Add files using upload-large-folder tool

002d081255dc240b062b581d7f24a1883dc21c197e49755a66f30280ecfa0f61/preprint/images_list.json

@@ -0,0 +1,66 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Design, synthesis, and in vitro characterization of an anti-albumin nanobody for site-selective conjugation of STING agonists. (a) Scheme depicting the concept of an albumin-hitchhiking nanobody-STING agonist conjugate for cancer immunotherapy. Anti-albumin nanobodies conjugated to STING agonists bind to circulating albumin in situ, resulting in improved pharmacokinetics and increased biodistribution to tumor sites that stimulates antitumor innate and adaptive immune responses. (b) Computational model of the anti-albumin nanobody (nAlb) binding at domain IIB of human serum albumin. (c) Isothermal calorimetry (ITC) traces (top) and binding isotherms (bottom) of nAlb binding to human and mouse serum albumin at pH 7.5. (d) Reaction scheme for generating molecularly homogeneous nAlb conjugates through site-selective enzymatic ligation of an amine-PEG3-azide followed by conjugation of agonist or dye cargo through copper-free click chemistry addition. (e) Structure of diABZI STING agonist conjugated to a DBCO-PEG11 handle for ligation to azide-functionalized nanobodies via click-chemistry. (f) Electrospray ionization mass spectrometry (ESI-MS) and (g) sodium dodecyl sulfate polyacrylamide electrophoresis (SDS-PAGE) demonstrating nanobody conjugate purity and molecular weight. (h-i) Dose-response curves in (h) A549-Dual (n=3) and (i) THP1-Dual IFN-I reporter cell lines (n=3) with estimated EC50 values indicated in the legends. (j) qPCR analysis of gene expression in murine bone marrow derived macrophages (BMDM) treated in vitro with 0.25 \u00b5M of free diABZI or nAlb-diABZI conjugate (n=3). P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to PBS control. Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Anti-albumin nanobodies increase cargo delivery to tumor sites to promote uptake by cancer cells and tumor-associated myeloid cells. (a) Dose-response curve for nanobody-Cy5 conjugate surface binding and intracellular uptake at 37 \u00b0C and 4 \u00b0C measured by flow cytometry in EGFR- (THP-1) in vitro. (b) Uptake of nAlb-Cy5 (2 \u00b5M) in RAW 264.7, EMT6, and BMDM cells with the addition of control PBS (-EIPA) or macropinocytosis inhibitor (+EIPA). (c) Colocalization of Cy5 (red) with lysotracker green (green) and Hoechst (blue) in RAW 264.7 cells with (d) percent colocalization determination for nAlb-Cy5 and nGFP-Cy5 in RAW 264.7 and EMT6 cells. (scale bar: 100 \u00b5m) (e) Pharmacokinetics of free DBCO-Cy5 dye and indicated nanobody-Cy5 conjugates injected intravenously at 2 mg/kg in healthy female C57BL/6 mice (n=5). Elimination phase half-life and area under the curve (AUC) are indicated in legend. (f) Representative IVIS fluorescent images of excised tumors and major organs and (g) quantification of average radiant efficiencies 24 h following intravenous administration of vehicle (PBS), DBCO-Cy5, nEGFR-Cy5, and nAlb-Cy5 at 2 mg/kg to female Balb/c mice with orthotopic EMT6 breast tumors (n=5-8). P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to the tumor. (h-i) Quantification of percent injected dose per gram of tissue (% ID/g) 24 h following intravenous administration of vehicle (PBS) and nAlb-Cy5 at 2 mg/kg to (h) female Balb/c mice with orthotopic EMT6 breast tumors (n=5) and (i) female C57BL/6 mice with subcutaneous B16.F10 tumors (n=5). P values determined by two-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001. (j) Representative fluorescent microscopy images of tumor sections stained for DAPI (blue), CD45 (green), and CD31 (red) 24 h following administration of nAlb-Cy5 (yellow) alone or in combination with nAlb-diABZI (scale bar: 200 \u00b5m). (k-l) Flow cytometric analysis of nAlb-Cy5 cellular uptake by in EMT6 tumors evaluated as (k) the percentage of indicated cell type comprising all Cy5+ live cells or (l) as the percentage of Cy5+ cells within an indicated live cell population 24 h following administration of vehicle (PBS), nAlb-Cy5 alone, or nAlb-Cy5 co-administered with nAlb-diABZI; median fluorescent intensities (MFI) for each cell population is shown in Fig. S12 (n=7-8). Inset: percentage of indicated cell population in the tumor as measured by flow cytometry. DC: dendritic cell; Mj: macrophage; MDSC: myeloid derived suppressor cell; NK: natural killer cell. P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to PBS control. Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Albumin-hitchhiking STING agonist inhibits breast tumor growth by shifting the immunocellular profile of the TME. (a) Schematic of EMT6 tumor inoculations, treatment schedule, and study end point for gene expression and flow cytometry analysis. (b) Tumor growth curves, and (c) spider plots of individual tumor growth curves for each mice with EMT6 tumors treated as indicated (n=8-9). SEM with P value determined by two-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; ****P<0.0001 on day 17 for all groups compared to PBS. (d-j) Flow cytometric analysis of breast tumors and spleen 24 h following final dose of nAlb-diABZI. (d) tSNE plots of live cells in EMT6 tumors colored by cell population with relative expression level of Ki67, CD69, and PD-1 as indicated on heat map. DC: dendritic cell; Mj: macrophage; NK: natural killer cell; MDSC: myeloid-derived suppressor cell. (e-f) Heat maps summarizing (e) the fold change in the percentage of indicated cell population and (f) fold change in the frequency of NK cells, CD8+ T cells, and CD4+ T cells expressing the indicated marker or marker combination in EMT6 breast tumors. (g) Quantification of Ki67+CD69+ and Ki67+PD1+ CD8+ and CD4+ T cells in EMT6 tumors following treatment with vehicle (PBS) or nAlb-diABZI. (h) Quantification of frequency of MHC-II+ and PD-L1+ macrophages in EMT-6 tumors following treatment with vehicle (PBS) or nAlb-diABZI. (i) Heat map summarizing fold change in the frequency of NK cells, CD8+ T cells, and CD4+ T cells expressing activation markers within splenic populations. *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 indicate a statistically significant difference for heat-maps between PBS and AP-diABZI treated groups as determined by two-way ANOVA. (j) Quantification of Ki67+CD69+ and Ki67+PD1+ CD8+ and CD4+ T cells in spleens. *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 indicate a statistically significant difference between PBS and nAlb-diABZI treated groups as determined by Student\u2019s t-test, n = 6 per group. Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Design, synthesis, and testing of bivalent nanobody-STING agonist conjugate for albumin-hitchhiking and targeting of PD-L1. (a) Scheme for the cloning, expression, and bioconjugation of small molecule cargo to generate the AP-diABZI conjugate. (b) SDS-PAGE and (c) ESI-MS confirming the purity and molecular weight of AP conjugates. (d-e) Dose-response curves for indicated nanobody-diABZI conjugate in (d) A549-Dual (n=3) and (e) THP1-Dual IFN-I reporter cell lines (n=3) with estimated EC50 values indicated in the legends. (f) qPCR analysis of genes associated with STING activation in bone marrow derived macrophages (BMDMs) in response to treatment at discrete time points (0, 0.5, 1, 2, 3, 4, 6, 8, 24 h) with indicated agonist at 0.25 \u00b5M (n=3). P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to PBS control. (g-h) Dose response curve for nAlb-Cy5 and AP-Cy5 conjugate surface binding and intracellular uptake at 37 \u00b0C and 4 \u00b0C measured by flow cytometry in (g) B16.F10 cells (n=2-3) and (h) EMT6 cells (n=3). (i) Mean fluorescent intensity (MFI) for nAlb-Cy5 and AP-Cy5 conjugate surface binding at 2 \u00b5M compared to PBS (0 \u00b5M) for EMT6 W.T. and EMT6 PD-L1 K.O. cell lines at 37 \u00baC. (j) Pharmacokinetics of indicated nanobody-Cy5 conjugate in healthy Balb/c female mice (n=5). Elimination phase half-life and area under the curve (AUC) are indicated in legend. (k) Representative IVIS fluorescent images of excised tumors and major organs (left) and quantification of average radiant efficiencies (right) of tumors and major organs 48 h after administration of nPD-L1-Cy5 and AP-Cy5 in mice with EMT6 breast tumors (n=3-4). P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to the tumor group. (l) Comparison of Cy5 radiant efficiencies in tumor tissue 48 h following administration of indicated nanobody-Cy5 conjugate. P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to the PBS control, as well as between nAlb-Cy5 and AP-Cy5. (m) Representative IVIS fluorescent images of excised tumors and major organs (left) and quantification of average radiant efficiencies (right) of tumors and major organs 48 h after administration of AP-Cy5 in mice with EMT6 W.T. and EMT6 PD-L1 K.O. breast tumors (n=5). P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to the tumor group. Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Systemic administration of AP-diABZI conjugates enhance antitumor immune and therapeutic responses in EMT6 breast cancer model. (a) Schematic of EMT6 tumor inoculation and treatment schedule (n=10). Anti-PD-L1 IgG (ICB) was injected I.P. at 100 \u00b5g and all nanobodies were injected I.V. at 1.25 \u00b5g of diABZI per injection. (b) Tumor growth curves, (c) spider plots of individual tumor growth curves, and (d) Kaplan-Meier survival plots for mice with EMT6 tumors treated as indicated. CR = complete responder; SEM with P value determined by two-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; ****P<0.0001 on day 22 for all groups compared to PBS. Endpoint criteria of 1500\u2009mm3\u00a0tumor volume with P value was determined by log-rank test; ****P<0.0001 compared to PBS control. (e) Spider plots of individual tumor growth curves and (f) Kaplan-Meier survival curves of mice challenged or re-challenged (complete responders after first treatment regimen) with EMT6 cells (n=9-10). (g) Scheme of EMT6 W.T. and EMT6 PD-L1 K.O. tumor inoculation and treatment schedule (n=5-13). AP-diABZI was injected I.V. at 1.25 \u00b5g of diABZI. (h) Kaplan-Meier survival plots for mice with EMT6 W.T. or PD-L1 K.O. tumors treated as indicated. (i-j) Volcano plots representing significance (-log10) and fold change (log2) for gene expression analysis in (i) nAlb-diABZI vs. PBS (n=4) and (j) AP-diABZI vs. PBS (n=4). (k-m) Heat maps of NanoString gene cluster matrices showing Z score fold changes for (k) functional gene annotations, (l) biological signatures, and (m) cell types. Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "AP-diABZI activates a tumoricidal NK and T cell response. Flow cytometric analysis of orthotopic EMT6 breast tumors 24 h following two intravenous doses of AP-diABZI (1.25 \u00b5g, n=8), or PBS (n=7). (a) tSNE plots of live cells in EMT6 tumors colored by cell population with relative expression level of Ki67, CD69, PD-1, and PD-L1 as indicated on heat map. DC: dendritic cell; Mj: macrophage; NK: natural killer cell; MDSC: myeloid-derived suppressor cell. (b) Heat map summarizing the fold change in the percentage of indicated cell populations in EMT6 tumors. (c) Bar plots showing an increase in CD8+ cells and the ratio of CD8+ to CD4+FoxP3+ cells (as precent of CD3+ tumor cells). (d) Quantification of Ki67+CD69+ and Ki67+PD1+ CD8+ T cells in EMT6 tumors. *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 indicate a statistically significant difference between PBS and AP-diABZI treated groups as determined by Student\u2019s t-test. (e) Spleen phenotyping heat map of frequency of NK cells, CD8+ T cells, and CD4+ T cells. *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 indicate a statistically significant difference for heat-maps between PBS and AP-diABZI treated groups as determined by two-way ANOVA. (f) Schematic of EMT6 tumor inoculation and treatment schedule with depletion antibodies (n=7-13). Anti-Asialo GM1 (NK) IgG, anti-CD8 IgG, and anti-CD4 IgG were injected I.P. at 100-200 \u00b5g and AP-diABZI was injected I.V. at 1.25 \u00b5g of diABZI per injection. (g) Tumor growth curves, and (h) Kaplan-Meier survival plots for mice with EMT6 tumors treated as indicated. CR = complete responder; SEM with P value determined by two-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; ****P<0.0001 on day 22 for all groups compared to PBS. Endpoint criteria of 1500\u2009mm3\u00a0tumor volume with P value was determined by log-rank test; ****P<0.0001 compared to PBS control. Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_7.png",
+        "caption": "Albumin-hitchhiking STING agonists stimulate antitumor immunity in B16.F10 melanoma tumor model. (a) Schematic of B16.F10 tumor inoculation and treatment schedule. (b) Tumor growth curves, (c) spider plots of individual tumor growth curves, and (d) Kaplan-Meier survival plots (n=10-15). Anti-PD-L1 IgG (ICB) was injected I.P. at 100 \u00b5g and all nanobodies were injected I.V. at 1.25 \u00b5g of diABZI per injection. (b) SEM with P value determined by two-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; ****P<0.0001 on day 18 for all groups compared to PBS. (d) Kaplan-Meier survival curves of mice treated with indicated formulation using 1500\u2009mm3\u00a0tumor volume as endpoint criteria with P value was determined by log-rank test; ****P<0.0001 compared to PBS control. (e) Schematic of B16.F10-OVA tumor inoculation, treatment schedule, and study end point for flow cytometry analysis (n=12). AP-diABZI was injected I.V. at 1.25 \u00b5g of diABZI per injection. (f) Tumor weight on day 15 for mice with B16.F10-OVA tumors treated with AP-diABZI or PBS. (g) Frequency of CD4+ and CD8+ T cells in the spleen at study endpoint. Flow cytometric analysis of the frequency of (h) CD69+ activated T cells, (i) CD44+CD62L- effector memory T cells, (j) CD44-CD62L+ na\u00efve T cells, and (k) CD44+CD62L+ central memory T cells. (l) SIINFEKL/H-2kB tetramer staining was performed to determine the frequency of OVA-specific CD8+ T cells in the spleen at study endpoint. (m) Representative flow cytometry dot plots demonstrating the distribution of CD8+ TEM (CD44+CD62L-) and TCM (CD44+CD62L+) within the OVA-specific (tetramer+) and non-OVA-specific (tetramer-) populations. *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 indicate a statistically significant difference between PBS and AP-diABZI treated groups as determined by Student\u2019s t-test. Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_8.png",
+        "caption": "Albumin-hitchhiking STING agonists improve immunotherapy responses in a model of lung metastatic melanoma and adoptive T cell transfer therapy. (a) Schematic of B16.F10-LUC I.V. tumor inoculation, treatment schedule, and study end point for analysis of lung tumor burden (n=11-15). Anti-PD-L1 IgG (ICB) was injected I.P. at 100 \u00b5g and all nanobodies were injected I.V. at 1.25 \u00b5g of diABZI per injection. (b) Representative images of lungs and (c) lung weights of mice treated as indicated. (d) Representative IVIS luminescent images and (e) quantification of average radiance from luciferase expressing B16.F10 within isolated lung tissue. P values determined by one-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; *P\u22640.05, **P\u22640.01, ***P\u22640.001, and ****P<0.0001 compared to the PBS control. (f-i) Evaluation of AP-diABZI as an adjuvant therapy for adoptive OT-I T cell transfer therapy in a B16.F10-OVA model (n=15). (f) Schematic of B16.F10-OVA tumor inoculation and of treatment schedule with OT-I transfer (0.5 million OT-I T cells) either on day 9 (OT-I alone or one dose (1.25 \u00b5g) AP-diABZI pre-treatment) or day 15 (three dose AP-diABZI pre-treatment). (g) Tumor growth curves, (h) spider plots of individual tumor growth curves, and (i) Kaplan-Meier survival curves. (g) P value determined by two-way ANOVA with post-hoc Tukey\u2019s correction for multiple comparisons; ****P<0.0001 on day 17 for all groups compared to PBS. (i) Kaplan-Meier survival curves of mice treated with indicated formulation using 1500\u2009mm3\u00a0tumor volume as endpoint criteria with P value was determined by log-rank test; ****P<0.0001 compared to PBS control. (CR = complete responder). Replicates are noted as biological, and data shown as mean \u00b1 SEM.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

02772d9f9a415f35d68ba0d93009adbd66773e0083b3ec1cb5c92a6cad853ac5/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Experimental approach. Nanoscale guest-host interactions were studied by: (A) synthesizing (defect-engineered) surface-anchored ZIF-8 microcrystals (Mim: 2-methylimidazole). The structure of ZIF-8 microcrystals is shown on the right, with color-coded crystal plane terminations; (B) applying in situ Photo-induced Force Microscopy (PiFM), a nano-infrared technique with a spatial resolution down to 5 nm, in the presence of formaldehyde vapor. Using hyperspectral imaging we mapped preferential sorption and conversion sites on ZIF-8 crystals.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Experimental evidence of structure sensitive formaldehyde sorption on ZIF-8. (A) An Atomic Force Microscopy (AFM) image of ZIF-8 shows the expression of different crystal planes. (B) Hyperspectral images, with a full IR spectrum at every pixel, of such ZIF-8 crystals were recorded during formaldehyde exposure. Overall spectrum intensity per pixel is denoted in mV. Individual planes within these images were analyzed by (C) creating morphology-based masks. (D) Crystal-averaged PiFM spectra at increasing formaldehyde (FA) pressure show the increase of three infrared bands indicating gas sorption. (E) Gas sorption occurs by breaking the C=O bond and forming a covalent CFA-NZIFbond. At the same time, FA is coordinated to the Zn atom through an O-adduct. Contour plots show the formaldehyde sorption-induced change in the IR spectrum for a corner (F), an edge (G), and a facet plane (H). (I) From the contour plots, response pressures to formaldehyde, i.e., the pressure at which a response was recorded, for these crystal planes were found, highlighting the structure sensitive sorption of formaldehyde.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Experimental and theoretical descriptions of heterogeneous gas sorption on various ZIF-8 surface terminations. (A) Representative infrared image mapping a chemisorbed formaldehyde vibration (1200 cm-1), showing the presence of nanoislands of alternate sorption behavior within the pristine {100} and {110} ZIF-8 facets. (B) Infrared map of a ZIF-8 framework vibration (1590 cm-1), showing an inverted image, ruling out topography-induced PiFM signal. (C) Point spectra taken on and off the nanoislands (markers in (A, B)) showcase non-homogenous sorption behavior, as well as the validity of the IR mapping technique. (D) Crystal surface termination chemistry depends on the orientation and height of the cut of the ZIF-8 crystal, indicated by colored lines (color coding on the right). (E) These terminations can be grouped into surface motifs describing Zn/linker density and positioning, where the surface energy is plane cut-dependent. Density Functional Theory (DFT) models of (formaldehyde sorption on) the most stable (F) {100}, (G) {110}, and (H) {310} planes show the plane-dependent energy and the corresponding formaldehyde adsorption energy. Furthermore, models show that formaldehyde binds to the framework via the formation of covalent OFA-ZnZIF and CFA-NZIFbonds.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Nanoscale visualization of heterogeneity within crystal planes of ZIF-8. (A) Representative AFM image of pyrrole-d5defect-engineered SURZIF-8, showing a higher ratio of {100}/{110} facets. (B) Spectroscopic evidence for defect linker incorporation was found when comparing the crystal-averaged spectra of a pristine (solid trace) and a defective (dashed trace) crystal. The defective ZIF-8 spectrum showed \u03b4CDand \u03bdC=N vibrations at 960 and 885 cm-1 corresponding to the defect linker, and \u03b4OH vibrations at 840 and 790 cm-1corresponding to defect-induced Zn-OH sites, see inset. (C) Peak ratio showing defects and Zn-OH concentrations by crystal plane in defective crystals, showing higher concentrations for high energy planes. Crystal-averaged peak ratios are given by solid lines. (D) A clustered 200x200 nm2 hyperspectral image and (E) corresponding cluster spectra of defective ZIF-8 under N2showing one {100} facet and two {110} facets. Clusters with similar chemical identities were calculated using Principal Component Analysis (PCA) and clustering, to visualize ZIF-8 defect linker distribution (\u03b4CD and \u03bdC=N in E). (F) bubble plot where bubble size corresponds to the fraction of a facet attributed to a certain cluster, and the bubble position on the y-axis indicates the relative defect concentration. Facet-averaged defect concentrations are marked with a star. This plot shows differences in relative defect concentrations among facets with comparable surface energies. However, it also shows that strong concentration differences between each facet and their highly localized intra-facet domains can be found.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Structure sensitive conversion of formaldehyde (FA) over defect-engineered ZIF-8. (A) Contour plot of the crystal surface-averaged in situ infrared spectra of a defective ZIF-8 crystal, showing the formation of formate, di-/poly-oxymethylene (DOM/POM), and methoxy surface species (relevant bands indicated by dashed lines in the contour plot). (B) A plot showing plane-dependent response pressures for intermediate surface species, thus showing structure sensitive FA conversion, as intermediates are observed at lower pressures on high-index planes (note: pyrrole-induced changes in crystal aspect ratios prevented the analysis of the {111} facets). Based on the observed species and in accordance with literature,[32,48] we propose two possible mechanisms for FA conversion over defective ZIF-8: a methoxy-mediated (C) and formate-mediated (D) mechanism.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

02772d9f9a415f35d68ba0d93009adbd66773e0083b3ec1cb5c92a6cad853ac5/preprint/preprint.md

