">3</span> G, spectra a and g). The sheet unit appears to be largely unaltered, as no hydration (e.g. 1.9 and 1.4 µm) bands and only a weak, possible Al/Si-OH (2.21 µm) band is consistently detected (Fig. <span class="InternalRef" refid="Fig3">3</span> G, spectrum a; Fig S2D and F). The pitted unit has similar spectral features (Fig S2F) as the sheet unit. In contrast, the fragmented breccia does not show pyroxene bands, and instead resembles olivine or glass with a broad band centered >1.0 µm, and a weak possible band at 1.9 µm, suggesting minimal hydration (Fig. <span class="InternalRef" refid="Fig3">3</span> G, spectrum e).
+
+While none of these observations independently and uniquely identify glass within the impactite stratigraphy, taken together they are consistent with a composition that is some combination of basement-derived LCP, glass, and glass recrystallized to LCP. The weak 2.21 µm hydroxylated absorption in the sheet unit commonly results from Al-OH or Si-OH bond (or both), and is one of the most typical features of glassy, silicate-rich impact melt such as the Ries suevite<sup>33</sup>—the type rock of impact-melt breccia on Earth<sup>39</sup>. The wide 0.9 and 1.9 µm LCP bands in the sheet unit (spectrum g in Fig. <span class="InternalRef" refid="Fig3">3</span> G) show some asymmetries that could be consistent with glass mixing, although the wide 1.9 µm band of crystallized LCP dominates over glassy basaltic materials. Judging from the depth and width of 0.9 and 1.9 µm LCP bands in the sheet unit, laboratory quantitative mixing experiments suggest the LCP component comprises at most 10–15% of the LCP-glass mixture<sup>23</sup>. A mixed composition of LCP and glass is also plausible for impact melt-bearing breccia, since significant recrystallization (devitrification) of the target material may have occurred when the melt material is cooling<sup>40</sup>. Using this logic, homogeneous resistant bedrock units with LCP-dominated spectra associated with large impact basins elsewhere on Mars have also been inferred to indicate partially recrystallized impact melt<sup>41</sup>.
+
+### 3 Alteration minerals and impact-induced hydrothermalism
+
+Strong signatures of alteration are detected in the ridged unit and altered bedrock. A clear 2.29–2.31 band and weak 1.39–1.42 band in the ridged unit suggest the presence of Fe/Mg-bearing phyllosilicates such as nontronite and saponite (Fig. 3 G, spectra c and d). These results were consistent with smectite-dominated spectra previously detected in the crater wall area<sup>21</sup>. Many outcrops also show evidence for Mg-carbonate based on an additional 2.51–2.53 band (e.g., Fig. 3 G, spectra c, d and f).
+
+Other localized alteration mineral signatures are also present in the crater rim. Serpentine was identified in the ridged network at the southern rim (Fig. 3 F) according to a diagnostic 2.12 band, along with 2.3 and 2.5 Fe/Mg-OH bands (Fig. 3 G, spectrum f). Furthermore, chlorite was detected based on 2.25, 2.00 and 2.3 band absorptions (Fig. 3 G, spectrum h) at the southern rim (Fig. 1 A). At the northern rim, possible prehnite was identified adjacent to a carbonate outcrop (Fig S2), based on the diagnostic 1.48 absorption and a 2.35 -OH band drop. Compared to the crater rims, the central uplift of Ritchey shows more limited diversity of alteration minerals, including Fe/Mg smectites, chlorite and hydrated silica, whereas serpentines and carbonates were not detected<sup>21</sup>.
+
+Both detailed spectral analysis (Fig. 3 G) and spectral parameter maps (Fig S2D, S2F) show a distinct lack of alteration in the sheet unit and breccia, which presents a stark contrast to the pervasive alteration detected in the underlying ridged unit and altered bedrock. The absence of hydration in the sheet unit indicates little-to-no aqueous alteration within the melt-rich breccia itself, possibly due to dehydration during emplacement of the melt, and only weak alteration in the underlying breccia unit (Fig. 3 E). Instead, most phyllosilicates and carbonate signatures are exposed within erosional windows on the ridged unit and altered bedrock (Fig. 3 C, 3 D).
+
+We hypothesize that phyllosilicate-carbonate assemblages beneath the impactite stratigraphy formed via hydrothermal alteration driven by the contact with the hot melt-rich breccia. Some bedrock outcrops exhibit carbonate-rich vein fillings in surface fractures (Fig. 1 C, see Fig. 3 G for spectrum), resembling the carbonate veins in the hydrothermally altered ultramafic rocks on Earth (e.g., listwanite in Oman<sup>42</sup>). Similar alkaline hydrothermal process at fractured mafic/ultramafic areas would also produce serpentine, saponite, chlorite and carbonate (and prehnite), indicating a wide range of temperatures and water-rock ratios<sup>11, 43, 44, 45, 46</sup>.
+
+Fluvio-lacustrine processes might have also led to the formation of phyllosilicates. However, although fluvial channels have dissected the entire impactite sequence (e.g., Fig. 3 A), there is no morphological evidence of fluvio-lacustrine origin for phyllosilicates or carbonates at the crater wall, such as bedded or layered features. Nor in the ejecta of Ritchey is there any clear indication of the same phyllosilicates-carbonate signatures detected in the rim, so these minerals are unlikely to be sourced from pre-impact bedrock (Fig S3). Thus, we suggest the alteration minerals at the crater rim of the Ritchey are most likely post-impact in-situ alteration products.
+
+### 4 Widespread habitable environments around the crater rim
+
+The preserved impactite and alteration stratigraphy (Fig S2A and S2B) at the rim of Ritchey only capture a fraction of the original hydrothermal activities triggered by the impact melt sheet. The initial volume of impact melt correlates with crater diameter, and can be calculated according to scaling laws<sup>47, 48</sup>. For a crater like Ritchey, models predict the production of ~ 70–80 km<sup>3</sup> melt<sup>49</sup>, with a thickness estimated between 333 and 500 meters in the deposited melt sheet<sup>8</sup>, much thicker than its present form (up to decameters thick). As the major heat source<sup>8, 9</sup>, the initial melt sheet of Ritchey likely possessed a greater volume than presently observed (Fig S2A), resulting in a basin-wide hydrothermal alteration across the crater rim (Fig S2B).
+
+The hydrothermal alteration minerals in the rim of Ritchey crater (Fig. <span class="InternalRef" refid="Fig3">3</span>) are largely associated with the ridged units and altered bedrock in close contact with the sheet units (Fig. <span class="InternalRef" refid="Fig3">3</span> and Fig. <span class="InternalRef" refid="Fig4">4</span>), suggesting that hydrothermal fluid flow was concentrated in these units. For complex impact craters, the crater wall commonly encompasses extensive normal and concentric faults due to shockwave propagation and gravitational collapse<sup>1, 50</sup>. The widespread, heavily-altered ridged units may represent extensional fault zones<sup>41, 51</sup>, which would enhance the permeability of the shallow crust and facilitate groundwater percolation<sup>52</sup>. Additionally, the degassing pipes in the melt-rich breccia and carbonate vein filling at the altered basement rock (Fig. <span class="InternalRef" refid="Fig1">1</span> C) further support convective hydrothermal percolation between the impact melt, the underlying ejecta and bedrocks. Permeability studies of terrestrial melt-rich breccia samples also reveal a low matrix permeability, suggesting that fluid flow through the melt-rich breccia itself would have been more limited<sup>53</sup> than in the fractured basement bedrocks.
+
+We hypothesize that the diverse alteration minerals in Ritchey track the evolution of the aqueous system following cooling of the impact melt. The impact melt glass can form at a temperature higher than 1870°C<sup>54</sup>. During the emplacement of the melt sheet, the incorporation of cooler fragmented materials into the melt-rich breccia on the rim leads to a lower temperature of > 750–900°C compared to the central peak and basin-filling melt sheet<sup>27, 55, 56</sup>. As the melt is cooling down, hydrothermal alteration would have driven a series of low-temperature alteration reactions below 400°C: (<span citationid="CR1" class="CitationRef">1</span>) Chlorite-related alteration between 300 to 400°C (e.g, chlorite-bearing assemblage at Sudbury impact structure on Earth<sup>57</sup>); (<span citationid="CR2" class="CitationRef">2</span>) Serpentinization (< 150 to 400°C), one of the most important reactions for astrobiological interests because it mobilizes critical elements and delivers chemicals (such as hydrogen and methane) to support critical microbial metabolism similar to the earliest biogeochemical systems on Earth<sup>58</sup>; (<span citationid="CR3" class="CitationRef">3</span>) low temperature (< 100°C) carbonation to form saponite and carbonate<sup>43</sup>.
+
+Such subsurface, low temperature hydrothermal sites at crater rims could have provided potential habitable environments and fueled the metabolisms of specific thermophiles/hyperthermophiles<sup>3</sup> or chemoautotrophs<sup>59</sup>. On Earth, colonizers in impact-induced hydrothermal systems include, but are not limited to prokaryotes such as sulfate reducing bacteria<sup>60, 61</sup> and cyanobacteria<sup>62</sup>, or even some microeukaryotes like fungi<sup>63</sup> (Fig. <span class="InternalRef" refid="Fig4">4</span>). Minerals like hydrated silica and carbonate precipitated from hydrothermal fluids could preserve biosignatures from these environments<sup>15</sup> (Fig. <span class="InternalRef" refid="Fig4">4</span>). Notably, the Perseverance rover on the Mars 2020 mission will soon begin exploring and collecting samples from the rim of Jezero, another large Noachian crater on Mars that likely also produced hydrothermal systems. Rock samples from impact hydrothermal deposits would be high priority as they would have high biosignature preservation potential<sup>64</sup>.