@@ -0,0 +1,176 @@
+# Abstract
+
+Many catalytic processes depend on the sorption and conversion of gaseous molecules on the surface of (porous) functional materials. These events often preferentially occur on specific, undercoordinated, external surface sites, a phenomenon known as structure sensitivity. However, so far, the study of such site-specific gas sorption/conversion behavior of porous functional materials has been limited due to a lack of nanoscale *in situ* characterization techniques. Here we show the combination of *in situ* Photo-induced Force Microscopy (PiFM) with Density Functional Theory (DFT) calculations to study the sorption and conversion of formaldehyde on the external surfaces of well-defined faceted ZIF-8 microcrystals with nanoscale resolution. We observed preferential adsorption of formaldehyde on high index planes, in accordance with surface energy stabilization criteria. Moreover, *in situ* PiFM allowed us to visualize unsaturated nanodomains within extended external crystal planes, showing enhanced sorption behavior on the nanoscale. After incorporation of defective linkers, structure sensitive conversion of formaldehyde through a methoxy- and a formate mechanism mediated by Lewis acidity was found. Strikingly, sorption and conversion were influenced more by the external surface termination than by the concentration of defects. DFT calculations showed that this is due to the presence of specific atomic arrangements on high-index crystal surfaces, reminiscent of enzymatic binding sites. With this research, we showcase the high potential of *in situ* PiFM for structure sensitivity studies on porous functional materials.
+
+# Main Text
+
+The performance of a functional material is often determined by only a small percentage of its surface sites and their atomic configuration. This phenomenon, known as structure sensitivity, describes the relationship between the fraction of exposed crystal surfaces and rate of conversion and is well known in the field of heterogeneous catalysis. [1–3] For example, in supported metal nanoparticle catalysts, the variation of metal nanoparticle size results in the exposure of a different fraction of active surface sites, leading to size-dependent performance.[4]
+
+Analogously, it is often observed for porous functional materials, such as metal-organic frameworks (MOFs), that their functionality is strongly dependent on their outer surface, despite their inner porosity.[5] As a result, evidence for structure sensitivity of porous functional materials can be found in literature.[6,7] For example, Pang and coworkers showed by ex situ Scanning Electron Microscopy (SEM) analysis that when ZIF-8 crystals were exposed to acidic SO₂ gas, the external surface of the {100} facet was more stable than the {110} facet.[8] However, such an ex situ technique is not able to describe the guest-host interaction during gas exposure. Additionally, the performance of undercoordinated crystal surface edges and corners was not considered,[8] while these high-energy crystal planes often present highly important sorption, and/or conversion, sites on functional materials.[9,10]
+
+Undercoordinated sites are purposely introduced in defect engineering approaches, for example through the incorporation of defective linkers.[11] The resulting undercoordinated metal centers can behave as Lewis acid sites with potentially increased activity towards gas sorption and/or conversion.[12] The localized integration of defects can be characterized by highly sensitive techniques, such as High-Resolution Transmission Electron Microscopy (HRTEM), Scanning Electron Diffraction (SED), or Atom Probe Tomography (APT).[13–17] However, to study the nanoscale guest-host interaction between (defective) external surface sites of functional porous materials, novel surface-sensitive in situ techniques are required.[18–21]
+
+Infrared spectroscopy is an ideal technique to describe guest-host interactions [22], but it suffers from low spatial resolution. Moreover, the low ratio of crystallite surface vs. bulk atoms will result in a loss of information on the external surface performance of porous functional materials. Tip-Enhanced Raman Spectroscopy (TERS) and Photo-induced Force Microscopy (PiFM), are AFM-based vibrational techniques able to circumvent these limitations, with a spatial resolutions down to the nanoscale (e.g. 5 nm for PiFM).[23,24] Both techniques are non-destructive, label free, can be performed in situ, and can be used to study a broad range of materials.[25] However, TERS suffers from arduous tip design- and preparation protocols, as well as measurement stability issues, which are not encountered when using PiFM. We recently showcased how PiFM can be used to study guest-host interactions in situ, looking at water sorption and induced defect formation on the surface of archetypical MOF thin films.[26] However, studying structure sensitivity requires the use of well-defined systems, where different crystal surface terminations can be imaged simultaneously under the same conditions.
+
+In this work, we used *in situ* PiFM, in combination with Density Functional Theory (DFT) calculations, to unravel structure sensitivity in gas sorption and conversion on the external surface of a microcrystalline MOF material in unprecedented detail (Fig. <span class="InternalRef" refid="Fig1">1</span>). Here, we show that with this toolbox we can purposely study highly functional, external surfaces of a MOF crystal planes during gas exposure, which is not possible through conventional, bulk techniques. We show that only data averaged over the entire crystal, including unavoidable defective fractions, is accessible via standard FTIR spectroscopy. We have used surface-anchored crystals of ZIF-8, a MOF with exceptional stability and potential applications in catalysis.[27,28] We used such ZIF-8 crystals with well-defined crystal plane terminations for nanoscale guest-host investigations with formaldehyde, a Volatile Organic Compound (VOC) indoor pollutant and model reactant for identifying C1-molecule conversion mechanisms.[29–32] Using this approach, we found that gas sorption preferentially occurred on high index crystal planes, such as corners and edges, of only 10s of nanometers in size within micrometer-sized crystals. Interestingly, we also found high-energy nanodomains within low-energy facets resulting in heterogeneous intra-facet sorption behavior. Furthermore, we observed that the incorporation of defects (i.e., pyrrole ligands) mainly took place at these nano-size high-energy planes and domains on the ZIF-8 surface. These defects led to the structure sensitive conversion of formaldehyde resembling the “type II” structure sensitivity in supported catalysts, favored on unsaturated sites.[1] Using our in situ nano-spectroscopy technique, two distinct formaldehyde conversion mechanisms on single crystal planes were evidenced. Overall, we found the sorption and conversion performance of the ZIF-8 crystals to be heavily dependent on structure sensitivity phenomena, and we expect the gathered insights and approach to be highly relevant for a wide range of micro-and nano-structured functional materials.
+
+# Results And Discussion
+
+## Structure sensitive formaldehyde sorption
+
+Surface anchored ZIF-8 microcrystals with well-defined facets were synthesized through a Layer-by-Layer (LbL) synthesis.[33,34] Pristine ZIF-8 crystals were synthesized using zinc nitrate and 2-methylimidazole linkers and AFM images showed that the crystals were 0.5-1 micrometer in size and expressed six well-defined crystal plane terminations, namely {100} and {110} facets, {210} and {211} edges, and {111} and {310} corners (Fig. <span class="InternalRef" refid="Fig2">2</span> A, S10).[35] These ZIF-8 crystals were used for nanoscale guest-host interaction studies using <em>in situ</em> Photo-induced Force Microscopy (PiFM, Fig. <span class="InternalRef" refid="Fig1">1</span> B). Hyperspectral (4-D) images of the crystals were recorded, composed of a full infrared spectrum (1-D) on every pixel (3-D: x, y plane, and z height), during stepped gas sorption from 0-480 ppm of formaldehyde (16 vol.% in H<sub>2</sub>O, Fig. <span class="InternalRef" refid="Fig2">2</span> B). In order to only absorb FA at the surface of the material, the maximum formaldehyde pressure was kept well below the range of critical pressures for the so-called “gate opening” of ZIF-8 pores, at which point, adsorption in the bulk of the material occurs (Figure S35). [36, 37]
+
+The surface-averaged <em>in situ</em> IR spectra for entire ZIF-8 crystals show the sorption of formaldehyde (FA) on the pristine ZIF-8 surface (Fig. <span class="InternalRef" refid="Fig2">2</span> D, full spectra and difference spectra found in Figure S18, 19). FA adsorption was observed through the increase of IR bands at 1280, 1200, and 895 cm<sup>-1</sup>, corresponding to δ<sub>CH2</sub>, rock, δ<sub>CH2</sub>, wag, and ν<sub>C-O</sub>, respectively (Fig. <span class="InternalRef" refid="Fig2">2</span> E).[38,39] The ν<sub>C=O</sub> vibration of physisorbed formaldehyde was not observed as the C=O double bond was broken due to strong chemisorption of formaldehyde on the ZIF-8 surface (<em>vide infra</em>).
+
+To study the sorption of FA as a function of crystal planes, we masked facet/edge/corner areas of the hyperspectral image based on morphology data (Fig. <span class="InternalRef" refid="Fig2">2</span> C, S11-S15). Subsequently, for each crystal plane, we created contour plots displaying their <em>in situ</em> nano-IR spectra (Fig. <span class="InternalRef" refid="Fig2">2</span> F-H, S16-S19, Table S4). Each of the planes showed the appearance of formaldehyde vibrations. However, the pressure at which these δ<sub>CH2, rock</sub>, δ<sub>CH2, wag</sub> bands appeared, i.e. the response pressure, differed for the different crystal planes (Fig. <span class="InternalRef" refid="Fig2">2</span> I). This revealed that sorption occurs first on corners, then edges, and finally on facets in the following order: {310} > {111} > {210} > {211} > {110} ~ {100}. This demonstrated the structure sensitive sorption of formaldehyde on ZIF-8 crystal planes, with the expected preference for high index facets.
+
+However, in most porous functional materials isolated and clustered defects are expected to be present in the crystal structure.[40] Therefore, we exploited the nanometer resolution of <em>in situ</em> PiFM to investigate intra-plane heterogeneities in gas sorption behavior. Figure <span class="InternalRef" refid="Fig3">3</span> A shows a 500x500 nm<sup>2</sup> intensity map of the 1200 cm<sup>-1</sup> band, corresponding to chemisorbed formaldehyde, over a pristine ZIF-8 crystal under 300 ppm formaldehyde, which was found to be sufficient for formaldehyde sorption on corners and edges, but insufficient for sorption on facets (Fig. <span class="InternalRef" refid="Fig2">2</span> I). This IR map evidences sorption on corners and edges, as well as enhanced FA signals on nanosized domains <em>within</em> the {100} and {110} planes.
+
+The nano-islands observed in pristine ZIF-8 microcrystals may be explained in three ways: (i) as an artifact induced by overall lower PiFM intensity; (ii) as a result of heterogeneity of ZIF-8 (e.g. a presence of different crystal surface terminations, <em>vide infra</em>); or (iii) as coverage dependent patterns of FA over homogeneous surfaces as a result of interactions between adsorbed FA molecules, similarly to what is observed for e.g. CO adsorption on metal surfaces.[41] To rule out factor (i), we took point spectra in and out of the nano-islands, which showed comparable ZIF-8 vibration intensities, but differences in the FA bands (Fig. <span class="InternalRef" refid="Fig4">4</span> C, S20). Furthermore, we performed IR mapping of the same area, but with the laser tuned to an aromatic vibration of the ZIF-8 framework (ν<sub>C=N</sub> 1590 cm<sup>-1</sup>, Fig. <span class="InternalRef" refid="Fig4">4</span> B, S21), which showed an inverted image. This showed both the surface sensitivity and validity of the PiFM method and that the external surface is intrinsically heterogeneous, thus supporting hypothesis (ii) over the others (which assume homogeneous surfaces).
+
+To study these heterogeneous planes and to link gas sorption behavior to specific external surface sites within these planes, we modeled three representative ZIF-8 crystallite surfaces using DFT calculations, namely the {100} and {110} facet planes, and a high index {310} corner plane (Fig. <span class="InternalRef" refid="Fig3">3</span> D-H). The {100} and {110} planes were chosen because of their high relative crystal surface coverage, and because of the nanoislands observed within these facets. The {310} plane was selected as representative of minority, high-energy terminations showing alternate behavior to formaldehyde vapor. It is crucial to realize that for each of the {100}, {110} and {310} planes, different surfaces can be constructed by varying the slicing height (Fig. <span class="InternalRef" refid="Fig3">3</span> D). These surfaces can be grouped into motifs according to the coordination environment of the Zn atom as well as the orientation of terminal ligands (Fig. <span class="InternalRef" refid="Fig3">3</span> E, S4). These motifs differ in the density of Zn atoms per surface area as well as in the number of Zn-N bonds cleaved, giving rise to various under-coordinated sites. The computed surface energies for the most stable {100}, {110}, and {310} terminations were found to be -3.21, -4.91, and −2.67 meV/Å<sup>2</sup>, respectively, confirming the higher-energy nature of higher index planes.[42]
+
+However, high-energy cuts of, for example, the {100} and {110} planes with energies of 3.45 meV/Å<sup>2</sup> and 0.62 meV/Å<sup>2</sup>, respectively, were also found. Because of the moderate differences in surface energies between low and high energy cuts of the {100} and {110} planes, multiple terminations may co-exist on the facet surfaces, for example as a result of entropic factors. Therefore, the formation of high-energy planes can be favored during synthesis conditions (rather than the 0K modeling conditions), leading to expression of the observed nanoislands.[43] Note that negative values for the most stable surface energies were found as a result of the use of a saturated surface model (SI). Using unsaturated surface models did not change our conclusions regarding relative stabilities of different surface orientations (Table S2).
+
+We subsequently modeled FA adsorption on these representative ZIF-8 terminations. In general, we found that the formaldehyde molecules chemisorb on ZIF-8 by forming a carbon adduct with a nitrogen atom from the linker, while the oxygen atom adsorbs on a Zn center.[5,44] Such chemisorption is consistent with the absence of molecularly adsorbed formaldehyde in the infrared spectra (Fig. <span class="InternalRef" refid="Fig2">2</span> D). Additionally, we found that (at 0 K) FA adsorption is an endothermic process (SI). Overall, the trend in FA binding on the most stable cleavages is consistent with experimental findings showing the order {310} > {110} > {100} (Fig. <span class="InternalRef" refid="Fig3">3</span> F-H, S5). As the most reactive surface termination we identify motif ІІІ, which expresses an isolated Zn site with two terminal linkers, one protonated and one unprotonated (Fig. <span class="InternalRef" refid="Fig3">3</span> E, S5), which can be found on {310} edges as well as on high energy cleavages of {100}. Further inspection of the high-energy cleavages showed their stronger FA binding behavior compared to their low-energy counterparts (e.g., 26.8 kJ/mol versus 74.1 kJ/mol for {100} and 47.1 kJ/mol versus 57.7 kJ/mol for {110}, Figure S5). Overall, our calculations supported the experimental observations of inter- and intra-plane structure sensitive sorption, dictated by the formation of nanodomains exposing high surface energy terminations, as described by hypothesis (ii).
+
+## Structure sensitive defect engineering
+
+To further explore site-specific FA adsorption, we purposefully incorporated 10% of pyrrole defect linker in the ZIF-8 crystals to enrich the surface chemistry with undersaturated metal sites (Fig. <span class="InternalRef" refid="Fig1">1</span> A). Atomic Force Microscopy (AFM) images of the defective crystals showed a change in the expression of its crystal planes: an increase in {100}:{110} facet aspect ratio was observed (Fig. <span class="InternalRef" refid="Fig4">4</span> A, S9, S10). This suggested a change in relative plane energies upon defect-engineering, which we attribute to strain relaxation of the external surface terminations.[45] These findings are supported by DFT results, which show that an exchange of imidazole linker with pyrrole is thermodynamically most favorable on the high energy cut of the {100} plane (Table S3), thereby effectively lowering a difference in surface energies between {100} and {110} planes, resulting in favored {100} expression.
+
+To experimentally and theoretically verify whether defect linker incorporation can be linked to surface energies, we measured a 1x1 µm<sup>2</sup> hyperspectral image of a defective crystal in N<sub>2</sub>. Crystal-averaged IR spectra showed the vibrational fingerprint of pyrrole-d<sub>5</sub> by two IR bands at 960 and 885 cm<sup>-1</sup>, corresponding to δ<sub>CD</sub> and ν<sub>C=N</sub>, respectively (Fig. <span class="InternalRef" refid="Fig4">4</span> B, S22).[46] Further inspection of the spectral data also revealed the presence of infrared bands at 840 and 790 cm<sup>-1</sup>, corresponding to two δ<sub>OH</sub> vibrations of Zn-OH<sub>2</sub> sites.[47] These bands indicate the enrichment of the external ZIF-8 surface chemistry with Lewis acid sites (Zn<sup>2+</sup>) and weakly basic sites (Zn-OH) upon incorporation of pyrrole. We then averaged the spectra of facet, edge, and corner-pixels and calculated their 885/1590 cm<sup>-1</sup> (pyrrole/imidazole) peak ratios. We plotted these ratios together with the crystal-averaged ratio in Fig. <span class="InternalRef" refid="Fig4">4</span> C. These ratios showed that expression of defect sites scaled with surface energy and is thus structure sensitive, with the {310}, {210}, and {211} planes having an above-average concentration of defects.
+
+Isotope-labeled deuterated pyrrole was used to pinpoint the distribution and clustering of pyrrole defects using their spectral fingerprint, and to correlate defect-specific guest-host interactions. To visualize pyrrole incorporation, we acquired another 200x200 nm<sup>2</sup> hyperspectral image of a defective ZIF-8 crystal in N<sub>2</sub>. Subsequently, we used Principal Component Analysis (PCA) and clustering of the hyperspectral image, to group pixels into clusters based on spectral similarities (Fig. <span class="InternalRef" refid="Fig4">4</span> D, SI). The spectra of the clusters revealed that clustering mainly occurred based on defect concentration, as all spectra showed a largely unchanged ZIF-8 spectrum yet showed variation in relative δ<sub>CD</sub> and ν<sub>C=N</sub> band intensities (Fig. <span class="InternalRef" refid="Fig4">4</span> E, S22).
+
+To deduce the location and degree of clustering of defects, we first sectioned the clustered image into one {100} and two {110} facets, based on the morphology map (Fig. <span class="InternalRef" refid="Fig4">4</span> D, S22). Then, we calculated the fraction of facet surface that was covered by each cluster (Figure S23-S26) which is represented by the bubble sizes in Fig. <span class="InternalRef" refid="Fig4">4</span> F. Subsequently, we quantified the relative defect linker concentration of each cluster by calculating the peak ratio between pristine and defective ZIF-8 vibrations (i.e. 885/1590 cm<sup>-1</sup>) for the 5 clusters (Fig. <span class="InternalRef" refid="Fig4">4</span> F, y-axis). The weighted average of defect linker concentration is also reported per facet (stars in Fig. <span class="InternalRef" refid="Fig4">4</span> F). In this way, we were able to distinguish between plane-averaged defect concentrations and nano-domains of high/low defect concentration. These results showed that defect concentrations varied more intra-facet, i.e., between parent facet and nanoislands, than inter-facet, an experimental insight that could only be gained because of the nanoscale resolution of the PiFM method.
+
+Finally, we modeled defect incorporation with DFT (Table S3). These calculations confirmed that pyrrole is preferentially incorporated into high-energy planes. Notably, this then suggests that the defect-rich nanodomains observed on {110} and {100} facets are indeed due to high energy terminations, further supporting hypothesis (ii) for observation of reactive nanoislands in pristine ZIF-8.
+
+## Structure sensitive formaldehyde conversion
+
+To discern between the effects of crystal plane terminations and pyrrole defects on the structure sensitive adsorption of formaldehyde, we performed <em>in situ</em> PiFM FA adsorption experiments on the defective crystals (Fig. <span class="InternalRef" refid="Fig5">5</span> A, S27). The rising intensity of 1280, 1200, and 895 cm<sup>-1</sup> bands with FA pressure (0-480 ppm) indicated the sorption of formaldehyde, similarly to the pristine crystals. However, in addition to these bands, signals at 1580, 1380, 1320, 1150, and 1060 cm<sup>-1</sup> were found. Such bands suggest the formation of different external surface adsorbates, such as formates, dioxymethylene (DOM) and polyoxymethylene (POM), and methoxy species (Fig. <span class="InternalRef" refid="Fig5">5</span> C, D, Table S4).[32,39] The systematic occurrence of these adsorbates showed that defect-mediated conversion of formaldehyde was taking place on the defective ZIF-8 surface.[32,48] The results were corroborated by bulk <em>in situ</em> FA adsorption experiments on commercial and synthesized ZIF-8 powders and films, using Attenuated Total Reflection FTIR spectroscopy, in which comparable spectral features and pressure-dependent trends in signal intensity were observed (Figures S36-S40).
+
+Based on our spectroscopic results, and in accordance with literature, we propose two possible formaldehyde conversion mechanisms that operate in parallel on the defective ZIF-8 surface: a methoxy and a formate mechanism (Fig. <span class="InternalRef" refid="Fig5">5</span> C, D).[49,50] In the methoxy mechanism, the chemisorbed FA undergoes a Cannizzaro-type reaction to form a methoxy species, a precursor for methanol and formic acid. In the formate mechanism, the FA is chemisorbed in the form of polymers such as dioxymethylene and polyoxymethylene, which are oxidized to form monodentate formates and converted to yield formic acid. Highly relevant for these mechanisms is the hydrolysis of the open zinc sites (<em>vide supra</em>), which provide an additional source of oxygen, which are continuously replenished by water from the aqueous formaldehyde solution. These findings are in line with bulk IR studies on oxide-supported metal nanoparticles, where oxygen from the support is instrumental to the conversion of formaldehyde.[51,52]
+
+To study the effect of plane terminations (structure sensitivity) on FA conversion, we first compared the response pressure of FA conversion over different crystal planes (Fig. <span class="InternalRef" refid="Fig5">5</span> B). This comparison showed that FA conversion, similar to FA adsorption, is structure sensitive as a function of crystal termination, with the following order: {310} > {210} > {211} > {110} ~ {100}, which is consistent with what was observed for pristine crystals (Fig. <span class="InternalRef" refid="Fig2">2</span> ). However, by using PCA and clustering within a single facet (e.g. {100}), we showed that defect-rich areas did show more responsive FA (conversion) behavior than the facet average (Figure S28, S29), similarly to what observed for nanoislands on pristine crystals (Fig. <span class="InternalRef" refid="Fig3">3</span> ).
+
+Since different crystal planes showed a different concentration of defects, dependent on the plane energy (Fig. <span class="InternalRef" refid="Fig4">4</span> C), we subsequently performed a control analysis to rule out a defect concentration effect on the trends observed in Fig. <span class="InternalRef" refid="Fig5">5</span> B. To this end, we used PCA and clustering to identify defect-rich and defect-poor areas spanning over the entire crystal surface (i.e. a combination of contributions from all planes). Both areas showed comparable FA adsorption and conversion behavior (Figure S28, S30). Therefore, we conclude that the defect concentration did not affect the trends observed in Fig. <span class="InternalRef" refid="Fig5">5</span> B, and that a higher defect concentration did not result in detection of FA conversion intermediates at lower pressures.
+
+To rationalize these findings, we modeled FA adsorption on the defective, low energy cuts of {100}, {110}, and {310} surfaces by DFT and compared the results to the previously discussed pristine ZIF-8 (Figure S31). Formaldehyde reaction energies (at 0 K) were found to be 97.1, 88.9, and 1.6 kJ/mol for the low-energy cuts of {100}, {110}, and {310}, respectively, again confirming experimental FA sorption (and conversion) trends. The models further showed that the absence of the second nitrogen atom on the pyrrole linker resulted in the formation of undersaturated Zn sites exhibiting Lewis acidity, in contrast to what was observed for pristine sites. We thus confirm that such Lewis acidity was responsible for the subsequent conversion of chemisorbed formaldehyde.
+
+Overall, the present in-depth nano-spectroscopic characterization campaign revealed that: (i) as expected, high-index crystal terminations outperform extended facets in FA adsorption due to exposure of undercoordinated moieties; (ii) seemingly extended and highly coordinated surfaces in MOFs microcrystals contain uncoordinated nanodomains which have similar behavior to high-index terminations, resulting in locally enhanced gas sorption; (iii) the inclusion of defects results in structure sensitive FA conversion, with higher defect density in high-energy surfaces; (iv) the facet-dependent structure sensitivity of FA adsorption is still observed in defect-engineered microcrystals, meaning that synthesis efforts for better gas sorption materials should be directed to the exposure of high-index crystal planes, rather than to higher defect density. These finding conclusively showed that not all defect sites are equal, adding a layer of complexity to structure sensitivity of FA sorption and conversion on ZIF-8, which could only be discovered using the proposed spatially resolved, <em>in situ</em> PiFM technique.
+
+# Conclusion
+
+Using a combination of well-defined ZIF-8 microcrystals, nanoscale *in situ* PiFM measurements and DFT calculations, we have shown the existence and importance of structure sensitivity in formaldehyde sorption and conversion over the porous functional material ZIF-8. We have shown that for (defect-engineered) ZIF-8 formaldehyde gas sorption preferentially occurs on the sterically isolated external surface sites of high energy planes, such as edges and corners ({310} > {111} > {210} > {211} > {110} ~ {100}), and that defect linker incorporation adheres to this surface-energy-based trend as well. Furthermore, as a result of the nanoscale infrared resolution, we found the co-existence of high-energy and low-energy crystal surface terminations within single facets, where the high-energy nanodomains showed enhanced formaldehyde sorption behavior. Similarly, such nanodomains show higher affinity to defect linker incorporation than their low-energy counterparts. Additionally, the incorporation of defects was found to lead to the formation of Lewis acidity and the consequent conversion of formaldehyde on the ZIF-8 surface through both a formate and a methoxy mechanism. Importantly, the performance of these defect sites for formaldehyde conversion was found to be heavily dependent on structure sensitivity phenomena, rather than on defect concentration. We believe *in situ* PiFM is a highly applicable analytical toolbox for establishing nanoscale structure-performance relationships for (porous) functional materials (e.g., MOFs, zeolites) and guest/probe molecules such as CO₂, NO, CO, etc. providing insights for the rational synthesis of improved nanostructured sorbents and catalysts. Outside the field of catalysis, we see high relevance to the understanding of nanostructured functional materials for sensing, gas separation (e.g. composite membranes), and drug delivery, and for environmental science problems such as understanding of micro/nano-plastic degradation and atmospheric nanoparticulate chemistry.
+
+# References
+
+1. A. Van Santen, R. Complementary Structure Sensitive and Insensitive Catalytic Relationships. *Acc. Chem. Res.* **42**, 57–66 (2008).
+
+2. Boudart, M. Heterogeneous catalysis by metals. *J. Mol. Catal.* **30**, 27–38 (1985).
+
+3. Bennett, C. O. & Che, M. Some geometric aspects of structure sensitivity. *J. Catal.* **120**, 293–302 (1989).
+
+4. Vogt, C. *et al.* Unravelling structure sensitivity in CO₂ hydrogenation over nickel. *Nat. Catal.* **1**, 127–134 (2018).
+
+5. Chizallet, C. *et al.* Catalysis of Transesterification by a Nonfunctionalized Metal−Organic Framework: Acido-Basicity at the External Surface of ZIF-8 Probed by FTIR and ab Initio Calculations. *J. Am. Chem. Soc.* **132**, 12365–12377 (2010).
+
+6. Pérez-Ramı́rez, J., Kapteijn, F. & Brückner, A. Active site structure sensitivity in N₂O conversion over FeMFI zeolites. *J. Catal.* **218**, 234–238 (2003).
+
+7. Leblebici, S. Y. *et al.* Facet-dependent photovoltaic efficiency variations in single grains of hybrid halide perovskite. *Nat. Energy* **1**, 16093 (2016).
+
+8. H. Pang, S., Han, C., S. Sholl, D., W. Jones, C. & P. Lively, R. Facet-Specific Stability of ZIF-8 in the Presence of Acid Gases Dissolved in Aqueous Solutions. *Chem. Mater.* **28**, 6960–6967 (2016).
+
+9. Vogt, C. *et al.* Understanding carbon dioxide activation and carbon–carbon coupling over nickel. *Nat. Commun.* **10**, 5330 (2019).
+
+10. Su, D., Dou, S. & Wang, G. Gold nanocrystals with variable index facets as highly effective cathode catalysts for lithium–oxygen batteries. *NPG Asia Mater.* **7**, e155–e155 (2015).
+
+11. Fang, Z., Bueken, B., De Vos, D. E. & Fischer, R. A. Defect-Engineered Metal–Organic Frameworks. *Angew. Chem. Int. Ed.* **54**, 7234–7254 (2015).
+
+12. Hu, Z. & Zhao, D. Metal–organic frameworks with Lewis acidity: synthesis, characterization, and catalytic applications. *CrystEngComm* **19**, 4066–4081 (2017).
+
+13. Ji, Z., Li, T. & Yaghi, O. M. Sequencing of metals in multivariate metal-organic frameworks. *Science* **369**, 674–680 (2020).
+
+14. N. Johnstone, D. *et al.* Direct Imaging of Correlated Defect Nanodomains in a Metal–Organic Framework. *J. Am. Chem. Soc.* **142**, 13081–13089 (2020).
+
+15. Shen, B., Chen, X., Shen, K., Xiong, H. & Wei, F. Imaging the node-linker coordination in the bulk and local structures of metal-organic frameworks. *Nat. Commun.* **11**, 2692 (2020).
+
+16. Gong, X. *et al.* Insights into the Structure and Dynamics of Metal–Organic Frameworks via Transmission Electron Microscopy. *J. Am. Chem. Soc.* **142**, 17224–17235 (2020).
+
+17. Liu, L. *et al.* Imaging defects and their evolution in a metal–organic framework at sub-unit-cell resolution. *Nat. Chem.* **11**, 622–628 (2019).
+
+18. Karst, J. *et al.* Watching in situ the hydrogen diffusion dynamics in magnesium on the nanoscale. *Sci. Adv.* **6**, eaaz0566 (2020).
+
+19. Eichhorn, J. *et al.* Nanoscale imaging of charge carrier transport in water splitting photoanodes. *Nat. Commun.* **9**, 2597 (2018).
+
+20. Wu, C.-Y. *et al.* High-spatial-resolution mapping of catalytic reactions on single particles. *Nature* **541**, 511–515 (2017).
+
+21. Hosono, N., Terashima, A., Kusaka, S., Matsuda, R. & Kitagawa, S. Highly responsive nature of porous coordination polymer surfaces imaged by in situ atomic force microscopy. *Nat. Chem.* **11**, 109–116 (2019).
+
+22. I. Hadjiivanov, K. *et al.* Power of Infrared and Raman Spectroscopies to Characterize Metal-Organic Frameworks and Investigate Their Interaction with Guest Molecules. *Chem. Rev.* **121**, 1286–1424 (2020).
+
+23. Otter, L. M. *et al.* Nanoscale Chemical Imaging by Photo-Induced Force Microscopy: Technical Aspects and Application to the Geosciences. *Geostand. Geoanalytical Res.* **45**, 5–27 (2021).
+
+24. Stöckle, R. M., Suh, Y. D., Deckert, V. & Zenobi, R. Nanoscale chemical analysis by tip-enhanced Raman spectroscopy. *Chem. Phys. Lett.* **318**, 131–136 (2000).
+
+25. Kumar, N., Weckhuysen, B. M., Wain, A. J. & Pollard, A. J. Nanoscale chemical imaging using tip-enhanced Raman spectroscopy. *Nat. Protoc.* **14**, 1169–1193 (2019).
+
+26. Delen, G., Monai, M., Meirer, F. & Weckhuysen, B. M. In Situ Nanoscale Infrared Spectroscopy of Water Adsorption on Nano-Islands of Surface-Anchored Metal-Organic Frameworks. *Angew. Chem. Int. Ed.* **60**, 1620–1624 (2020).
+
+27. Park, K. S. *et al.* Exceptional chemical and thermal stability of zeolitic imidazolate frameworks. *Proc. Natl. Acad. Sci. U. S. A.* **103**, 10186–10191 (2006).
+
+28. Chen, B., Yang, Z., Zhu, Y. & Xia, Y. Zeolitic imidazolate framework materials: recent progress in synthesis and applications. *J. Mater. Chem. A* **2**, 16811–16831 (2014).
+
+29. Bian, Y. *et al.* Metal–organic framework-based nanofiber filters for effective indoor air quality control. *J. Mater. Chem. A* **6**, 15807–15814 (2018).
+
+30. Decoste, J. B. & Peterson, G. W. Metal-organic frameworks for air purification of toxic chemicals. *Chem. Rev.* **114**, 5695–5727 (2014).
+
+31. Li, P. *et al.* Metal-organic frameworks with photocatalytic bactericidal activity for integrated air cleaning. *Nat. Commun.* **10**, 2177 (2019).
+
+32. Lavalley, J.-C., Lamotte, J., Busca, G. & Lorenzelli, V. Fourier transform i.r. evidence of the formation of dioxymethylene species from formaldehyde adsorption on anatase and thoria. *J. Chem. Soc. Chem. Commun.* 1006–1007 (1985).
+
+33. Shekhah, O. *et al.* Step-by-step route for the synthesis of metal-organic frameworks. *J. Am. Chem. Soc.* **129**, 15118–15119 (2007).
+
+34. Delen, G., Ristanović, Z., Mandemaker, L. D. B. & Weckhuysen, B. M. Mechanistic Insights into Growth of Surface-Mounted Metal-Organic Framework Films Resolved by Infrared (Nano-) Spectroscopy. *Chem. Eur. J.* **24**, 187–195 (2018).
+
+35. Shen, K. *et al.* Ordered macro-microporous metal-organic framework single crystals. *Science* **359**, 206–210 (2018).
+
+36. Yim, C. *et al.* Adsorption and desorption characteristics of alcohol vapors on a nanoporous ZIF-8 film investigated using silicon microcantilevers. *Chem. Commun.* **51**, 6168–6171 (2015).
+
+37. Cousin Saint Remi, J. *et al.* Biobutanol Separation with the Metal–Organic Framework ZIF-8, *ChemSusChem* **4**, 1074–1077 (2011).
+
+38. Popova, G. Y., Budneva, A. A. & Andrushkevich, T. V. Identification of Adsorption Forms by IR Spectroscopy For Formaldehyde and Formic Acid on K₃PMo₁₂O₄₀. *React. Kinet. Catal. Lett.* **61**, 353–362 (1997).
+
+39. Lochař, V. FT-IR study of methanol, formaldehyde and methyl formate adsorption on the surface of Mo/Sn oxide catalyst. *Appl. Catal. A Gen.* **309**, 33–36 (2006).
+
+40. Rivera-Torrente, M. *et al.* Spectroscopy, Microscopy, Diffraction and Scattering of Archetypal MOFs: Formation, Metal Sites in Catalysis and Thin Films. *Chem. Soc. Rev.* **49**, 6694–6732 (2020).
+
+41. Imbihl, R. & Ertl, G. Oscillatory Kinetics in Heterogeneous Catalysis. *Chem. Rev.* **95**, 697–733 (2002).
+
+42. Tran, R. *et al.* Surface energies of elemental crystals. *Sci. Data* **3**, 160080 (2016).
+
+43. Cheetham, A. K., Bennett, T. D., Coudert, F.-X. & Goodwin, A. L. Defects and disorder in metal organic frameworks. *Dalt. Trans.* **45**, 4113–4126 (2016).
+
+44. W. Alley, P. Imidazole-formaldehyde reaction. Formation of 1-imidazolemethanol. *J. Org. Chem.* **40**, 1837–1838 (2002).
+
+45. Wu, J. *et al.* Icosahedral Platinum Alloy Nanocrystals with Enhanced Electrocatalytic Activities. *J. Am. Chem. Soc.* **134**, 11880–11883 (2012).
+
+46. Geidel, E. & Billes, F. Vibrational spectroscopic study of pyrrole and its deuterated derivatives: comparison of the quality of the applicability of the DFT/Becke3P86 and the DFT/Becke3LYP functionals. *J. Mol. Struct.* **507**, 75–87 (2000).
+
+47. Smirnov, E. P. & Tsyganenko, A. A. CNDO Study of the Properties of the OH Groups. *React. Kinet. Catal. Lett.* **7**, 425–429 (1977).
+
+48. F. Edwards, J. & L. Schrader, G. Methanol, formaldehyde, and formic acid adsorption on methanol synthesis catalysts. *J. Phys. Chem.* **89**, 782–788 (2002).
+
+49. Nie, L., Yu, J., Jaroniec, M. & Tao, F. F. Room-temperature catalytic oxidation of formaldehyde on catalysts. *Catal. Sci. Technol.* **6**, 3649–3669 (2016).
+
+50. Ye, J. *et al.* Room-temperature formaldehyde catalytic decomposition. *Environ. Sci. Nano* **7**, 3655–3709 (2020).
+
+51. Busca, G., Lamotte, J., Claude Lavalley, J. & Lorenzelli, V. FT-IR study of the adsorption and transformation of formaldehyde on oxide surfaces. *J. Am. Chem. Soc.* **109**, 5197–5202 (2002).
+
+52. Trillo, J. M., Munuera, G. & Criado, J. M. Catalytic Decomposition of Formic Acid on Metal Oxides. *Catal. Rev.* **7**, 51–86 (1972).
+
+# Supplementary Files
+
+- [DelenetalSINatCommun.pdf](https://assets-eu.researchsquare.com/files/rs-2011018/v1/8a0718d937208c358c761a91.pdf)  
+  Supplementary information

027a56a016f6f13e3412ceda5a7a820fce8448e7e77bf429de0785a3ff3db1e4/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Hydrogen-induced structural phase transition in [Gd2C]2+\u22192e\u2212 electride and its hydrides. a-c, Crystal structure of [Gd2C]2+\u22192e\u2212 (a), Gd2CHx (x\u22641.0) (b) and Gd2CHy (y>2.0) (c), where the IAEs (red) and hydrogens (blue) occupy octahedral (O) and tetrahedral (T) sites of interlayer space between Gd cationic layers. The Wyckoff positions of IAEs and hydrogens are derived from the previous reports29,30 and powder ND analysis of hydrogenated [Y2C]2+\u22192e\u2212 (Supplementary Fig. 3). d-f, Rietveld refinement of powder XRD patterns for Gd2CHx (x\u22641.0) (d) and Gd2CHy (y>2.0) (e, f). Bragg position of the pristine R3\u0305m structured [Gd2C]2+\u22192e\u2212 is shown for a comparison in d. The P3\u0305m1 structure (No. 164) and P3\u03051m structure (No. 162) were used for the refinements of Gd2CHy (y>2.0) in e and f, respectively.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Hydrogen-induced transitions of electrical and magnetic properties in [Gd2C]2+\u22192e\u2212 electride and its hydrides. a, Temperature dependence of normalized resistance, R/R400K, for pristine [Gd2C]2+\u22192e\u2212 and hydrogenated Gd2CHx (x\u22641.0) and Gd2CHy (y>2.0) samples. b, Temperature dependence of magnetization (M) measured under 0.1 T for the three samples. c, Magnetic field (H) dependence of M measured at 2 K for the three samples.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "AC susceptibility measurements for hydrogenated Gd2CHx (x\u22641.0) and Gd2CHy (y>2.0). a,b, Real part (\u03c7\u2019) (a) and imaginary part (\u03c7\u201d) (b) of AC susceptibility for Gd2CHx (x\u22641.0). c,d, \u03c7\u2019 (c) and \u03c7\u201d (d) of AC susceptibility for Gd2CHy (y>2.0). Insets in a and c show the constant and increasing behavior of the \u03c7\u2019 near the phase transition points under different frequencies, respectively. Insets in b and d are the plot of maximum \u03c7\u201d depending on the frequency, showing the constant and increasing behavior, respectively.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Reversible structural and magnetic transitions between [Gd2C]2+\u22192e\u2212 electride and its hydrides. a, Temperature dependence of magnetization (M) for pristine [Gd2C]2+\u22192e\u2212, hydrogenated Gd2CHx (x\u22641.0) and Gd2CHy (y>2.0), and dehydrogenated samples of hydrogenated Gd2CHy (y>2.0) at different temperatures. Magnetic transition temperatures obtained by Curie-Weiss fitting are shown in Supplementary Fig. 12. b, Powder XRD patterns of dehydrogenated samples of hydrogenated Gd2CHy (y>2.0) at different temperatures. c, M-T curves of pristine [Gd2C]2+\u22192e\u2212, hydrogenated Gd2CHy (y>2.0) and sample dehydrogenated at 1300 K. M-T curves are shown in Supplementary Fig. 13.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Tunable magnetocaloric effect between [Gd2C]2+\u22192e\u2212 electride and its hydrides. a-c, Isothermal magnetization of pristine [Gd2C]2+\u22192e\u2212 electride (a), hydrogenated Gd2CHx (x\u00a31.0) (b) and Gd2CHy (y>2.0) (c). d, ln(M)-ln(H) plot at each magnetic transition temperature for the three samples. e, Difference in magnetic entropy (\u0394SM) under magnetic field of 1 ~ 5 T. f, ln(-\u0394SM)-ln(H) plot for the [Gd2C]2+\u22192e\u2212 electride and its hydrogenated samples.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

027a56a016f6f13e3412ceda5a7a820fce8448e7e77bf429de0785a3ff3db1e4/preprint/preprint.md

@@ -0,0 +1,168 @@
+# Abstract
+
+In electrides, interstitial anionic electrons (IAEs) in the quantized energy levels at cavities of positively charged lattice framework possess their own magnetic moment and interact with each other or surrounding cations, behaving as quasi-atoms and inducing diverse magnetism. Here, we report the reversible structural and magnetic transitions by the substitution of the quasi-atomic IAEs in the ferromagnetic two-dimensional [Gd₂C]²⁺×2e⁻ electride with hydrogens and subsequent dehydrogenation of the canted antiferromagnetic Gd₂CHᵧ (y > 2.0). It is demonstrated that structural and magnetic transitions are strongly coupled by the presence or absence of the magnetic quasi-atomic IAEs and non-magnetic hydrogen anions in the interlayer space, which dominate exchange interactions between out-of-plane Gd–Gd atoms. Furthermore, the magnetic quasi-atomic IAEs are inherently conserved by the hydrogen desorption from the P3̅1m structured Gd₂CHᵧ, restoring the original ferromagnetic state of the R3̅m structured [Gd₂C]²⁺×2e⁻ electride. This variable density of magnetic quasi-atomic IAEs enables the quantum manipulation of floating electron phases on the electride surface.
+
+Physical sciences/Materials science/Condensed-matter physics/Magnetic properties and materials  
+Physical sciences/Materials science/Condensed-matter physics/Phase transitions and critical phenomena
+
+# Main
+
+Interstitial anionic electrons (IAEs) construct an ionic crystal as an essential ingredient together with positively charged lattice framework, forming an electride, which is distinguished from conventional ionic compounds with defect color centers trapping electrons<sup>1−3</sup>. Benefited from the intriguing nature of IAEs, exotic physical and chemical properties of electrides, such as low work function<sup><span class="CitationRef">4</span></sup>, high electronic mobility<sup><span class="CitationRef">5</span></sup>, excellent electron reservoir<sup><span class="CitationRef">6</span>, <span class="CitationRef">7</span></sup>, efficient catalytic activity<sup><span class="CitationRef">8</span>, <span class="CitationRef">9</span></sup>, and quantum properties of magnetism<sup><span class="CitationRef">10</span>, <span class="CitationRef">11</span></sup>, superconductivity<sup><span class="CitationRef">12</span></sup>, and topology<sup><span class="CitationRef">13</span></sup>, have attracted considerable interests in both fundamental science and practical applications. This triggers the exploratory research for the discovery of a new class of electrides over the past decades, leading to the success in finding several two-dimensional (2D) electrides; [Ca<sub>2</sub>N]<sup>+</sup>⋅e<sup>−</sup> and [Re<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> (Re = Y, Sc and Gd) with IAEs at interlayers<sup><span class="CitationRef">5</span>, <span class="CitationRef">11</span>, <span class="CitationRef">14</span></sup> and van der Waals [ReCl]<sup>2+</sup>⋅2e<sup>−</sup> (Re = Y and La) with IAEs at intralayers<sup>15−17</sup>. Emergent quantum properties of such 2D electrides are governed by the localization degree of IAEs and their hybridization with neighboring cations. In particular, topological Weyl and ferromagnetic states in [Gd<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> electride with strong hybridization of IAEs and Gd cations are of great interest in the field of quantum materials<sup><span class="CitationRef">11</span>, <span class="CitationRef">18</span></sup>.
+
+Meanwhile, the magnetism of 2D electrides originates from the existence of strongly localized IAEs at interlayer space, which have own magnetic moments and facilitate the magnetic interaction with neighboring cations<sup><span class="CitationRef">10</span>, <span class="CitationRef">11</span>, <span class="CitationRef">17</span></sup>. For example, the IAEs in the [Y<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> play as ferromagnetic particles in the lattice framework composed of only paramagnetic elements, exhibiting the superparamagnetism<sup><span class="CitationRef">10</span></sup>. The [Gd<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> was found to be the first room-temperature ferromagnetic electride with <em>T</em><sub>c</sub> of 350 K due to the exchange interactions of interlayer Gd cations across IAEs<sup><span class="CitationRef">11</span></sup>. Importantly, the IAEs in [Gd<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> are considered as magnetic quasi-atomic electrons with substantive magnetic moments that are responsible for the occurrence of ferromagnetism from the antiferromagnetic [Gd<sub>2</sub>C]<sup>2+</sup> lattice framework. The concept of interstitial quasi-atomic electrons (IQEs), suggested by Miao and Hoffmann, has been esteemed to understand the nature of elemental and compound electrides on the basis of theoretical ground<sup><span class="CitationRef">19</span></sup>. Indeed, potassium electride under high pressure is stabilized by ferromagnetic ordering of the IQEs<sup><span class="CitationRef">20</span></sup>. Furthermore, the mixed-cation [YGdC]<sup>2+</sup>·2e<sup>−</sup> electride exhibited the ferrimagnetic state, which was attributed to the direct exchange interactions between magnetic IQEs at different crystallographic positions<sup><span class="CitationRef">21</span></sup>. In addition to the magnetic ordering of IAEs in the electrides, the IAEs on the cleaved surface of 2D [Gd<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> electride are found to be spin-polarized Fermi liquid and crystallized into the hexatic phase by decreasing their density on the surface<sup><span class="CitationRef">22</span></sup>.
+
+The IQEs in the magnetic electrides can be regarded as analogous to the substituted or doped magnetic elements in typical magnetic alloys<sup><span class="CitationRef">23</span></sup>, indicating that the presence of the IQEs can provide a freedom to tune the magnetic properties and study their role in magnetic phase transitions of electrides. In the previous report<sup><span class="CitationRef">11</span></sup>, the substitution of chlorine atoms for the IAEs in ferromagnetic [Gd<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> electride resulted in the transition to antiferromagnetic Gd<sub>2</sub>CCl and proved the presence of magnetic IQEs. On the other hand, hydrogen-substituted electrides, in which hydrogens absorb the IQEs and form the hydrogen anions, have been examined to find the crystallographic positions of IQEs and elucidate their contribution to the electronic density of state<sup><span class="CitationRef">24</span>, <span class="CitationRef">25</span></sup>. Recent computational studies suggested that the hydrogenation for the monolayers of 2D [Ca<sub>2</sub>N]<sup>+</sup>⋅e<sup>−</sup> and [Gd<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> electrides significantly altered their magnetic properties<sup><span class="CitationRef">26</span>, <span class="CitationRef">27</span></sup>. However, an experimental investigation of hydrogenation of the magnetic 2D electrides has been hardly ever reported in spite of a possibility to identify the critical role of magnetic IQEs for triggering the magnetism and tuning the magnetic properties. This might come from the experimental difficulty in handling the chemically unstable electrides as well as a common expectation for the hydrogen-induced embrittlement.
+
+Here, we report the hydrogenation and dehydrogenation of the ferromagnetic [Gd<sub>2</sub>C]<sup>2+</sup>⋅2e<sup>−</sup> electride, which simultaneously induced the magnetic and structural phase transitions. Depending on the relative concentration between IQEs and hydrogen anions, crystal structure and magnetic phase exhibited the strongly coupled reversible transitions between the ferromagnetic layered structure of the <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> space group to the canted antiferromagnetic layered strucutre of the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> space group, providing an experimental proof on the magnetic nature of IQEs.
+
+## Hydrogen-induced phase transition in [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride
+
+The pristine [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride is crystalized in anti-CdCl<sub>2</sub>-type layered structure belonging to the rhombohedral <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> space group, where the IQEs are occupying the interlayer space (Fig. <span class="InternalRef">1</span> a). Note that the occupancy of IQEs can be found in both octahedral and tetrahedral sites of Gd-sublattice at the interlayer space. From the electron localization function (ELF) obtained by theoretical calculations of [Y<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup>, [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> and [YGdC]<sup>2+</sup>·2e<sup>−</sup> electrides<sup>10,11,21</sup>, it was revealed that most IQEs occupy the octahedral sites with a minor occupancy at tetrahedral sites. Substitution of IQEs with hydrogens in the isostructural 2D electrides has been experimentally investigated for [Ca<sub>2</sub>N]<sup>+</sup>·e<sup>−</sup> and [Y<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup>, exhibiting a distinct difference in structural transition<sup>28−30</sup>. Hydrogenation of [Ca<sub>2</sub>N]<sup>+</sup>·e<sup>−</sup> electride with fully delocalized IAEs leads to the transition from 2D layered structure to 3D-like a cubic structure of <em>Fd3m</em> (No. 227) space group due to the penetration of excess hydrogens into the positively charged [Ca<sub>2</sub>N]<sup>+</sup> layers<sup>28</sup>. On the contrary, hydrogens only occupied the interlayer space between positively charged [Y<sub>2</sub>C]<sup><span class="CitationRef">2</span>+</sup> layers, leading to the structural phase transition to the different layered structures of <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m1</em> and <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em><sup>2<span class="CitationRef">9</span>, <span class="CitationRef">30</span></sup> depending on the hydrogen concentration at octahedral and tetrahedral sites as shown in Figs. <span class="InternalRef">1</span> b,c. The <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m1</em> structure (No. 164) is derived from the occupancy of 2 moles of hydrogen at only tetrahedral sites and the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structure (No. 162) is derived from the excess occupancy at octahedral sites with additional one mole of hydrogen. Thus, hydrogens can substitute the IQEs at octahedral sites within the concentration of one mole, maintaining the pristine <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> structure.
+
+Pristine [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride was annealed at 600 K and 1000 K under hydrogen pressure of 10<sup>−1</sup> Torr with 4% H<sub>2</sub> mixed Ar gas (Supplementary Fig. 1). In order to ensure a homogeneity of hydrogen distribution in the samples, the pulverized powders of the electride were annealed for 24 hours. The hydrogenated samples were investigated by X-ray diffraction (XRD) pattern measurement and analyzed by Rietveld refinement method. The powder XRD pattern of the sample hydrogenated at 1000 K (hereafter referred as Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0)) shows a good consistency with the simulated pattern of <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> structured pristine [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride (Fig. <span class="InternalRef">1</span> d). On the other hand, the sample hydrogenated at 600 K clearly showed a structural transition to <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structure (Figs. <span class="InternalRef">1</span> e,f). The XRD pattern of the sample hydrogenated at 600 K was well refined with the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structure rather than <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m1</em> structure (see Supplementary Table 1 for the comparison of Rietveld refinement results with <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m1</em> and <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> and Supplementary Fig. 2 for the refinement results with <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structure of another sample). This structural phase transition is reminecent of the hydrogenation of the isostructural [Y<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride, yielding the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structure of Y<sub>2</sub>CH<sub>2.55</sub>, where 2 moles of hydrogen occupy all tetrahedral sites and additional 0.55 moles of hydrogen occupy the octahedral sites, as confirmed from neutron diffraction (ND) study<sup><span class="CitationRef">29</span>, <span class="CitationRef">30</span></sup>. Although the ND measurements of hydrogenated Gd<sub>2</sub>CH<sub>x</sub> and Gd<sub>2</sub>CH<sub>y</sub> compounds are not possible due to a high absorption nature of neutron beam by Gd atoms, our hydrogenation condition was verified by the hydrogenation of the isostructural [Y<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride and its ND measurements followed by Rietveld refinements, ensuring the hydrogen concentration over 2 moles and structural phase transition to <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> of Gd<sub>2</sub>CH<sub>y</sub> (y > 2.0) for the sample hydrogenated at 600 K (Supplementary Fig. 3). Crystal structure data derived from the Rietveld refinements of XRD and ND measurements are given in Supplementary Tables 1,2.
+
+## Hydrogen-induced magnetic transition in [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride
+
+The IAEs have played a key role in governing the electronic and magnetic behavior of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride, showing the anisotropy in the metallic conduction and magnetic property. In particular, the ferromagnetism of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride was explained from the strongly localized IAEs with their own magnetic moments, which facilitate the exchange interaction between Gd−Gd atoms across the IAEs in the interlayer space<sup><span class="CitationRef">11</span></sup>. This quasi-atomic nature of the IAEs can be verified by the substitution with other elements and subsequent change in electrical and magnetic properties. Figure <span class="InternalRef">2</span> a shows the temperature dependence of electrical resistivity for the pristine [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride and hydrogenated samples. The normalized resistance, <em>R/R</em><sub><em>400</em>K</sub> of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride and the hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) decreased with the decrease in temperature, showing the same metallic behavior. The increased <em>R/R</em><sub><em>400</em>K</sub> and enhanced electron-phonon scattering in the hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) imply the reduced concentration of the IQEs (Supplementary Fig. 4). However, the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) showed the typical behavior of semiconductors, indicating that most IQEs are substituted by the hydrogens and metal-insulating transition occurs according to their concentration.
+
+A profound effect of hydrogenation was found in the magnetic properties. Figure <span class="InternalRef">2</span> b shows the temperature dependence of magnetic susceptibility (<em>χ</em>) for the pristine [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride and its hydrogenated samples. Compared to the ferromagnetic transition temperature (<em>T</em><sub>c</sub>) of 350 K of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride, the hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) sample exhibits the decreased <em>T</em><sub>c</sub> of 220 K, which was determined from the temperature dependence of <em>dM</em>/<em>dT</em> curve (Supplementary Fig. 5a). This <em>T</em><sub>c</sub> decrease is obviously attributed to the reduced IQE concentration by the substitution with hydrogen atoms. Furthermore, this behavior is reminiscent of well-known strategy for tuning the <em>T</em><sub>c</sub> of ferromagnets by the substitution with non-magnetic elements<sup><span class="CitationRef">23</span></sup>, proving the magnetic nature of the IQEs. An interesting feature is the enhanced <em>χ</em> and the bifurcation between ZFC and FC curves in the hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) sample. From the fact that the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m1</em> structure (No. 164) is derived from the occupancy of 2 moles of hydrogen at only tetrahedral sites, the IQEs at tetrahedral sites are preferably substituted with hydrogens, leading to the stronger localization of the IQEs at octahedral sites and enhancement of the susceptibility. Because the hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) sample was crystallized into the <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> structure, it is expected that the IQEs in the tetrahedral sites are randomly substituted.
+
+This random distribution of IQEs and hydrogens probably results in the spin canting, showing the bifurcation between ZFC and FC curves. Importantly, it should be noted that the magnetic transition occurs by the substitution of most IQEs with hydrogens. The <em>M−T</em> curve in the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structured Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) exhibits a typical feature of antiferromagnetic materials with ZFC downturn behavior. The isothermal magnetization (<em>M−H</em> curve) was also measured at 2 K. No hysteresis loop was observed in every sample, discarding a hard ferromagnetic ground state. Although the <em>M−H</em> curves exhibited the saturation of <em>M</em> at small <em>H</em> in the <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> structured [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride and hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) sample, the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structured Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) showed no saturation even at 5 T, suggesting an antiferromagnetic ground state. In the Curie-Weiss (CW) analysis on the inverse susceptibility curve (<em>χ</em><sup>−<span class="CitationRef">1</span></sup><em> vs. T</em>) for all the samples (Supplementary Fig. 5b), the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride and hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) sample followed the CW fit. However, a slight deviation from the linear fit was observed in the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) sample, suggesting the antiferromagnetic nature (see also Supplementary Fig. 6).
+
+## Canted antiferromagnetism induced by hydrogenation
+
+To clearly understand the antiferromagnetic properties of the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0), we further examined AC susceptibility. Figure <span class="InternalRef">3</span> shows the results of AC susceptibility measurements under 1.5 Oe of <em>H</em><sub>AC</sub> with different frequencies. Both the real part (<em>χ’</em>) and imaginary part (<em>χ”</em>) extracted from the AC susceptibility of ferromagnetic [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> and hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) showed the constant temperature of peak position (<em>T</em><sub>p</sub>, indicated by arrow) regardless of frequency (Figs. <span class="InternalRef">3</span> a,b, Supplementary Fig. 7). On the other hand, both <em>T</em><sub>p</sub> of the <em>χ’</em> and <em>χ”</em> for the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) showed a systematic shift towards higher temperatures with the increase in frequency (Figs. <span class="InternalRef">3</span> c,d). This increasing <em>T</em><sub>p</sub> behavior and non-zero net magnetic moment of <em>χ”</em> confirms the canted spin ordering in the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0).
+
+In contrary to the antiferromagnetic Gd<sub>2</sub>CCl, where the IQEs are substituted by Cl atoms and the 2D array of gadolinium cations was responsible for the magnetism without splitting between ZFC and FC curves<sup><span class="CitationRef">11</span></sup>, the Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) demonstrates non-zero net magnetization and weak hysteresis in <em>M−H</em> curve (Supplementary Fig. 8). These behaviors indicate that the exchange interaction of Gd–IQEs–Gd was largely suppressed by the substitution of IQEs with hydrogens, but the presence of hydrogen anions probably leads to the canted spin structure between the 2D gadolinium arrays. Critical behavior analysis around magnetic transition temperature provides a plausible canted antiferromagnetic structure of the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0), where the XY model is not available below the <em>T</em><sub>N</sub> (Supplementary Figs. 9 and 10f,i), indicating that the hydrogen anions trigger the antiferromagnetic spin canting for the out-of-plane Gd–Gd atoms across the non-magnetic hydrogen anions as shown in the schematic illustration of Supplementary Fig. 11.
+
+## Reversible structural and magnetic phase transitions
+
+A strong coupling between structural and magnetic phase transitions in the hydrogenation of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride is also observed in the dehydrogenation of the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0). Dehydrogenation was conducted by heating the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y≥2.0) under a vacuum of 10<sup>−5</sup> Torr. Figure <span class="InternalRef">4</span> a shows the <em>M−T</em> curves of the dehydrogenated samples of Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) at different temperatures together with the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> (grey) and its hydrogenated samples (blue and red). The decreased magnetic transition temperature upon the hydrogenation, from <em>T</em><sub>c</sub> ~ 350 K of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride to <em>T</em><sub>N</sub> ~ 20 K of the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0), is perfectly reversed by the dehydrogenation, which increases the magnetic transition temperatures with the increase in dehydrogenation temperature. Notably, the magnetic transition temperature increases up to <em>T</em><sub>c</sub> ~ 350 K, which is exactly same as that of the pristine [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride. It is also noted that the dehydrogenated samples show the magnetic transition from antiferromagnetism (dehydrogenated at 500 K and 800 K) to ferromagnetism (dehydrogenated at 1000 K and 1500 K) with proceeding the dehydrogenation. This is the reverse of the hydrogenation of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride accompanied by the coupled structural and magnetic phase transitions. Indeed, it was confirmed from the XRD measurements of dehydrogenated samples (Fig. <span class="InternalRef">4</span> b) that the structural transition occurs from the <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structure of hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) and its dehydrogenated sample at 500 K to <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> structure of dehydrogenated samples at 800 K, 1100 K and 1300 K. Most of all, the <em>T</em><sub>c</sub> as well as <em>M−T</em> curve of the dehydrogenated sample at 1300 K indicate that the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> electride is re-formed by the re-generated magnetic IQEs at the original Wyckoff positions, re-inducing the ferromagnetic Gd−IQEs−Gd exchange interactions (Fig. <span class="InternalRef">4</span> c).
+
+## Desorption of hydrogens and conservation of IAEs
+
+It is worthwhile to consider that the magnetic IQEs are inherently conserved from the dehydrogenation of the hydrogenated electride. When both processes of hydrogenation and dehydrogenation were completed, the positively charged [Gd<sub>2</sub>C]<sup>2+</sup> layers in the <em>R</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>m</em> and <em>P</em><span class="InlineEquation"><span class="mathinline">\\(\\stackrel{-}{3}\\)</span></span><em>1m</em> structures were compensated by the IQEs and hydrogen anions, respectively. Furthermore, the charge neutrality of hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) and dehydrogenated samples at 500 K, 800 K and 1100 K are also maintained by the coexistence of IQEs and hydrogen anions in the positively charged [Gd<sub>2</sub>C]<sup>2+</sup> layered lattice framework. These results indicate that the continuous substitution between IQEs with hydrogens occurs during the both processes. Although the reaction of IQEs with hydrogens, which produces the hydrogen anions in the hydrogenated samples, can be reasonably expected, the conservation of IQEs by the desorption of hydrogens in the dehydrogenated samples is an unprecedented phenomenon when considering that the formation of vacancy at the site of hydrogen ions is a general feature in the dehydrogenated materials<sup><span class="CitationRef">31</span>, <span class="CitationRef">32</span></sup>. Many theoretical studies have explored the nature of electride by the substitution of IQEs with hydrogen anions and experimental results have been reported to provide evidence on the existence of the IQEs in the electrides<sup><span class="CitationRef">33</span>, <span class="CitationRef">34</span></sup>. However, to our best knowledge, the dehydrogenation producing the electride has never been reported yet. Indeed, proof-of-demonstrations on the magnetic IQEs of the electrides have been rare in experiments. In contrary to the well-known intercalation and deintercalation processes of elements in 2D materials<sup><span class="CitationRef">35</span>. <span class="CitationRef">36</span></sup>, the reversible transitions depending on the stoichiometric balance between IQEs and hydrogen anions strongly indicates that the IAEs at a specific Wyckoff position behave as quasi-atoms to keep the charge neutrality and to ensure the thermodynamic and electronic stability of the electrides<sup><span class="CitationRef">11</span>, <span class="CitationRef">19</span>, <span class="CitationRef">37</span></sup>. In particular, because the continuous change in magnetic transition temperatures is exclusively ascribed to the relative concentration of magnetic IQEs and hydrogens, the nature of IAEs can be regarded as magnetic IQEs, allowing the freedom to tune the magnetic properties of electrides.
+
+### Magnetocaloric effect upon hydrogenation.
+
+A magnetocaloric effect (MCE) can also provide an insight into the nature of magnetic phase transition and the existence of magnetic IQEs, which are related to the difference in the degree of freedom of spin alignment. The MCE of a system with a weaker exchange interaction shows a larger magnetic entropy (<em>S</em><sub>M</sub>) near the magnetic transition temperature. The Δ<em>S</em><sub>M</sub> can be derived from isothermal magnetization as shown in Figs. <span class="InternalRef">5</span> a−c. These measurements give the nature of spin exchange interaction from the critical exponent <em>δ</em> value in the relation of <em>M</em>(<em>H,T</em>) = D<em>H</em><sup>1/<em>δ</em></sup>, (<em>T</em> = <em>T</em><sub>c</sub>), which can be obtained from the slope of ln (<em>H</em>) − ln (<em>M</em>) plot (Fig. <span class="InternalRef">5</span> d). The <em>δ</em> value of 2.1 for the canted antiferromagnetic hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) is well-matched with the spin-disordered state<sup>38−42</sup>. For the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> and hydrogenated Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0), the <em>δ</em> values of 3.3 and 4.9 are of ferromagnetic systems. Furthermore, the temperature dependence of the Δ<em>S</em><sub>M</sub> from the following equation,
+
+$$
+{{\\Delta }S}_{M}=\\underset{0}{\\overset{H}{\\int }}{\\left(\\frac{\\partial M}{\\partial T}\\right)}_{H}dH \\left(1\\right)
+$$
+
+is shown in Fig. <span class="InternalRef">5</span> e. The isothermal Δ<em>S</em><sub>M</sub> showed a broad maximum around <em>T</em><sub>c</sub> for the ferromagnetic [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> and Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0), whereas the canted antiferromagnetic Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) exhibited a sharp increase around <em>T</em><sub>N</sub>. The maximum entropy change at <em>T</em><sub>c</sub> was found to be as high as 17.3 J⋅Kg<sup>−1</sup> K<sup>−1</sup> for the Gd<sub>2</sub>CH<sub>y</sub> (y>2.0), which is almost three times higher than those of [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> and Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) samples, indicating that a weaker exchange interaction is present for the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0). In addition, the power-law fitting of −Δ<em>S</em><sub>M</sub> ∝ <em>H</em><sup>n</sup> (Fig. <span class="InternalRef">5</span> f) also implies that the hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) with n greater than 1.0 follows the antiferromagnetic behavior, which is clearly distinguished from the ferromagnetic [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> and Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0). Besides of the identification of the change in the magnetism according to the concentration of magnetic IQEs, the magnetic properties such as relative cooling power (RCP) can be also controlled by the magnetic IQEs. We calculated the RCP with the equation of |ΔS<sub>M</sub><sup>max</sup>| ⋅ δ<em>T</em><sub>FWHM</sub> (Supplementary Fig. 14), where the δ<em>T</em><sub>FWHM</sub> is the full width at half maxima of the peak of the Δ<em>S</em><sub>M</sub> versus <em>T</em> plot in the Fig. <span class="InternalRef">5</span> e. Clearly, the RCP value of hydrogenated Gd<sub>2</sub>CH<sub>y</sub> (y>2.0) sample is quite larger (~ 350 J⋅Kg<sup>−1</sup> at 5 T) than those of the [Gd<sub>2</sub>C]<sup>2+</sup>·2e<sup>−</sup> and Gd<sub>2</sub>CH<sub>x</sub> (x≤1.0) samples, showing a comparable capability to that of magnetocaloric AlFe<sub>2</sub>B, Gd<sub>5</sub>Si<sub>4</sub> and Gd<sub>2</sub>FeAlO<sub>6</sub> materials<sup><span class="CitationRef">43</span>, <span class="CitationRef">44</span></sup>. These exotic properties based on the magnetic IQEs can provide a possibility to explore a new refrigerant electride material.
+
+# Conclusion
+
+In summary, we explored the non-magnetic hydrogen substitution for the magnetic IQEs in the 2D [Gd₂C]²⁺·2e⁻ electride and found the strongly coupled structural and magnetic phase transitions. This coupling was also observed by the dehydrogenation of the Gd₂CHy (y>2.0), which perfectly conserved the IQEs. The structural phase transition between higher symmetric R$\overline{3}$m structure of the [Gd₂C]²⁺·2e⁻ electride and lower symmetric P$\overline{3}$1m structure of the hydrogenated Gd₂CHy (y>2.0) is accompanied with the magnetic phase transition in a wide temperature range between ferromagnetism at the Tc of 350 K to canted antiferromagnetism at the TN of 20 K. The reversible magnetic transition is governed by the spin exchange interactions in the out-of-plane Gd–Gd atoms, which are mediated across the magnetic IQEs or non-magnetic hydrogen anions. Our results clearly identified the nature of magnetic IQEs and proved their critical role in tuning the magnetic properties, providing the IQEs as a new ingredient in magnetic materials. These IQEs, which can have interactions with each other or surrounding cations, can thus trigger antiferromagnet, ferromagnet, or permanent magnet, all the magnetism. Finally, the reversible substitution and conservation between magnetic IQEs and hydrogen anions can provide a possible platform to study the exotic magnetic state of quantum electron phases such as Wigner crystal on the electrides<sup><span citationid="CR22" class="CitationRef">22</span>, <span citationid="CR45" class="CitationRef">45</span></sup>.
+
+# Methods
+
+Synthesis of [Gd₂C]²⁺·2e⁻ electride and its hydrides.
+
+All samples were handled in glove boxes filled with high-purity argon gas (Ar 99.999%) to prevent the oxidation of raw materials and synthesized samples. The synthesis method of a polycrystalline ingot of [Gd₂C]²⁺∙2e⁻ electrides is performed by the arc-melting process with mixed Gd metal pieces and graphite pieces in a 2:1 molar ratio under high-purity Ar atmosphere. Before synthesizing the hydrogenated sample, we pulverized the polycrystalline [Gd₂C]²⁺∙2e⁻ electrides to powder and pelletize for handling. As displayed in Supplementary Fig. 1, hydrogenation is performed in the quartz tube furnace under Ar-based 4% H₂ mixed gas. To synthesize the Gd₂CHy (y ≥ 2.0) composition, pelletized [Gd₂C]²⁺∙2e⁻ electrides were heat treatment of around 600 K under a 1 atm environment. The Gd₂CHx (x ≤ 1.0) was synthesized around 1000 K and under 10⁻¹ Torr pressure, which is created by flowing the H₂-mixed gas and vacuum pumping at the same time. The dehydrogenation process was performed by pelletized Gd₂CHy (y ≥ 2.0) under 10⁻⁵ Torr by vacuum pumping with different temperatures.
+
+Structural characterization by X-ray and neutron powder diffraction.
+
+The crystal structure of the [Gd₂C]²⁺∙2e⁻ electride, hydrogenated samples (Gd₂CHx (x ≤ 1.0) and Gd₂CHy (y ≥ 2.0)), and dehydrogenated samples were investigated by XRD using a Rigaku SmartLab diffractometer with monochromatic Cu Kα radiation (8.04 keV) at room temperature. The well-ground powder samples were prepared in glove boxes and measured in a plastic dome-type stage filled with Ar gas to avoid oxidation during measurements. A high resolution neutron powder diffraction of hydrogenated Y₂CHy (y ≥ 2.0) sample was measured at the HANARO, a research reactor of the Korea Atomic Energy Research Institute. The wavelength of the neutron beam of HRPD is λ = 0.1834528 nm, and the measurement error of the lattice change rate using this beam is about ± 0.004%. General Structure Analysis System (GSAS) software package was applied to perform Rietveld refinement.
+
+Magnetic and electrical properties characterization.
+
+The sampling for resistivity and magnetic properties measurements were performed in the high-purity Ar-filled glove boxes. The temperature-dependent resistivity measurements were performed by the physical property measurement system (PPMS DynaCool, Quantum Design). The four-electrode is made by silver epoxy on the pelletized samples. After that, samples were covered with Apiezon N grease to block the oxidation during the sample loading to PPMS and measurements. The measurement of magnetic properties used a vibrating sample magnetometer (VSM, Quantum Design) and Squid magnetometer (MPMS3, Quantum Design) for AC magnetic susceptibility. A plastic capsule copula containing a weighted sample was coated with N grease to prevent the oxidation of samples.
+
+# References
+
+1. Dye, J. L. Electrides: ionic salts with electrons as the anions. *Science* **247**, 663–668 (1990).
+
+2. Dye, J. L. Electrides: early examples of quantum confinement. *Acc. Chem. Res.* **42**, 1564–1572 (2009).
+
+3. Markham, J. J. *F-Centers in Alkali Halides* (Academic Press, New York, 1966).
+
+4. Toda, Y. et al. Field emission of electron anions clathrated in subnanometer-sized cages in [Ca₂₄Al₂₈O₆₄]⁴⁺ (4e⁻). *Adv. Mater.* **16**, 685−689 (2004).
+
+5. Lee, K., Kim, S. W., Toda, Y., Matsuishi, S. & Hosono, H. Dicalcium nitride as a two-dimensional electride with an anionic electron layer. *Nature* **494**, 336−340 (2013).
+
+6. Kim, Y. J., Kim, S. M., Hosono, H., Yang, J. W. & Kim, S. W. The scalable pinacol coupling reaction utilizing the inorganic electride [Ca₂N]⁺·e⁻ as an electron donor. *Chem. Commun.* **50**, 4791 (2014).
+
+7. Kim, Y. J. et al. Two dimensional inorganic electride-promoted electron transfer efficiency in transfer hydrogenation of alkynes and alkenes. *Chem. Sci.* **6**, 3577−3581, (2015).
+
+8. Kitano, M. et al. Ammonia synthesis using a stable electride as an electron donor and reversible hydrogen store. *Nat. Chem.* **4**, 934−940 (2012).
+
+9. Kang, S. H. et al. Water- and acid-stable self-passivated dihafnium sulfide electride and its persistent electrocatalytic reaction. *Sci. Adv.* **6**, 23 (2020).
+
+10. Park, J. et al. Strong localization of anionic electrons at interlayer for electrical and magnetic anisotropy in two-dimensional Y₂C electride. *J. Am. Chem. Soc.* **139**, 615−618 (2017).
+
+11. Lee, S. Y. et al. Ferromagnetic quasi-atomic electrons in two-dimensional electride. *Nat. Commun.* **11**, 1526 (2020).
+
+12. Hosono, H. et al. Superconductivity in room-temperature stable electride and high-pressure phases of alkali metals. *Phil.Trans. R. Soc.* A **373**, 2037 (2015).
+
+13. Hirayama, M., Matsuishi, S., Hosono, H. & Murakami, S. Electrides as a new platform of topological materials. *Phys. Rev. X* **8**, 031067 (2018).
+
+14. Park, J. et al. Tuning the spin-alignment of interstitial electrons in two-dimensional Y₂C electride via chemical pressure. *J. Am. Chem. Soc.* **139**, 17277–17280 (2017).
+
+15. Wan, B. et al. Identifying quasi-2D and 1D electrides in yttrium and scandium chlorides via geometrical identification. *npj Comput. Mater.* **4**, 77 (2018).
+
+16. Liu, S., Li, W., Kim, S. W. & Choi, J. H. Decisive role of interlayer ionic couplings for the electronic properties of two-dimensional layered electrides. *J. Phys. Chem. C.* **124**, 1398−1404 (2020).
+
+17. Song, H. Y. et al. Van der Waals electride: toward intrinsic two-dimensional ferromagnetism of spin-polarized anionic electrons. *Mater. Today Phys.* **20**, 100473 (2021).
+
+18. Liu S. et al. Ferromagnetic Weyl fermions in two-dimensional layered electride Gd₂C. *Phys. Rev. Lett.* **125**, 187203 (2020).
+
+19. Miao, M. & Hoffmann, R. High pressure electrides: a predictive chemical and physical theory. *Acc. Chem. Res.* **47**, 1311–1317 (2014).
+
+20. Pickard, C. J. & Needs, R. J. Predicted pressure-induced s-band ferromagnetism in alkali metals. *Phys. Rev. Lett.* **107**, 087201 (2011).
+
+21. Lee, S. Y. et al. Mixed-cation driven magnetic interaction of interstitial electrons for ferrimagnetic two-dimensional electride. *npj Quantum Mater.* **6**, 21 (2021).
+
+22. Kim, S. et al. Quantum electron liquid and its possible phase transition. *Nat. Mater.* **21**, 1269–1274 (2022).
+
+23. O’handley, R. C. *Modern magnetic materials: Principles and applications.* (Wiley, 1999)
+
+24. Lu, Y. et al. Water durable electride Y₅Si₃: electronic structure and catalytic activity for ammonia synthesis. *J. Am. Chem. Soc.* **138**, 3970−3973 (2016).
+
+25. Cao, Y. et al. Topochemical synthesis of Ca₃CrN₃H involving a rotational structural transformation for catalytic ammonia synthesis. *Angew. Chem. Int. Ed.* **61**, e202209187 (2022).
+
+26. Qiu, X. L., Zhang, J. F., Lu, Z. Y. & Liu, K. Manipulating the electronic and magnetic properties of monolayer electride Ca₂N by hydrogenation. *J. Phys. Chem. C.* **123**, 24698−24704 (2019).
+
+27. Xu, D., Zhang, J. F., Lu, Z. Y. & Liu, K. Magnetic and electronic transitions in monolayer electride Gd₂C induced by hydrogenation: a first-principles study. *Phys. Rev. B* **106**, 045138 (2022).
+
+28. Verbraeken, M. C., Suard E. & Irvine, J. T. S. Order and disorder in Ca₂ND₀.₉₀H₀.₁₀–a structural and thermal study. *J. Solid State Chem.* **184**, 2088-2096 (2011).
+
+29. Maehlen, J.P., Yartys, V. A. & Hauback, B.C. Structural studies of deuterides of yttrium carbide. *J. Alloys Comp.* **351**, 151–157 (2003).
+
+30. Maehlen, J. P., Yartys, V. A. & Hauback B. C. Structural studies of the deuterides of carbon containing yttrium alloy. *J. Alloys Comp.* **356–357**, 475–479 (2003).
+
+31. German. E. & Gebauer, R. Improvement of hydrogen vacancy diffusion kinetics in MgH₂ by niobium- and zirconium-doping for hydrogen storage applications. *J. Phys. Chem. C* **120**, 4806–4812 (2016).
+
+32. Kitano, M. et al. Essential role of hydride ion in ruthenium-based ammonia synthesis catalysts. *Chem. Sci.*, **7**, 4036–4043 (2016).
+
+33. Zheng, Q. et al. Direct visualization of anionic electrons in an electride reveals inhomogeneities. *Sci. Adv.* **7**, eabe6819 (2021).
+
+34. Venkatraman, K., Hachtel, J. & Chi, M. Correlating inhomogeneity in anionic electron density with hydrogen incorporation in Y₅Si₃ electrides. *Microsc. Microanal.* **27**, 146–147 (2021).
+
+35. Stark, M. S., Kuntz, K. L., Martens, S. J. & Warren, S. C. Intercalation of layered materials from bulk to 2D. *Adv. Mater.* **31**, 1808213 (2019).
+
+36. Xue, Y., Opening two-dimensional materials for energy conversion and storage: a concept. *Adv. Energy Mater.* **7**, 1602684 (2017).
+
+37. Miao, M. & Hoffmann R. High-pressure electrides: the chemical nature of interstitial quasiatoms. *J. Am. Chem. Soc.* **137**, 3631–3637 (2015)
+
+38. Li, D. et al. Percolation of spatially constraint network. *EPL* **93**, 68004 (2011).
+
+39. Karube, K. et al. Universality and critical behavior at the critical endpoint in the itinerant-electron metamagnet UCoAl. *Phys. Rev. B* **86**, 024428 (2012).
+
+40. Belayachi, A., Dormann, J. L. & Noguès, M. Critical analysis of magnetically semi-disordered systems: critical exponents at various transitions. *J. Phys.: Condens. Matter.* **10**, 1599–1619 (1998).
+
+41. Herrero, A. et al. Tailoring the magnetocaloric, magnetic and thermal properties of Dy₆(Fe,Mn)X₂ intermetallics (X=Sb, Te, Bi). *J. Alloys Comp.* **890**, 161849 (2021).
+
+42. Jia, Y., Chenga, Y., Wang, H., Zhanga, Z. & Li, L. Magnetocaloric properties and critical behavior in double perovskite RE₂CrMnO₆ (RE = La, Pr, and Nd) compound. *Ceram. Int.* **46**, 25043–25049 (2020).
+
+43. GschneidnerJr, K. A., Pecharsky, V. K. & Tsokol, A. O. Recent developments in magnetocaloric materials. *Rep. Prog. Phys.* **68**, 1479 (2005).
+
+44. Tan, X., Chai, P., Thompson, C. M. & Shatruk, M. Magnetocaloric effect in AlFe₂B₂: toward magnetic refrigerants from earth-abundant elements. *J. Am. Chem. Soc.* **135**, 9553 (2013).
+
+45. Nitin K., Nicolás M-. D., Allan H. M. & Elbio D. Magnetic ground states of honeycomb lattice Wigner crystals. *Commun. Phys.* **5**, 289 (2022).
+
+# Supplementary Files
+
+- [Supplementary.docx](https://assets-eu.researchsquare.com/files/rs-2825044/v1/0631952501b04414a748c83c.docx)