+
+# Methods
+
+Visible images and topography  
+Landform investigation was mainly based on the enhanced false-colour greyscale (RED) and IRB (infrared-blue) HiRISE (High Resolution Imaging Experiment<sup>25</sup>) visible images up to 25 cm/pixel. The 20 m/pixel Context Camera mosaic<sup>66</sup> (see also <span class="ExternalRef"><span class="RefSource">https://murray-lab.caltech.edu/CTX/</span><span address="https://murray-lab.caltech.edu/CTX/" class="RefTarget" targettype="URL"></span></span>) was also used as a lower-resolution base-map to substantiate the areas where HiRISE are not covered. Digital elevation model (DEM) data up to 20m/pixel were acquired from HRSC (High Resolution Stereo Camera<sup>67</sup>), supplemented by the ~300m/pixel-resolution MOLA (Mars Orbiter Laser Altimeter<sup>68</sup>) DEM data. A three-dimensional terrain model was generated using the commercial software ArcGIS Pro (Esri).
+
+CRISM data  
+Determination of mineralogy was carried out using CRISM visible/near-infrared (~0.3–2.6 µm) hyperspectral image data (Compact Reconnaissance Imaging Spectrometer for Mars<sup>24, 69</sup>) at the most recently calibrated MTRDR level (Mapped Targeted Reduced Data Records<sup>70, 71</sup>). Eight FRT (full resolution targeted, 18 m/pixel) images and two HRL (half resolution long, 36 m/pixel) images were analysed using the CAT (CRISM analysis toolkit) in ENVI. A set of spectral parameters containing spectral reflectance at diagnostic wavelengths is compiled in the delivered MTRDR data and these parameters are combined as RGB components to highlight particular minerals<sup>70</sup> such as carbonates (CR2 browse product) and phyllosilicates with Fe and Mg (PFM browse product), as shown in Fig<span class="InternalRef" refid="MOESM1">S1</span>. All parameters in selected browse products were viewed using a linear stretch of 50–98% to enhance spectral features and suppress noises. Average reflectance spectra were extracted from each Region of Interest (ROI) containing several tens to thousands of pixels (Supplementary Information 2). These spectra were then ratioed against spectrally neutral, bland areas at similar elevations (Supplementary Information 2) to further suppress atmospheric and instrumental effects. Spectra were compared to standard laboratory spectra from NASA RELAB facility at Brown University: magnesite (F1CC06B by Jack Mustard), saponite (CASA58, Edward A. Cloutis), serpentine (LALZ01, Takahiro Hiroi), chlorite (LACL14, Takahiro Hiroi) and prehnite (LAZE03, Edward A. Cloutis), and published results (glass-pyroxene mixtures from<sup>72</sup>). Additional mineral mapping for ejecta was carried out using the newest released version of CRISM multispectral reduced data records (MRDR version 4<sup>71</sup>).
+
+# References
+
+1. Melosh HJ. Impact cratering: A geologic process. New York: Oxford University Press; Oxford: Clarendon Press 1989.
+2. Pierazzo E, Melosh HJ. Melt Production in Oblique Impacts. Icarus 2000, 145 (1) : 252-261.
+3. Nisbet EG, Sleep NH. The habitat and nature of early life. Nature 2001, 409 (6823) : 1083-1091.
+4. Nisbet EG, Sleep NH. 1 - The physical setting for early life. In: Rothschild LJ, Lister AM (eds). Evolution on Planet Earth. Academic Press: London, 2003, pp 3-24.
+5. Yasui M, Tazawa T, Hashimoto R, Arakawa M, Ogawa K. Impacts may provide heat for aqueous alteration and organic solid formation on asteroid parent bodies. Communications Earth & Environment 2021, 2 (1) : 95.
+6. Osinski GR, Cockell CS, Pontefract A, Sapers HM. The Role of Meteorite Impacts in the Origin of Life. Astrobiology 2020, 20 (9) : 1121-1149.
+7. Rathbun JA, Squyres SW. Hydrothermal Systems Associated with Martian Impact Craters. Icarus 2002, 157 (2) : 362-372.
+8. Abramov O, Kring DA. Impact-induced hydrothermal activity on early Mars. Journal of Geophysical Research: Planets 2005, 110 (E12).
+9. Ivanov BA, Pierazzo E. Impact cratering in H2O-bearing targets on Mars: Thermal field under craters as starting conditions for hydrothermal activity. Meteoritics & Planetary Science 2011, 46 (4) : 601-619.
+10. Naumov MV. Principal features of impact-generated hydrothermal circulation systems: mineralogical and geochemical evidence. Geofluids 2005, 5 (3) : 165-184.
+11. Schwenzer SP, Kring DA. Impact-generated hydrothermal systems capable of forming phyllosilicates on Noachian Mars. Geology 2009, 37 (12) : 1091-1094.
+12. Cockell CS. Trajectories of Martian Habitability. Astrobiology 2014, 14 (2) : 182-203.
+13. Cockell CS, Bush T, Bryce C, Direito S, Fox-Powell M, Harrison JP, et al. Habitability: A Review. Astrobiology 2016, 16 (1) : 89-117.
+14. Bosak T, Moore KR, Gong J, Grotzinger JP. Searching for biosignatures in sedimentary rocks from early Earth and Mars. Nature Reviews Earth & Environment 2021, 2 (7) : 490-506.
+15. Hays LE, Graham HV, Des Marais DJ, Hausrath EM, Horgan B, McCollom TM, et al. Biosignature Preservation and Detection in Mars Analog Environments. Astrobiology 2017, 17 (4) : 363-400.
+16. Carrozzo FG, Di Achille G, Salese F, Altieri F, Bellucci G. Geology and mineralogy of the Auki Crater, Tyrrhena Terra, Mars: A possible post impact-induced hydrothermal system. Icarus 2017, 281: 228-239.
+17. Michalski JR, Onstott TC, Mojzsis SJ, Mustard J, Chan QHS, Niles PB, et al. The Martian subsurface as a potential window into the origin of life. Nature Geoscience 2018, 11 (1) : 21-26.
+18. Marzo GA, Davila AF, Tornabene LL, Dohm JM, Fairén AG, Gross C, et al. Evidence for Hesperian impact-induced hydrothermalism on Mars. Icarus 2010, 208 (2) : 667-683.
+19. Turner SMR, Bridges JC, Grebby S, Ehlmann BL. Hydrothermal activity recorded in post Noachian-aged impact craters on Mars. Journal of Geophysical Research: Planets 2016, 121 (4) : 608-625.
+20. Ehlmann BL, Mustard JF, Murchie SL, Bibring J-P, Meunier A, Fraeman AA, et al. Subsurface water and clay mineral formation during the early history of Mars. Nature 2011, 479 (7371) : 53-60.
+21. Sun VZ, Milliken RE. The geology and mineralogy of Ritchey crater, Mars: Evidence for post-Noachian clay formation. Journal of Geophysical Research: Planets 2014, 119 (4) : 810-836.
+22. Pan L, Ehlmann BL, Carter J, Ernst CM. The stratigraphy and history of Mars' northern lowlands through mineralogy of impact craters: A comprehensive survey. Journal of Geophysical Research: Planets 2017, 122 (9) : 1824-1854.
+23. Horgan BHN, Anderson RB, Dromart G, Amador ES, Rice MS. The mineral diversity of Jezero crater: Evidence for possible lacustrine carbonates on Mars. Icarus 2020, 339: 113526.
+24. Murchie S, Arvidson R, Bedini P, Beisser K, Bibring JP, Bishop J, et al. Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) on Mars Reconnaissance Orbiter (MRO). Journal of Geophysical Research: Planets 2007, 112 (E5).
+25. McEwen AS, Eliason EM, Bergstrom JW, Bridges NT, Hansen CJ, Delamere WA, et al. Mars Reconnaissance Orbiter's High Resolution Imaging Science Experiment (HiRISE). Journal of Geophysical Research: Planets 2007, 112 (E5).
+26. Newsom HE. Hydrothermal alteration of impact melt sheets with implications for Mars. Icarus 1980, 44 (1) : 207-216.
+27. Osinski GR, Tornabene LL, Grieve RA. Impact ejecta emplacement on terrestrial planets. Earth and Planetary Science Letters 2011, 310 (3-4) : 167-181.
+28. Mader MM, Osinski GR. Impactites of the Mistastin Lake impact structure: Insights into impact ejecta emplacement. Meteoritics & Planetary Science 2018, 53 (12) : 2492-2518.
+29. Sacks LE, Tornabene LL, Osinski GR, Sopoco R. Hargraves Crater, Mars: Insights into the internal structure of layered ejecta deposits. Icarus 2022, 375: 114854.
+30. Pilles EA, Tornabene LL, Osinski GR. Impactite dykes in impact crater central uplifts: Insights from Negril crater, Mars. Icarus 2021, 355: 114153.
+31. Spudis PD, Martin DJP, Kramer G. Geology and composition of the Orientale Basin impact melt sheet. Journal of Geophysical Research: Planets 2014, 119 (1) : 19-29.
+32. Siegert S, Branney MJ, Hecht L. Density current origin of a melt-bearing impact ejecta blanket (Ries suevite, Germany). Geology 2017, 45 (9) : 855-858.
+33. Caudill C, Osinski GR, Greenberger RN, Tornabene LL, Longstaffe FJ, Flemming RL, et al. Origin of the degassing pipes at the Ries impact structure and implications for impact-induced alteration on Mars and other planetary bodies. Meteoritics & Planetary Science 2021, 56 (2) : 404-422.
+34. Kring DA, Tikoo SM, Schmieder M, Riller U, Rebolledo-Vieyra M, Simpson SL, et al. Probing the hydrothermal system of the Chicxulub impact crater. Science Advances 2020, 6 (22) : eaaz3053.