05cdd05c8033f50c5d58d58f81996f649b1fce4c208692e95be55fcdb4c8a5aa/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Spatial distribution of intracranial electrode sampling. (A) The figure shows the number of patients whose artifact-free, nonepileptic intracranial EEG data were available at each cortical site. (B) The cerebral cortex was divided into four lobes using FreeSurfer (https://surfer.nmr.mgh.harvard.edu/fswiki/CorticalParcellation; Desikan et al., 2006). Green represents the frontal lobe, white represents the temporal lobe, pink represents the parietal lobe, and blue represents the occipital lobe.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "The developmental changes of cortical MI and HFO at given lobes. (A) MI\u226580 Hz & 0.5-1 Hz and (B) HFOHIL \u226580 Hz occurrence rate (/min). In each dot plot, a regression line is depicted based on a statistical model that incorporates the square root of age (\u221aage) as an independent variable. MI: modulation index. MI\u226580 Hz and 0.5-1 Hz denotes the strength of coupling between the amplitude of HFO\u226580 Hz and the phase of slow-wave0.5-1 Hz. HFOHIL: high-frequency oscillation defined by Cr\u00e9pon et al., 2010.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Snapshots of the developmental atlases of cortical MI and HFO. (A) Model-predicted MI\u226580 Hz & 0.5-1 Hz and (B) HFOHIL \u226580 Hz occurrence rate. Each snapshot presents model-predicted values at ages 1, 5, 10, and 20 years. MI: modulation index. MI\u226580 Hz and 0.5-1 Hz denotes the strength of coupling between the amplitude of HFO\u226580 Hz and the phase of slow-wave0.5-1 Hz. \u00a0HFOHIL: high-frequency oscillation defined by Cr\u00e9pon et al., 2010. Videos S1-S3 present the longitudinal values spanning from 1 to 21 years of age across multiple generations.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Effect of slow-wave spectral frequency bands on modulation index (MI). (A) frontal lobe. (B) Temporal lobe. (C) Parietal lobe. (D) Occipital lobe. MI\u226580 Hz and s Hz denotes the strength of coupling between the amplitude of HFO\u226580 Hz and the phase of slow-waves Hz. MI\u226580 Hz & s Hz at a given electrode site is plotted as a function of s being 0.5-1, 1-2, 2-3, 3-4, 4-5, 5-6, 6-7, or 7-8 Hz. Light blue dots: younger individuals. Magenta dots: older individuals.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Dynamic tractography. The video snapshots present the varying intensity of (A) co-growth of MI\u226580 Hz & 0.5-1 Hz and (B) co-diminution of HFOHIL\u226580 Hz at ages 1, 5, 10, and 20 years, as estimated by univariate regression analysis incorporating \u221aage as an independent variable. Videos S4-S5 show the data across generations from 1 to 21 years.\u00a0",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

05cdd05c8033f50c5d58d58f81996f649b1fce4c208692e95be55fcdb4c8a5aa/preprint/preprint.md

@@ -0,0 +1,290 @@
+# Abstract
+
+The cortex generates high-frequency oscillations (HFO) nested in slow waves during sleep, and these signals are especially elevated in the seizure onset zone. Thus, HFO occurrence rate and Modulation Index (MI), which quantifies the strength of coupling between HFO amplitude and slow-wave phase, represent promising epilepsy biomarkers. However, their diagnostic utility may be suboptimal because the endogenous developmental distributions are unknown. To improve age-appropriate localization of the epileptogenic zone, we hence constructed normative brain atlases demonstrating the developmental changes in MI and HFO rate.
+
+Our study investigated extraoperative, intracranial EEG data from 114 patients with focal epilepsy (ages 1.0 to 41.5 years) who achieved International League Against Epilepsy class I outcomes following resective surgery. We analyzed 20-minute slow-wave sleep epochs at 8,251 nonepileptic electrode sites (those outside the seizure onset zone, interictal spike zone, or MRI-visible lesions). Each electrode was transposed onto a standard brain template, and we then calculated its MI and HFO occurrence rate using four different detector toolboxes. Linear and nonlinear regression models determined the developmental slope of MI and HFO rate at each cortical mesh point. Mixed model analysis established the significance of MI and HFO rate developmental changes in each region of interest, while accounting for the independent effects of patient and epilepsy profiles. Finally, we created a dynamic tractography movie visualizing white matter pathways connecting cortical regions showing developmental co-growth in MI.
+
+We found that the occipital lobe exhibited enhanced MI compared to other lobes in both children and adults. Increased age, square root of age, and log base 10 of age were independently associated with elevated MI exclusively in the occipital lobe. The cortical regions showing developmental co-growth in MI were connected via the vertical occipital fasciculi and posterior callosal fibers. In contrast, we did not observe any significant association between age measures and HFO rate in the occipital lobe, but rather noted an inverse relationship between age and HFO rate in the temporal, frontal, and parietal lobes.
+
+Our study suggests that phase-amplitude coupling between physiologic HFO and delta waves, as rated by MI, is strengthened during development, in the occipital lobe particularly during toddlerhood and preschool. Given that occipital delta-nested HFO are believed to support visual memory consolidation, our observations imply that process may be significantly strengthened during early childhood. The data is publicly available to provide investigators with a crucial reference for MI and HFO-based presurgical evaluation of the epileptogenic zone.
+
+Health sciences/Diseases/Neurological disorders/Epilepsy  
+Biological sciences/Neuroscience/Diseases of the nervous system/Epilepsy
+
+# Introduction
+
+In patients with drug-resistant focal epilepsy, surgical resection of the epileptogenic zone results in long-term seizure control, and clinicians employ intracranial EEG (iEEG) recording via subdural and/or depth electrodes to localize such areas. The occurrence rate of spontaneous interictal high-frequency oscillations (HFO) - defined as transient bursts of ≥ 80 Hz activity (Zweiphenning et al., 2022 a) - has garnered significant attention as an iEEG-based epilepsy biomarker, since it is generally increased in the seizure onset zone (SOZ) (Gloss et al., 2014; Höller et al., 2015; Frauscher et al., 2017; Remakanthakurup Sindhu et al., 2020). Furthermore, a randomized clinical trial involving 78 patients with extratemporal lobe epilepsy recently reported that the efficacy of HFO-guided resection in controlling seizures was non-inferior to that of conventional resection guided by interictal spike discharges (Zweiphenning et al., 2022 a).
+
+In addition, HFO are generally nested within slow waves during sleep, and modulation index (MI) is a quantitative measure of the coupling strength between HFO amplitude and slow-wave phase (Canolty et al., 2006; Miyakoshi et al., 2013). MI represents a potentially valuable iEEG-based epilepsy biomarker; in particular, it is an effective proxy measure for epileptiform spike-and-wave discharges, since each spike component is associated with a transient increase in HFO that is stereotypically coupled with a delta wave (Motoi et al., 2018; Kural et al., 2020). Retrospective analysis of iEEG recordings from human epilepsy patients demonstrated that MI is elevated in the SOZ, and complete resection of high-value MI sites was associated with better postoperative seizure outcomes (Nonoda et al., 2016; Iimura et al., 2018; Motoi et al., 2018; Kural et al., 2020; Kuroda et al., 2021; Ma et al., 2021). Notably, MI has advantages over HFO rate as a biomarker that include being a continuous measure and requiring significantly less computational time (Kuroda et al., 2021).
+
+In clinical practice, caution is required when using MI and HFO rate as iEEG biomarkers because these signals are endogenously present and exhibit topographical variation across the cortex (Guragain et al., 2018; Bernabei et al., 2022; Taylor et al., 2022). The occipital lobe is an area that naturally exhibits elevated MI and HFO rate, regardless of epileptogenicity (Nagasawa et al., 2012; Alkawadri et al., 2014; Frauscher et al., 2018; Kuroda et al., 2021). However, the developmental implications of these physiologic distributions remain unknown, and gleaning this knowledge is expected to significantly improve interpretation of iEEG biomarker clinical significance, in the pediatric brain.
+
+The aim of the present study was to create a novel normative atlas that visualizes developmental changes in MI and HFO rate across artifact-free nonepileptic iEEG electrode sites (defined as those outside the SOZ, interictal spike zone, and MRI-visible lesions; Frauscher et al., 2018; Kuroda et al., 2021; Bernabei et al., 2022; Taylor et al., 2022). To optimize the generalizability of our observations, we employed four different detector toolboxes to calculate HFO rate (Staba et al., 2002; Gardner et al., 2007; Crépon et al., 2010; Zelmann et al., 2010) and used mixed model analysis to determine whether the observed developmental slope of iEEG biomarkers remained significant after accounting for the independent effects of patient and epilepsy profiles (Sakakura et al., 2022).
+
+Invasive studies of healthy cortices in animal models and nonepileptic cortices of patients with focal epilepsy have reported that spontaneous HFO are nested in delta waves and involve large-scale cortical networks during slow-wave sleep (Steriade et al., 1993; Contreras et al., 1996; Sanchez-Vives and McCormick, 2000; Csercsa et al., 2010; Hangya et al., 2011; Nagasawa et al., 2012; Arnulfo et al., 2020). This nested HFO during slow-wave sleep is believed to replay sensory-related neural communications across two regions occurring during wakefulness and facilitate the consolidation of long-term memory related to given sensory representations. For example, spontaneous HFO ≥ 80 Hz nested in delta waves 0.5−1 Hz in the primary visual cortex during slow-wave sleep are thought to be involved in the consolidation of visual memory (Ji and Wilson, 2007; Mehta, 2007; Sasaki et al., 2010; Buzsáki and da Silva, 2012; Nagasawa et al., 2012).
+
+Given that preverbal infants are known to encode and retain visual information (Werner and Perlmutter, 1979) and this process likely involves HFO-delta phase-amplitude coupling, we hypothesized that young children would have already developed the spatial profile of MI ≥ 80 Hz & 0.5−1 Hz: namely physiological enhancement in the nonepileptic occipital lobe, as seen in older children and adults. In addition, since pattern recognition memory is known to improve most rapidly before six years of age (Toornstra et al., 2019), we further hypothesized that the developmental slope of MI ≥ 80 Hz & 0.5−1 Hz would be steepest during young childhood. Here, MI ≥ 80 Hz & 0.5−1 Hz denotes the strength of coupling between the amplitude of HFO ≥ 80 Hz and the phase of slow-wave 0.5−1 Hz.
+
+In addition to modeling cortical ontogenic changes, the present study also visualized white matter pathways supporting the development of neural communications via delta-nested HFO (as rated by MI ≥ 80 Hz & 0.5−1 Hz), during slow-wave sleep. To accomplish this, we utilized our novel imaging technique, dynamic tractography, which combines iEEG signals with MRI tractography (Sonoda et al., 2021; Kitazawa et al., 2023; Ono et al., 2023). Previous studies in animal models and patients with focal epilepsy have demonstrated that neural circuits engaging in high-frequency communication are susceptible to use-dependent modifications of synaptic transmission that enhance effective connectivity via monosynaptic white matter tracts in the developing brain (Buonomano and Merzenich, 1998; Singer, 2018; Sonoda et al., 2021). Additionally, white matter development, as rated by MRI tractography, is generally drastic during early childhood and modest during late adolescence (Asato et al., 2010; Lebel et al., 2018; Baum et al., 2022). Using the dynamic tractography method, we visualized intra- and inter-hemispheric, monosynaptic white matter tracts that connect cortical regions exhibiting significant developmental co-growth of MI ≥ 80 Hz & 0.5−1 Hz.
+
+# Materials And Methods
+
+## Patients
+
+We studied a consecutive series of 114 patients with focal epilepsy (ages 1.0 to 41.5 years) who met the following eligibility criteria (Table 1; Figure S1). Inclusion criteria consisted of: [a] simultaneous video-iEEG recording between January 2007 and November 2020, as part of presurgical assessment at Children's Hospital of Michigan or Harper University Hospital, Detroit Medical Center, [b] iEEG sampling rate of 1,000 Hz (Davis et al., 2018), [c] iEEG contained an artifact-free 20-minute slow-wave sleep epoch at least two hours apart from clinical seizure events (Bagshaw et al., 2009; Nagasawa et al., 2012), and [d] International League Against Epilepsy (ILAE) class 1 outcome in the last follow-up after focal resection (Kuroda et al., 2021). Exclusion criteria included [a] a history of previous resective epilepsy surgery, [b] undergoing hemispherotomy or hemispherectomy, and [c] lacking artifact-free, nonepileptic electrode sites (defined as those outside the SOZ [Asano et al., 2009], interictal spike zone [Kural et al., 2020], and MRI-visible lesions [van Klink et al., 2021]). The Institutional Review Board of Wayne State University approved the current study, and we obtained written informed consent from the patients' legal guardians and written assent from pediatric patients aged 13 years or older.
+
+**Table 1**  
+Patient Profile. In Figure S1, we have provided the distribution of patient ages.
+
+| Number of patients | 114 |
+|--------------------|-----|
+| Mean age in years (range) | 11.4 (1.0–41.5) |
+| Proportion of female (%) | 47.4 |
+| Sampled hemisphere (%) | |
+| Left | 45.6 |
+| Right | 43.0 |
+| Both | 11.4 |
+| Seizure onset zone (%) | |
+| Frontal | 24.6 |
+| Temporal | 47.4 |
+| Parietal | 27.2 |
+| Occipital | 19.3 |
+| MRI-visible structural lesion (%) | 66.7 |
+| Mean number of antiseizure medications (range) | 2.0 (1–5) |
+
+## Extraoperative video-iEEG
+
+We acquired extraoperative video-iEEG data using the same protocols as reported in our previous studies (Nagasawa et al., 2012; Kuroda et al., 2021; Sakakura et al., 2022). A licensed neurosurgeon surgically implanted platinum subdural disk electrodes (3 mm diameter and 10 mm center-to-center distance) on the hemisphere suspected to contain the epileptogenic zone, followed by the placement of surface electromyographic electrodes on the deltoid muscles and electrooculographic electrodes (2.5 cm below and 2.5 cm lateral to the outer canthi) to assess body movements during iEEG recording (Nariai et al., 2011; Nagasawa et al., 2012). Intracranial EEG recording aimed to determine the boundaries of the presumed epileptogenic zone, as well as functionally important cortices, and the spatial extent and duration of iEEG sampling were solely determined by clinical needs specific to each patient. Following implantation, iEEG data was continuously recorded at the bedside with a sampling rate of 1,000 Hz, for 2–7 days. To ensure the accuracy of the developmental iEEG atlases, we only included artifact-free iEEG segments from 8,251 nonepileptic electrode sites (mean: 72.4 per patient; range: 12 to 121; Fig. 1 A). This approach minimized the unwanted effects of pathological HFO events (Guth et al., 2021; Dimakopoulos et al., 2022).
+
+## MRI
+
+We obtained preoperative 3-tesla MRI data, including T1-weighted spoiled gradient-echo volumetric and fluid-attenuated inversion recovery images (Nakai et al., 2017; Sakakura et al., 2022). The FreeSurfer software package was used to reconstruct the MRI surface image of patients aged two and above (http://surfer.nmr.mgh.harvard.edu; Desikan et al., 2006; Ghosh et al., 2010). In cases where the software failed to detect the pial surface accurately due to insufficient cerebral myelination, a board-certified neurosurgeon (K.S.) manually delineated the pial surface using the Control Point function (https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/ControlPoints_freeview/; Deoni et al., 2015; Croteau-Chonka et al., 2016; Remer et al., 2017; Sakakura et al., 2022). For patients younger than two, we used the Infant FreeSurfer software package to reconstruct the surface image (https://surfer.nmr.mgh.harvard.edu/fswiki/infantFS; Zöllei et al., 2020; Sakakura et al., 2022).
+
+We displayed electrode sites on the pial surface of the brain using preoperative MRI and a post-implant CT image (Stolk et al., 2018; Sakakura et al., 2022). Two board-certified neurosurgeons (K.S. and N.K.) visually assessed intraoperative photographs to confirm the spatial accuracy of electrode locations co-registered to the MRI surface image (Pieters et al., 2013). In order to pool sites from all patients, it was necessary to normalize the electrode locations to the standardized FSaverage brain surface (http://surfer.nmr.mgh.harvard.edu). In a previous iEEG study of 32 patients, we found that the mean coregistration error ranged below 0.4 mm, and there was no significant correlation between patient age and the severity of coregistration error (Sakakura et al., 2022).
+
+## Focal resection and postoperative seizure outcome
+
+In our previous studies (Asano et al., 2009; Sonoda et al., 2022), we described the guiding principle for determining the boundary of cortical resection. Our aim was to remove the SOZ and any adjacent MRI lesions, while preserving functionally-important cortex; this procedure is intended to maximize seizure control and minimize development of cognitive and/or sensorimotor deficits. Importantly, none of the data from this study was available to inform clinical decision making. Consistent with the study design, all 114 patients achieved ILAE Class-1 outcome at the last follow-up, which occurred at least one year after surgery.
+
+## Modulation index (MI)
+
+We identified artifact-free, 20-minute slow-wave sleep iEEG recordings that were at least 2 hours apart from ictal events using visual assessment (Bagshaw et al., 2009; Motoi et al., 2018; Kuroda et al., 2021). At each artifact-free nonepileptic site, we used the EEGLAB winPACT toolbox (https://sccn.ucsd.edu/wiki/WinPACT; Miyakoshi et al., 2013) to automatically compute MI; this program transforms all iEEG data points into Hilbert spectra and quantifies the strength of coupling between HFO amplitude and the instantaneous phase of slow waves. In the notation MI≥f Hz & s Hz denotes the strength of coupling between the amplitude of HFO≥f Hz and the phase of slow-wave s Hz.
+
+Below, we report the normative developmental changes primarily in MI≥80 Hz & 0.5−1 Hz (Fig. 2), which is considered a surrogate marker of the delta-nested HFO that are believed to underly consolidation of long-term memory related to given sensory representations (Ji and Wilson, 2007; Mehta, 2007; Sasaki et al., 2010; Buzsáki and da Silva, 2012; Nagasawa et al., 2012; Nonoda et al., 2016). We also report the normative developmental changes in MI≥80 Hz & 3−4 Hz, which are suggested to be increased in iEEG traces showing frequent interictal spike-and-wave discharges (Motoi et al., 2018; Kural et al., 2020). The normative spatial variability of this biomarker value could be a valuable reference in epilepsy presurgical evaluation.
+
+## High-frequency oscillation (HFO)
+
+During the same 20-minute slow-wave sleep iEEG epochs described above, we used the RIPPLELAB software (https://github.com/BSP-Uniandes/RIPPLELAB/; Navarrete et al., 2016) to compute the HFO rate at each artifact-free nonepileptic site, using four different detection algorithms: [1] Short Time Energy (STE) method (Staba et al., 2002), [2] Short Line Length (SLL) method (Gardner et al., 2007), [3] Hilbert (HIL) method (Crépon et al., 2010), and [4] Montreal Neurological Institute (MNI) method (Zelmann et al., 2010). The STE method defines a HFOSTE≥f Hz event as an iEEG segment presenting successive root mean square values greater than five standard deviations above the overall mean of root mean squares and containing more than six peaks greater than three standard deviations on the f-Hz high-pass filtered iEEG trace. The SLL method defines a HFOSLL≥f Hz as an iEEG segment presenting SLL amplitude augmentation greater than the 97.5th percentile of the empirical cumulative distribution function computed on the f-Hz high-pass filtered iEEG trace. The HIL method defines a HFOHIL≥f Hz as an iEEG segment presenting Hilbert transform-based envelope augmentation greater than five standard deviations on the f-Hz high-pass filtered iEEG trace. The MNI method defines a HFOMNI≥f Hz as an iEEG segment presenting root mean square energy above the 99.9999th percentile compared to the baseline. If the baseline was absent due to persistent high-frequency activity, the MNI method treated an iEEG segment presenting the ≥95th percentile of the cumulative distribution function as computed on the 1-minute f-Hz high-pass filtered iEEG trace. We used the default parameter settings in the RIPPLELAB software, as reported previously (Kuroda et al., 2021). In the present study, we focused on reporting the developmental changes in the occurrence rate of HFOHIL≥80 Hz (/min) as HFO event was defined commonly on Hilbert transformed iEEG traces when quantifying the HFOHIL occurrence rate and computing MI.
+
+An ancillary analysis was performed on the rates of HFO≥150 Hz and HFO≥250 Hz for interested readers.
+
+### Statistical analysis to confirm physiological enhancement of MI and HFO in the occipital lobe during young childhood
+
+In a previous study, we created normative atlases of MI and HFO based on data from 47 patients aged between 4 and 19 years, which revealed a general enhancement of these measures in the occipital lobe (Kuroda et al., 2021). In the current study, we aimed to replicate this finding in a cohort of 14 children aged between 1.0 and 3.9 years and 100 individuals aged 4 years or older. We used mixed model analysis with electrode location in the occipital lobe (yes = 1) as the fixed effect predictor variable and either MI≥80 Hz & 0.5−1 Hz, MI≥80 Hz & 3−4 Hz, HFOSTE≥80 Hz, HFOSLL≥80 Hz, HFOHIL≥80 Hz, or HFOMNI≥80 Hz as the dependent variable. The random effect factors included intercept and patient. We considered an FDR-corrected p-value of less than 0.05 as significant, for comparisons of six iEEG measures. We reported the mixed model effect and 95% confidence interval (95% CI) to highlight the impact of topography on normative iEEG biomarker measures. All statistical analyses were performed using Matlab R2020a (MathWorks Inc., Natick, MA).
+
+### Statistical analysis to visualize the developmental slope of cortical MI and HFO at given mesh points
+
+We presented iEEG biomarker measures at each nonepileptic electrode site on a standardized brain surface image using FreeSurfer and interpolation within 10 mm from the electrode center at the individual patient level (Sakakura et al., 2022, Ono et al., 2023). Using linear and nonlinear univariate regression models at cortical surface mesh points consisting of 20 neighboring FreeSurfer vertex finite elements (Desikan et al., 2006; Sakakura et al., 2022), we then determined the developmental slope of the iEEG biomarker, at the whole brain level. We used age, square root of age (√age), and log10 age as an independent variable, and MI≥80 Hz & 0.5−1 Hz, MI≥80 Hz & 3−4 Hz, HFOSTE≥80 Hz, HFOSLL≥80 Hz, HFOHIL≥80 Hz, or HFOMNI≥80 Hz as dependent variables. We evaluated the goodness of fit of each regression model using Akaike Information Criterion (AIC). A biomarker value was considered to increase with age if the regression slope was significantly greater than zero. We identified the mesh points where the developmental change (enhancement or diminution) of a given iEEG biomarker measure survived an FDR correction (for 18 comparisons: six iEEG measures × three types of age measures) in the resulting normative atlas. We also created video atlases, each displaying MI≥80 Hz & 0.5−1 Hz, MI≥80 Hz & 3−4 Hz, or HFOHIL≥80 Hz values predicted by a given regression model at cortical mesh points (Fig. 3; Videos S1-S3).
+
+### Statistical analysis to determine the independent effects of development on cortical MI and HFO in each lobe
+
+We used mixed model analysis (Sonoda et al., 2021; Sakakura et al., 2022) to determine the lobe where a developmental change of a given iEEG biomarker remained significant, after controlling for the independent effects of clinical profile and epilepsy-related variables. The aim was to account for potential confounders that could affect MI and HFO measures.
+
+To assess the developmental change of MI at each lobe, we utilized the MATLAB fitlme command (https://www.mathworks.com/help/stats/fitlme.html) to fit a mixed model specified by the following formula: 'MI ~ 1 + age + sex + SOZ + MRI + hemisphere + number of antiseizure medications + (1|patient)'. Here, the dependent variable was MI at a given analysis mesh point, and the fixed effect predictors included [1] age at surgery (e.g., √year), [2] sex (female = 1), [3] presence of SOZ in a given lobe (yes = 1), [4] presence of MRI-visible structural lesion (yes = 1), [5] sampled hemisphere (left = 1), and [6] number of oral antiseizure medications taken immediately before the initiation of iEEG recording. We considered a larger number of antiseizure medications as a surrogate of a more severe seizure-related cognitive burden since polytherapy is associated with more disabling seizures and seizure-related cognitive impairment (Kwan and Brodie, 2001; Kuroda et al., 2021). We employed this approach since no single neuropsychological assessment can quantify the severity of cognitive impairment across all age ranges. Our random effect factors included the intercept and patient. We deemed an FDR-corrected p-value of 0.05 (for 18 comparisons: six iEEG measures × three types of age measures) as the significance threshold.
+
+### Statistical analysis to determine if the spectral frequency band of slow waves nesting HFO would change with development
+
+Some might hypothesize that slow waves with a higher spectral frequency band would facilitate a higher rate of spontaneous neural communication through nested HFO. As an ancillary analysis, we, therefore, sought to determine whether the spectral frequency band of slow waves nesting HFO would change with development. To accomplish this, we measured MI≥80 Hz & s Hz, where s was 0.5-1, 1–2, 2–3, 3–4, 4–5, 5–6, 6–7, or 7–8 (Fig. 4) and calculated the regression slope of MI≥80 Hz & s Hz with respect to s Hz, in each lobe. We then assessed whether this regression slope was dependent on patient age in each lobe using the Spearman correlation. A higher regression slope in older individuals, compared to younger individuals, would suggest that HFO≥80 Hz in older individuals were preferentially coupled with slow waves of higher spectral frequency bands. We set an FDR-corrected p-value of 0.05 (for testing at four lobes) as the significance threshold.
+
+## Statistical analysis: white matter tracts between cortices showing developmental MI co-growth
+
+We visualized white matter tracts directly connecting cortical mesh points that showed significant developmental co-growth (or co-reduction) of MI≥80 Hz & 0.5−1 Hz. To this end, we used the regression slope of MI≥80 Hz & 0.5−1 Hz as a function of √age at each cortical mesh point, as computed in a regression analysis mentioned above. We declared that the developmental enhancement (or reduction) of nested HFO-based neural communications took place between two cortical mesh points only if [1] two distinct mesh points showed significantly positive (or negative) regression slopes, and [2] these mesh points were accompanied by direct tractography streamlines on diffusion-weighted imaging (DWI) analysis (Kitazawa et al., 2023; Ono et al., 2023).
+
+We delineated white matter DWI streamlines using open-source data from 1,065 healthy participants (http://brain.labsolver.org/diffusion-mri-templates/hcp-842-hcp-1021; Yeh et al., 2018), as previously reported (Mitsuhashi et al., 2021, 2022; Sonoda et al., 2021; Kitazawa et al., 2023; Ono et al., 2023). Our group validated the use of open-source DWI data by demonstrating that the inferred velocity of neural propagations induced by single-pulse electrical stimulation was similar whether using open-source or individual patient DWI data (Sonoda et al., 2021). We placed seeds (4-mm radius) at cortical mesh points demonstrating significant developmental enhancement (or reduction) of nested HFO-based neural communications (as rated by MI≥80 Hz & 0.5−1 Hz). Using DSI Studio (http://dsi-studio.labsolver.org/), we visualized DWI streamlines directly connecting the mesh points within Montreal Neurological Institute standard space. For fiber tracking, we utilized the following parameters: a quantitative anisotropy threshold of 0.05, a maximum turning angle of 70°, and a streamline length of 20 to 250 mm. In this investigation, we exclusively visualized DWI streamlines with at least 50% of their coordinates in one of the following white matter tracts: arcuate fasciculus, cingulum, corpus callosum, extreme capsule, frontal aslant tract, inferior fronto-occipital fasciculus, inferior longitudinal fasciculus, middle longitudinal fasciculus, superior longitudinal fasciculus, uncinate fasciculus, or vertical occipital fasciculus, as defined in DSI Studio (as previously performed in Sonoda et al., 2021). We excluded streamlines involving the brainstem, basal ganglia, thalamus, or cerebrospinal fluid space.
+
+The resulting dynamic tractography video atlases highlighted the intensity of MI≥80 Hz & 0.5−1 Hz developmental co-growth (or co-reduction) via given tractography streamlines for every 0.1 years; thereby, the intensity was defined as (√│regression slope at a mesh point│ × √│regression slope at another mesh point│), at a given streamline connecting a pair of mesh points (Fig. 5).
+
+## Data and code availability statement
+
+The iEEG data are available at https://openneuro.org/ (doi: 10.18112/openneuro.ds004551.v1.0.2) and https://nemar.org/. The analysis codes are available at https://github.com/kaz1126/flatten_map_and_tractography.
+
+# Results
+
+## Developmental atlas of cortical MI
+
+Figure 2A illustrates the developmental changes in cortical MI ≥80 Hz & 0.5-1 Hz at specific lobes using dot plots. The regression slope, uncorrected p-value, regression t-value, and degrees of freedom (DF) are also shown. Video S1 provides a comprehensive summary of normative development of MI ≥80 Hz & 0.5-1 Hz, estimated by univariate regression models with age, √age, or log₁₀ age as independent variables. Linear and nonlinear regression models consistently indicate significant developmental growth in MI ≥80 Hz & 0.5-1 Hz prominently in the occipital lobe during young childhood and beyond. The assessment of AIC suggests that the developmental growth in occipital MI ≥80 Hz & 0.5-1 Hz was drastic during young childhood and modest during older childhood. The mean regression slope was +0.0072 /year (95%CI: +0.0050 to +0.0093; AIC: -368.0) in the linear regression model, +0.047 /√year (95%CI: +0.034 to +0.060; AIC: -374.3) and +0.14 /log₁₀ year (95%CI: +0.10 to +0.18; AIC: -376.6) in the nonlinear models. The regression slopes of MI ≥80 Hz & 0.5-1 Hz in the parietal, temporal, and frontal lobes were much flatter, compared to the occipital lobe (Figure 2 and Video S1).
+
+As detailed in Figure S2 and Video S2, the developmental changes of MI ≥80 Hz & 0.5-1 Hz and MI ≥80 Hz & 3-4 Hz were qualitatively similar. The developmental growth of MI ≥80 Hz & 3-4 Hz was highest in the occipital lobe. The mean regression slope was +0.019 /√year (95%CI: +0.014 to +0.023) in the occipital lobe but not significantly different from zero in the remaining lobes.
+
+Figure 3A displays snapshots of normative developmental atlases for MI ≥80 Hz & 0.5-1 Hz at the whole brain level, emphasizing the increased values in cortical regions near the calcarine sulcus.
+
+## Developmental atlas of cortical HFO rate
+
+The regression models, applied to all 114 patients, did not demonstrate any significant developmental growth or diminution of occipital HFO ≥80 Hz as defined by any detector (uncorrected p-value > 0.05; Figure 2B; Figure S2). In contrast, the rate of extra-occipital HFO ≥80 Hz generally showed significant developmental diminution (uncorrected p-value < 0.001). Such developmental diminution of extra-occipital HFO resulted in a relatively maintained HFO occurrence rate in the occipital lobes, particularly those adjacent to the calcarine cortex, during adolescence and beyond (Figure 3B). Below, we describe the findings of HFO HIL ≥80 Hz in detail.
+
+Video S3 provides a comprehensive summary of normative development of HFO HIL≥80 Hz rate. Occipital HFO HIL ≥80 Hz rate failed to show significant developmental change (uncorrected p-value > 0.05; Figure 2B). In contrast, extra-occipital HFO HIL ≥80 Hz rate showed significant developmental diminution (uncorrected p-value < 0.001; Figure 2B). The assessment of AIC suggested that the developmental diminution in extra-occipital HFO HIL ≥80 Hz was comparable during young and older childhood (AIC in extra-occipital lobes: 2.60 x 10⁴ in age-incorporated model; 2.59 x 10⁴ in √age model; 2.59 x 10⁴ in log₁₀ age model). The regression slope was -0.41/√year (95%CI: -0.46 to -0.36) in the frontal lobe, -0.30/√year (95%CI: -0.36 to -0.25) in the temporal lobe, -0.33/√year (95%CI: -0.40 to -0.26) in the parietal lobe, and +0.081/√year (95%CI: -0.075 to +0.24) in the occipital lobe.
+
+Due to the infrequent occurrence of events, regression models failed to fit the data of HFO ≥150 Hz and HFO ≥250 Hz rates in a meaningful manner.
+
+## MI enhancement in the occipital lobe during young childhood and after
+
+The mixed model analysis of iEEG data of 14 children aged from 1.0 and 3.9 suggested that, as compared to the other lobes, the occipital lobe showed higher MI ≥80 Hz & 0.5-1 Hz (mixed model effect: +0.087 [95%CI: +0.078 to +0.097]; uncorrected p-value: < 0.001) and MI ≥80 Hz & 3-4 Hz (mixed model effect: +0.033 [95%CI: +0.028 to +0.038]; uncorrected p-value: < 0.001).
+
+An ancillary mixed model analysis of iEEG data of 100 individuals aged from 4.0 and 41.5 suggested that, as compared to the other lobes, the occipital lobe likewise showed higher MI ≥80 Hz & 0.5-1 Hz (mixed model effect: +0.17 [95%CI: +0.17 to +0.18]; uncorrected p-value: < 0.001) and MI ≥80 Hz & 3-4 Hz (mixed model effect: +0.063 [95%CI: +0.061 to +0.066]; uncorrected p-value: < 0.001). These findings indicate that the nonepileptic occipital lobe exhibits higher MI in comparison to the other regions, across all age groups (Figure 3A). Moreover, the difference in MI between the occipital and extra-occipital lobe regions was found to be approximately two times greater in older individuals.
+
+## Topographical variations of HFO rate during young childhood and after
+
+The mixed-model analysis of iEEG data from 14 children between the ages of 1.0 and 3.9 did not show a higher HFO rate in the occipital lobe compared to extra-occipital lobes, except when considering the HFO rate defined by the Hilbert method (HFO STE ≥80 Hz mixed model effect: +0.022 [95%CI: -0.039 to +0.082]; uncorrected p-value: 0.48; HFO SLL ≥80 Hz mixed model effect: +0.036 [95%CI: -0.46 to +0.53]; uncorrected p-value: 0.89; HFO HIL ≥80 Hz mixed model effect: +0.84 [95%CI: +0.57 to +1.1]; uncorrected p-value: < 0.001; HFO MNI ≥80 Hz mixed model effect: -0.044 [95%CI: -0.90 to +3.0 x 10⁻³]; uncorrected p-value: 0.067). The model suggests that the nonepileptic occipital lobe of children under four years old generates Hilbert method-defined HFO ≥80 Hz occurrences at a rate of 0.84/min more frequently than extra-occipital lobe regions.
+
+Another mixed-model analysis of iEEG data from 100 individuals between the ages of 4.0 and 41.5 demonstrated that the occipital lobe had a higher HFO rate compared to the extra-occipital lobes, except when considering the HFO rate defined by the MNI method (HFO STE ≥80 Hz mixed model effect: +0.14 [95%CI: +0.12 to +0.16]; uncorrected p-value: < 0.001; HFO SLL ≥80 Hz mixed model effect: +0.59 [95%CI: +0.43 to +0.75]; uncorrected p-value: < 0.001; HFO HIL ≥80 Hz mixed model effect: +1.8 [95%CI: +1.7 to +1.9]; uncorrected p-value: < 0.001; HFO MNI ≥80 Hz mixed model effect: +0.043 [95%CI: -0.014 to +0.099]; uncorrected p-value: 0.14). The model suggests that the non-epileptic occipital lobe of individuals of four years old or above generates Hilbert method-defined HFO ≥80 Hz occurrences at a rate of 1.8/min more frequently than extra-occipital lobe regions (Figure 3B).
+
+## Independent effects of development on cortical MI in each lobe
+
+The mixed model analysis, employed to all 114 patients, confirmed that an increase in √age was associated with an increase in occipital MI ≥80 Hz & 0.5-1 Hz (mixed model effect: +0.046 [95%CI: +0.019 to +0.072]; uncorrected p-value: 7.43 x 10⁻⁴; t-value: +3.39; DF: 868) and MI ≥80 Hz & 3-4 Hz (mixed model effect: +0.019 [95%CI: +0.0090 to +0.029]; uncorrected p-value: 1.22 x 10⁻⁴; t-value: +3.86; DF: 868), independent of patient and epilepsy profiles. In contrast, mixed model analysis failed to confirm that an increase in √age was associated with an increase in extra-occipital MI ≥80 Hz & 0.5-1 Hz or MI ≥80 Hz & 3-4 Hz (uncorrected p-value: >0.05).
+
+## Independent effects of development on cortical HFO rate in each lobe
+
+The mixed model analysis confirmed that an increase in √age was independently associated with a diminution of HFO HIL≥80 Hz rate in the frontal (mixed model effect: -0.53; uncorrected p-value: 5.68 x 10⁻⁷; t-value: -4.97; DF: 2973), temporal (mixed model effect: -0.34; uncorrected p-value: 2.0 x 10⁻³; t-value: -3.09; DF: 2388), and parietal lobes (mixed model effect: -0.38; uncorrected p-value: 5.43 x 10⁻⁴; t-value: -3.46; DF: 1994). However, this phenomenon was not observed in the occipital lobe (mixed model effect: +0.069; uncorrected p-value: 0.69; t-value: +0.40; DF: 868).
+
+## Developmental changes of the spectral frequency band of slow waves nesting HFO
+
+Looking at MI ≥80 Hz & varying slow waves, we failed to find a significant developmental change of the frequency band of slow waves nesting HFO (Figure 4). On average, the regression slope of MI ≥80 Hz & s Hz as a function of slow waves (s Hz) was -0.014 /Hz in the frontal lobe, -0.013 /Hz in the parietal lobe, -0.010 /Hz in the temporal lobe, and -0.027 /Hz in the occipital lobe. In other words, MI ≥80 Hz values were generally higher when using lower frequency slow waves. The Spearman correlation failed to demonstrate a significant correlation between this regression slope and patient age in any lobe (uncorrected p-value: >0.05). Thus, the present study failed to demonstrate that the spectral frequency band of slow waves nesting HFO would change with development.
+
+## White matter tracts connecting cortices with developmental MI co-growth
+
+Univariate regression models, incorporating √age as an independent variable, found that 159 out of the 9464 cortical mesh points (1.7%) showed significantly positive regression slopes, whereas no cortical mesh points showed significantly negative regression slopes. DWI analysis revealed that the vertical occipital fasciculi and posterior callosal fibers directly connected mesh point pairs showing significant developmental co-growth of MI ≥80 Hz & 0.5-1 Hz (Figure 5A and Video S4).
+
+## White matter tracts connecting cortices with developmental HFO co-diminution
+
+Univariate regression models incorporating √age as an independent variable found that 902 out of the 9464 cortical mesh points (9.5%) showed significantly negative regression slopes, whereas no cortical mesh points showed significantly positive regression slopes. DWI analysis revealed that the arcuate fasciculus, corpus callosum, extreme capsule, frontal aslant tract, inferior fronto-occipital fasciculus, inferior longitudinal fasciculus, middle longitudinal fasciculus, and superior longitudinal fasciculus directly connected mesh point pairs showing significant developmental co-diminution of HFO HIL≥80 Hz rate (Figure 5B and Video S5).
+
+# Discussion
+
+## Clinical significance of enhanced MI in the occipital lobe in epilepsy presurgical evaluation
+
+For the first time, the current study created a developmental atlas of MI at the whole brain level, including the medial and inferior surfaces of the cerebral cortex. It demonstrated physiologically enhanced MI in the nonepileptic occipital lobe of children younger than 4 years and in the older patient group. The occipital lobe showed higher MI ≥ 80 Hz & 3−4 Hz than the other lobes by 0.033 on average in children younger than 4 years and by 0.063 on average in older patients. Although MI ≥ 80 Hz & 3−4 Hz is an excellent surrogate marker of interictal spike-and-wave discharges (Motoi et al., 2018; Kural et al., 2020), the occurrence of physiological HFO nested in slow-wave background activity (Steriade et al., 1993; Contreras et al., 1996; Sanchez-Vives and McCormick, 2000, Nagasawa et al., 2012; Nonoda et al., 2016) inflates MI ≥ 80 Hz & 3−4 Hz in the nonepileptic occipital lobe. Therefore, iEEG investigators should cautiously interpret the significance of high-value, occipital MI in presurgical evaluation for young and older patients with drug-resistant focal epilepsy. Nonepileptic occipital lobe sites, especially those proximal to the calcarine sulcus, are expected to show MI values higher than nonepileptic extra-occipital lobe sites (Video S1 and Fig. 3 A). With awareness of such topographic variations of MI, iEEG investigators can reduce the risk of incorrect localization of the epileptogenic zone.
+
+Investigators previously reported the utility of normative atlases of MI, HFO, and other spectral frequency bands, in the presurgical evaluation of patients with the mean age ranging between adolescence and adulthood (Frauscher et al., 2018; Guragain et al., 2018; Kuroda et al., 2021; Bernabei et al., 2022; Taylor et al., 2022; Zweiphenning et al., 2022 b). A prospective study is warranted to investigate whether a more inclusive resection of cortical sites, whose MI values deviate from the age-specific normal range, would predict better postoperative seizure control in young children.
+
+## Clinical significance of HFO atlases in epilepsy presurgical evaluation
+
+For the first time, the current study investigated the developmental changes of HFO occurrence rate defined by four open-source detectors (Fig. 2; Figure S2). Commonly across detectors, the HFO rate diminished in the frontal, parietal, and temporal lobes with development. Thus, clinicians need to be aware that young children may show nonepileptic HFO in the extra-occipital lobe regions more frequently than adults. As best demonstrated in Video S3, HFO HIL ≥80 Hz rate was physiologically enhanced in the occipital regions proximal to the calcarine cortex, but such occipital HFO enhancement was only evident during adolescence and adulthood and not during infancy or toddlerhood.
+
+HFO MNI ≥80 Hz rate was very low in the occipital lobes throughout all ages; this MNI method-specific observation can be attributed to its detector being designed to be agnostic to persistent forms of high-frequency activity (Zelmann et al., 2010), as often seen in the nonepileptic occipital lobe (Nagasawa et al., 2012).
+
+## Mechanistic significance of developmental changes of MI and HFO
+
+What is responsible for the developmental enhancement of MI in the occipital lobe? The current study found that MI ≥ 80 Hz & 0.5−1 Hz increases rapidly during early childhood and modestly thereafter. However, the occipital HFO rate did not increase significantly with development, suggesting that rate changes may not sufficiently account for the developmental enhancement of MI (Fig. 2 B). Furthermore, an assessment of the spectral frequency bands of slow-waves nesting HFO (Fig. 4) did not provide evidence that such nesting slow rhythms become faster with development. In contrast, our white matter connectivity assessment revealed that occipital regions exhibiting a developmental enhancement of MI were accompanied by monosynaptic white matter streamlines, including the vertical occipital fasciculi and posterior callosal fibers (Fig. 5 A). These white matter structures are essential for object recognition (Baynes et al., 1998; Schulte et al., 2010; Yeatman et al., 2014; Herbet et al., 2018); thus, one possible explanation for the developmental enhancement of occipital MI is that the cerebral cortex learns to generate HFO at a preferred phase of background rhythms in an experience-dependent manner, optimizing visual memory consolidation during slow-wave sleep (Ji and Wilson, 2007; Mehta, 2007; Sasaki et al., 2010; Buzsáki and da Silva, 2012; Nagasawa et al., 2012). This notion is supported by the observation that monocular sight deprivation during a critical period altered delta-gamma phase-amplitude coupling in the mouse visual cortex (Malik et al., 2022).
+
+## Innovation
+
+The methodological innovations in the current study include the generation of developmental atlases based on iEEG signals from 114 patients, including infants and toddlers. We previously validated our group-level analysis of iEEG across generations from infancy to adulthood (Sakakura et al., 2022). A pivotal procedure to ensure the spatial accuracy of electrode-brain surface coregistration was a manual delineation of the pial surface in the temporal lobe regions that are frequently unmyelinated in young children (Sakakura et al., 2022).
+
+Visualization of the direct white matter streamlines connecting cortical regions with a developmental enhancement of MI is an innovative application of dynamic tractography technology (Sonoda et al., 2021; Kitazawa et al., 2023; Ono et al., 2023). We interpret the visualized streamlines, including vertical occipital fasciculi and posterior callosal fibers, to reflect the ongoing development of large-scale neural communications based on slow wave-nested HFO before reaching the adult level.
+
+## Methodological considerations
+
+Inevitable limitations in the present study include limited spatial sampling, though 8,251 nonepileptic electrode sites were available for analysis. We are only able to place electrodes that are clinically indicated. Neither Heschl nor insular gyri were analyzed in the present study using subdural electrodes. Thus, our study was not designed to assess the developmental enhancement of MI or HFO in much of the primary auditory cortex. We did not sample iEEG signals from the thalamus either because there was no clinical indication. Since stereotactic-EEG depth electrodes are used to assess the thalamic nuclei to determine the optimal target for deep brain stimulation and responsive neurostimulation therapy, further studies are expected to determine the developmental changes of MI and HFO in such deep brain structures in the future.
+
+Another limitation included cross-sectional measurement due to the invasive nature of iEEG recording. Chronic iEEG recording is available through eight intracranial electrode contacts in patients undergoing responsive neurostimulation. Yet, longitudinal assessment of nonepileptic iEEG signals using these contacts may not be feasible because these electrodes are implanted in the area proximal to the epileptogenic zone.
+
+Our ancillary analysis on the rates of HFO ≥ 150 Hz and HFO ≥ 250 Hz revealed that the infrequent occurrence of these events made it impossible for regression models to meaningfully fit the data. To detect sufficient numbers of HFO ≥ 150 Hz and HFO ≥ 250 Hz events for analysis of developmental changes, interictal slow-wave sleep epochs much longer than 20 minutes would be necessary. However, the duration of iEEG recording is determined solely by clinical needs, making it unrealistic to expect that all study participants would have prolonged iEEG epochs that meet the eligibility criteria employed in this study. Although the spatial distribution of the rates of HFO ≥ 80 Hz, HFO ≥ 150 Hz, and HFO ≥ 250 Hz are generally similar to each other, some studies suggest that HFO ≥ 250 Hz may be more specific than HFO ≥ 80 Hz in localizing the epileptogenic zone; however, others imply that HFO ≥ 250 Hz may not be as practical as HFO ≥ 80 Hz due to its rarity and its presence in the nonepileptic eloquent cortex (Gloss et al., 2014; Höller et al., 2015; Frauscher et al., 2017; Roehri et al., 2018; Kuroda et al., 2021; Zweiphenning et al., 2022 a).
+
+One drawback of using iEEG is that we are unable to include healthy controls, because it is unethical to implant subdural electrodes without clinical indication. Thus, we excluded electrode sites affected by either SOZ, interictal spike zone, and MRI-visible lesions from the analysis. Furthermore, we employed mixed model analysis to control the effects of potential confounding factors on MI/HFO measures, including antiseizure medications, presence of MRI-visible lesions, and locations of SOZ. Despite those issues, iEEG involves benefits over scalp recording such as, > 100 times better signal fidelity and sampling from the medial or inferior cerebral cortex surface (Ball et al., 2009).
+
+# References
+
+1. Zweiphenning W, Klooster MAV', van Klink NEC, et al. Intraoperative electrocorticography using high-frequency oscillations or spikes to tailor epilepsy surgery in the Netherlands (the HFO trial): a randomised, single-blind, adaptive non-inferiority trial. Lancet Neurol. 2022;21(11):982–993.
+2. Gloss D, Nolan SJ, Staba R. The role of high-frequency oscillations in epilepsy surgery planning. Cochrane Database Syst Rev. 2014;1(1):CD010235.
+3. Höller Y, Kutil R, Klaffenböck L, et al. High-frequency oscillations in epilepsy and surgical outcome. A meta-analysis. Front Hum Neurosci. 2015;9:574.
+4. Frauscher B, Bartolomei F, Kobayashi K, et al. High-frequency oscillations: The state of clinical research. Epilepsia. 2017;58(8):1316–1329.
+5. Remakanthakurup Sindhu K, Staba R, Lopour BA. Trends in the use of automated algorithms for the detection of high-frequency oscillations associated with human epilepsy. Epilepsia. 2020;61(8):1553–1569.
+6. Canolty RT, Edwards E, Dalal SS, et al. High gamma power is phase-locked to theta oscillations in human neocortex. Science. 2006;313(5793):1626–1628.
+7. Miyakoshi M, Delorme A, Mullen T, Kojima K, Makeig S, Asano E. Automated detection of cross-frequency coupling in the electrocorticogram for clinical inspection. Annu Int Conf IEEE Eng Med Biol Soc. 2013;2013:3282–3285.
+8. Kuroda N, Sonoda M, Miyakoshi M, et al. Objective interictal electrophysiology biomarkers optimize prediction of epilepsy surgery outcome. Brain Commun. 2021;3(2):fcab042.
+9. Nonoda Y, Miyakoshi M, Ojeda A, et al. Interictal high-frequency oscillations generated by seizure onset and eloquent areas may be differentially coupled with different slow waves. Clin Neurophysiol. 2016;127(6):2489–2499.
+10. Iimura Y, Jones K, Takada L, et al. Strong coupling between slow oscillations and wide fast ripples in children with epileptic spasms: Investigation of modulation index and occurrence rate. Epilepsia. 2018;59(3):544–554.
+11. Ma H, Wang Z, Li C, Chen J, Wang Y. Phase-Amplitude Coupling and Epileptogenic Zone Localization of Frontal Epilepsy Based on Intracranial EEG. Front Neurol. 2021;12:718683.
+12. Motoi H, Miyakoshi M, Abel TJ, et al. Phase-amplitude coupling between interictal high-frequency activity and slow waves in epilepsy surgery. Epilepsia. 2018;59(10):1954–1965.
+13. Kural MA, Duez L, Sejer Hansen V, et al. Criteria for defining interictal epileptiform discharges in EEG: A clinical validation study. Neurology. 2020;94(20):e2139-e2147.
+14. Guragain H, Cimbalnik J, Stead M, et al. Spatial variation in high-frequency oscillation rates and amplitudes in intracranial EEG. Neurology. 2018;90(8):e639-e646.
+15. Bernabei JM, Sinha N, Arnold TC, et al. Normative intracranial EEG maps epileptogenic tissues in focal epilepsy [published correction appears in Brain. 2023 Jan 5;146(1):e5]. Brain. 2022;145(6):1949–1961.
+16. Taylor PN, Papasavvas CA, Owen TW, et al. Normative brain mapping of interictal intracranial EEG to localize epileptogenic tissue. Brain. 2022;145(3):939–949.
+17. Nagasawa T, Juhász C, Rothermel R, Hoechstetter K, Sood S, Asano E. Spontaneous and visually driven high-frequency oscillations in the occipital cortex: intracranial recording in epileptic patients. Hum Brain Mapp. 2012;33(3):569–583.
+18. Alkawadri R, Gaspard N, Goncharova II, et al. The spatial and signal characteristics of physiologic high frequency oscillations. Epilepsia. 2014;55(12):1986–1995.
+19. Frauscher B, von Ellenrieder N, Zelmann R, et al. High-Frequency Oscillations in the Normal Human Brain. Ann Neurol. 2018;84(3):374–385.
+20. Staba RJ, Wilson CL, Bragin A, Fried I, Engel J Jr. Quantitative analysis of high-frequency oscillations (80–500 Hz) recorded in human epileptic hippocampus and entorhinal cortex. J Neurophysiol. 2002;88(4):1743–1752.
+21. Gardner AB, Worrell GA, Marsh E, Dlugos D, Litt B. Human and automated detection of high-frequency oscillations in clinical intracranial EEG recordings. Clin Neurophysiol. 2007;118(5):1134–1143.
+22. Crépon B, Navarro V, Hasboun D, et al. Mapping interictal oscillations greater than 200 Hz recorded with intracranial macroelectrodes in human epilepsy. Brain. 2010;133(Pt 1):33–45.
+23. Zelmann R, Mari F, Jacobs J, Zijlmans M, Chander R, Gotman J. Automatic detector of high frequency oscillations for human recordings with macroelectrodes. Annu Int Conf IEEE Eng Med Biol Soc. 2010;2010:2329–2333.
+24. Sakakura K, Sonoda M, Mitsuhashi T, et al. Developmental organization of neural dynamics supporting auditory perception. Neuroimage. 2022;258:119342.
+25. Steriade M, Nuñez A, Amzica F. A novel slow (< 1 Hz) oscillation of neocortical neurons in vivo: depolarizing and hyperpolarizing components. J Neurosci. 1993;13(8):3252–3265.
+26. Contreras D, Timofeev I, Steriade M. Mechanisms of long-lasting hyperpolarizations underlying slow sleep oscillations in cat corticothalamic networks. J Physiol. 1996;494 (Pt 1)(Pt 1):251–264.
+27. Sanchez-Vives MV, McCormick DA. Cellular and network mechanisms of rhythmic recurrent activity in neocortex. Nat Neurosci. 2000;3(10):1027–1034.
+28. Csercsa R, Dombovári B, Fabó D, et al. Laminar analysis of slow wave activity in humans. Brain. 2010;133(9):2814–2829.
+29. Hangya B, Tihanyi BT, Entz L, et al. Complex propagation patterns characterize human cortical activity during slow-wave sleep. J Neurosci. 2011;31(24):8770–8779.
+30. Arnulfo G, Wang SH, Myrov V, et al. Long-range phase synchronization of high-frequency oscillations in human cortex. Nat Commun. 2020;11(1):5363. Published 2020 Oct 23.
+31. Ji D, Wilson MA. Coordinated memory replay in the visual cortex and hippocampus during sleep. Nat Neurosci. 2007;10(1):100–107.
+32. Mehta MR. Cortico-hippocampal interaction during up-down states and memory consolidation. Nat Neurosci. 2007;10(1):13–15.
+33. Sasaki Y, Nanez JE, Watanabe T. Advances in visual perceptual learning and plasticity. Nat Rev Neurosci. 2010;11(1):53–60.
+34. Buzsáki G, Silva FL. High frequency oscillations in the intact brain. Prog Neurobiol. 2012;98(3):241–249.
+35. Werner JS, Perlmutter M. Development of visual memory in infants. Adv Child Dev Behav. 1979;14:1–56.
+36. Toornstra A, Hurks PPM, Van der Elst W, Kok G, Curfs LMG. Measuring visual, spatial, and visual spatial short-term memory in schoolchildren: studying the influence of demographic factors and regression-based normative data. Journal of Pediatric Neuropsychology. 2019;5:119–131.
+37. Kitazawa Y, Sonoda M, Sakakura K, et al. Intra- and inter-hemispheric network dynamics supporting object recognition and speech production. Neuroimage. 2023;270:119954.
+38. Ono H, Sonoda M, Sakakura K, et al. Dynamic cortical and tractography atlases of proactive and reactive alpha and high-gamma activities. Brain Commun. 2023. https://doi.org/10.1093/braincomms/fcad111
+39. Buonomano DV, Merzenich MM. Cortical plasticity: from synapses to maps. Annu Rev Neurosci. 1998;21:149–186.
+40. Singer W. The role of oscillations and synchrony in the development of the nervous system. In: Benasich AA, Ribary U, Lupp J, eds. Emergent Brain Dynamics. Prebirth to Adolescence. Strungmann Forum Reports. MIT Press; 2018:15–32.
+41. Sonoda M, Silverstein BH, Jeong JW, et al. Six-dimensional dynamic tractography atlas of language connectivity in the developing brain. Brain. 2021;144(11):3340–3354.
+42. Asato MR, Terwilliger R, Woo J, Luna B. White matter development in adolescence: a DTI study. Cereb Cortex. 2010;20(9):2122–2131.
+43. Lebel C, Deoni S. The development of brain white matter microstructure. Neuroimage. 2018;182:207–218.
+44. Baum GL, Flournoy JC, Glasser MF, et al. Graded Variation in T1w/T2w Ratio during Adolescence: Measurement, Caveats, and Implications for Development of Cortical Myelin. J Neurosci. 2022;42(29):5681–5694.
+45. Davis KA, Devries SP, Krieger A, et al. The effect of increased intracranial EEG sampling rates in clinical practice. Clin Neurophysiol. 2018;129(2):360–367.
+46. Bagshaw AP, Jacobs J, LeVan P, Dubeau F, Gotman J. Effect of sleep stage on interictal high-frequency oscillations recorded from depth macroelectrodes in patients with focal epilepsy. Epilepsia. 2009;50(4):617–628.
+47. Asano E, Juhász C, Shah A, Sood S, Chugani HT. Role of subdural electrocorticography in prediction of long-term seizure outcome in epilepsy surgery. Brain. 2009;132(Pt 4):1038–1047.
+48. van Klink NEC, Zweiphenning WJEM, Ferrier CH, et al. Can we use intraoperative high-frequency oscillations to guide tumor-related epilepsy surgery?. Epilepsia. 2021;62(4):997–1004.
+49. Nariai H, Nagasawa T, Juhász C, Sood S, Chugani HT, Asano E. Statistical mapping of ictal high-frequency oscillations in epileptic spasms. Epilepsia. 2011;52(1):63–74.
+50. Guth TA, Kunz L, Brandt A, et al. Interictal spikes with and without high-frequency oscillation have different single-neuron correlates. Brain. 2021;144(10):3078–3088.
+51. Dimakopoulos V, Gotman J, Stacey W, et al. Protocol for multicentre comparison of interictal high-frequency oscillations as a predictor of seizure freedom. Brain Commun. 2022;4(3):fcac151.
+52. Nakai Y, Jeong JW, Brown EC, et al. Three- and four-dimensional mapping of speech and language in patients with epilepsy. Brain. 2017;140(5):1351–1370.
+53. Desikan RS, Ségonne F, Fischl B, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006;31(3):968–980.
+54. Ghosh SS, Kakunoori S, Augustinack J, et al. Evaluating the validity of volume-based and surface-based brain image registration for developmental cognitive neuroscience studies in children 4 to 11 years of age. Neuroimage. 2010;53(1):85–93.
+55. Deoni SC, Dean DC 3rd, Remer J, Dirks H, O'Muircheartaigh J. Cortical maturation and myelination in healthy toddlers and young children. Neuroimage. 2015;115:147–161.
+56. Croteau-Chonka EC, Dean DC 3rd, Remer J, Dirks H, O'Muircheartaigh J, Deoni SCL. Examining the relationships between cortical maturation and white matter myelination throughout early childhood. Neuroimage. 2016;125:413–421.
+57. Remer J, Croteau-Chonka E, Dean DC 3rd, et al. Quantifying cortical development in typically developing toddlers and young children, 1–6 years of age. Neuroimage. 2017;153:246–261.
+58. Zöllei L, Iglesias JE, Ou Y, Grant PE, Fischl B. Infant FreeSurfer: An automated segmentation and surface extraction pipeline for T1-weighted neuroimaging data of infants 0–2 years. Neuroimage. 2020;218:116946.
+59. Stolk A, Griffin S, van der Meij R, et al. Integrated analysis of anatomical and electrophysiological human intracranial data. Nat Protoc. 2018;13(7):1699–1723.
+60. Pieters TA, Conner CR, Tandon N. Recursive grid partitioning on a cortical surface model: an optimized technique for the localization of implanted subdural electrodes. J Neurosurg. 2013;118(5):1086–1097.
+61. Sonoda M, Rothermel R, Carlson A, et al. Naming-related spectral responses predict neuropsychological outcome after epilepsy surgery. Brain. 2022;145(2):517–530.
+62. Navarrete M, Alvarado-Rojas C, Le Van Quyen M, Valderrama M. RIPPLELAB: A Comprehensive Application for the Detection, Analysis and Classification of High Frequency Oscillations in Electroencephalographic Signals. PLoS One. 2016;11(6):e0158276.
+63. Kwan P, Brodie MJ. Neuropsychological effects of epilepsy and antiepileptic drugs. Lancet. 2001;357(9251):216–222.
+64. Yeh FC, Panesar S, Fernandes D, et al. Population-averaged atlas of the macroscale human structural connectome and its network topology. Neuroimage. 2018;178:57–68.
+65. Mitsuhashi T, Sonoda M, Sakakura K, et al. Dynamic tractography-based localization of spike sources and animation of spike propagations. Epilepsia. 2021;62(10):2372–2384.
+66. Mitsuhashi T, Sonoda M, Firestone E, et al. Temporally and functionally distinct large-scale brain network dynamics supporting task switching. Neuroimage. 2022;254:119126.
+67. Zweiphenning WJEM, von Ellenrieder N, Dubeau F, et al. Correcting for physiological ripples improves epileptic focus identification and outcome prediction. Epilepsia. 2022;63(2):483–496.
+68. Baynes K, Eliassen JC, Lutsep HL, Gazzaniga MS. Modular organization of cognitive systems masked by interhemispheric integration. Science. 1998;280(5365):902–905.
+69. Schulte T, Müller-Oehring EM. Contribution of callosal connections to the interhemispheric integration of visuomotor and cognitive processes. Neuropsychol Rev. 2010;20(2):174–190. doi: 10.1007/s11065-010-9130-1
+70. Yeatman JD, Weiner KS, Pestilli F, Rokem A, Mezer A, Wandell BA. The vertical occipital fasciculus: a century of controversy resolved by in vivo measurements. Proc Natl Acad Sci U S A. 2014;111(48):E5214-E5223.
+71. Herbet G, Zemmoura I, Duffau H. Functional Anatomy of the Inferior Longitudinal Fasciculus: From Historical Reports to Current Hypotheses. Front Neuroanat. 2018;12:77.
+72. Malik A, Eldaly ABM, Agadagba SK, et al. Neuromodulation in the developing visual cortex after long-term monocular deprivation [published online ahead of print, 2022 Nov 17]. Cereb Cortex. 2022;bhac448.
+73. Roehri N, Pizzo F, Lagarde S, et al. High-frequency oscillations are not better biomarkers of epileptogenic tissues than spikes. Ann Neurol. 2018;83(1):84–97.
+74. Ball T, Kern M, Mutschler I, Aertsen A, Schulze-Bonhage A. Signal quality of simultaneously recorded invasive and non-invasive EEG. Neuroimage. 2009;46(3):708–716.
+
+# Supplementary Files
+
+- [HFOsakakuradevelopmentsupplementarydocumentea7ks5.docx](https://assets-eu.researchsquare.com/files/rs-2799931/v1/dd697e8fe978096340b9f1f3.docx)
+- [nreditorialpolicychecklistks1.pdf](https://assets-eu.researchsquare.com/files/rs-2799931/v1/ac68728689e21d5e133063c8.pdf)
+- [nrreportingsummaryks2ea1.pdf](https://assets-eu.researchsquare.com/files/rs-2799931/v1/4e998922f8723d862b1b3d6f.pdf)
+- [NCOMMS2315354rs.pdf](https://assets-eu.researchsquare.com/files/rs-2799931/v1/4bb0c80e0d830136fa423d75.pdf)  
+  Reporting Summary
+- [VideoS1MI800.5ks2.mp4](https://assets-eu.researchsquare.com/files/rs-2799931/v1/61cdc36570b7c71308e9f23d.mp4)  
+  Video S1
+- [VideoS2MI803ks1.mp4](https://assets-eu.researchsquare.com/files/rs-2799931/v1/f8c6658e05433ccc8600c928.mp4)  
+  Video S2
+- [VideoS3HIL80ks2.mp4](https://assets-eu.researchsquare.com/files/rs-2799931/v1/8de032ac2b6b2c536b9e0546.mp4)  
+  Video S3
+- [VideoS4MI800.5tractks1.mp4](https://assets-eu.researchsquare.com/files/rs-2799931/v1/6519361a5ca10e1a370b3c9e.mp4)  
+  Video S4
+- [VideoS5HIL80tractks3.mp4](https://assets-eu.researchsquare.com/files/rs-2799931/v1/ffb25194e2a1a637ea20a836.mp4)  
+  Video S5
+- [nrreportingsummaryea2.pdf](https://assets-eu.researchsquare.com/files/rs-2799931/v1/9c6222be14f542eeb082f121.pdf)  
+  Reporting Summary