+35. Boyce JM, Wilson L, Mouginis-Mark PJ, Hamilton CW, Tornabene LL. Origin of small pits in martian impact craters. Icarus 2012, 221 (1) : 262-275.
+36. Tornabene LL, Osinski GR, McEwen AS, Boyce JM, Bray VJ, Caudill CM, et al. Widespread crater-related pitted materials on Mars: Further evidence for the role of target volatiles during the impact process. Icarus 2012, 220 (2) : 348-368.
+37. Cannon KM, Mustard JF. Preserved glass-rich impactites on Mars. Geology 2015, 43 (7) : 635-638.
+38. Pieters CM. Strength of mineral absorption features in the transmitted component of near-infrared reflected light: First results from RELAB. Journal of Geophysical Research: Solid Earth 1983, 88 (B11) : 9534-9544.
+39. Stöffler D, Artemieva NA, Wünnemann K, Reimold WU, Jacob J, Hansen BK, et al. Ries crater and suevite revisited—Observations and modeling Part I: Observations. Meteoritics & Planetary Science 2013, 48 (4) : 515-589.
+40. Reimold WU, Hansen BK, Jacob J, Artemieva NA, Wünnemann K, Meyer C. Petrography of the impact breccias of the Enkingen (SUBO 18) drill core, southern Ries crater, Germany: New estimate of impact melt volume. GSA Bulletin 2012, 124 (1-2) : 104-132.
+41. Scheller EL, Ehlmann BL. Composition, Stratigraphy, and Geological History of the Noachian Basement Surrounding the Isidis Impact Basin. Journal of Geophysical Research: Planets 2020, 125 (7) : e2019JE006190.
+42. Rajendran S, Nasir S, Kusky TM, Ghulam A, Gabr S, El-Ghali MAK. Detection of hydrothermal mineralized zones associated with listwaenites in Central Oman using ASTER data. Ore Geology Reviews 2013, 53: 470-488.
+43. Kelemen PB, Hirth G. Reaction-driven cracking during retrograde metamorphism: Olivine hydration and carbonation. Earth and Planetary Science Letters 2012, 345-348: 81-89.
+44. Ehlmann BL, Mustard JF, Murchie SL, Poulet F, Bishop JL, Brown AJ, et al. Orbital Identification of Carbonate-Bearing Rocks on Mars. Science 2008, 322 (5909) : 1828-1832.
+45. Ehlmann BL, Mustard JF, Murchie SL. Geologic setting of serpentine deposits on Mars. Geophysical Research Letters 2010, 37 (6).
+46. Aupart C, Morales L, Godard M, Jamtveit B. Seismic faults triggered early stage serpentinization of peridotites from the Samail Ophiolite, Oman. Earth and Planetary Science Letters 2021, 574: 117137.
+47. Grieve RAF, Cintala MJ. An analysis of differential impact melt-crater scaling and implications for the terrestrial impact record. Meteoritics 1992, 27 (5) : 526-538.
+48. Cintala MJ, Grieve RAF. Scaling impact melting and crater dimensions: Implications for the lunar cratering record. Meteoritics & Planetary Science 1998, 33 (4) : 889-912.
+49. Abramov O, Wong SM, Kring DA. Differential melt scaling for oblique impacts on terrestrial planets. Icarus 2012, 218 (2) : 906-916.
+50. Kenkmann T, Poelchau MH, Wulf G. Structural geology of impact craters. Journal of Structural Geology 2014, 62: 156-182.
+51. Head JW, Mustard JF. Breccia dikes and crater-related faults in impact craters on Mars: Erosion and exposure on the floor of a crater 75 km in diameter at the dichotomy boundary. Meteoritics & Planetary Science 2006, 41 (10) : 1675-1690.
+52. Bense VF, Gleeson T, Loveless SE, Bour O, Scibek J. Fault zone hydrogeology. Earth-Science Reviews 2013, 127: 171-192.
+53. Parnell J, Taylor CW, Thackrey S, Osinski GR, Lee P. Permeability data for impact breccias imply focussed hydrothermal fluid flow. Journal of Geochemical Exploration 2010, 106 (1) : 171-175.
+54. El Goresy A. Baddeleyite and its significance in impact glasses. Journal of Geophysical Research (1896-1977) 1965, 70 (14) : 3453-3456.
+55. Engelhardt W, Arndt J, Fecker B, Pankau H. Suevite breccia from the Ries crater, Germany: Origin, cooling history and devitrification of impact glasses. Meteoritics 1995, 30 (3) : 279-293.
+56. Osinski GR, Tornabene LL, Banerjee NR, Cockell CS, Flemming R, Izawa MRM, et al. Impact-generated hydrothermal systems on Earth and Mars. Icarus 2013, 224 (2) : 347-363.
+57. Molnár F, Watkinson DH, Jones PC. Multiple Hydrothermal Processes in Footwall Units of the North Range, Sudbury Igneous Complex, Canada, and Implications for the Genesis of Vein-Type Cu-Ni-PGE Deposits. Economic Geology 2001, 96 (7) : 1645-1670.
+58. Russell MJ, Hall AJ, Martin W. Serpentinization as a source of energy at the origin of life. Geobiology 2010, 8 (5) : 355-371.
+59. Brazelton WJ, Schrenk MO, Kelley DS, Baross JA. Methane- and Sulfur-Metabolizing Microbial Communities Dominate the Lost City Hydrothermal Field Ecosystem. Applied and Environmental Microbiology 2006, 72 (9) : 6257-6270.
+60. Parnell J, Boyce A, Thackrey S, Muirhead D, Lindgren P, Mason C, et al. Sulfur isotope signatures for rapid colonization of an impact crater by thermophilic microbes. Geology 2010, 38 (3) : 271-274.
+61. Parnell J, Boyce AJ, Osinski GR, Izawa MRM, Banerjee N, Flemming R, et al. Evidence for life in the isotopic analysis of surface sulphates in the Haughton impact structure, and potential application on Mars. International Journal of Astrobiology 2012, 11 (2) : 93-101.
+62. Schaefer B, Grice K, Coolen MJL, Summons RE, Cui X, Bauersachs T, et al. Microbial life in the nascent Chicxulub crater. Geology 2020, 48 (4) : 328-332.
+63. Drake H, Ivarsson M, Heim C, Snoeyenbos-West O, Bengtson S, Belivanova V, et al. Fossilized anaerobic and possibly methanogenesis-fueling fungi identified deep within the Siljan impact structure, Sweden. Communications Earth & Environment 2021, 2 (1) : 34.
+64. Beaty DW, Grady MM, McSween HY, Sefton-Nash E, Carrier BL, Altieri F, et al. The potential science and engineering value of samples delivered to Earth by Mars sample return. Meteoritics & Planetary Science 2019, 54 (S1) : S3-S152.
+65. Lima-Zaloumis, J., Neubeck, A., Ivarsson, M., Bose, M., Greenberger, R., Templeton, A.S., et al. Microbial biosignature preservation in carbonated serpentine from the Samail Ophiolite, Oman. Communications Earth & Environment 2022, 3, 231.
+66. Malin MC, Bell Iii JF, Cantor BA, Caplinger MA, Calvin WM, Clancy RT, et al. Context Camera Investigation on board the Mars Reconnaissance Orbiter. Journal of Geophysical Research: Planets 2007, 112 (E5).
+67. Neukum G, Jaumann R, Team tHC-IaE. HRSC: The high resolution stereo camera of Mars Express. In: Mars Express: the scientific payload Ed by Andrew Wilson, scientific coordination: Agustin Chicarro ESA SP-1240, Noordwijk, Netherlands: ESA Publications Division, ISBN 92-9092-556-6, 2004, p 17-35 2004, 1240: 17-35.
+68. Smith DE, Zuber MT, Frey HV, Garvin JB, Head JW, Muhleman DO, et al. Mars Orbiter Laser Altimeter: Experiment summary after the first year of global mapping of Mars. Journal of Geophysical Research: Planets 2001, 106 (E10) : 23689-23722.
+69. Murchie SL, Seelos FP, Hash CD, Humm DC, Malaret E, McGovern JA, et al. Compact Reconnaissance Imaging Spectrometer for Mars investigation and data set from the Mars Reconnaissance Orbiter's primary science phase. Journal of Geophysical Research: Planets 2009, 114 (E2).
+70. Viviano CE, Seelos FP, Murchie SL, Kahn EG, Seelos KD, Taylor HW, et al. Revised CRISM spectral parameters and summary products based on the currently detected mineral diversity on Mars. Journal of Geophysical Research: Planets 2014, 119 (6) : 1403-1431.
+71. Seelos FP, Seelos KD, Murchie SL, Novak MAM, Hash CD, Morgan MF, et al. The CRISM investigation in Mars orbit: Overview, history, and delivered data products. Icarus 2023 : 115612.
+72. Horgan BHN, Cloutis EA, Mann P, Bell JF. Near-infrared spectra of ferrous mineral mixtures and methods for their identification in planetary surface spectra. Icarus 2014, 234: 132-154.