06bffe0e66fe9b7bc88c34f124ac1a5025fa53d58d4be56435974a852e35e681/preprint/images_list.json

@@ -0,0 +1,58 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpg",
+        "caption": "Group differences for PD-VH vs. PD-noVH. Shown are regions whereby PD-noVH had decreased (a) cortical thickness and (b) surface area (SA). Regions are colour coded by p value (S3). (a) Widespread decreased thickness was found in PD-VH; the regions with the greatest effect size were medial occipital parietal and frontal regions. (b) SA was reduced in PD-VH in the left and right medial occipital and in the left insular gyrus, and in the medial central and superior frontal regions. Results are corrected for multiple comparisons.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpg",
+        "caption": "Receptors density profiles: methods and main results.  a) Procedure and rationale of the regression models. Both the independent receptor density maps and our participants\u2019 MRI scans were parcellated with the Destrieux atlas. Cortical thickness and SA values were extracted for each region of the atlas for the participants\u2019 scans, and binding potential was extracted for each region of the atlas for the receptor density maps. Each receptor\u2019s binding potential was used in separate models as a predictor of difference of the means of thickness/SA between PD-VH and PD-noVH.  b) Results of regression models. Shown are the results of the models with the regions which were different between groups as dependent variable. Results are reported for 5-HT2A, 5-HT1A and D2/D3 receptors binding potential and thickness on the left and binding potential and SA on the right.  ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpg",
+        "caption": "Graphical representation of the regions contribution to each of the Dimensions resulting from the PCA. a) Regions contributing to dimension1 and 2, cortical thickness. Dimension 1 (pink): left superior frontal gyrus, the left middle frontal gyrus and the bilateral precentral gyrus. Dimension 2 (green): bilateral cuneus and occipital superior gyrus. B) Regions contributing to dimension 1 and 2 for surface area. Dimension 1 (pink): left and right calcarine sulci, right occipitotemporal lingual gyrus, right occipital pole, Dimension 2 (green) left central insular area, anterior and superior portions of the circular sulcus of the insula. ",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpg",
+        "caption": "Regions with the most significant difference in inter-regional correlations of cortical thickness between groups: the inter-regional correlations for these regions were significantly greater for VH patients. Shown in the circular plot, only the inter-regional correlations with a difference greater than 0.3 in the r2 (for details and z scores on all differences, see S8). \nLegend: IPL = inferior parietal lobule; LOG = lat. Occipital gyrus; MTG = middle temporal gyrus; PHG = parahippocampal gyrus; paraC = paracentral gyrus; IFG opercularis. = inferior frontal gyrus; SFG = superior frontal gyrus; SMAR = supramarginal gyrus; FP = frontoparietal thickness; TT =temporal transverse; FUS = fusiform gyrus; postC = postcentral gyrus. The two vertical lines separate L and R hemisphere regions (left on left).\n",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.jpg",
+        "caption": "Hubs and communities: cortical thickness. a) Hubs identified based on efficiency, betweenness centrality and degree. Regions in bold are common hubs for both VH and noVH. b) Communities identified for each group. Legend: red = 1st community, green =2nd, blue= 3rd, pink = 4th, yellow = 5th. Only the first five communities are represented as they are the most informative ones. In bold the regions identified for that same community also in the surface area analysis. The regions underlined are the same regions presented in the figure with the regions with the greatest difference in inter-regional covariance.\nLegend: BSTS = banks superior temporal; IPL = inferior parietal lobule; SPL = superior parietal lobule; LOG = lat. Occipital gyrus; MTG = middle temporal gyrus; STG = superior temporal gyrus; cMFG = caudal middle frontal gyrus; PHG = parahippocampal gyrus; paraC = paracentral gyrus; preC = precentral gyrus; postC = postcentral gyrus; IFG = inferior frontal gyrus; SFG = superior frontal gyrus; SMAR = supramarginal gyrus; FP = frontoparietal thickness; TT =temporal transverse; FUS = fusiform gyrus; CUN= cuneus.\n",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.jpg",
+        "caption": "Regions with the most significant difference in inter-regional correlations of surface area between the groups: these correlations were significantly greater for PD-VH. Only the inter-regional correlations with a difference greater than 0.3 in the r2 are shown (for details see S8).\nLegend: IPL = inferior parietal lobule; SPL = superior parietal lobule; LOG = lat. Occipital gyrus; MTG = middle temporal gyrus; STG = superior temporal gyrus; cMFG = caudal middle frontal gyrus; PHG = parahippocampal gyrus; paraC = paracentral gyrus; preC = precentral gyrus; IFG = inferior frontal gyrus; SFG = superior frontal gyrus; SMAR = supramarginal gyrus; FP = frontoparietal thickness; TT = temporal transverse; FUS = fusiform gyrus; CUN= cuneus; PCUN = precuneus; MOF = middle orbitofrontal gyrus. The two vertical lines separate L and R hemisphere regions (left on left)\n",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_7.jpg",
+        "caption": "Hubs and communities: surface area. a) SA hubs identified based on efficiency, betweenness centrality and degree. Regions in bold are common hubs for both VH and noVH patients. b) Communities identified for each group. Legend: red = 1st community, green = 2nd, blue= 3rd, pink = 4th, yellow = 5th. In bold the regions identified for that same community also in the SA analysis. The regions underlined are the same regions presented in the figure with the regions with the greatest difference in inter-regional covariance.\nLegend: BSTS = banks superior temporal; IPL = inferior parietal lobule; SPL = superior parietal lobule; LOG = lat. Occipital gyrus; MTG = middle temporal gyrus; STG = superior temporal gyrus; ITG = inferior temporal gyrus; cMFG = caudal middle frontal gyrus; PHG = parahippocampal gyrus; paraC = paracentral gyrus; preC = precentral gyrus; postC = postcentral gyrus; IFG = inferior frontal gyrus; SFG = superior frontal gyrus; SMAR = supramarginal gyrus; FP = frontoparietal thickness; TT = temporal transverse; FUS = fusiform gyrus; CUN= cuneus\t\n",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