+
+# Supplementary Files
+
+- [SupplementaryTableSpectra.xlsx](https://assets-eu.researchsquare.com/files/rs-4370272/v1/c83d6c9dff60913894bd82da.xlsx)  
+  Dataset 1
+
+- [SupplementaryInformation.docx](https://assets-eu.researchsquare.com/files/rs-4370272/v1/058744688f555bba37cd4481.docx)  
+  Supplementary Figures

1da78bbf060fc36775b0461ad1ea4eaef3411f009bb962072bfc328250ad8ee5/preprint/images_list.json

@@ -0,0 +1,50 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Recurrent and emerging SST trend patterns in the tropical Pacific. The SST trend (\u00b0C per decade) based on HadISST during (a)1980-2022 and (b) 1958-2022. The left insets of (b) show the timeseries of raw (black dashed lines) and 15-year running mean (red solid lines) annual-mean SST anomalies in the cold tongue (top inset; 5\u00b0S\u20135\u00b0N, 190\u00b0W\u2013270\u00b0W) and the warm pool (bottom inset; 5\u00b0S\u20135\u00b0N, 140\u00b0E\u2013170\u00b0W). The SST anomalies were calculated relative to the climatology of the first 50 years (1870-1919). Dots in (a-b) indicate the trend exceeding the 95% confidence level. (c) Pattern correlations of 43-year SST trends in historical period with the trend during 1980-2022 in the tropical Pacific region (30\u00b0S-30\u00b0N, 120\u00b0E-270\u00b0W) based on HadISST (black solid line), ERSSTv5 (green dashed line), Kaplan (red dotted line) and COBE (blue dashed line). (d) Similar to (c) but for 65-year SST trends with the trend during 1958-2022. Arrows labeled P1, P2 and P3 indicate the periods with strongest positive and negative correlations in (c) and P4 with strongest negative correlation in (d).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Quantification of internally-generated and emerging SST trends. (a) Histograms of pattern correlations of 43-year SST trend patterns over the tropical Pacific (30\u00b0S-30\u00b0N, 120\u00b0E-270\u00b0W) in historical period with that during 1980-2022. The Gaussian distributions are inferred for 63 samples in earlier (blue line) and 48 samples in later (red line) periods, separately, based on the averages and variances of the pattern correlations for the trend patterns. (b) Similar to (a) but for the 65-year SST trend. The Gaussian distributions are inferred based on 41 samples for the earlier period and 48 long-term for the later period. (c) Quantification of the IPO\u2019s contribution to the trend of zonal SST gradient in the equatorial Pacific (see Methods). The red solid line indicates the 43-year trend, and the red dashed line indicates the IPO\u2019s contribution. Similarly, the blue solid line and dashed line indicate the 65-year trend and the IPO\u2019s contribution, respectively. (d) The 43-year SST trend (\u00b0C per decade) in 1969-2012 with near-zero IPO contribution. Dots in (d) indicate the trend exceeding the 95% confidence level.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Recurrent and emerging signals in thermocline depth, surface wind stress and SSH. The thermocline depth trend (m/decade) based on ORAs5 subsurface temperature during (a) 1980-2022 and (b) 1958-2022. The surface wind stress (vectors; N/m2) and SSH (contours; m/decade) trend during (c) 1980-2022 and (d) 1958-2022. Dots in (a-b) and (c-d) indicate the trend in thermocline depth and SSH exceeding the 95% confidence level, respectively. (e) Pattern correlations of 43-year trend patterns in historical period for the thermocline depth (red lines), SSH (blue lines) and zonal wind stress (green lines) in the tropical Pacific region (30\u00b0S-30\u00b0N, 120\u00b0E-270\u00b0W) based on SODA (solid lines) and ORAs5 (dashed lines) with the corresponding trend during 1980-2022 based on ORAs5. (f) Similar to (e) but for 65-year trend with the trend during 1958-2022.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "See image above for figure legend\u00a0",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Ocean dynamics for differences in ocean current trend patterns. Zonal mean trend of (a) surface zonal current (m/s per decade), (b) upper 50-m averaged meridional current (m/s per decade) and (c) vertical current at the 50-m depth (m/day per decade) over the tropical central-to-eastern Pacific (150\u00b0E-90\u00b0W) during 1980-2022 (blue lines) and 1958-2022 (red lines). The IPO-related trend in surface zonal current (m/s per decade), upper 50-m averaged meridional current (m/s per decade) and vertical current at the 50-m depth (m/day per decade) are also shown in grey lines for comparison (see details in Methods for IPO-related trend). (d) The IPO-related trend in surface zonal current (shadings; m/s per decade) and its geostrophic component (contours; m/s per decade, at 0.01 m/s intervals with the zero line omitted) calculated based on Eqs. (7-8) using ORAs5 data. (e) The surface zonal current trend (shadings; m/s per decade) and its geostrophic component during 1958-2022 (contours; m/s per decade, at 0.01 m/s per decade intervals with the zero line omitted). Dots in (d-e) indicate the observed trend exceeding the 95% confidence level. (f-g) Similar to (d-e) but for the Ekman pumping calculated based on Eqs. (9-11). Dots in (f-g)indicate the values exceeding the 95% confidence level.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Ocean dynamics for the eastern Pacific cooling linked to the IPO and climate change. Heat budget terms averaged over the eastern equatorial Pacific (5\u00b0S\u20135\u00b0N, 190\u00b0W\u2013270\u00b0W; left box in Fig. 1b)including ocean current change (UcTa, VcTa, WcTa), temperature gradient change (UcTa, VcTa, WcTa), nonlinear terms (UaTa, VaTa, WaTa) and their sum (SUM) related to (a) the IPO and (b) the emerging SST trend (\u2103/month per decade; see Methods for detailed explanation). Dotted bars indicate heat budget terms exceeding 90% significance tests.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

1da78bbf060fc36775b0461ad1ea4eaef3411f009bb962072bfc328250ad8ee5/preprint/preprint.md

@@ -0,0 +1,197 @@
+# Abstract
+
+Recent debates have centered around whether the La Niña-like sea surface temperature (SST) trend pattern in the tropical Pacific in the past several decades is a response to anthropogenic forcings or internal variability, particularly the Interdecadal Pacific Oscillation (IPO). This study identifies an emerging SST warming pattern in the tropical Pacific featuring a narrow equatorial cooling band, in stark contrast to the meridionally broad SST trend pattern shaped by the IPO. The emerging SST trend pattern is associated with changes in subsurface temperature structure and sea level height that are distinct from those related to the recurrent IPO. The differences are primarily driven by their different surface wind stress patterns. The emerging wind stress pattern also drives distinctive ocean dynamical processes, fostering the unique eastern Pacific cooling. Our findings set a path to distinguish the often-tangled tropical Pacific climate change signals from internal variability through the underlying dynamics of each.
+
+**MAIN TEXT**
+
+Earth and environmental sciences/Climate sciences/Climate change  
+Earth and environmental sciences/Ocean sciences/Physical oceanography
+
+# Introduction
+
+The fact that under greenhouse gas-driven global warming some parts of Earth do not warm, or even cool, presents a fascinating problem. The surface tropical eastern Pacific stands out as one such region that resists warming<sup>1</sup>, associated with cooling and shoaling of the tropical Pacific Ocean thermocline over past decades<sup>2, 3</sup>. Evaluated up to now, whether over recent decades or back a century or more, the tropical Pacific exhibited a La Niña-like trend pattern with pronounced warming in the west and lack of warming in the cold tongue region<sup>1–6</sup>. In contrast, state-of-the-art climate models respond to anthropogenic greenhouse gas (GHG) forcing by generally simulating enhanced warming in the eastern Pacific and an El Niño-like warming pattern<sup>7, 8</sup>. Numerous studies have sought to reconcile this discrepancy and two primary arguments have been proposed. One suggests that the climate models exhibit systematic biases in simulating the tropical Pacific SST response to external forcing<sup>2, 3, 9, 10</sup>, and that correcting the cold tongue bias would align the models' simulations with observations, which are argued to be the forced response<sup>2</sup>. The other suggests that the internal variability significantly contributes to the recent observed trend, and the observed trend would fall within the range of climate model simulations once internal variabilities were sufficiently sampled in models<sup>11, 12</sup>.
+
+Although the evidence is not definitive, the Coupled Model Intercomparison Projects (CMIP) models and Large Ensembles rarely even come close to the observed long-term trends, which casts doubts on the ability of climate models to correctly respond to rising GHGs with the proper dynamic and thermodynamic balances<sup>9</sup>. Alternatively, we may consider using observations to estimate the forced climate response. However, observational analyses, which inherently rely on a single realization, come with their own limitations. Beyond the inescapable observational error/uncertainty<sup>13, 14</sup>, a major challenge arises from the fact that the tropical Pacific is dynamically active on decadal to multi-decadal timescales, so that the background internally-generated long-term variability can sometimes overwhelm the climate change signal<sup>15, 16</sup>. For example, the phase transition from positive to negative phases in the Interdecadal Pacific Oscillation (IPO) — the dominant climate mode in the Pacific ocean on decadal to multi-decadal time scales and closely related to the Pacific Decadal Oscillation<sup>17–20</sup> — is argued to have contributed to the observed La Niña-like SST trend pattern over the past four decades<sup>15, 21, 22</sup>. In contrast, some other studies highlight that the long-term change in the tropical Pacific in response to external forcing exhibits distinctive patterns unlike those on decadal timescales. For instance, while SST anomalies associated with the IPO in the tropical Pacific typically feature a meridionally broad pattern<sup>20</sup>, the long-term lack of warming is found to be narrowly confined to the cold tongue<sup>2–4, 6, 13</sup>. Seager et al.<sup>17</sup> further elaborate that some individual model simulations, containing both internal variability and the model’s forced response, come close to the observations but nonetheless fail to represent several fundamental elements of the observed long-term trends, including the distinctive narrowness of the lack of warming and the underlying subsurface ocean dynamical processes. These findings suggest that the tropical Pacific’s responses to anthropogenic forcing might present unique characteristics that set them apart from those typically associated with internal decadal variabilities.
+
+In this study, we seek to answer whether there is a detectable climate change signal in the tropical Pacific that can be clearly separated from internal variability and is robust across different datasets. This is crucial for setting a benchmark against which to assess the ability of climate models to reproduce the observed forced response in the presence of strong internal variability. In subsequent sections, we will demonstrate a distinctive climate change signal across various atmosphere-ocean fields in the tropical Pacific that is emerging from the internally-generated decadal variability and investigate its underlying dynamics.
+
+# Results
+
+Recurrent and emerging SST trend patterns.  