06bffe0e66fe9b7bc88c34f124ac1a5025fa53d58d4be56435974a852e35e681/preprint/preprint.md

@@ -0,0 +1,339 @@
+# Abstract
+
+Parkinson’s psychosis (PDP) describes a spectrum of symptoms that may arise in Parkinson’s disease (PD) including visual hallucinations (VH). Imaging studies investigating the neural correlates of PDP have been inconsistent in their findings, due to differences in study design and limitations of scale. Here we use empirical Bayes harmonisation to pool together structural imaging data from multiple research groups into a large-scale mega-analysis, allowing us to apply new methodological approaches to identify cortical regions and networks involved in VH and their relation to receptor binding. Differences of cortical thickness and surface area show a wider cortical involvement underlying VH than previously recognised, including primary visual cortex and its surrounds, and the hippocampus, independent of its role in cognitive decline. Structural covariance analyses point to a strong involvement of the attentional control networks in PD-VH, while associations with receptor density maps suggest neurotransmitter loss may drive the cortical changes.
+
+[Cognitive Neuroscience](/browse?subjectArea=Cognitive%20Neuroscience) | [Neurology](/browse?subjectArea=Neurology) | Parkinson’s disease | Parkinson’s psychosis | MRI | structural imaging | structural covariance | neurotransmitters | visual hallucinations
+
+# 1. Introduction
+
+Patients with Parkinson’s disease (PD), aside the typical motor symptoms<sup>1</sup>, commonly experience a variety of non-motor symptoms, including psychiatric ones<sup>2</sup>. Among these, visual hallucinations (VH) and related visual phenomena form a spectrum of symptoms referred to as <em>Parkinson’s psychosis</em><sup>3</sup> (PDP). There is a continuum of experiences with patients initially experiencing minor hallucinations (perception of presence or passage) and illusions that progress to formed hallucinations (initially with insight preserved), then hallucinations in other modalities and delusions<sup>4</sup>. Such symptoms may affect up to 70% of PD patients in more advanced stages of the illness<sup>5</sup> in the context of dopamine therapy, but do not show a clear relationship between medication introduction or dose suggesting the symptoms of PDP are not simply medication side-effects<sup>4</sup>. VH predict a range of poor outcomes including more rapid cognitive decline and development of dementia<sup>6</sup>–<sup>8</sup> and nursing home placement<sup>9</sup>. It is difficult to determine how VH might be related to these poor outcomes without a clear understanding of the brain systems involved in VH<sup>4</sup>.
+
+Imaging studies of VH in PD to date have been based on relatively small samples and have used differing designs that variously control for the degree of cognitive decline, stage of PD and dopamine medication. This makes it difficult to disentangle brain changes related specifically to VH mechanisms as distinct from those related to cognitive decline, PD stage or medication effects. As a result, a heterogeneous array of structural differences has been reported. Depending on whether or not cognition is controlled for, some studies have found volume reductions in specific regions that have not been replicated in other studies including: hippocampus<sup>10</sup>, cerebellum<sup>10</sup>–<sup>11</sup>, lateral, superior and medial frontal cortex<sup>11</sup>–<sup>13</sup>, thalamus<sup>14</sup> and different subregions of visual association cortex, broadly defined to include the lateral occipital cortex, ventral occipito-temporal cortex (ventral stream) and visual parietal lobe (dorsal stream)<sup>11</sup>, <sup>15</sup>, <sup>16</sup>.
+
+A meta-analysis utilising the previously reported regional differences demonstrated very little consistency across studies, suggesting this may be due to heterogeneity in structural brain correlates of VH, varying sensitivity in multiple small studies, or the involvement at different locations of a unifying brain network whose dysfunction results in VH<sup>17</sup>. While meta-analytical techniques can be useful to collate findings from different studies and help understand the consistency of brain regions involved, there are limitations in their ability to include variables such as cognition, medication dose, PD stage and duration as covariates, given that these are usually incorporated into the analyses at the study level and each study contributes a different set of regions to the meta-analysis. In contrast, mega-analyses bring together subject-level data across sites in one analysis, which presents a number of advantages. These include methodological rigour, with shared quality control and pre-processing pipelines, including software version control and the ability to include unpublished data or published data that was not used in the primary analysis (e.g. structural data collected for functional imaging studies). The same experimental design model and covariates can be applied uniformly across the data set helping address design variations in previous studies. Another advantage of the increased sample size is the additional power to explore morphometric features such as cortical thickness and cortical surface area along with undertaking complex analyses, such as structural covariance. Cortical thickness and surface area are considered as orthogonal components, which are genetically unrelated<sup>18</sup> and can be considered separate morphometric components in ageing and disease<sup>1</sup>, <sup>19</sup>.The main correlate of cortical volume is cortical surface area, but volume loss is best captured by cortical thickness<sup>19</sup>, <sup>20</sup>. Separate measurement and analysis of these two components thus offer a better understanding of the underlying cortical changes associated with VH in PD than volume measures alone. Finally, mega-analyses create a valuable resource that can evolve and be made available to the wider neuroimaging community, especially important in PDP given that such patients are difficult to recruit and scan.
+
+Several neurotransmitter systems have been associated with VH in PD. Initially, VHs were proposed to be a side effect of <em>dopaminergic</em> medication<sup>21</sup>, but later evidence has led to a revision of this view. Current consensus is that dopaminergic medication interacts with disease-related susceptibility factors in PD to cause VH, rather than as a simple side effect<sup>3</sup>. Cholinergic pathways have also been implicated in VH<sup>22</sup>, <sup>23</sup>, with neurodegeneration in brainstem and forebrain cholinergic nuclei<sup>22</sup> and electrophysiological measures of cholinergic function reduced in patients with VH<sup>24</sup>. Recently, a role for serotonergic dysfunction in VH has been suggested<sup>25</sup>, linked to alterations of 5-HT<sub>A</sub> receptor density<sup>26</sup>, <sup>27</sup> (for a review<sup>28</sup>).
+
+In summary, our mega-analysis of PD with VH compared to PD without VH enables analyses that are not available to smaller scale studies to help explore the mechanisms of VH. Specifically, we are able to determine the regional cortical thickness and surface area changes associated with VH and relate these morphometric features to measures of symptom severity in a subgroup where finer-grain clinical detail is available. We perform a principal component analysis to identify smaller-scale morphometric differences within a high dimensional set of regions. In addition, we perform an exploratory structural network analysis to highlight associations between regions and clusters of connections linked to VH. Structural covariance allows us to assay covariation of differences in grey matter morphology between different brain structures, providing information on which regions similarly change in thickness or surface area. In order to understand the neurochemical associations of these changes, we also test the hypothesis that structural differences are related to the spatial variation in subtypes of receptors for which high resolution PET atlases are available (dopamine and serotonin).
+
+# 2. Results
+
+## 2.1 Patient characteristics
+
+The final dataset consisted of 493 participants (193 F), of which 135 were PD-VH. Each individual study had matched their participants for age, gender, disease onset, MMSE, UPDRS-III and levodopa equivalent daily dose (LED), with few exceptions (Table 1). We also included the unpublished data in separate ANOVAs to check group similarity (Table 1), in meta-analyses (S2) and in an ANOVA including the whole mega-analysis sample. While the ANOVAs and the meta-analysis demonstrated we have good matching on the criteria, the mega-analysis ANOVA shows that there is a difference of 2.19 years in age (F(1,491) = 6.56, p = .01) (PD-VH = 67.85, SD = 7.74; 62 F, and PD-noVH = 65.66, SD = 8.71; 131 F) and there is a greater proportion of females in the PD-VH group (χ² = 3.585, p = .06).
+
+Morphometrics were harmonised (S1) and we did not find significant differences in total intracranial volume (TIV) (F(1,493) = .043, p = .84) and total brain volume (F(1,493) = 2.488, p = .115), but in total gray matter volume (F(1,493) = 5.41, p = .02) (see S2). For the subsample of 146 patients we had additional information and a subsample analysis was performed (see 2.3).
+
+### Table 1
+
+| Study | N Patients | Age | Onset | MMSE | UPDRS-III | LEDD |
+|---|---|---|---|---|---|---|
+| Shin et al., 2012 (Yonsey University) | 46 PD-VH (23F)<br>64 PD-noVH<br>(38 F) | 71.3 ± 5.9<br>70.7 ± 5.7<br>p = ns | 3.3 ± 3.0<br>2.8 ± 3.0<br>p = ns | 25.3 ± 3.0<br>25.7 ± 2.9<br>p = ns | 24.1 ± 10.4<br>21.6 ± 11.0<br>p = ns | 482.4 ± 252.6<br>501.4 ± 167.5<br>p = ns |
+| Shine et al., 2012 + unpublished data (University of Sidney ) | 26 PD-VH<br>(12 F)<br>48 PD-noVH<br>(12 F) | 66.6 ± 7.2<br>66.4 ± 8.6<br>p = .9 | 6.0 ± 3.9<br>5.4 ± 3.5<br>p = .5 | 28.7 ± 1.7<br>29.6 ± 1.7<br><span class="BoldItalic">p = .04</span> | 32.0 ± 13.4<br>27.8 ± 16.2<br>p = .2 | 664.3 ± 495.2<br>706.8 ± 502.7<br>p = .7 |
+| Firbank et al., 2018 (only non-dementia data retained) (University of Newcastle) | 11 PD-VH (2F),<br>11 PD-noVH (2F) | 75.0 ± 3<br>71.7 ± 5.3<br>p = .2 | 10.2 ± 8.2<br>10.1 ± 7.6<br>p = .9 | 25.9 ± 1.6<br>27.2 ± 2.4<br>p = .2 | 51.7 ± 22.2<br>30.50 ± 14.73<br>p = .05 | 469.9 ± 311.3<br>693.4 ± 474.1<br>p = .2 |
+| Yao et al., 2014 (University of Hong Kong) | 12 PD-VH,<br>(9F)<br>12 PD-noVH<br>(8F) | 67.6 ± 7.4<br>73.4 ± 7.4<br>p = .2 | 10.0 ± 3.5<br>8.4 ± 5.1<br>p = .4 | 27.6 ± 2.4<br>28.5 ± 1.7<br>p = .09 | 20.9 ± 10.6<br>18.0 ± 12.9<br>p = .5 | 978.7 ± 361.3<br>704.9 ± 519.4<br>p = .2 |
+| Lefebvre et al., 2018 Unpublished structural data (University of Lille) | 18 PD-VH<br>(7F)<br>16 PD-noVH<br>(4F) | 62.9 ± 6.0<br>63.8 ± 2.2<br>p = .2 | 8.2 ± 5.3<br>7.9 ± 4.2<br>p = .2 | 28.0 ± 1.24<br>28.8 ± 1.20<br>p = .2 | 25.0 ± 8.4<br>21.8 ± 7.9<br>p = .2 | 859.7 ± 411.1<br>804.3 ± 297.4<br>p = .7 |
+| ffytche and Lawn, 2021 (King’s College London) | 7 PD-VH<br>(4 F)<br>9 PD-noVH<br>(3 F) | 66.1 ± 6.5<br>68.7 ± 7.2<br>p = .3 | 8.3 ± 5.2<br>5.8 ± 2.5<br>p = .2 | 29.7 ± 0.5<br>26.8 ± 4.1<br>p=. 06 | 25.6 ± 6.6<br>40 ± 13.4<br><span class="BoldItalic">p=. 01</span> | 759.4 ± 529.2<br>746.2 ± 487.1<br>p = .9 |
+| Oxford Discovery Cohort, unpublished * (Baig et al., 2015; Griffanti et al. 2020) | 7 PD-VH<br>(5F)<br>103 PD-noVH<br>(36F) | 63.86 ± 10.4<br>60.35 ± 10.4<br>p = .4 | 2.0 ± 1.0<br>2.4 ± 1.6<br>p = .5 | 28.7 ± 1.4<br>28.6 ± 1.3<br>p = .2 | 23.0 ± 12.7<br>23.8 ± 10.3<br>p = .9 | 978.7 ± 361.3<br>323.8 ± 244.3<br>p = .2 |
+| T1 data submitted, demographics in Zarkali et al., 2020 (University College London) | 19 PD-VH<br>(13F)<br>86 PD-noVH<br>(35F) | 64.6 ± 8.2<br>64.5 ± 7.9<br>p = .9 | 4.2 ± 2.4<br>4.1 ± 2.5<br>p = .8 | 28.9 ± 1.6<br>28.9 ± 1.1<br>p = .9 | 24.1 ± 13.1<br>21.7 ± 11.0<br>p = .4 | 415.6 ± 162.5<br>461.5 ± 269.2<br>p = .5 |
+
+* non-motor symptoms (Baig et al., 2015), T1 data (Griffanti et al., 2020) separately published, but not in a publication studying them together.
+
+## 2.2 Hallucinators (PD-VH) vs. non-hallucinators (PD-noVH) multivariate analysis of variance.
+
+**Cortical Thickness.** Lower thickness in PD-VH was present in a widespread set of regions (see Fig. 1 and S3). No regions showed greater cortical thickness in PD-VH. A main effect of age (F(1,492) = 3.38, η² = .60, p < .001), gender (F(1,492) = 1.51, η² = .40, p < .001) and TIV (F(1,492) = 2.38, η² = .51, p < .001) was observed.
+
+**Surface area.** We found reduced area in PD-VH mainly in frontal and occipital regions (see Fig. 1) (for all tables and details see S3). A significant main effect of age (F(1,492) = 2.08, η² = .47, p < .001), gender (F(1,492) = 1.50, η² = .39, p < .001) and TIV (F(1,492) = 6.32, η² = .73, p < .001) was observed.
+
+**Subcortical volumes.** We found a lower volume for PD-VH in the bilateral amygdala (see S3). A significant main effect of age (F(1,492) = 11.87, η² = .62, p < .001), gender (F(1,492) = 2.64, η² = .26, p < .001) and TIV (F(1,492) = 255.89, η² = .97, p < .001) was observed.
+
+## 2.3 Subgroup analysis.
+
+We created a subsample for which we have Neuropsychiatric Inventory (NPI) hallucinations subscale scores (frequency * distress), focussing on VH. The sample consists of 146 patients (67 PD-VH, 79 PD-noVH), matched for age, gender, TIV, medication, cognition, onset and PD severity (UPDRS-III) (but see S4 for detailed comparisons). Results from the custom multivariate ANCOVAs were overall consistent with those found for the main sample (see S4).
+
+When correlating the NPI score with morphometrics, inverse correlations were significant for right hemisphere cortical thickness in the intraparietal sulcus (r = − .24, p = .05), the superior temporal sulcus (r = − .26, p = .03), the Jensen sulcus (between the anterior and posterior rami of the IPS) (r = − .27, p = .03) and the cingulum (marginalis) (r = − .25, p = .05), and a positive correlation was found with the right frontomarginal gyrus (r = .26, p = .04). Results did not change when carrying out partial correlations between NPI score and morphometrics, and with levodopa equivalent dose (LED) as covariate (see S4).
+
+## 2.4 Receptors density maps regression models.
+
+Receptor densities maps of D2/D3, 5-HT₂A and 5-HT₁A were parcellated using the Destrieux atlas to ensure that density and morphometric data were aligned. We explored the relationship between the differences in cortical thickness and surface area between PD-VH and PD-noVH (Fig. 2) with separate linear models for each density map. We carried out i) a model including the morphometric difference values only of regions where we found a significant difference, and ii) a model including the morphometric difference values in all regions. The maps used were independent atlases built on healthy subjects’ PET data (see Methods).
+
+**Thickness.** The model with 5-HT₂A binding potential as predictor and the mean difference as dependent variable was significant for the subset of regions where the groups differed (β = -.252, t = -2.2, p = .03), whereas no relationship was observed when considering all the atlas regions (β = -.02, t = − .31, p = .75). A similar result was observed for 5-HT₁A (significant regions: β = − .26, t = -2.25, p = 0.03; all regions: β = .008, t = .092, p = .504) and for D2/D3 receptors (significant regions: β = -.35, t = -3.14, p = 0.002; all regions: β = .09, t = .95, p = .34) (see Fig. 2 for methods and results). In addition, we compared the slopes of the models, finding no difference between 5-HT₂A, 5-HT₁A and D2/D3 for significant regions) (S6).
+
+**Surface area.** The models with 5-HT₂A binding potential per region as predictor and the mean difference per region as dependent variable was significant, but only for regions which differed between groups (β = -.22, t = 2.1, p = .038). No relationship was observed with all regions included (β = .15, t = 1.75, p = .08). The models with 5-HT₁A binding potential as predictor was significant for differing regions (β = .27, slope = 0.22, t = 2.2, p = .01) and with lower significance for all regions (β = .181, t = 2.07, p = .04). When using D2/D3 as a predictor, the model was significant for significantly differing regions (β = .318, t = 2.5, p = .01) and the model for all regions showed greater significance (β = .277, t = 3.24, p = .001). In all cases, the greater the mean difference, the lower the binding potential. However, when estimating the confidence intervals of the models, the model for D2/D3 was no longer significant (S6). 5-HT₂A, 5-HT₁A and D2D3 slopes for significant regions did not differ (S6). See Fig. 2, for methods and results and S6 for further details.
+
+The same models were carried out also for subcortical volumes not yielding significant results for all receptors (see S6).
+
+## 2.5 Principal components analysis (PCA).
+
+We performed PCA to reduce the dimensionality of the dataset while preserving variability, to identify underlying clusters to clarify the results from the group-level analyses.
+
+**Cortical thickness.** The PCA returned two dimensions with eigenvalues > 1 explaining 67.58% of total variance. The regions best representing Dimension 1 (eigenvalue = 4.47, 49.69% of variance) as assessed with the cosine squared index were the left superior frontal gyrus, the left middle frontal gyrus and the bilateral precentral gyrus. The regions best representing Dimension 2 (eigenvalue = 1.61, 17.89% of variance) were the cuneus, and the occipital superior gyrus, bilaterally (see Fig. 3 a; for scree plots see S7).
+
+**Surface area.** The PCA returned two components with eigenvalues > 1. Dimension 1 (eigenvalue = 4.48, 56.01% of variance) and Dimension 2 (eigenvalue = 1.38, 17.21% % of variance) for a total cumulative variance of 73.23% of explained variance. For Dimension 1, the contributing regions were visual regions: left and right calcarine sulci, the right occipitotemporal lingual gyrus and the right occipital pole. For Dimension 2, the contributing regions were the left central insular area, the anterior and superior portions of the circular sulcus of the insula (Fig. 3. b, S7).
+
+For cortical thickness only, we found a significant inverse correlation of Dimension 1 individual contributions and NPI score (r = .-138, p = .049). In addition, the thickness for the Dimension 1 regions (left SFG, MFG, precentral) negatively correlated with NPI score (r = -.15, p = .046, one-tailed Pearson correlation), with the individuals having higher pathological score having also the lower thickness in these regions.
+
+## 2.6 Structural covariance analysis.
+
+To explore and characterise the gray matter morphology covariation and network-level organisation of PD-VH and PD-noVH patients for cortical thickness and surface area we performed structural covariance analyses.
+
+After specifying a general linear model for each region, the structural covariance matrices (68x68) for each group was defined by estimating the inter-regional correlation between model residuals of thickness and area (in separate models).
+
+**Cortical thickness.** Significant difference of the two covariance matrices (PD-VH, PD-noVH) was first tested (χ² = 3010.82, df = 2278, z of differences = 3.83). The cell-by-cell comparisons of residuals’ inter-regional correlation coefficients highlighted differences in interregional covariance, in particular in the left inferior temporal gyrus, supramarginal gyrus and inferior parietal lobe (IPL), superior frontal (SFG) and inferior frontal gyrus (IFG) pars opercularis and the fusiform and lateral occipital gyri on the right (Fig. 4). Overall, inter-regional correlations were greater for the PD-VH group (but see also S8).
+
+Hubs, that is nodes (here regions) that are thought to strongly contribute to the global network function, were identified in frontal, parietal and occipital regions for the PD-noVH group, and in frontal, temporal and parietal regions for the PD-VH group (Fig. 5 a). Permutation tests for vertex-level measures returned differences in betweenness centrality, which was greater in PD-VH in the left and right lingual gyrus, in the left lateral occipital gyrus and the right SPL (p FDR < .05). Communities are sets of brain regions characterised by denser and stronger relations among themselves, if compared with regions of other communities. Structural covariance-based communities have been found to replicate neighbourhoods observed with seed-based approaches in fMRI and DTI (see Methods for details). The first community in the PD-VH group comprised mainly occipitotemporal regions, with the second involving parietal and some frontal regions. In the PD-noVH group, the first community consisted of mostly frontoparietal regions whereas the second comprised occipitoparietal regions (Fig. 5 b). In addition, the PD-noVH group showed higher modularity, as assessed with bootstrapping (mean = 0.29 SD = 0.02, CI 0.25, 0.36 at density 13%) (for communities by lobe, see S8).
+
+**Surface area.** As for thickness, the two covariance matrices were different (χ² = 5347.2, df = 2278, z of differences = 6.8). In addition, among the others, significant differences in interregional covariance were observed bilaterally in the rostral MFG, STS, fusiform gyrus, and IPL; in the left caudal MFG, lateral occipital gyrus, SPL, and insula and in the right anterior and posterior cingulate, and IFG pars opercularis, with a pattern very similar to the one observed for thickness (see Fig. 6 and S8).
+
+Hubs were identified mainly in occipitotemporal and frontal regions for the PD-noVH group and in frontal, temporal and occipital regions for the PD-VH group (Fig. 7). In accordance with this result, vertex-level permutation tests returned differences in betweenness centrality the left fusiform gyrus; in addition, differences were observed for the middle orbitofrontal gyrus, IFG orbitalis and triangularis, and in the bilateral anterior cingulate (p < .003, p FDR < .09), whereby centrality was greater for PD-noVH in these regions, but greater for PD-VH in the left caudal MFG and in the right SFG. The first community in the PD-VH group is characterised by occipitotemporal and frontal and the second community by occipito-parietal and parietal regions only (Fig. 7 b; representation by lobe is in S8). In addition, PD-noVH showed greater modularity, as assessed with bootstrapping (0.29, CI 0.21, 0.36 density 13%).
+
+Finally, we found a significant positive correlation between difference of the surface area means in the NPI subsample and with the difference in local efficiency (r = .24, p = 0.02), whereby the greater the difference in the surface area, the greater the difference in the local efficiency. The regions with both the greatest area differences and efficiency differences were the bilateral lingual gyrus, lateral occipital gyrus, right cuneus and right insula.
+
+# 3. Discussion
+
+We have presented a mega-analysis of patients with Parkinson’s disease with and without visual hallucinations, demonstrating widespread alterations in brain structure, with differential effects for cortical thickness and surface area and examined their relationship to receptor distributions and network-level effects. Below we discuss the implications of the findings and their relationship to current theories of VH.
+
+## Cortical thickness and surface area
+
+Cortical thickness and surface area (SA) are considered two separate components in ageing and disease, reflecting different aspects of the neurodegenerative process. Cortical thickness loss relates to cortical layering and, by inference, cytoarchitecture, while surface area relates to gyral anatomy and, by inference, underlying white matter. Widespread reductions in cortical thickness in hallucinators were identified in the occipital, parietal, temporal, frontal and limbic lobes. The regions of reduced thickness encompassed all cortical regions identified in previous structural imaging studies (for a review), suggesting previous variability may relate to stochastic effects introduced by relatively smaller samples and design differences. With the larger sample of the mega-analysis, the extent of cortical regions involved appears wider than previously suspected. However, not all regions are equally affected and, notably, there appears to be a posterior asymmetry with relative sparing of the left ventral visual stream (ventral occipito-temporal cortex) compared to the homologous region in the right hemisphere. This region plays a key role in all models of VH in PD but a greater involvement of the right hemisphere has not been noted previously. The PCA analysis helped define key sub-regions within the extensive areas of cortical thinning that contributed most to the group difference, identifying a frontal and an occipital dimension. Of these, the cuneus bilaterally and left dorso-medial aspect of the superior frontal gyrus emerged as the dominant components. These regions have been reported in previous studies but do not play a prominent role in accounts of VH in PD. The cuneus is one of the earliest regions to show cortical atrophy in PDP, while cortical thinning in the dorso-medial superior frontal gyrus has been reported in patients, months to years prior to the development of VH. It may be that the prominence of these regions in the mega-analysis relates to the longer duration of these changes compared to other brain regions resulting in a greater consistency of thickness reduction between patients.
+
+For SA, the differences between groups were more circumscribed with bilateral medial occipital SA reduction for patients with VH in a region corresponding to the primary visual cortex and its surrounds (striate and extra-striate cortex) and the left insula. This is the first-time such extensive structural changes have been identified in the primary visual cortex and its surrounds in PD patients with VH and helps account for wide-ranging low-level visual deficits found (for a review). These regions also have reduced cortical thickness but their prominence in the SA analysis may imply additional gyral atrophy, sulcal widening and a reduction of underlying white matter.
+
+The mega-analysis also allowed us to move beyond a binary comparison of VH versus noVH to examine brain regions linked to VH severity (NPI hallucination subscale) and taking into account any variability associated with age, gender, TIV, medication, cognition, disease onset and PD severity. Regions with reduced thickness for higher severity scores were found in posterior parietal, posterior cingulate and superior temporal cortex. Previous studies have associated these regions with mental rotation and visuospatial transformation for the IPS, and imagery for the STS. These processes are altered in patients with PD and VH, thus one can infer an involvement of these processes and these regions in VH severity. In addition, the IPS is also part of the dorsal attentional network, previously implicated in VH in PD (see discussion below). These regions were also identified as hubs in the structural covariance analysis, discussed further below. We cannot disentangle whether these correlations are driven primarily by the frequency or distress of VH as these measures were only available for part of the subsample. However, this is the first time a link between cortical structural changes and phenomenological aspects of VH severity has been identified.
+
+## Subcortical regions, hippocampus and cerebellum
+
+In addition to the detailed analysis of the cerebral cortex we were able to examine the volumes of subcortical structures as well. Bilateral volume reduction was found in the amygdalae. Lewy bodies have been found in the basolateral nucleus of the amygdala associated with VH in PD patients at a similar level of cognitive impairment to those studied here that may account for this finding. Unlike the amygdala, there are only sparse Lewy bodies in the hippocampus at this disease stage and volume changes in this structure are more difficult to interpret. Since the prevalence of VH increases as PD progresses, it is difficult to disentangle brain changes related primarily to cognitive decline from those related primarily to VH or that may contribute equally to both. Reductions of hippocampal volume have been found in some, but not all, studies of VH in PD, depending on whether patients are matched for cognitive decline. Here we found smaller left (and a trend for right) hippocampus in the NPI sample where we were able to covary for age, gender, TIV, onset, LED, PD severity and cognition. The volume reductions in the NPI analysis cannot be explained by differences in cognition or PD progression between groups, confirming a role for the hippocampus in the mechanism of VH that is independent of cognition, thus highlighting the need to carefully design studies and control for cognitive and disease factors when examining hippocampal contributions to VH. The thalamus has been suggested as a key hub linking several cortical networks associated with VH in PD. We did not find altered thalamic volumes, suggesting that any functional changes in this structure are not associated with volume loss. Finally, reduced volume in cerebellar lobules VIII, IX/VII and Crus 1 is associated with VH in PD. Freesurfer does not segment specific cerebellar subfields but volume changes were found in cerebellar white matter that may relate to these cerebellar cortical changes.
+
+## Neurotransmitter receptor density and structural imaging changes
+
+There is only sparse Lewy body pathology in the cortex of PD patients with VH at the disease stage included in our analysis, raising the question of what causes the extensive cortical changes found in this and previous studies. One possibility is that such cortical changes represent synaptic loss secondary to degeneration in neurotransmitter inputs to the cortex. Previous studies have found changes in cholinergic, serotonergic, dopaminergic and GABAergic systems in PD patients with VH; however, the relationship between regions of cortex with volume loss and the cortical distribution of these neurotransmitter systems has yet to be examined. We were able to investigate this relationship for subtypes of dopamine and serotonin receptors for which high resolution maps are available and found that cortical regions with higher binding had increased cortical volume loss. The association, in particular for 5-HT₂A, was confined to regions linked to VH rather than the cortex as a whole, suggesting the neurotransmitter effects were specific to VH, consistent with the possibility that degeneration in these neurotransmitter systems in PD underlies synaptic loss and cortical thinning. 5-HT₂A and 5-HT₁A binding maps were correlated so the same cortical regions are likely to have contributed to both serotonin findings. While increased binding was associated with increased thickness loss, the opposite association was found for SA, with higher binding exhibiting less SA reduction. This finding was not specific to VH regions for dopamine and 5-HT₁A so may reflect a different process to the thickness alterations found. It is also unclear what causes low binding regions to be associated with increased loss of SA.
+
+## Structural covariance
+
+The examination of inter-regional correlations, with areas sharing reductions in thickness or SA considered part of a functionally connected network, showed that regions of greater inter-regional thickness correlation in PD-VH overlap with those of the dorsal and ventral attention networks (DAN and VAN), with the notable addition of para-hippocampal regions. Most of these regions of higher covariance have reduced thickness in PD-VH, suggesting the covariance is driven by correlated reductions in thickness. Dysregulation of VAN, DAN and default mode networks (DMN) have been implicated in models of VH in PD with reduced activity in the DAN of PD-VH, and the inter-regional covariance findings support this view. In contrast, the inter-regional SA covariance findings highlight key DMN regions in medial frontal and posterior cingulate cortex. These regions were not found to have reduced SA in PD-VH, suggesting a relative preservation of the DMN compared to VAN and DAN. Indeed, results from dynamic fMRI have indicated active coupling between the DMN and the visual network, which correlated with the frequency of misperceptions, as opposed to reduced connectivity between the DMN, VAN and DAN. Hub metrics for thickness in the occipital lobe and parietal lobe were stronger in patients with VH, suggesting cortical thinning has a wider impact on the network in these patients, highlighting the importance of functional alterations in early visual areas in VH. One could argue that VH may not only depend upon on areas presenting neural pathology, but also on areas that may be relatively unaffected but operate in a network where there is pathology elsewhere, thus becoming functionally pathological while structurally intact. Indeed, all the regions where richness of connections was either lower or higher for PD-VH fell outside areas of reduced SA in VH, suggestive of a more functional pathology which needs to be further explored with functional connectivity. Finally, of particular note was the extent, in the hub analysis, of interconnected areas in the ventral, lateral and medial temporal lobe that was larger in the PD-VH group. These regions had reduced thickness in PD-VH implying the local extent of thickness reduction is greater in PD-VH.
+
+## Strengths and limitations
+
+This is the first mega-analysis of VH in PD, pooling data to create the largest sample of PD patients with and without VH tested to date. While this is a major strength of the study, it also introduces complexities that smaller studies do not have to address. One is the variability of clinical data available for each site, limiting the analyses we could perform with the full dataset of 493 participants. This means that some of the key analyses, for example those related to clinical covariates, could only be carried out in a smaller sample of 146 participants, but this is still substantially larger than any previous study. Another complexity is the need to address variance in the data caused by scanning at different sites and scanner types. Previous studies have typically used voxel-based methods to examine structural differences between PD-VH and PD-noVH. We used a different method to allow us to harmonise data between sites and examine cortical thickness and SA separately, but this means our findings are not directly comparable to those of previous studies. The primary focus of the study is on the cerebral cortex so we have not attempted to examine the detailed anatomy of regions such as the basal ganglia, hippocampus, cerebellum and thalamus that may have a role in VH. Finally, we do not have access to high resolution density maps for cholinergic receptor subtypes which limits the range of neurotransmitter analyses we can perform.
+
+## Conclusions
+
+The mega-analysis has allowed us to resolve several uncertainties in the previous literature and describe new features of the VH phenotype in PD. With a sufficiently large sample, more widely distributed cortical involvement emerges than previously suspected with the important novel finding of involvement of the primary visual cortex and its surrounds. Structural covariance modelling has helped dissect out networks linked to attentional control within the widespread cortical regions affected, adding further evidence for the role of these networks in PD-VH. The findings also help resolve ambiguities between structural correlates of general cognitive decline or PD progression and those specifically related to VH. Patients at the same stage of PD and general cognitive impairment who experience VH have lower hippocampal volumes than those who do not. The hippocampus does not currently play a central role in models of VH in PD and our findings suggest this needs to be reconsidered. We can argue that the hippocampus represents part of an extended DMN composed of functional hubs, a dorsal medial subsystem and a medial temporal subsystem, which includes the hippocampus. Thus, structural covariance, graph-level analyses and structural hippocampal imaging point to the involvement of attentional control networks in PD-VH. Finally, the findings shed light on why widespread cortical changes occur at a stage of PD with only sparse cortical neuropathology. The associations between dopaminergic and serotonergic receptor binding and cortical thickness provide the first evidence that the cortical changes may be driven by neurotransmitter reductions, raising the possibility of novel interventions to mitigate these effects at an earlier stage of disease.