+The SST trend pattern after the satellite data become available (1980–2022) is often used as the observational reference for assessing climate model simulations of the forced response (Fig. 1a). This pattern is characterized by meridionally broad negative anomalies in the tropical Pacific, particularly south of the equator, and positive anomalies in the northwestern and southwestern Pacific. As noted by previous studies 21, 22, and shown in Figure S1a, the pattern is akin to that of the typical internal IPO variability on decadal to multi-decadal scale 17, 19. However, should we posit that the decadal variability (i.e., the phase transition of IPO from positive to negative phases) dominates this trend pattern, it would follow that this pattern is not unique in historical periods, given that the IPO has experienced several phase transitions in the observed record. Thus, we assess whether this short-term trend is a manifestation of decadal variability by calculating pattern correlations for tropical Pacific (30°S–30°N, 120°E–270°W) SST trends across equivalent time spans of 43 years but ending in different years to the most recent one (e.g., 1979 to 2021, 1978 to 2020 and so on all the way back to 1870 to 1912; Fig. 1c). The pattern correlations range from around −0.8 to 1, demonstrating an oscillatory behavior matching the IPO's phase transitions over the last century (Fig. S1b). Analyses for the whole Pacific (60°S–60°N, 120°E–270°W) or the equatorial Pacific (10°S–10°N, 120°E–270°W) yield similar conclusion (not shown).
+
+In comparison, a different SST trend pattern beginning in the mid-1950s is shown in Fig. 1b. The exact start year 1958 is chosen to approximate the time when the tropical Pacific warm pool starts to warm up (insets of Fig. 1b), as well as facilitating the analysis of subsurface ocean processes based on the Ocean ReAnalysis System-5 (ORAs5) data which become available in that year 2, 3, 23. The lack of warming in the tropical Pacific is still evident in this long-term trend pattern, however, it is markedly confined meridionally to the equatorial region (Fig. 1b), in contrast with the much broader meridional distribution that is more indicative of the IPO (Fig. 1a). Also, the warm anomalies in the northwestern and southwestern Pacific are less pronounced. This long-term trend pattern shows an emerging feature as evidenced in the pattern correlations of the latest trend with historical trends for equivalent 65-year spans. These show a rapid decline when the end year of the trend is adjusted backward and lingering around zero in earlier periods.
+
+We further identify periods with the strongest positive (P1:1870–1912 and P3:1942–1984) and strongest negative (P2: 1914–1956) pattern correlations with the most recent short-term trend, and the strongest negative pattern correlation with the most recent long-term trend (P4: 1870–1934) (Fig. S1c–f). The SST trend patterns in P1, P2 and P3 in the tropical Pacific broadly match that in the most recent 43-year trend, showing an IPO-like spatial distribution (Fig. S1c–e). Notwithstanding, the southward-displaced center of the cooling in the eastern Pacific (Fig. 1a) is less apparent in these historical periods, implying that the feature is potentially indicative of a climate change signal rather than a result of internal variability 24, 25. In contrast, the trend in P4 (Fig. S1f) appears to be a weak reflection of the IPO pattern rather than a negative analogue of the latest 65-year trend pattern, again confirming the distinctiveness of the emerging SST trend pattern recently. The recurring feature of the short-term trend and the emerging feature of the long-term trend based on HadISST were consistently identified across different SST datasets including ERSSTv5, Kaplan, COBE (Figs. 1c,d and S2).
+
+The emerging feature is apparent when examining the histogram and probability distributions of pattern correlations for short-term and long-term trends in the historical records (Fig. 2a,b). While the overall distribution of the short-term trends closely resembles a symmetrical distribution around zero over time, and there are earlier periods that had trends with the opposite spatial pattern to the most recent one, the long-term trend has increasingly skewed towards positive values in recent decades. There are no past ant-analogs of the long-term trend. These trends were further categorized into two subsets, an earlier period with presumably weak climate change signal and a later period subject to pronounced anthropogenic forcing, based on whether they end before or after 1975. This cutoff year was selected to roughly split the periods evenly, and qualitative conclusions remain the same with slight changes in the cutoff year. The probability distributions for earlier and later periods elucidate that while the short-term trends in recent decades cannot be separated from those in earlier periods, the long-term trends in recent decades are notably separated from the background intrinsic decadal variability in earlier period. We further quantified the IPO’s contribution to the SST trend (Fig. 2c) by comparing the trend in the tropical Pacific zonal SST gradient and the IPO’s contribution, which is calculated using the trend in the IPO index in different time span and the IPO-related zonal SST gradient based on Figure S1. Consistent with Fig. 2a,b, the internal variability of the IPO predominantly shapes both the short-term and long-term trends during the earlier period when the climate change signal is relatively weak. In the recent period, however, the imprints of climate change are distinctly reflected in both short-term and long-term trends (Fig. 2c). However, as noted above, the spatial pattern of the most recent short-term trend is overwhelmed by the IPO (Fig. 1) with the IPO accounting for over half of the observed zonal SST gradient trend, while the impact of the IPO on the most recent long-term trend is minor. Moreover, when examining short-term trends with a negligible IPO impact, such as that from 1969–2012, the short-term trend pattern closely resembles the emerging SST pattern detected in the extended period (Figs. 1c and 2d).
+
+## Distinct ocean dynamics for emerging climate change signal and decadal variability
+
+Our analysis has established that there is a distinctive SST trend pattern detected in an extended period emerging from the short-term SST trend pattern primarily related to the IPO. A closed Bjerknes feedback loop necessitates corresponding changes in the surface wind stress and the subsurface thermocline depth, in accordance with SST variations 26, 27. In Fig. 3, we show that there exists a recurring trend pattern of thermocline depth with a dipole-like structure (Fig. 3a,e). In contrast, the emerging trend pattern is characterized with an overall shoaling throughout the tropical Pacific (Fig. 3b,e). Also, the recurring surface wind stress pattern features a broad strengthening of the zonal wind stress across the tropical Pacific (Fig. 3c,e), while the emerging wind stress trend pattern displays a dipole structure with strengthened wind stress in the central equatorial Pacific and weakened wind stress in the eastern equatorial Pacific (Fig. 3d,e). We also examined changes in the surface wind stress and the subsurface thermocline depth that are directly linked to the IPO (Fig. S3). These IPO-related patterns with dipole-like thermocline change and same-signed wind stress change in the tropical Pacific bear a close resemblance to the short-term trends, which again substantiate the argument that the short-term trends over the span of approximately an IPO cycle predominantly reflect the recurrent influence of the IPO manifested in various ocean-atmosphere fields in the tropical Pacific.
+
+In the tropical Pacific, the thermocline depth and SSH are closely linked to each other and exhibit similar interannual fluctuations associated with El Niño-Southern Oscillation 27, 28. On decadal timescales, the dipole-like pattern characterized with much stronger sea level rise in the tropical western Pacific compared to the east, and thermocline deepening in the west and shoaling in the east, are also present in the short-term trends (Fig. 3a,c,f). In contrast to the shorter-term trend, the emerging long-term trend exhibits an overall, but small, sea level fall in the central-to-eastern equatorial Pacific region, accompanied by sea level rise in the off-equatorial regions (Fig. 3d). The long-term thermocline depth trend is also different from the shorter-term one in that it has shoaling or little change across the basin, even though the shoaling is greater in the east. The thermocline depth-SSH-wind stress changes associated with the short-term trend oscillate back and forth together (Fig. 3e) but those associated with the long-term trend are collectively emerging over time for the long-term trend (Fig. 3f).
+
+These observed SSH and thermocline trends were previously demonstrated to be some combination of wind-driven, thermal expansion and the influence of changes in the Earth’s gravity field due to loss of land ice 29–31. While there is general consensus on zonal dipole-like SSH change related to internal variability in both observations 32 and CMIP models 33, the anthropogenic contribution to SSH change remains elusive 31, 33, 34. Our subsequent analyses will focus on examining SSH. A 1.5-layer reduced-gravity model is used here to investigate the dynamic linkages between the surface and subsurface components (see details in Methods; note that the same dynamics can be formulated as describing a single baroclinic vertical mode and hence have more general applicability). The reduced-gravity model can simulate the wind stress driven changes in SSH and thermocline depth 35, 36; here we solve analytically an equilibrium version of the reduced gravity system (see Methods, Eq. (6)). As shown in Fig. 4a,b, both the short-term IPO-related trend and the long-term emerging trend can be quite realistically simulated by prescribing their corresponding surface wind stress trend patterns. The pattern correlations between observations and model for the observed and wind stress-driven trend patterns reach as high as 0.87 for the short-term trend and 0.72 for the long-term in the tropical Pacific, suggesting the dominant role of the wind-driven redistribution of the heat content in the tropical Pacific upper ocean by the surface wind stress for both decadal variability and the emerging climate change signal. According to Eq. (6), variations in SSH at each longitude are determined by the impact of surface wind stress and its horizontal gradients zonally integrated from the eastern boundary to that longitude. The spatial distribution of the wind stress effects (B in Eq. (6); Fig. 4c,d) underscores that it is the wind stress and its horizonal gradients in the central tropical Pacific that are most important in redistributing the heat content and driving the SSH changes in the tropical Pacific, for both climate change and decadal variability.