+
+# 4. Methods
+
+## 4.1 Studies selection
+
+Based on the literature, we identified N=17 studies of VH in patients with PD that included acquisition of T1-weighted structural MRI scan, as part of a structural or functional data analyses, and with patients meeting our inclusion criteria (see below). We contacted the research groups responsible for the studies and among those N=8 groups took part in the project, offering previously published and/or unpublished data: Prof. Simon Lewis<sup>44</sup> (University of Sydney), Prof. Phil Hyu Lee and Dr. Chung<sup>47</sup> (Yonsei University), Prof. Henry Mak, Prof. Grainne McAlonan and Prof. S.L. Ho<sup>40</sup> (King’s College London and The University of Hong Kong,), Prof. Kathy Dujardin, Prof. Renaud Jardri and Dr. Delphine Pins<sup>48</sup> (University of Lille), Prof. John-Paul Taylor and Dr. Michael Firbank<sup>36</sup> (Newcastle University), Dr. Rimona Weil<sup>49</sup> (University College London,), Prof. Michele Hu, Prof. Clare Mackay and Dr. Ludovica Griffanti<sup>50,51</sup> (Oxford Parkinson’s Centre Discovery Cohort), Dr. Dominic ffytche<sup>41</sup> (King’s College London) (see <strong>Table 1</strong> in the Results section for details). Only data from participants diagnosed as dementia-free were included to minimise the contribution to the study of global cortical changes in patients with PD dementia. The study (LRS-19/20-17680) was given ethical approval by King’s College London Research Ethics Office, Psychiatry, Nursing and Midwifery (PNM) Research Ethics Panel on the 25/03/2020 and was subsequently pre-registered on the Open Science Framework site on 04/05/2020 (<u>https://osf.io/nzatk</u>). The methods follow the pre-registered plan with the addition of exploratory graph theoretical analyses based on structural covariance (section 4.3.4 and results in section 2.6).
+
+## 4.2 Participants
+
+Raw T1-weighted MRI scans were obtained from 8 different groups for a total of 519 subjects. We used 493 MRI scans in the analysis after discarding N=20 participants who did not meet the criteria in terms of diagnosis (e.g. healthy controls, with diagnosis of dementia) or whose scan did not segment well during pre-processing and subsequent troubleshooting steps or was not suitable for analysis (e.g. motion) (N=6). Patients with a MMSE score below 24 (raw) were retained (N=8) only when part of a published work in which the absence of dementia was specifically stated. The final sample comprised 493 participants, 135 with VH, 358 without VH (further details in <em>Results</em> section and in <strong>Table 1</strong> and <strong>S2</strong>). Hallucination data collection varied across groups, as several used a different scale to screen for VH. Each group had previously divided patients into PD-VH and PD-noVH and we retained these original groupings for the mega-analysis.
+
+## 4.3 MRI data pre-processing and harmonisation
+
+MRI data was pre-processed with Freesurfer 6.0.0<sup>52</sup> to estimate cortical thickness, surface area and subcortical volumes. Data was processed on King’s College London HPC infrastructure Rosalind (https://rosalind.kcl.ac.uk), with the standard recon-all procedure, consisting of motion correction, skull-stripping, affine registration to Talairach atlas, segmentation, smoothing, and parcellation mapping. In order to screen for possible errors in the segmentation process, mean cortical thickness measures and manual slice by slice inspection were used to identify possible errors in the white-grey matter boundary and pial reconstruction steps. For subjects that did not segment properly the failed processing steps were re-run (autorecon3) after performing the appropriate corrections. Low quality scans (e.g. with excessive motion, n= 4) or scans that did not segment well upon troubleshooting (n =2) were discarded. Individual cortical thickness, subcortical volumes and surface area measures were extracted based on the Destrieux atlas<sup>53</sup>. In order to explore structural differences between patients with and without VH across the different cohorts minimising variance due to different recruitment sites and, therefore, different scanners, we used a harmonisation method. ComBat is an empirical Bayesian algorithm aiming at minimising the variance due to the scanner features and to maintain the variance related to biological features and has been previously successfully used in studies of cortical thickness<sup>54-55</sup>. In this study, this method has been also used to harmonise volume and surface area for each participant (see <strong>Supplemental information S1</strong> for more details about this method and plotted results).
+
+### 4.3.1 Group differences analysis
+
+First, we conducted a meta-analysis with R package ‘metafor’<sup>56</sup> to check whether patients were matched on the relevant demographic and clinical variables. Results are mentioned in the main text and reported with forest plots and a detailed description in <strong>Supplemental Information S2</strong>.
+
+Then, we conducted separate exploratory ANOVAs and MANCOVAs for cortical thickness, surface area and subcortical volumes to screen for group differences between hallucinators and non-hallucinators, with age, gender and total intracranial volume (TIV) as covariates (when appropriate upon checking assumptions; see <em>Results</em> and Supplemental Information <strong>S2</strong>). Multiple comparisons were Bonferroni corrected. The models were calculated using SPSS 24.0.0.0 (IBM corp. 2016) and R 4.0.0 (R core team, 2017). Results are presented in <strong>Figure 1</strong>, created with a custom colour coding based on <em>p</em> values and by overlaying region labels on a brain render.
+
+We used Tukey’s method programmed in R with the 1.5*IQR rule to identify outliers other than those excluded upon unsuccessful pre-processing. This allowed the careful inspection of the identified subjects in order to verify whether the outlier value depended upon measure errors (e.g. harmonisation bugs) or incorrectly entered data, or on the subject, with the purpose of retaining outliers depending on the subject (e.g. intrinsic features of the subject). No participants were discarded upon this check for this analysis.
+
+### 4.3.2 Sensitivity and Subgroup analysis
+
+Of the eight original groups, three used the Neuropsychiatric Inventory (NPI) to score visual hallucinations. For this subgroup of studies, patients were matched for age, gender, onset, levodopa equivalent daily dose (LEDD), and Mini Mental State Examination (MMSE) score. Within each of the 3 studies, patients were also matched in terms of motor symptoms severity (UPDRS-III). We also ran a one-way ANOVA to check whether the subsample was matched for UPDRS-III but data was missing for 20 participants. We computed the group mean and used that to fill the missing value for the between groups multivariate ANOVA. We carried out Pearson’s product moment correlation coefficient between NPI score and the cortical thickness, surface area and subcortical volume data; we computed the same analysis with LED as a covariate in order to address its potential role in VH severity. In addition, we compared the PD-VH and PD-noVH in the data set using the original VH binary scores to check for consistency in the results with the larger data set, including age, gender, disease onset, LED, PD severity (UPDRS-III) and MMSE as covariates (<strong>Supplemental Information S4</strong>). We also conducted analyses of variance with a larger subgroup and with graded VH scores (mild, moderate, severe), together with an ordinal logistic regression (for details on the sample, methods and results see <strong>Supplemental Information S5</strong>).
+
+### 4.3.3 Receptor density profiles
+
+Regression models with the difference of the means (VH – noVH) of morphometrical features (thickness, surface area, subcortical volume) as dependent variable and receptors density profiles as predictors were carried out, a methodology previously used on brain volumes<sup>57</sup>. Specifically, receptors density profiles were obtained for D2/D3, 5-HT1<sub>A</sub> and 5-HT2<sub>A</sub> based on a [<sup>18</sup>F] Fallypride template<sup>58</sup> and a [11C] Cumi-101 5-HT<sub>1A</sub> and a [11C] Cimbi-36 5-HT<sub>2A</sub> templates<sup>59</sup>, respectively. We have focussed on DA and 5-HT as high resolution templates are available for these receptors of interest at the moment. Including cholinergic maps in the analysis would greatly enrich this approach given the importance of cholinergic transmission in VH in PD (as described in the introduction) and will be done once high resolutions templates will be available. [<sup>18</sup>F] Fallypride is a D2/D3 receptor antagonist with a high signal to noise ratio<sup>60</sup>. [11C] Cumi-101 and [11C] Cimbi-36 are high affinity PET radioligands for 5-HT1<sub>A</sub> and 5-HT2<sub>A</sub> receptors<sup>59</sup> (Beliveau et al., 2017). Parametric modelling of the binding potential used the cerebellum as reference region<sup>61</sup> and thus the vertices corresponding to the cerebellum were excluded from the regression analyses. Each of these templates was registered to the Talairach space using the <em>fsaverage</em> template subject and parcellated with the Destrieux atlas, to ensure alignment with the parcellated structural data of our participants. For each of the vertices we extracted the binding potential using <em>fslmeants</em>. Regression models were carried out to estimate the relationship between cortical thickness and surface area differences of the mean between VH and noVH patients (regions resulting from the first group-level MANCOVAs and ANOVAs, see <strong>2.1</strong>, <strong>S3</strong>) and receptor density profiles. For surface area, we used regions that resulted different in PD-VH vs. PD-noVH from an exploratory one-way ANOVA, as the number of regions resulting different in the basis of the MANCOVAs performed and reported in S3 were too small in number to carry out a more powered model. We ran separate models for each receptor and for thickness, surface area and volume. In addition, for each receptor we ran three different models. First, we examined the relationship between the receptor’s binding potential in the regions with significant differences in cortical thickness/surface area between PD-VH and PD-noVH. The sloopes for these models were also compared (<strong>Supplemental Information S6</strong>). Then, in order to better investigate such relationship, we also assessed whether the receptor’s binding potential could predict thickness/area values for all regions; finally, with the same purpose, we ran models considering only regions where the difference between the groups was not significant (<strong>S6</strong>). Linear regression models were coded in R using the packages rstatix<sup>62</sup> (Kassambara, 2020) and MASS<sup>63</sup>. For each regression model, in order to identify outliers, Cook’s distance was computed and any data point with a Cook’s distance >1 was marked as highly influential, explored and if appropriate discarded<sup>64</sup>. In addition, the confidence intervals of the significant regression models were estimated with the bootstrapping technique<sup>65</sup> with 100,000 cycles (<strong>S6</strong>). Methods and results for thickness and surface area are graphically represented in Figure 3, for results on volume and further details see <strong>Supplemental Information S6</strong>).
+
+### 4.3.4 Principal component analysis (PCA)
+
+Results from the MANCOVAs comparing PD-VH and PD-noVH highlighted the involvement of widespread cortical regions in a high dimensional dataset. We used principal component analysis (PCA), in order to reduce the dimensionality of the dataset and to identify putative latent dimensions underlying the differences in structure in PD-VH versus PD-noVH patients while retaining as much variance as possible<sup>66</sup>. Data from both hemispheres was entered in each model (one for cortical thickness, one for surface area). Analyses were carried out with R packages <em>factominer</em><sup>67</sup> and <em>factoextra</em><sup>68</sup>. The scree plots for the PCA are reported in <strong>S7</strong>. Separate PCAs were carried out for thickness and surface area. PCA inputs comprised the significantly different regions from the MANCOVAs (<strong>S3</strong>). Results are presented in Figure 3, created with a custom colour coding based on the components and by overlaying region labels on a brain render. To further explore a possible relationship of PCA components and hallucination severity, we carried out correlational analyses (Pearson product-moment) between the individual contributions to the different PCA dimensions, the NPI scores in the NPI subsample and the mean thickness and surface area of each dimension/component, that is the mean thickness/area across the regions constituting that component.
+
+### 4.3.5 Structural covariance and graph theory analysis
+
+In order to investigate inter-regional properties to explore and characterise the gray matter network-level organisation of PD-VH, we built networks based on structural covariance, a technique that assays covariation of differences in grey matter morphology between different brain structures across a specific population<sup>69</sup>. Since the most widely used atlas for this kind of analysis is the Desikan-Killiany atlas<sup>70</sup> (see also <sup>71</sup>), we extracted morphometric features (thickness, surface area) at the 68 vertices of this atlas. The dataset was harmonised for multi-site effects with the same procedure described in section <em>4.3.1</em>. The dataset was reduced to 467 cases as the design matrix based on the full dataset was not invertible due to high collinearity of some columns. We discarded N=26 subjects coming from the smallest datasets and the problem was overridden. Homogeneity of groups was verified with a Levene’s test<sup>72,73</sup>.
+
+The dataset counts 467 subjects, 118 PD-VH, 349 PD-noVH, with participants being matched for age. <em>Age</em> and <em>gender</em> were used as covariates in the models. Analyses were carried out with R package <em>braingraph</em><sup>74</sup> and <em>igraph</em><sup>75</sup>. To construct the networks, first we specified a general linear model for each region (thickness/area as outcome variable, age and gender as covariates). The structural covariance matrices of the two groups were defined by estimating the inter-regional correlation between model residuals of thickness and area (in separate models) (e.g. <sup>76</sup>) to build a 68x68 matrix. A density-based threshold<sup>77</sup> was applied to the matrix in order to retain a percentage of the most positive correlations as non-zero elements in a binary adjacency matrix. Different densities ranging from 0.05 to 0.20 with a 0.01 step size were explored. The differences between PD-VH and PD-noVH covariance matrices were then computed, first to establish that the two matrices differed significantly from one another; secondly, a cell by cell comparison was carried out to establish which covariance patterns were significantly greater for the PD-VH group compared to the PD-noVH group. Random undirected and unweighted graphs were created for each group, and vertex- and graph-level metrics were computed for each group. For visualisation purposes a density of 0.13 was selected. Vertex importance was assessed using degree, betweenness centrality and nodal efficiency. A hub was categorised as such if its betweenness centrality was greater than the mean plus 1 standard deviation - calculated on all vertices at the same density<sup>78-82</sup>. To assess network segregation in order to better understand the communities observed, we used modularity, which is a measure of the strength of network partitions. High modularity is a measure of how much vertices from the same community are more connected to each other. Modularity was computed with the Louvain algorithms, which also partitioned the network in communities<sup>83</sup>. Cortical thickness-based networks have been shown to have distinct modules/communities of regions, similar to those derived from fMRI and DTI data<sup>78</sup>. Network analyses were performed with permutation tests (5000 cycles) and bootstrapping analyses to compare vertex-level measures. Results were false discovery rate corrected.
+
+Finally, to further assess the relationship between graph level metrics and visual hallucinations in the full sample, we computed Pearson’s correlation coefficients between the difference of the means of graph metrics of interest (vulnerability, transitivity, local and nodal efficiency, path length, betweenness centrality, eccentricity, distance) for the models on thickness and surface area separately, and the difference of the means in thickness and in surface area, respectively, with the NPI subsample, for which we have all clinical and demographic information and in which participants are matched on all those variables.
+
+# References
+
+1. Dickerson, B. C., Feczko, E., Augustinack, J. C., Pacheco, J., Morris, J. C., Fischl, B., & Buckner, R. L. (2009). Differential effects of aging and Alzheimer's disease on medial temporal lobe cortical thickness and surface area. *Neurobiology of aging*, *30*(3), 432–440.
+
+2. Schapira, A. H., Chaudhuri, K. R., & Jenner, P. (2017). Non-motor features of Parkinson disease. *Nature Reviews Neuroscience*, *18*(7), 435.
+
+3. Ravina, B., Marder, K., Fernandez, H. H., Friedman, J. H., McDonald, W., Murphy, D., … Goetz, C. (2007). Diagnostic criteria for psychosis in Parkinson's disease: report of an NINDS, NIMH work group. *Movement Disorders*, *22*(8), 1061–1068.
+
+4. ffytche, D.H., Creese, B., Politis, M., Chauhuri, K.R., Weintraub, D., Ballard, C. and Aarsland, D. (2017). The psychosis spectrum in Parkinson disease. *Nature Reviews*, 13(2), pp.81–95.
+
+5. Levin et a Levin, J., Hasan, A., & Höglinger, G. U. (2016). Psychosis in Parkinson’s disease: identification, prevention and treatment. *Journal of Neural Transmission*, *123*(1), 45–50
+
+6. Aarsland, D., Andersen, K., Larsen, J. P., & Lolk, A. (2003). Prevalence and characteristics of dementia in Parkinson disease: an 8-year prospective study. *Archives of neurology*, *60*(3), 387–392.
+
+7. Anang, J. B., Gagnon, J. F., Bertrand, J. A., Romenets, S. R., Latreille, V., Panisset, M., … Postuma, R. B. (2014). Predictors of dementia in Parkinson disease: a prospective cohort study. *Neurology*, *83*(14), 1253–1260.
+
+8. Uc, E. Y., McDermott, M. P., Marder, K. S., Anderson, S. W., Litvan, I., Como, P. G., … Growdon, J. C. (2009). Incidence of and risk factors for cognitive impairment in an early Parkinson disease clinical trial cohort. *Neurology*, *73*(18), 1469–1477.
+
+9. Goetz, C. G., & Stebbins, G. T. (1993). Risk factors for nursing home placement in advanced Parkinson's disease. *Neurology*, *43*(11), 2222–2222.
+
+10. Ibarretxe-Bilbao, N., Ramirez-Ruiz, B., Tolosa, E., Marti, M. J., Valldeoriola, F., Bargallo, N., & Junque, C. (2008). Hippocampal head atrophy predominance in Parkinson’s disease with hallucinations and with dementia. *Journal of neurology*, *255*(9), 1324–1331.
+
+11. Watanabe, H., Senda, J., Kato, S., Ito, M., Atsuta, N., Hara, K., … Sobue, G. (2013). Cortical and subcortical brain atrophy in Parkinson's disease with visual hallucination. *Movement Disorders*, *28*(12), 1732–1736.
+
+12. Sanchez-Castaneda, C., Rene, R., Ramirez‐Ruiz, B., Campdelacreu, J., Gascon, J., Falcon, C., … Junque, C. (2010). Frontal and associative visual areas related to visual hallucinations in dementia with Lewy bodies and Parkinson's disease with dementia. *Movement Disorders*, *25*(5), 615–622.
+
+13. Gama, R. L., Bruin, V. M. S., Távora, D. G. F., Duran, F. L., Bittencourt, L., & Tufik, S. (2014). Structural brain abnormalities in patients with Parkinson’s disease with visual hallucinations: a comparative voxel-based analysis. *Brain and cognition*, *87*, 97–103.
+
+14. Shin, S., Lee, J. E., Hong, J. Y., Sunwoo, M. K., Sohn, Y. H., & Lee, P. H. (2012). Neuroanatomical substrates of visual hallucinations in patients with non-demented Parkinson's disease. *Journal of Neurology, Neurosurgery & Psychiatry*, *83*(12), 1155–1161.
+
+15. Goldman, J. G., Stebbins, G. T., Dinh, V., Bernard, B., Merkitch, D., deToledo-Morrell, L., & Goetz, C. G. (2014). Visuoperceptive region atrophy independent of cognitive status in patients with Parkinson’s disease with hallucinations. *Brain*, *137*(3), 849–859.
+
+16. Ramirez-Ruiz, B., Martí, M. J., Tolosa, E., Gimenez, M., Bargallo, N., Valldeoriola, F., & Junque, C. (2007). Cerebral atrophy in Parkinson's disease patients with visual hallucinations. *European journal of neurology*, *14*(7), 750–756.
+
+17. Weil, R. S., Hsu, J. K., Darby, R. R., Soussand, L., & Fox, M. D. (2019). Neuroimaging in Parkinson’s disease dementia: connecting the dots. *Brain communications*, *1*(1), fcz006.
+
+18. Panizzon, M. S., Fennema-Notestine, C., Eyler, L. T., Jernigan, T. L., Prom-Wormley, E., Neale, M., … Xian, H. (2009). Distinct genetic influences on cortical surface area and cortical thickness. *Cerebral cortex*, *19*(11), 2728–2735.
+
+19. Storsve, A. B., Fjell, A. M., Tamnes, C. K., Westlye, L. T., Overbye, K., Aasland, H. W., & Walhovd, K. B. (2014). Differential longitudinal changes in cortical thickness, surface area and volume across the adult life span: regions of accelerating and decelerating change. *Journal of Neuroscience*, *34*(25), 8488–8498.
+
+20. Im, K., Lee, J. M., Lyttelton, O., Kim, S. H., Evans, A. C., & Kim, S. I. (2008). Brain size and cortical structure in the adult human brain. *Cerebral cortex*, *18*(9), 2181–2191.
+
+21. Goodwin, F. K. (1971). Psychiatric side effects of levodopa in man. *Jama*, *218*(13), 1915–1920.
+
+22. Janzen, J., Van‘t Ent, D., Lemstra, A. W., Berendse, H. W., Barkhof, F., & Foncke, E. M. J. (2012). The pedunculopontine nucleus is related to visual hallucinations in Parkinson’s disease: preliminary results of a voxel-based morphometry study. *Journal of neurology*, *259*(1), 147–154.
+
+23. Collerton, D., Perry, E., & McKeith, I. (2005). Why people see things that are not there: a novel perception and attention deficit model for recurrent complex visual hallucinations. *Behavioral and Brain Sciences*, *28*(6), 737–757.
+
+24. Manganelli, F., Vitale, C., Santangelo, G., Pisciotta, C., Iodice, R., Cozzolino, A., … Santoro, L. (2009). Functional involvement of central cholinergic circuits and visual hallucinations in Parkinson's disease. *Brain*, *132*(9), 2350–2355.
+
+25. Ballanger, B., Strafella, A. P., van Eimeren, T., Zurowski, M., Rusjan, P. M., Houle, S., & Fox, S. H. (2010). Serotonin 2A receptors and visual hallucinations in Parkinson disease. *Archives of neurology*, *67*(4), 416–421.
+
+26. Yasue, I., Matsunaga, S., Kishi, T., Fujita, K., & Iwata, N. (2016). Serotonin 2A receptor inverse agonist as a treatment for Parkinson’s disease psychosis: a systematic review and meta-analysis of serotonin 2A receptor negative modulators. *Journal of Alzheimer's Disease*, *50*(3), 733–740.
+
+27. Huot, P., Johnston, T. H., Darr, T., Hazrati, L. N., Visanji, N. P., Pires, D., … Fox, S. H. (2010). Increased 5-HT2A receptors in the temporal cortex of parkinsonian patients with visual hallucinations. *Movement disorders*, *25*(10), 1399–1408
+
+28. Powell A, Ireland C, Lewis SJG. Visual Hallucinations and the Role of Medications in Parkinson's Disease: Triggers, Pathophysiology, and Management. (2020) J Neuropsychiatry Clin Neurosci. Fall;32(4):334–343. doi: 10.1176/appi.neuropsych.19110316. Epub 2020 May 6. PMID: 32374649.
+
+29. Lenka, A., Jhunjhunwala, K. R., Saini, J., & Pal, P. K. (2015). Structural and functional neuroimaging in patients with Parkinson's disease and visual hallucinations: a critical review. *Parkinsonism & related disorders*, *21*(7), 683–691.
+
+30. Pagonabarraga, J., Soriano-Mas, C., Llebaria, G., López-Solà, M., Pujol, J., & Kulisevsky, J. (2014). Neural correlates of minor hallucinations in non-demented patients with Parkinson's disease. *Parkinsonism & related disorders*, *20*(3), 290–296.
+
+31. ffytche, D.H., Pereira, J.B., Ballard, C., Chaudhuri, K.R., Weintraub, D. & Aarsland, D. (2017), 'Risk factors for early psychosis in PD: insights from the Parkinson's Progression Markers Initiative', *Journal of neurology, neurosurgery, and psychiatry*, vol. 88, no. 4, pp. 325–331. https://doi.org/10.1136/jnnp-2016-314832
+
+32. Rimona S. Weil, Anette E. Schrag, Jason D. Warren, Sebastian J. Crutch, Andrew J. Lees, Huw R. Morris, Visual dysfunction in Parkinson’s disease (2016) *Brain*, Volume 139, Issue 11, Pages 2827–2843. https://doi.org/10.1093/brain/aww175
+
+33. Papadopoulos, A., Sforazzini, F., Egan, G., & Jamadar, S. (2018). Functional subdivisions within the human intraparietal sulcus are involved in visuospatial transformation in a non-context‐dependent manner. *Human brain mapping*, *39*(1), 354–368.
+
+34. Tian, X., Zarate, J. M., & Poeppel, D. (2016). Mental imagery of speech implicates two mechanisms of perceptual reactivation. *Cortex*, *77*, 1–12.
+
+35. Sokolov, A. A., Zeidman, P., Erb, M., Ryvlin, P., Friston, K. J., & Pavlova, M. A. (2018). Structural and effective brain connectivity underlying biological motion detection. *Proceedings of the National Academy of Sciences*, *115*(51), E12034-E12042.
+
+36. Firbank, M. J., Parikh, J., Murphy, N., Killen, A., Allan, C. L., Collerton, D., … Taylor, J. P. (2018). Reduced occipital GABA in Parkinson disease with visual hallucinations. *Neurology*, *91*(7), e675-e685.
+
+37. Shine, J. M., Keogh, R., O'Callaghan, C., Muller, A. J., Lewis, S. J., & Pearson, J. (2015). Imagine that: elevated sensory strength of mental imagery in individuals with Parkinson's disease and visual hallucinations. *Proceedings of the Royal Society B: Biological Sciences*, *282*(1798), 20142047.
+
+38. Shine JM, O'Callaghan C, Halliday GM, Lewis SJ. (2014a) Tricks of the mind: Visual hallucinations as disorders of attention. Prog Neurobiol. 2014 May;116:58–65. doi: 10.1016/j.pneurobio.2014.01.004. Epub 2014 Feb 10. PMID: 24525149.
+
+39. Harding, A. J., Stimson, E., Henderson, J. M. & Halliday, G. M. Clinical correlates of selective pathology in the amygdala of patients with Parkinson’s disease. *Brain* 125, 2431–2445
+
+40. Yao, N., Shek-Kwan Chang, R., Cheung, C., Pang, S., Lau, K. K., Suckling, J., … Ho,S. L. (2014). The default mode network is disrupted in Parkinson's disease with visual hallucinations. *Human brain mapping*, *35*(11), 5658–5666.
+
+41. Lawn and ffytche 2020 Cortex (in press)
+
+42. Corbetta, M., & Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in the brain. *Nature reviews neuroscience*, *3*(3), 201–215.
+
+43. Shine, J. M., Halliday, G. M., Gilat, M., Matar, E., Bolitho, S. J., Carlos, M., …Lewis, S. J. (2014b). The role of dysfunctional attentional control networks in visual misperceptions in Parkinson's disease. *Human Brain Mapping*, *35*(5), 2206–2219.
+
+44. Shine JM, Muller AJ, O'Callaghan C, Hornberger M, Halliday GM, Lewis SJ. (2015) Abnormal connectivity between the default mode and the visual system underlies the manifestation of visual hallucinations in Parkinson's disease: a task-based fMRI study. NPJ Parkinsons Dis. 1:15003. doi: 10.1038/npjparkd.2015.3. PMID: 28725679; PMCID: PMC5516559.
+
+45. Ffytche, D. H. (2008). The hodology of hallucinations. *Cortex*, *44*(8), 1067–1083.
+
+46. Qi, H., Liu, H., Hu, H., He, H., & Zhao, X. (2018). Primary disruption of the memory-related subsystems of the default mode network in Alzheimer’s disease: resting-state functional connectivity MRI study. *Frontiers in aging neuroscience*, *10*, 344.
+
+47. Shin, S., Lee, J. E., Hong, J. Y., Sunwoo, M. K., Sohn, Y. H., & Lee, P. H. (2012). Neuroanatomical substrates of visual hallucinations in patients with non-demented Parkinson's disease. *Journal of Neurology, Neurosurgery & Psychiatry*, *83*(12), 1155–1161.
+
+48. Lefebvre, S., Baille, G., Jardri, R., Plomhause, L., Szaffarczyk, S., Defebvre, L.,… Dujardin, K. (2016). Hallucinations and conscious access to visual inputs in Parkinson’s disease. *Scientific reports*, *6*(1), 1–10.
+
+49. Zarkali, A., McColgan, P., Leyland, L. A., Lees, A. J., Rees, G., & Weil, R. S. (2020). Fiber-specific white matter reductions in Parkinson hallucinations and visual dysfunction. *Neurology*, *94*(14), e1525-e1538.
+
+50. Baig, F., Lawton, M., Rolinski, M., Ruffmann, C., Nithi, K., Evetts, S. G., … Hu,M. T. (2015). Delineating nonmotor symptoms in early Parkinson's disease and first-degree relatives. *Movement Disorders*, *30*(13), 1759–1766.
+
+51. Griffanti, L., Klein, J. C., Szewczyk-Krolikowski, K., Menke, R. A., Rolinski, M.,Barber, T. R., … Rumbold, J. (2020). Cohort profile: the Oxford Parkinson’s Disease Centre Discovery Cohort MRI substudy (OPDC-MRI). *BMJ open*, *10*(8), e034110.
+
+52. Fischl, B. (2012). FreeSurfer. *Neuroimage*, *62*(2), 774–781.
+
+53. Destrieux CE, Halgren E, Dale AM, Fischl B, Sereno MI (1998) Variability of the human brain studied on the flattened cortical surface. Soc Neurosci Abstr 24:1164.
+
+54. Fortin, J. P., Cullen, N., Sheline, Y. I., Taylor, W. D., Aselcioglu, I., Cook, P.A., … McInnis, M. (2018). Harmonization of cortical thickness measurements across scanners and sites. *Neuroimage*, *167*, 104–120.
+
+55. Radua, J., Vieta, E., Shinohara, R., Kochunov, P., Quidé, Y., Green, M., … Nenadic,I. (2020). Increased power by harmonizing structural MRI site differences with the ComBat batch adjustment method in ENIGMA. *NeuroImage*, 116956.
+
+56. Viechtbauer W (2010). “Conducting meta-analyses in R with the metafor package.” *Journal of Statistical Software*, *36*(3), 1–48. https://www.jstatsoft.org/v36/i03/.
+
+57. Selvaggi, P., Hawkins, P. C., Dipasquale, O., Rizzo, G., Bertolino, A., Dukart, J.,… Zelaya, F. (2019). Increased cerebral blood flow after single dose of antipsychotics in healthy volunteers depends on dopamine D2 receptor density profiles. *NeuroImage*, *188*, 774–784.
+
+58. Dunn, J. T., Clark-Papasavas, C., Marsden, P., Baker, S., Cleij, M., Kapur, S., …Reeves, S. J. (2013). Establishing test–retest reliability of an adapted [18F] fallypride imaging protocol in older people. *Journal of Cerebral Blood Flow & Metabolism*, *33*(7), 1098–1103.
+
+59. Beliveau, V., Ganz, M., Feng, L., Ozenne, B., Højgaard, L., Fisher, P. M., … Knudsen,G. M. (2017). A high-resolution in vivo atlas of the human brain's serotonin system. *Journal of Neuroscience*, *37*(1), 120–128.
+
+60. Mukherjee, J., Christian, B. T., Dunigan, K. A., Shi, B., Narayanan, T. K., Satter, M., & Mantil, J. (2002). Brain imaging of 18F-fallypride in normal volunteers: Blood analysis, distribution, test‐retest studies, and preliminary assessment of sensitivity to aging effects on dopamine D‐2/D‐3 receptors. *Synapse*, *46*(3), 170–188.
+
+61. Ichise, M., Liow, J. S., Lu, J. Q., Takano, A., Model, K., Toyama, H., … Carson, R.E. (2003). Linearized reference tissue parametric imaging methods: application to[11C] DASB positron emission tomography studies of the serotonin transporter in human brain. *Journal of Cerebral Blood Flow & Metabolism*, *23*(9), 1096–1112.
+
+62. Kassambara, A. (2020). rstatix: Pipe-Friendly Framework for Basic Statistical Tests. R package version 0.5.0. https://CRAN.R-project.org/package=rstati
+
+63. Ripley, B., Venables, B., Bates, D. M., Hornik, K., Gebhardt, A., Firth, D., & Ripley, M. B. (2013). Package ‘mass’. *Cran r*, *538*, 113–120.
+
+64. Cook, R. D., & Weisberg, S. (1982). Criticism and influence analysis in regression. *Sociological methodology*, *13*, 313–361.
+
+65. Efron, B., & Tibshirani, R. (1986). Bootstrap methods for standard errors, confidence intervals, and other measures of statistical accuracy. *Statistical science*, 54–75.
+
+66. Jolliffe, I. T., & Cadima, J. (2016). Principal component analysis: a review and recent developments. *Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences*, *374*(2065), 20150202.
+
+67. Lê, S., Josse, J. & Husson, F. (2008). FactoMineR: An R Package for Multivariate Analysis. *Journal of Statistical Software*. 25(1). pp. 1–18.
+
+68. Kassambara, A., & Mundt, F. (2017). Package ‘factoextra’. *Extract and visualize the results of multivariate data analyses*, *76*.
+
+69. Lerch, J. P., van der Kouwe, A. J., Raznahan, A., Paus, T., Johansen-Berg, H., Miller,K. L., … Sotiropoulos, S. N. (2017). Studying neuroanatomy using MRI. *Nature Neuroscience*, *20*(3), 314–326.
+
+70. Desikan, R. S., Ségonne, F., Fischl, B., Quinn, B. T., Dickerson, B. C., Blacker,D., … Albert, M. S. (2006). An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. *Neuroimage*, *31*(3), 968–980.
+
+71. Carmon, J., Heege, J., Necus, J. H., Owen, T. W., Pipa, G., Kaiser, M., … Wang, Y.(2020). Reliability and comparability of human brain structural covariance networks. *NeuroImage*, *220*, 117104.
+
+72. Nimon, K. F. (2012). Statistical assumptions of substantive analyses across the general linear model: a mini-review. *Frontiers in psychology*, *3*, 322.
+
+73. Cheung, M. (2019). Four covariance structure models for canonical correlation analysis: A COSAN modeling approach.
+
+74. Watson, C. G. (2019). brainGraph: Graph Theory Analysis of Brain MRI Data. R package version 2.7.2. https://github.com/cwatson/brainGraph
+
+75. Gabor Csardi and Tamas Nepusz. The igraph software package for complex network research. (2016) InterJournal, Complex Systems, 1695(5):1
+
+76. He Y, Chen ZJ, Evans AC (2007) Small-world anatomical networks in the human brain revealed by cortical thickness from MRI. Cereb Cortex 17: 2407–2419. Medline
+
+77. Fornito, A., Zalesky, A., & Bullmore, E. (2016). *Fundamentals of brain network analysis*. Academic Press.
+
+78. Watson, C. G., Stopp, C., Newburger,J.W., and Rivkin, M. J.,. (2018) Graph theory analysis of cortical thickness networks in adolescents with d-transposition of the great arteries. *Brain and Behavior*, 8(2). ISSN 2162–3279. doi: 10.1002/brb3.834.
+
+79. Bernhardt, B. C., Chen, Z., He, Y., Evans, A. C., & Bernasconi, N. (2011). Graph-theoretical analysis reveals disrupted small-world organization of cortical thickness correlation networks in temporal lobe epilepsy. Cerebral Cortex, 21(9), 2147–2157.
+
+80. Hosseini, S. H., & Kesler, S. R. (2013). Influence of choice of null network on small-world parameters of structural correlation networks. PLoS One, 8(6), e673
+
+81. Tijms, B. M., Möller, C., Vrenken, H., Wink, A. M., de Haan, W., van der Flier, W. M., & Barkhof, F. (2013). Single-subject grey matter graphs in Alzheimer’s disease. PLoS One, 8(3), e58921.
+
+82. Wang, D., Shi, L., Liu, S., Hui, S. C., Wang, Y., Cheng, J. C., & Chu, W. C. (2013). Altered topological organization of cortical network in adolescent girls with idiopathic scoliosis. PLoS One, 8(12), e83767
+
+83. Blondel, V. D., Guillaume, J.-L., Lambiotte, R., & Lefebvre, E. (2008). Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment,2008(10), P10008
+
+# Supplementary Files
+
+- [NCOMMS2107350.pdf](https://assets-eu.researchsquare.com/files/rs-270425/v1/d1522ce0b162410177606d5e.pdf)  
+  Reporting Summary
+
+- [MVignandoSupplementalInformation.docx](https://assets-eu.researchsquare.com/files/rs-270425/v1/e4ae170cf7efd76711526155.docx)