+
+The different wind stress patterns also drive different ocean circulation changes. Figure 5 presents the zonally averaged ocean current trends over the central-to-eastern Pacific associated with the recurrent short-term and emerging long-term trends. We also display the IPO-related ocean currents, which again exhibit a strong consistency with the short-term trends. The most striking difference between the decadal variability-related and climate change-related trends is the opposite-signed surface zonal currents in the equatorial and north off-equatorial regions (Fig. 5a; compare Fig. 5d,e). There is a significant strengthening of the surface westward zonal currents, except in the central equatorial Pacific, for the decadal variability (contours in Fig. 6a). In contrast, the emerging pattern shows a weakening of westward surface currents in the central-to-eastern Pacific (Fig. 6b). The changes in surface zonal currents in the tropical Pacific are predominantly governed by its geostrophic component (Eqs. (7–8); shadings in Fig. 6a,b), which follows the spatial pattern of the SSH that has been established to be connected to the surface wind stress trend patterns (Fig. 4). Wind stress impacts can also be detected in the Ekman pumping change (Eq. (11)). While the decadal upwelling pattern is approximately symmetrical around the equator, the meridional center of the emerging upwelling pattern is displaced towards near 5 degrees south. The overall strengthening of zonal wind stress across the equatorial Pacific linked to decadal variability fosters pronounced upwelling from the western to eastern equatorial Pacific (Fig. 6c). In comparison, the emerging dipole-like wind stress pattern contributes to enhanced upwelling in the central equatorial Pacific and weakened upwelling to the east, while stronger trade winds south of the equator contribute to increased local upwelling (Fig. 6d), thereby accounting for the different meridional locations of upwelling change. The meridional currents averaged in the mixed layer, indicative of the strength of the shallow overturning circulation, are quite similar for the short-term and long-term trends, showing a consistent strengthening of the poleward transport in both hemispheres despite different magnitudes.
+
+These changes in wind-driven ocean currents, in turn, account for the IPO-related and emerging climate change related temperature change in the equatorial Pacific via different ocean dynamical processes (Fig. 6). The IPO-related cooling in the equatorial eastern Pacific (Fig. 1a) is primarily driven by the cooling effect of the zonal advection (UcTa) (Fig. 6a) resulting from strengthened zonal current (Fig. 5d). The meridional advection (VaTc) related to enhanced poleward transport and the change due to thermocline shoaling (WcTa) also contribute to the IPO-related cooling in the equatorial region. In contrast, the emerging cooling signal is relatively muted, primarily because the zonal advective warming effect due to weakened zonal current largely offsets the cooling effect related to the thermocline feedback due to the thermocline shoaling (Fig. 6b). The mean meridional advection (VcTa) also contributes to the emerging cooling trend in the central-to-eastern Pacific (Fig. 6b) via the strengthened meridional temperature gradient (Fig. 1b). Such effect is not observed for the IPO-related variability (Fig. 6a), where broader eastern Pacific cooling leads to insignificant changes in the meridional temperature gradient. Although the vertical upwelling changes are evident for the equatorial region, the contributions of the Ekman pumping term to the wider equatorial Pacific (5°S–5°N) temperature change, either related to the IPO or the emerging climate change, are rather minor due to the immediate opposite-signed effect off the equator for both variabilities.
+
+# Discussion
+
+In this study, we identify an emerging climate change signal in the tropical Pacific across different observational datasets, which exhibits distinctive ocean-atmosphere dynamics that differ from those typically associated with IPO-related decadal variability. The emerging SST trend pattern features a narrow band of cooling in the eastern equatorial Pacific, linked to thermocline shoaling/SSH decreases in the central-to-eastern Pacific and dipole-like changes in zonal surface wind stress. In contrast, the recurrent IPO-driven SST trend pattern is characterized by a meridionally broader cooling in the eastern Pacific, corresponding to zonal dipole-like thermocline/SSH changes and an overall strengthening of tropical Pacific zonal wind stress. The different changes in wind stress pattern lead to distinct ocean circulation changes. These oceanic responses to the surface wind stress account for their surface cooling in the eastern Pacific, with the thermocline shoaling playing a dominant role in the emerging cooling and enhanced zonal advective cooling mainly driving the IPO-related cooling.
+
+While basic geophysical fluid dynamics underpin our argument that the observed oceanic changes can be interpreted as adjustments to variations in surface wind stress, further investigations including targeted ocean model experiments are required to comprehensively assess the relative contributions of local versus remote wind effects<sup>37</sup>, as well as to understand the initial wind response to GHGs. The climatological settings of the tropical Pacific may inherently predispose it to different initial SST response in the warm pool and cold tongue region, and a corresponding trade wind response<sup>2, 38</sup>. Due to the increased atmospheric static stability in response to GHG forcings<sup>39, 40</sup> related to stronger temperature change in the upper troposphere compared to the surface (Fig. S4), this initial response to rising GHGs might not be amplified as efficiently via Bjerknes feedback as those observed for the internal modes on interannual to decadal timescales. Additionally, climate variations outside of the tropical Pacific have been argued to influence the tropical Pacific trade winds through teleconnections<sup>24, 25, 41–44</sup>. Further, it has been argued that pronounced decadal-to-multidecadal SST changes in the Atlantic Ocean are also dominated by the response to the same external forcing that the tropical Pacific encounters<sup>45</sup>, suggesting an alternative explanation for the co-occurrence of these long-term variabilities across different regions, and the potential for an inter-basin interaction in the pattern of SST response to rising GHGs. More work is needed to disentangle causal relationships among the long-term changes in different basins<sup>46, 47</sup>.
+
+It is also critical to acknowledge that while we aim to distinguish between the recurrent IPO-related decadal variability and the climate change signal, these two may have become coupled together. We have emphasized the differences between the ocean-atmosphere dynamics of each, however, they do share much in common: shoaling of the thermocline in the east, enhanced upwelling somewhere in the central-to-eastern equatorial Pacific and an enhanced zonal SST gradient across the equatorial Pacific. It seems reasonable to postulate that if the response to radiative forcing is the emerging pattern seen here, then it will initiate coupled ocean-atmosphere feedbacks that favor a negative IPO state that also has an enhanced SST gradient. This might explain why the most recent IPO swing has been extreme and robust (as our analysis shows in Figs. 1 and 2). If so, this suggests that in nature forcing is projecting onto natural modes of variability, while it is not clear whether climate models can reproduce that kind of physical behavior. This would require a new perspective on how internal variability interacts with the climate change signal in future studies.
+
+# Materials and Methods
+
+## Datasets
+
+The SST data used here are the Hadley Centre data HadISST version 1.1 with horizontal resolutions of 1° × 1°<sup>48</sup>, the National Oceanic and Atmospheric Administration ERSSTv5 data with horizontal resolutions of 2° × 2°<sup>49</sup>, the Centennial in Situ Observation Based Estimates of SST (COBE) from the Japanese Meteorological Agency with horizontal resolutions of 1° × 1°<sup>50</sup>, and Kaplan Extended SST version 2 with horizontal resolutions of 5° × 5°<sup>51</sup>. HadISST, ERSSTv5, and Kaplan were used from 1870 to 2022 and COBE from 1890 to 2022. We utilized subsurface temperature, surface wind stress, SSH, zonal and meridional currents from ORAs5 with horizontal resolutions of 0.25° × 0.25° and 75 vertical levels, the latest ocean reanalysis products provided by the European Centre for Medium-Range Weather Forecasts<sup>23</sup>. The vertical velocity for ORAs5 was derived from zonal and meridional currents based on mass continuity<sup>52</sup>. We also used subsurface temperature and surface wind stress from the Simple Ocean Data Assimilation (SODA), version 2.2.4 with horizontal resolutions of 0.25° × 0.25° and 40 vertical layers during 1871–1979<sup>53</sup>, in conjunction with the version 3.3.2 with horizontal resolutions of 0.25° × 0.25° and 50 vertical layers during 1980–2018<sup>54</sup>. The air temperature data was obtained from the National Centers for the Environmental Prediction–National Center for the Atmospheric Research (NCEP–NCAR) reanalysis 1 with horizontal resolutions of 2.5°×2.5° and 17 vertical layers<sup>55</sup>. All datasets were interpolated onto a horizontal grid of 1° × 1° to enable comparison among datasets.
+
+## Statistical methods and definitions of indices
+
+Anomalies for all variables were calculated as departures from the monthly climatology unless specified otherwise. Statistical significance tests were performed based on the two-tailed Student’s t-test with n–2 degrees of freedom, where n is the sample size. The thermocline depth was identified as the depth of the 20℃ isotherm. The qualitative conclusion remains similar based on the thermocline depth defined by the maximum vertical temperature gradient. The zonal SST gradient in the tropical Pacific was defined as the temperature difference between the western Pacific (5° S–5° N, 140° E–170° E, indicated by the left box in Fig. 1 b) and the eastern Pacific (5° S–5° N, 190° W–270° W, indicated by the right box in Fig. 1 b). The IPO index was calculated based on the difference between the SST anomalies averaged over the central equatorial Pacific (10°S–10°N, 170°E–90°W ) and the average of the SST anomalies in the northwest (25°N–45°N, 140°E–145°W) and southwest Pacific (50°S–15°S, 150°E–160°W) following Henley et al.<sup>17</sup>. Then a 13-year low-pass filter based on Fast Fourier Transform was applied to extract the decadal-scale component of IPO variability. To assess the IPO's contribution to the short-term trend of variable $x$, we calculate the IPO-related trend by regressing the detrended $x$ against the IPO index and then multiplying this by the IPO index's linear trend from 1980 to 2022.
+
+## Reduced gravity system linking SSH and thermocline depth to surface wind stress
+
+A 1.5-layer reduced-gravity system is considered here following the formulation of Veronis<sup>36</sup> to establish the relationship between the change in the surface wind stress and change in the SSH and thermocline depth over the tropical Pacific. We made several modifications to Veronis' framework by including the meridional wind stress component ${\tau }^{y}$ (previously set to zero), the zonally-varying zonal wind stress ${\tau }^{x}$ (previously assumed to be zonally-uniform), and a damping term (previously not considered) and using an equatorial $\beta$-plane ($f=\beta y$, in which $\beta =$ 2.3 * 10<sup>−11</sup> m<sup>−1</sup> s<sup>−1</sup>). We also adopt a linear system with a specified spatially-uniform climatological upper layer thickness in the tropical Pacific ($\stackrel{-}{h}=$ 150 m)<sup>56</sup>, Taking $\varDelta \rho = 2.7$ kg/m<sup>3</sup> as the density contrast between upper and bottom layers yields a first baroclinic mode gravity wave speed of c ~ 2.0 m/s, where c<sup>2</sup> = ${g}^{{\prime }}\stackrel{-}{h}$, ${g}^{{\prime }}=g\frac{\varDelta \rho }{{\rho }_{0} },$ and ${\rho }_{0}= 1025$ kg/m<sup>3</sup> is the reference density. The governing equations on an equatorial $\beta$-plane are:
+
+|  |  |
+|---|---|
+| $-fV=-{g}^{{\prime }}\stackrel{-}{h}\frac{\partial h}{\partial x}+\frac{{\tau }^{x}}{{\rho }_{0} }$ | Eq. (1) |
+| $fU= -{g}^{{\prime }}\stackrel{-}{h}\frac{\partial h}{\partial y}+\frac{{\tau }^{y}}{{\rho }_{0} }$ | Eq. (2) |
+| $\frac{\partial U}{\partial x}+\frac{\partial V}{\partial y}=-rh$ | Eq. (3) |
+
+in which $h$ is the upper layer thickness between SSH ($h_{1};\text{m})$ and thermocline depth ($h_{2};\text{m})$, $r=$ 1/5.5 year<sup>−1</sup> is the damping coefficient, $U={\int }_{{h}_{1}}^{{h}_{2}}udz$ ($u$ the zonal current; m/s), and $V={\int }_{{h}_{1}}^{{h}_{2}}vdz$ ($v$ the meridional current; m/s). Cross-differentiating Eqs. (1–2) and using Eq. (3), we obtain the linkage between layer thickness change and wind stress change:
+
+|  |  |
+|---|---|
+| $\frac{\partial h}{\partial x}- \frac{\beta {y}^{2}r}{{g}^{{\prime }}\stackrel{-}{h}}h=\frac{y}{{{\\rho }_{0}g}^{{\prime }}\stackrel{-}{h}}(\frac{{\tau }^{x}}{y}+\frac{\partial {\\tau }^{y}}{\partial x}-\frac{\partial {\\tau }^{x}}{\partial y})$ | Eq. (4) |
+
+Let $A=-\frac{\beta {y}^{2}r}{{g}^{{\prime }}\stackrel{-}{h}}h$ and Eq. (4) ($B=\frac{y}{{{\\rho }_{0}g}^{{\prime }}\stackrel{-}{h}}(\frac{{\tau }^{x}}{y}+\frac{\partial {\\tau }^{y}}{\partial x}-\frac{\partial {\\tau }^{x}}{\partial y}),$) can be solved as:
+
+|  |  |
+|---|---|
+| $h={e}^{A({x}_{e}-x)}{h}_{e}+{\int }_{{x}_{e}}^{x}B{e}^{A({x}^{{\prime }}-x)}dx{\\prime }$ | Eq. (5) |
+
+in which ${x}_{e}$ indicates the eastern boundary, ${h}_{e}$ the layer thickness at the eastern boundary. The change in the SSH can then be directly linked to the change in the surface wind stress if the change of SSH near the eastern boundary is neglected (which is approximately justified on the basis of the changes in Fig. 3):
+
+|  |  |
+|---|---|
+| ${h}_{1}=\frac{\varDelta \rho }{\rho +\varDelta \rho }{\int }_{{x}_{e}}^{x}B{e}^{A({x}^{{\prime }}-x)}dx{\\prime }$ | Eq. (6) |
+
+## Estimation of geostrophic zonal current and Ekman pumping
+
+The geostrophic component of the surface current can be determined by considering the balance between the Coriolis force and the pressure gradient force. In spherical coordinates, the geostrophic zonal current $(u_{g})$ outside of the equatorial region is expressed as:
+
+|  |  |
+|---|---|
+| ${u}_{g}=-\frac{g}{f}\frac{\\partial h}{\\partial y}$ | Eq. (7) |
+
+At the equator where $f=0$, an estimate of the equatorial semi-geostrophic zonal current $(u_{sg})$ is derived by calculating the second derivative of the SSH on an equatorial $\beta$-plane, which are suggested to be in good agreement with measured velocities<sup>57</sup>,<sup>58</sup>:
+
+|  |  |
+|---|---|
+| ${u}_{sg}=-\frac{g}{\\beta }\frac{{\\partial }^{2}h}{\\partial {y}^{2}}$ | Eq. (8) |
+
+Following the approach of Cane and Zebiak<sup>59</sup>, the Ekman transport ($U_{E}, V_{E}$) in the tropical region is formulated by incorporating a frictional component as:
+
+|  |  |
+|---|---|
+| ${U}_{E}=({r}_{s}{\\tau }^{x}+f{\\tau }^{y})/{\\rho }_{0}({f}^{2}+{{r}_{s}}^{2})$ | Eq. (9) |
+| ${V}_{E}=({r}_{s}{\\tau }^{y}-f{\\tau }^{x})/{\\rho }_{0}({f}^{2}+{{r}_{s}}^{2})$ | Eq. (10) |
+
+where $r_{s}$ indicates the surface layer friction coefficient (1/2 day<sup>−1</sup>). The Ekman transport away from the equator is consistent with classical Ekman theory. At the equator where $f=0$, the friction allows an Ekman transport in the direction of the wind stress. Ekman pumping velocity ($w_{E}$) is thus derived from the divergence of the Ekman transport:
+
+|  |  |
+|---|---|
+| ${w}_{E}=\frac{\partial {U}_{E}}{\partial x}+\frac{\partial {V}_{E}}{\partial y}$ | Eq. (11) |
+
+## Mixed layer heat budget analysis for the long-term SST Change
+
+The heat budget for the mixed layer temperature<sup>60</sup> can be expressed as
+
+|  |  |
+|---|---|
+| $\frac{\\partial T}{\\partial t}={-u_{a}{\\frac{\\partial T}{\\partial x}}_{ c}}{-u_{c}{\\frac{\\partial T}{\\partial x}}_{ a}}{-u_{a}{\\frac{\\partial T}{\\partial x}}_{ a}}{-v_{a}{\\frac{\\partial T}{\\partial y}}_{ c}}{-v_{c}{\\frac{\\partial T}{\\partial y}}_{ a}}{-v_{a}{\\frac{\\partial T}{\\partial y}}_{ a}}{-w_{a}{\\frac{\\partial T}{\\partial z}}_{ c}}{-w_{c}{\\frac{\\partial T}{\\partial z}}_{ a}}{-w_{a}{\\frac{\\partial T}{\\partial z}}_{ a}}+R.$ | Eq. (12) |
+
+in which $a$ denotes anomaly and $c$ denotes climatology. The heat budget terms include changes in the mean current ($UaTc$, $VaTc$, $WaTc$), changes in the mean temperature gradient ($UcTa$, $VcTa$, $WcTa$), and their nonlinear interaction ($UaTa$, $VaTa$, $WaTa$). The zonal advection and meridional advection terms were averaged over a uniform mixed layer depth of 50 m. The vertical velocity was calculated at the bottom of the mixed layer, and the vertical advection between the 50–100 m and the upper 50 m layers was calculated only in the presence of upwelling. The residual term ($R$) for the mixed layer indicates the surface heat flux and subgrid/submonthly processes.
+
+To evaluate the heat budget related to the long-term emerging temperature changes, we identified two sub-periods during 1958–2022: the first 20 years (1958–1977) as a reference period, and the most recent 20 years (2003–2022) as the period of climate change. We then calculated the averages of each heat budget term in the quasi-equilibrium period P1 and climate change period P2 based on Eq. (12), and estimated the contributions of each term to the observed temperature changes by calculating their differences:
+
+|  |  |
+|---|---|
+| ${\\stackrel{-}{\\frac{\\partial T}{\\partial t}}}_{P2}\\approx {\\stackrel{-}{\\frac{\\partial T}{\\partial t}}}_{P2}-{\\stackrel{-}{\\frac{\\partial T}{\\partial t}}}_{P1}={\\overline{UaTc}}_{P2}-{\\overline{UaTc}}_{P1}+ \\dots +{\\overline{WaTa}}_{P2}-{\\overline{WaTa}}_{P1}+ {\\stackrel{-}{R}}_{P2}-{\\stackrel{-}{R}}_{P1}$ | Eq. (13) |
+
+To reflect the contributions of these terms to the temperature change over one decade, we normalized their units to °C/month per decade by dividing by a factor of 6.5. In addition, to analyze the IPO's impact on the temperature change on decadal timescales, the detrended heat budget terms were regressed against the IPO index. The regression coefficients were then scaled by the linear trend in the IPO index from 1980 to 2022.
+
+# References
+
+1. Cane MA et al (1997) Twentieth-Century Sea Surface Temperature Trends. Science 275:957–960
+2. Seager R et al (2019) Strengthening tropical Pacific zonal sea surface temperature gradient consistent with rising greenhouse gases. Nat Clim Chang 9:517–522
+3. Seager R, Henderson N, Cane M (2022) Persistent Discrepancies between Observed and Modeled Trends in the Tropical Pacific Ocean. J Clim 35:4571–4584
+4. Zhang W, Li J, Zhao X (2010) Sea surface temperature cooling mode in the Pacific cold tongue. J Geophys Res 115:2010JC006501
+5. Karnauskas KB, Seager R, Kaplan A, Kushnir Y, Cane MA (2009) Observed Strengthening of the Zonal Sea Surface Temperature Gradient across the Equatorial Pacific Ocean*. J Clim 22:4316–4321
+6. Solomon A, Newman M (2012) Reconciling disparate twentieth-century Indo-Pacific ocean temperature trends in the instrumental record. Nat Clim Change 2:691–699
+7. Knutson TR, Manabe S (1995) Time-Mean Response over the Tropical Pacific to Increased C0<sub>2</sub> in a Coupled Ocean-Atmosphere Model. J Clim 8:2181–2199
+8. Meehl GA, Washington WM (1996) El Niño-like climate change in a model with increased atmospheric CO<sub>2</sub> concentrations. Nature 382:56–60
+9. Wills RCJ, Dong Y, Proistosecu C, Armour KC, Battisti DS (2022) Systematic Climate Model Biases in the Large-Scale Patterns of Recent Sea‐Surface Temperature and Sea‐Level Pressure Change. *Geophysical Research Letters* 49, eGL100011 (2022)
+10. Luo J-J, Wang G, Dommenget D (2018) May common model biases reduce CMIP5’s ability to simulate the recent Pacific La Niña-like cooling? Clim Dyn 50:1335–1351
+11. Watanabe M, Dufresne J-L, Kosaka Y, Mauritsen T, Tatebe H (2021) Enhanced warming constrained by past trends in equatorial Pacific sea surface temperature gradient. Nat Clim Chang 11:33–37
+12. Olonscheck D, Rugenstein M, Marotzke J (2020) Broad Consistency Between Observed and Simulated Trends in Sea Surface Temperature Patterns. *Geophysical Research Letters* 47, e2019GL086773
+13. Coats S, Karnauskas KB (2017) Are Simulated and Observed Twentieth Century Tropical Pacific Sea Surface Temperature Trends Significant Relative to Internal Variability? Geophys Res Lett 44:9928–9937
+14. Deser C, Phillips AS, Alexander MA (2010) Twentieth century tropical sea surface temperature trends revisited. Geophys Res Lett 37:2010GL043321
+15. Dong L, Zhou T (2014) The formation of the recent cooling in the eastern tropical Pacific Ocean and the associated climate impacts: A competition of global warming, IPO, and AMO. JGR Atmos 119
+16. Hansen J et al (1997) Forcings and chaos in interannual to decadal climate change. J Geophys Res 102:25679–25720
+17. Henley BJ et al (2017) Spatial and temporal agreement in climate model simulations of the Interdecadal Pacific Oscillation. Environ Res Lett 12:044011
+18. Mantua NJ, Hare SR, Zhang Y, Wallace JM, Francis RC (1997) A Pacific Interdecadal Climate Oscillation with Impacts on Salmon Production. Bull Amer Meteor Soc 78:1069–1079
+19. Power S, Casey T, Folland C, Colman A, Mehta V (1999) Inter-decadal modulation of the impact of ENSO on Australia. Clim Dyn 15:319–324
+20. Zhang Y, Wallace JM, Battisti DS (1997) ENSO-like Interdecadal Variability: 1900–93. J Clim 10:1004–1020
+21. Wang B, Liu J, Kim H-J, Webster PJ, Yim S-Y (2012) Recent change of the global monsoon precipitation (1979–2008). Clim Dyn 39:1123–1135
+22. Kosaka Y, Xie S-P (2013) Recent global-warming hiatus tied to equatorial Pacific surface cooling. Nature 501:403–407
+23. Zuo H, Balmaseda MA, Tietsche S, Mogensen K, Mayer M (2019) The ECMWF operational ensemble reanalysis–analysis system for ocean and sea ice: a description of the system and assessment. Ocean Sci 15:779–808
+24. Kang SM, Ceppi P, Yu Y, Kang I-S (2023) Recent global climate feedback controlled by Southern Ocean cooling. Nat Geosci 16:775–780
+25. Kang SM et al (2023) Global impacts of recent Southern Ocean cooling. *Proc. Natl. Acad. Sci. U.S.A.* 120, e2300881120
+26. Bjerknes J (1969) Atmospheric teleconnections from the equatorial Pacific. Mon Wea Rev 97:163–172
+27. Wyrtki K (1975) El Niño—The Dynamic Response of the Equatorial Pacific Ocean to Atmospheric Forcing. J Phys Oceanogr 5:572–584
+28. Rebert JP, Donguy JR, Eldin G, Wyrtki K (1985) Relations between sea level, thermocline depth, heat content, and dynamic height in the tropical Pacific Ocean. J Geophys Res 90:11719–11725
+29. Bamber JL, Westaway RM, Marzeion B, Wouters B (2018) The land ice contribution to sea level during the satellite era. Environ Res Lett 13:063008
+30. Wigley TML, Raper SCB (1987) Thermal expansion of sea water associated with global warming. Nature 330:127–131
+31. Timmermann A, McGregor S, Jin F-F (2010) Wind Effects on Past and Future Regional Sea Level Trends in the Southern Indo-Pacific*. J Clim 23:4429–4437
+32. Merrifield MA (2011) A Shift in Western Tropical Pacific Sea Level Trends during the 1990s. J Clim 24:4126–4138
+33. Marcos M, Amores A (2014) Quantifying anthropogenic and natural contributions to thermosteric sea level rise: Marcos and Amores: Anthropogenic ocean warming. Geophys Res Lett 41:2502–2507
+34. McGregor S, Gupta AS, England MH (2012) Constraining Wind Stress Products with Sea Surface Height Observations and Implications for Pacific Ocean Sea Level Trend Attribution*. J Clim 25:8164–8176
+35. Sarachik ES, Cane MA (2010) The El Niño-Southern Oscillation Phenomenon. Cambridge University Press
+36. Veronis G (1973) Model of world ocean circulation: I. Wind-driven, two-layer. J Mar Res 31
+37. Emile-Geay J, Cane MA (2009) Pacific Decadal Variability in the View of Linear Equatorial Wave Theory*. J Phys Oceanogr 39:203–219
+38. Clement AC, Seager R, Cane MA, Zebiak SE (1996) An Ocean Dynamical Thermostat. J Clim 9:2190–2196
+39. Manabe S, Wetherald RT (1975) The Effects of Doubling the CO<sub>2</sub> Concentration on the climate of a General Circulation Model. J Atmos Sci 32:3–15
+40. Fu Q, Manabe S, Johanson CM (2011) On the warming in the tropical upper troposphere: Models versus observations. Geophys Res Lett 38:2011GL048101
+41. Meehl GA et al (2021) Atlantic and Pacific tropics connected by mutually interactive decadal-timescale processes. Nat Geosci 14:36–42
+42. McGregor S et al (2014) Recent Walker circulation strengthening and Pacific cooling amplified by Atlantic warming. Nat Clim Change 4:888–892
+43. Luo J-J, Sasaki W, Masumoto Y (2012) Indian Ocean warming modulates Pacific climate change. *Proc. Natl. Acad. Sci. U.S.A.* 109, 18701–18706
+44. Dong Y, Armour KC, Battisti DS, Blanchard-Wrigglesworth E (2022) Two-Way Teleconnections between the Southern Ocean and the Tropical Pacific via a Dynamic Feedback. J Clim 35:6267–6282
+45. He C et al (2023) Tropical Atlantic multidecadal variability is dominated by external forcing. Nature 622:521–527
+46. Deser C, Phillips AS (2023) Spurious Indo-Pacific Connections to Internal Atlantic Multidecadal Variability Introduced by the Global Temperature Residual Method. Geophys Res Lett 50, e2022GL100574
+47. Fenske T, Clement A (2022) No Internal Connections Detected Between Low Frequency Climate Modes in North Atlantic and North Pacific Basins. *Geophysical Research Letters* 49, e2022GL097957
+48. Rayner NA (2003) Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J Geophys Res 108:4407
+49. Huang B et al (2017) Extended Reconstructed Sea Surface Temperature, Version 5 (ERSSTv5): Upgrades, Validations, and Intercomparisons. J Clim 30:8179–8205
+50. Ishii M, Shouji A, Sugimoto S, Matsumoto T (2005) Objective analyses of sea-surface temperature and marine meteorological variables for the 20th century using ICOADS and the Kobe Collection. Intl J Climatology 25:865–879
+51. Kaplan A et al (1998) Analyses of global sea surface temperature 1856–1991. J Geophys Res 103:18567–18589
+52. Vidard A, Bouttier P-A, Vigilant F (2015) NEMOTAM: tangent and adjoint models for the ocean modelling platform NEMO. Geosci Model Dev 8:1245–1257
+53. Carton JA, Giese BS (2008) A Reanalysis of Ocean Climate Using Simple Ocean Data Assimilation (SODA). Mon Weather Rev 136:2999–3017
+54. Carton JA, Chepurin GA, Chen L (2018) SODA3: A New Ocean Climate Reanalysis. J Clim 31:6967–6983
+55. Kalnay E et al (1996) The NCEP/NCAR 40-Year Reanalysis Project. Bull Amer Meteor Soc 77:437–471
+56. Zebiak SE (1985) *Tropical Atmosphere–Ocean Interaction and the El Niño/Southern Oscillation Phenomenon.* Ph.D. thesis, Massachusetts Institute of Technology, 260 pp
+57. Moore DWH, Philander SGH (1978) Modeling of the tropical oceanic circulation. The Sea. Wiley-Interscience, pp 319–361
+58. Picaut J, Hayes SP, McPhaden MJ (1989) Use of the geostrophic approximation to estimate time-varying zonal currents at the equator. J Geophys Res 94:3228–3236
+59. Zebiak SE, Cane MA (1987) A Model El Niño–Southern Oscillation. Mon Wea Rev 115:2262–2278
+60. Jin F, An S, Timmermann A, Zhao J (2003) Strong El Niño events and nonlinear dynamical heating. Geophys Res Lett 30
+
+# Supplementary Files
+
+- [SummplementaryFileContrast0628.docx](https://assets-eu.researchsquare.com/files/rs-4656683/v1/1b9839165e12107f09057f0e.docx)