08aebf361b3e6167428945ee0469ab84c7465d77a1439e9b4f1c63b73651dcd4/preprint/images_list.json

@@ -0,0 +1,50 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Swarming behavior of enzymatic nanomotors viewed from the side. (a) Schematics illustrating the preparation of enzymatic nanomotors and the mechanism of solutal buoyancy resulting in swarming behavior. (b) A time-lapse sequence of images that show the directional and collective movement of enzymatic nanomotors in fuel. The fluid flow is analyzed by adding tracer particles and is shown in black arrows. Scale bar: 1 mm. (c) A time-lapse sequence of snapshots of computational results according to the assumed mechanism. The color bar indicates the nanomotor concentration and white arrows display the fluid velocity.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Control factors that affect swarming behavior. (a) A time-lapse sequence of images of enzymatic nanomotors\u2019 movement in fuel showing three different stages of swarming behavior, i.e. (1) ascending, (2) ascending and spreading, (3) ascending, spreading, and sinking. Scale bar: 4 mm \u00d7 4 mm. The center of mass tracking of UrNM swarms in y-axis under various conditions: (b) UrNM concentration, (d) urea concentration, and (f) HA concentration. (1)-(3) are chosen and displayed in (a). Velocity analysis of active UrNM swarms and passive MSNP swarms shown in (c), (e) and (g) correspond to (b), (d) and (f), respectively. (h) Velocity analysis in y-axis at different heights during particulates\u2019 upward movement in varied UrNM concentration, urea concentration, and fuel with diverse HA concentration. Significant difference is analyzed by students\u2019 t-test: ***=P\u2009<\u20090.001; **=P\u2009<\u20090.01; *=P\u2009<\u20090.05; ns = not significant (P\u2009>\u20090.05). N=5. ND means velocity is lower than detectable value.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Verification of the products of UrNMs catalysis reaction accelerate directional movement. Real-time pH changes with time when adding varied amount of UrNMs into (a) urea dissolved in PBS buffer and (b) urea dissolved in acetate buffer. (c) UrNMs or urease were added in urea dissolved in PBS buffer or acetate buffer (300 mM or 150 mM). Nothing was added in the control group. The color change of phenol red indicates the production of ammonia. (d) Real-time monitoring of NH3. (e) Real-time monitoring of the generated CO2. (f) The specific enzymatic activity of urease and UrNMs in different concentrations of urea dissolved in PBS or acetate buffer. N=3.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Swarming behavior shaped by vertical confinement. Intensity maps and particle image velocimetry (PIV) of UrNM swarms in 300 mM urea solutions in microfluidic chips with varied heights of (a) 1.6 mm, (b) 0.5 mm, and (c) 0.25 mm. The average pixel intensity was calculated over 40 s periods from video recordings (left panels). A zoomed-out view in the right panels shows corresponding PIV measurements. Scale bars in the small panel: 1 mm, in the enlarged panel: 0.5 mm.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Computational modeling shows UrNM swarms move upward under different conditions. (a) Simulation snapshots show two representative collective behaviors of UrNM swarms in low fuel concentration (left) and in viscous fuel (right). The color bar depicts the particulate density \u03c1, and white arrows display the fluid flow velocity. Snapshots of videos at dimensionless time 2.8 under different conditions: (b) various concentrations of particulate (\u03c1 = 1~4), (c) particulate in different fuel concentrations (c = 0.6~1.2), and (d) particulate in fuel with different viscosity (\u03b7 = 0.6~1.2). The domain of integration size 160\u00d740 dimensionless units, number of grid points 1024\u00d7256. Panels (e)-(f) correspond to mean velocity quantification during the upward motion. The dimensionless unit of length corresponds to 0.1 mm, and the dimensional unit of time corresponds to 1-10 seconds of the experiment depending on the parameter choice.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Computational results describing the swarming behavior shaped by confinement. Snapshots at dimensionless time 0, 3 and 6 in (a) relatively less confined space (large height), for \u03b2 = 12, \u03b5 = 0.05, the domain of integration size 60\u00d760 dimensionless units, number of grid points 512\u00d7512; (b) relatively confined space (medium height) for \u03b2 = 12, \u03b5 = 0.03; and (c) smaller height, for \u03b2 = 20, \u03b5 = 0.01, respectively. Density p is shown in colors.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

08aebf361b3e6167428945ee0469ab84c7465d77a1439e9b4f1c63b73651dcd4/preprint/preprint.md

@@ -0,0 +1,227 @@
+# Abstract
+
+Enzymatic nanomotors harvest kinetic energy through the catalysis of chemical fuels. When a group of self-propelled nanomotors is placed in a fuel-rich environment, they assemble into ordered groups and exhibit intriguing swarming behaviors akin to the self-organization observed in bacterial colonies, bioconvection of aerobic microorganismal suspensions, and the coordinated movements of fish, ants, and birds. This swarming behavior presents numerous advantages compared to individual nanomotors, including expanded coverage and prolonged propulsion duration. However, the physical mechanisms underlying the swarming have yet to be fully elucidated. Our study investigates the formation of enzymatic swarms using experimental analysis and computational modeling. We show that the directional movement of enzymatic nanomotor swarms is due to their solutal buoyancy. We investigated various factors that impact the movement of nanomotor swarms, such as particle concentration, fuel concentration, fuel viscosity, and vertical confinement. We examine the effects of these factors on swarm self-organization to gain a deeper understanding. In addition, the urease catalysis reaction produces ammonia and carbon dioxide, accelerating the directional movement of active swarms in urea compared with passive ones in the same conditions. The numerical analysis agrees with the experimental findings. Our findings are crucial for the potential biomedical applications of enzymatic nanomotor swarms, ranging from enhanced diffusion in bio-fluids and targeted delivery to high-efficiency cancer therapy.
+
+Physical sciences/Nanoscience and technology/Nanoscale materials/Nanoparticles  
+Physical sciences/Physics/Fluid dynamics  
+Biological sciences/Biochemistry/Biocatalysis
+
+# Introduction
+
+Swarming behavior is widespread in nature. While individual units of a group obey simple rules, they present complex and intriguing collective behavior when assembling into highly ordered structures<sup>1</sup>. Living organisms use distributed or swarm intelligence to accomplish sophisticated tasks to survive. Examples range from collective cell migration<sup>2</sup>, honeybees adapting to repeated shaking to maintain mechanical stability of the swarm<sup>3</sup>, to emperor penguins packing in a huddle in a highly coordinated manner to survive cold winter<sup>4</sup>. Multiple synthetic swarming systems have been developed with inspiration from nature. The commonly used strategies for the assembly and control of swarms include: (1) applying one or multiple external forces, such as magnetic fields<sup>5–8</sup>, light<sup>9,10</sup>, ultrasound<sup>11,12</sup>, electric fields<sup>13,14</sup>, (2) utilizing chemicals as signals<sup>15–17</sup>, (3) combining biological microswimmers, such as sperm cells and algae, into artificial moieties as a hybrid integration<sup>18–20</sup>, (4) exploiting DNA base-pair interactions<sup>21,22</sup>. These well-designed swarms show many advantages compared to single-unit functionalities, like enhanced coverage and fluid mixing, intelligent multitasking, and environmental adaptation.
+
+Micro/nanomotors (MNMs) are artificial active devices achieving self-propulsion through converting various types of energy into mechanical motion<sup>23</sup>. Enzyme-powered MNMs harnessing energy from enzymatic catalysis of chemical fuel have been proven to show directional mobility. For instance, enzyme-modified liposome motors move directionally up or down by interacting with chemical gradients<sup>24</sup>; catalase-containing microcapsules exhibit enzyme-powered oxygen gas bubble-dependent buoyancy in the presence of hydrogen peroxide, a phenomenon also known as gravitaxis<sup>25</sup>; enzyme-attached macroscale sheets display clockwise or counter-clockwise rotation by a solutal buoyancy mechanism arising from enzyme pump<sup>26</sup>; enzyme catalysis causes fluid flow and directional transport of tracer particles on lipid membranes<sup>27</sup>. Recent findings of the enzymatic nanomotors swarming behavior highlight their potential in biomedical applications, such as accelerated diffusion in viscous media<sup>28</sup>, enhanced targeted payload delivery<sup>29</sup>, and more effective cancer therapy<sup>30</sup>. These enhancements stem from chemical reactions and the resulting asymmetrical distribution of ions in solutions. However, understanding and controlling the collective movement of enzymatic MNMs remains a challenge.
+
+Bioconvection is a self-organized and self-sustained vortex motion that arises naturally in suspensions of microorganisms<sup>31</sup>. It visually resembles the Rayleigh–Bénard convection in fluid heated from below<sup>32</sup>. The bioconvection emerges due to the unstable density gradients resulting from the accumulation of buoyant microorganisms<sup>33</sup>. Each microorganism plays a pivotal role in driving accumulation and fluid flow. Certain gravitactic algae or aerotactic bacteria exhibit upward swimming. In the presence of an upper surface, they form a thin boundary layer of microorganism-rich heavier fluid, which becomes unstable, leading to the formation of falling plumes<sup>34</sup>. Inspired by the mechanisms of bioconvection, we describe enzymatic nanomotor swarms that show gravitactic behavior and collective movement in three dimensions (3D). We model a swarm as light particulates immersed in a denser fluid environment. The particulate swarm moves upward, creating a convective flow in a closed fuel-filled space. This dynamic resembles bioconvection and sheds light on the mechanism of swarming behavior observed in enzymatic nanomotors.
+
+# Results and Discussion
+
+In our study, we view the swarming behavior of enzymatic nanomotors from the side or from the top. Control factors for directional and collective mobility of enzymatic nanomotors, such as particle and fuel concentration and media viscosity, are investigated. In addition, their swarming behavior in vertically confined chambers is studied. The enzymatic nanomotors are based on mesoporous silica nanoparticles with urease attached (UrNMs) and dispersed in phosphate buffer saline (PBS) buffer (for the characterization of UrNMs see the Supplementary Materials Fig. S1 and the Methods). From the side, upon introducing a drop of particulate in a fuel-filled chamber, the drop shows upward motion against gravity, generating a convective flow within the closed space, Fig. 1a. As the particulate reaches the upper boundary, it expands to balance the mean upward force, forming a layer of unstable particle-rich fluid. The layer then sinks in the form of falling plumes. The upward movement of a nanomotor swarm is due to buoyancy arising from the density difference between the reaction product-rich particulate and the media with fuel. We state that individual nanomotors perform urease catalysis reactions. They generate ammonia and carbon dioxide, making the particulates less dense. We conduct computational modeling based on two-fluid hydrodynamics and compare the computational results to the experiments, demonstrating a good agreement, Fig. 1b, c.
+
+## Controlling swarming behavior of UrNMs.
+
+Swarming behavior can be viewed from the side. We studied the influence of three main control factors, UrNM concentration, urea concentration, and viscosity mediated by hyaluronic acid (HA) concentration, on the swarming behavior. As illustrated in Fig. 2a, there are three stages of the swarming behavior of enzymatic nanomotors, i.e., ascending (1, 2 and 3), spreading (2 and 3), and sinking (3). When a swarm of nanomotors seeded in the bottom of the chamber that is filled with fuel, they show directional mobility against gravity. Under different conditions, they display various forms of collective behavior and velocity. Figure 2b shows the *z*-component of the particulate center of mass as a function of time within 32 s. The velocity difference can be deducted from Fig. S2 showing that for higher UrNM concentrations, particulates reach a lower *z* position in 5 s. As the UrNM concentration increases to 20 mg/mL, the majority of the nanomotor swarms cannot get to the upper boundary due to gravity (Fig. 2b, c and video S1). The velocity field was analyzed by front tracking the particulate based on custom Python code. As expected, compared with passive nanoparticles (MSNPs), active nanomotors show enhanced upward speeds, Fig. S3 and video S2. We assume that enzymatic catalysis of urea produces microbubbles<sup>35</sup> and the product, ammonia, makes this particulate less dense. Although the product quantity may be larger with higher UrNM concentration, the density of particulate increases as well when we increase the concentration of nanoparticles. We suggest that there should be a competition between the two opposite conditions, after which the effect of increased particulate density takes the lead, and the upward particulate velocity decreases with the increased UrNM concentration. Buoyancy, the main driving force, is strongly influenced by fuel concentration. Figures 2d, e show that the upward speeds increase with the fuel concentration. One can clearly observe the upward motion of particulates at concentrations of 150 mM urea and above. However, in the presence of 100 mM urea concentration, particulate almost stays at the seeding point, and there is no difference between the upward motion of active and passive swarms. We argue that this is because in low urea concentration, density difference resulting in a buoyancy force is not sufficient to lift the particulate. We added hyaluronic acid into the fuel to change the media viscosity observing that the upward speeds of particulate decreases with the increasing concentration of hyaluronic acid, Fig. 2f, g. As it was shown above, active swarms show enhanced speed compared to passive swarms in viscous media. When the concentration of hyaluronic acid increases to 3 mg/mL, both active and passive swarms remain at the seeding point because higher viscosities inhibit fluid convection. We conducted particulate velocity analysis at elevated heights in the middle of the chamber. In Fig. 2h, active particles move slightly faster in the middle of their paths and decrease their speeds when approaching the upper boundary in different groups, while passive particles keep decreasing their speeds (Fig. S4). For instance, a particulate of 5 mg/mL UrNPs moves upward at 1.74 ± 0.09 mm/s at 4 mm height, 1.93 ± 0.14 mm/s at 5 mm height, and 1.70 ± 0.08 mm/s at 7 mm/s, while a particulate of the same concentration of passive nanoparticles moves at 0.96 ± 0.03 mm/s at 4 mm height, 0.68 ± 0.02 mm/s at 5 mm height, and 0.47 ± 0.02 mm/s at 7 mm height. The acceleration process of active particulate could be due to the density changes caused by chemical reaction products.
+
+## Products of UrNMs catalysis reaction accelerate directional movement.
+
+Urease catalyzes the decomposition of urea into ammonia (NH<sub>3</sub>) and carbon dioxide (CO<sub>2</sub>). On the one hand, NH<sub>3</sub> is highly soluble in water due to the formation of hydrogen bonds with water molecules. This interaction results in a smaller density of the solution<sup>36,37</sup>. On the other hand, the released NH<sub>3</sub> dissolves in water, resulting in an alkaline solution (Fig. 3a) and promoting CO<sub>2</sub> to dissolve. Under proper fuel concentration, the formation of NH<sub>3</sub> and CO<sub>2</sub> microbubbles can be observed<sup>35</sup>. However, in acidic buffers, such as acetate buffer (pH = 4.6, Fig. 3b), CO<sub>2</sub> may exist because the abundant hydrogen ions inhibit the dissolution of CO<sub>2</sub> and the ionization of carbonic acid. Since the temperature change during chemical reactions is hard to detect (Fig. S5), we rule out heat effect on the upward movement. We conducted experiments to verify the existence of NH<sub>3</sub> and CO<sub>2</sub>. In Fig. 3c, cover papers were pre-dipped in phenol red solutions, a pH indicator. Upon adding urease or UrNMs into urea solution, NH<sub>3</sub> is produced and evaporates until it dissolves in the cover paper that contains phenol red, the color change of which from light yellow to pink indicates the presence of NH<sub>3</sub>. The production of CO<sub>2</sub> can be observed in acetate buffer, which maintains an acidic environment during the urease catalysis reaction, Fig. 3b. CO<sub>2</sub> bubbles produced by UrNMs reacting with urea dissolving in acetate buffer can be observed on the wall of a cuvette (video S3). In addition, the produced NH<sub>3</sub> and CO<sub>2</sub> in acetate buffer can be directly detected by a gas sensor, an optoelectronic analysis equipment that is able to accurately detect low concentration gases at ppm level, as shown in Fig. 3d, e. The enzymatic activity of UrNMs in urea solutions in both PBS buffer (Fig. 3f, S6) and acetate buffer (Fig. S7) was examined. In PBS buffer, the specific enzymatic activity of UrNMs increases from 4.08 ± 0.02 U/mg in 50 mM urea solutions to 4.82 ± 0.41 U/mg in 300 mM urea solutions. In acetate buffer, the specific enzymatic activity of UrNMs is slightly weaker, with 1.94 ± 0.05 U/mg in 50 mM urea solutions and 3.36 ± 0.45 U/mg in 300 mM urea solutions. This is because the known optimum pH for urease catalytic activity is around 7 ~ 8<sup>38</sup>. The above results indicate that urease catalysis reaction produces NH<sub>3</sub> and dissolved CO<sub>2</sub> in PBS buffer, and NH<sub>3</sub> and CO<sub>2</sub> gas in acetate buffer, which are the main reasons that cause accelerated directional movement of active UrNM swarms in urea.
+
+## Vertical confinement shapes swarming behavior.
+
+Since buoyancy is the primary force that drives the self-organization of active particulates, we studied the influence of vertical confinement on their swarming behavior. As shown in Fig. 4, microfluidic chips with three different heights (1.6 mm, 0.5 mm, and 0.25 mm) were designed and filled with urea in the vertically confined chamber. Then active UrNMs were introduced and entered the chamber from the side by capillary force. In Fig. 4a and video S4, these active UrNM swarms exhibit collective movement in the chamber of 1.6 mm height. The density maps, observed from the top, show that the swarms aggregate, coarsen, and change their patterns over time. Particle image velocimetry (PIV) also confirms that the fluid flow is initially faster when the nanomotors are injected into the chamber. Fig. S8-S10 show the PIV results at 25 s time intervals in confinement with different heights. After 50 s, nanomotors keep moving and swarming behavior is still transient. After 100 s, the fluid flow keeps a relatively high speed, 1.5 µm/s on average. However, the fluid flow direction remains the same according to the arrows. As a comparison, without fuel UrNMs sink to the bottom in a confined chamber and expand along the bottom plane, Fig. S11. The convective flow is also weaker than that caused by UrNMs with fuel, Fig. S12-14. When the vertical confinement is changed to 0.5 mm, the movement of active UrNMs becomes localized. In Fig. 4b, the density map shows that the pattern of UrNMs only slightly changes over time. The PIV reveals that fluid flow velocity decreases compared to larger height values. After 50 s, the swarms barely move. When the height is further reduced to 0.25 mm, the swarms’ movement is hindered, as displayed by the unchanged shape of swarms over time and the decreased velocity of fluid flow in PIV, Fig. 4c. Active UrNMs in PBS solutions also show decreased velocity when the chamber height decreases (Fig. S14). However, compared with the active UrNMs in fuel, there are no significant differences. We also analyzed the swarm dynamics by pixel intensity distribution. A time-lapse sequence of snapshots at 12 s time intervals from video recordings is selected. As shown in Fig. S15, in a 1.6 mm-high chamber, the pixel intensity of active UrNMs in fuel is broadly distributed in the region of interest (ROI) in the initial 60 seconds, and gradually changes to narrowly distributed in 2 min. However, for the 0.5 mm-high chamber and the 0.25 mm-high chamber, pixel intensities are monodispersed in the ROI within the time durations. As a comparison, the pixel intensities of active UrNMs in PBS solutions are highly monodispersed in the three different chambers, Fig. S16. These results indicate that the vertical confinement shapes the swarms by affecting fluid convective flows.
+
+## Computational modeling shows similarity with experiments.
+
+Our starting point is two-fluid hydrodynamics<sup>39</sup>. One fluid is a solvent with the kinematic viscosity *η*, flow velocity **v**, solvent pressure *p*, and solvent density *ρ*<sub>0</sub>. Another fluid is the particulate with the volume density *ρ*, coarse-grained particulate velocity **u**, and pressure *P = qρ*, and the factor *q* depends on the temperature (as for gases). We describe the dynamics by the simplified Navier-Stokes Eq. (1), coupled to the reaction-advection equation for the concentration of chemical fuel *c*, Eq. (2), and a mass transport equation for the particulate density, Eq. (3):
+
+$$
+{\rho }_{0}\left({\partial }_{t}\mathbf{v}+\mathbf{v}\nabla \mathbf{v}\right)= \eta {\nabla }^{2}\mathbf{v}-\nabla p-{\mathbf{z}}_{0}\rho \left(g\alpha -\epsilon c\right)
+$$
+1
+
+$$
+{\partial }_{t}c+\nabla \cdot \left(\mathbf{v}c\right)= {D}_{c}{\nabla }^{2}c-\gamma \rho c
+$$
+2
+
+$$
+{\partial }_{t}\rho +\nabla \cdot \left(\mathbf{v}\rho \right)=\left(q {\nabla }^{2}\rho +\alpha g{\partial }_{z}\rho \right)/{\kappa }_{1}
+$$
+3
+
+where z<sub>o</sub> *ρεc* is the volume buoyancy force due to gas generation, **z**<sub>o</sub> is the unit vector in the z-direction, the gas is produced due to the reaction between fuel *c* and particulate *r* with the reaction rate *g*. Other parameters: fuel diffusion *D*<sub>c</sub>, gravity acceleration *g*, relative particulate/solvent density contrast *a*, *ε* is the relative buoyancy coefficient that depends on the density of reaction products, and *k*<sub>1</sub> is the normalized drag coefficient. The details of model derivation are presented in Supplementary Note 1. Equations (1)-(3) were solved by the finite difference method using Matlab. We considered a two-dimensional rectangular integration domain (corresponding to the size view) with periodic boundary conditions in the *x*-direction and non-slip conditions in the *z*-direction. The primary difference with models of enzyme-generated solutal buoyancy mechanisms considered in Ref. <sup>40</sup> is that the enzyme distribution is not fixed but dynamically updated by the reaction-generated flow.
+
+When buoyancy is not sufficient to counterbalance the gravity of particulates, like in the cases of high concentration of particles and low concentration of fuel, the particulate is not able to rise to the top plane and sink to the bottom after seeding, Fig. 5a, left panel. On the contrary, in the cases of low concentration of particles and high concentration of fuel, particulates rise and spread along the top plane, then descend, experiencing a similar process as in experiment, Fig. 5a, right panel, and video S5. In simulations, the volume density *ρ* changes from 1 to 4, chemical fuel *c* ranges from 0.6–1.2, and kinematic viscosity *η* varies from 0.1-1.0 to simulate different concentrations of particles, fuel, and HA, respectively. In Figs. 5b-d, frames at dimensionless time 2.8 are chosen from computer videos for different parameters. Figure 5b shows that in the same time frame, particulate with smaller density *ρ* enters the sinking stage, while particulate with larger *ρ* is still in the ascending or spreading stage, indicating that lighter particulates move faster. This observation agrees with the experimental results and can be further verified by Fig. 5e. The mean velocity of particulate during upward movement decreases with the increase of density *ρ*. In Fig. 5c, particulate settles to the bottom when chemical fuel concentration *c* is low (*c* = 0.6). Increasing the *c* value (*c* = 0.8) triggers particulate’s upward movement, yet it settles before reaching the top plane. Only relatively high fuel concentrations force the particulate to go through the three stages, and its upward speed increases with the increase of *c* value. In Fig. 5f, the gradual increase of the mean particulate velocity with the fuel concentration from simulations agrees with that observed in the experiments. The effect of viscosity is shown in Fig. 5d, g. Particulate in lower viscosity media enters the sinking stage earlier than for higher viscosity. Computational modeling confirms that the increased fuel viscosity slows down the particulate motion.
+
+## Computational modeling of the vertical confinement effects.
+
+We performed computational modeling of the effect of vertical confinement on swarming behavior. The details are presented in Supplementary Note 2. The model is derived from Eqs. (1)-(3) by height-averaging using the approach like in Ref. <sup>41</sup>. The corresponding two-dimensional equations in the *x-y* plane are solved by the quasi-spectral method in the periodic square domain using Matlab.
+
+Parameter *β* is proportional to the reaction rate and parameter *e* ~ *h*<sup>2</sup>, where *h* is the height of the chamber. We adjust the value of these two control parameters to describe the fluid flow slowdown caused by confinement. In Fig. 6a, numerical results show that in vertical confinement, particulate moves dynamically and form aggregates in the center area of the cell. A similar phenomenon has been observed in experiment, Fig. 4a. However, when the chamber’s height is reduced, the fluid flow slows, and the reaction rate decreases. As a result, particulate movement becomes more localized, and the shape formed by a particulate remains almost unchanged within the time durations, as shown in Fig. 6b-c and video S6. Furthermore, there is no significant difference between the swarm dynamics in two highly confined chambers because fluid convection is inhibited by vertical confinement.
+
+# Conclusion
+
+We investigated the swarming behavior of enzymatic nanomotors from the side and from the top. We attribute their collective behavior to buoyancy-induced convection. When introducing a drop of UrNMs, dispersed in PBS buffer, into a fuel medium (high concentration of urea dissolved in PBS), the UrNMs exhibit directional upward movement due to buoyancy arising from the density difference between the particulate and the fuel medium. UrNMs decompose urea and generate carbon dioxide and ammonia, with the latter dissolving in water, further reducing the particulate density and enhancing its upward movement. When reaching the solid-air interface, UrNMs spread along the interface, form an unstable layer of front, and then sink in the form of finger-like aggregates. The process resembles natural bioconvection in microorganismal suspensions.
+
+Particle concentration, fuel concentration, and viscosity are crucial parameters to control enzymatic swarming behavior. Specifically, increasing particle concentration, decreasing fuel concentration, or increasing viscosity can decrease the density difference between the particulate and the fuel, impeding the initiation of upward movement and subsequent convection. This phenomenon explains the settlement of nanoparticles to the bottom when observed under inverted microscopy. Furthermore, the movement of UrNMs in vertical confinement also serves as a demonstration of buoyancy-induced convection. Confinement hinders fluid convection, indicating that the swarming behavior of enzymatic nanomotors requires vertical spaces to overcome dissipation.
+
+We performed computational modeling based on the buoyancy-driven convection mechanisms; the results align well with experimental findings. In computational modeling, particulate ascends due to buoyancy, spreads upon reaching the top, and consequently descends because of gravity. Consistent with the experimental observations, an increase in particulate density ($\rho$), a decrease in fuel concentration ($c$), or an increase in fuel viscosity ($\eta$) decreases the mean particulate velocity. Computational modeling also agrees with experimental observations for particulate moving in vertical confinement. By adjusting the parameters $\beta$ and $\kappa$, corresponding to the reaction rate and the chamber height, respectively, the computational model predicts that vertical confinement shapes the swarms by controlling fluid convection.
+
+# Methods
+
+## Synthesis of MSNPs-NH₂
+
+Mesoporous silica nanoparticles (MSNPs) serving as chassis for urease-propelled nanomotors were synthesized by the sol-gel procedure according to our previous report<sup>42</sup>. The surface of MSNPs was then modified for further functionalization. Briefly, 20 mg MSNPs in ethanol 99% (Panreac Applichem cat. no. 131086-1214) and 100 µL 3-aminopropyltriethoxysilane (APTES) 99% (Sigma-Aldrich cat. no. 440140) were mixed and placed in an end-to-end shaker at room temperature for 24 h. The resulting nanoparticles were then collected and washed in water by centrifugation (4000 rpm, 5 min) four times to remove residual APTES. The collected MSNPs-NH₂ nanoparticles were dried for further use.
+
+## Synthesis of UrNMs
+
+The prepared MSNPs-NH₂ nanoparticles (2.5 mg) were resuspended in 1mL PBS 1× (Thermo Fisher Scientific cat. no. 70011-036) and activated with 100 µL GA 25 wt% (Sigma-Aldrich cat. no. G6257) in an end-to-end shaker for 2.5 hours at room temperature. The activated MSNPs-NH₂ were then collected and washed four times in PBS 1× by centrifugation (4000 rpm, 5 min), then resuspended in 1mL PBS 1× with 3 mg urease from Canavalia ensiformis (Sigma-Aldrich cat. no. U4002). The mixture reacted at room temperature in an end-to-end shaker overnight. The resulting urease-nanomotors (UrNMs) were collected and washed thrice in PBS 1× by centrifugation (4000 rpm, 5 min). Keep the supernatant of centrifugation for further quantification of the enzyme linkage. Finally, resuspend the collected UrNMs in PBS 1× (0.5 mL) and store them in the fridge at 4℃ for future use.
+
+## DLS measurements of UrNPs
+
+Malvern Nanosizer (Zetasizer Nano ZSP) was used to measure the diffusion coefficient of UrNPs across a range of urea concentrations (0, 50, 100, 150, and 300 mM) and the surface charge of MSNPs, MSNPs-NH₂, and UrNMs. We analyzed the diffusion coefficient of UrNPs (20 µg/mL) at each urea concentration and zeta potential values of each type of nanoparticles (20 µg/mL) with three runs per experiment. Nine measurements per type of particle were performed to obtain statistically relevant data.
+
+### UrNMs characterization.
+
+The synthesized MSNPs were characterized by scanning electron microscope (SEM), which shows a uniform particle size distribution centered at 450 nm (Fig. S1 a, b). Amino groups were then grafted on the surface of MSNPs, facilitating further modification of urease on the MSNPs surface by linking GA molecules between amino groups. The surface modification process was characterized by zeta potential measurements (Fig. S1 c). The introduction of amino groups results in a negative surface charge of MSNPs reversed from −38.7 ± 4.61 mV to a positive surface charge of 27.77 ± 7.9 mV. The subsequent linking of GA molecules and urease is confirmed by particle surface charge changes due to the presence of abundant aldehyde groups and carboxyl groups, with negative surface charge reverses to -14.5 ± 9.13 mV and −9.02 ± 4.34 mV for GA molecule and urease, respectively. DLS measurement indicates that the prepared UrNMs show an enhanced diffusion coefficient in elevated urea concentrations (Fig. S1 d).
+
+## Optical video recording
+
+The swarming behavior of UrNMs in a vertically confined space was recorded using a Leica DMi8 microscope equipped with a high-speed cooled charge-coupled device (CCD) camera from Hamamatsu and a 2.5× objective lens. Two pieces of cover glasses were separated by spacers (Silicone isolators from Grace Bio-Labs) with varying heights: 1.6 mm, 0.5 mm, and 0.25 mm. The confined space was then filled either with PBS or with a 300 mM urea solution in PBS and positioned under the microscope. A drop of UrNMs or MSNPs (3 µL) was added to the liquid-filled chamber, and videos (25 fps, 2 min) were recorded. Optical videos from side view were recorded using a digital camera (Thorlabs, DCC1240M-GL) equipped with a lens (FUJINON, HF35HA-1S). A 24×1.6×8 mm chamber was prepared by separating two pieces of cover glasses with spacers. A drop of UrNMs or MSNPs (3 µL) was added to the liquid-filled chamber and videos (15 fps, 2 min) were recorded.
+
+## Bicinchoninic acid (BCA) assay
+
+The amount of urease linked onto the MSNP surface was quantified by BCA analysis (Thermo Fisher Scientific cat. no. 23227), table S1. Bovine serum albumin (BSA) was used as the standard for quantifying the concentration of protein concentrations. First, a series of BSA concentrations 2000, 1500, 1000, 750, 500, 250, 125, 25, 0 µg/mL were prepared, and BSA solutions and the as-prepared supernatant of UrNMs samples were added separately into a 96-multi-well plate, 25 µL for each well. Then 200 µL of working reagent (light sensitive), made with 50 parts of reagent A and 1 part of reagent B, were added to each well that has been used, either with the BSA standards or the sample. Next, shake the plate for 30 s to mix the solutions, and incubate the reaction for 30 min at 37℃. After that, the absorption of both BSA solutions and the samples after the reaction was measured at 562 nm wavelength. Comparing the protein quantity left in the supernatant, the initial quantity of proteins added, and the standard concentrations of BSA.
+
+## Enzymatic activity measurement
+
+Urease activity was detected before and after being linked on the surface of MSNPs and was compared in PBS and acetate buffer. 0.025 mM Phenol red (Sigma-Aldrich cat. no. 114529) was added to different concentrations of urea solutions (0, 50, 100, 200, and 300 mM). The solvent of these solutions could be either PBS or acetate buffer. 2 µL PBS solution or UrNMs or urease (130 µg/mL) were added in a 96-well plate separately, followed by the addition of urea solutions. The 96-well plate was immediately placed in a multimode microplate reader (BioTek Synergy HTX). The absorption changes of phenol red were measured in real-time at 560 nm wavelength for 60 min for measurements in PBS buffer and 100 min for acetate buffer. The incubation temperature is at room temperature, the minimal time interval for measurements is 30 seconds, and the sample was orbitally shaken at a minimal frequency.
+
+## Estimation of specific enzymatic activity
+
+The specific enzymatic activity was determined by calculating the slope of the enzymatic activity curve. According to the Beer-Lambert law,
+
+$$A=klc$$
+
+where *A* is the absorbance, *k* is the molar attenuation coefficient of phenol red, and *l* is the path length of 0.5 cm, the concentration changes *c* of phenol red per minute were computed. Subsequently, the specific enzymatic activity was derived based on the quantity of enzyme utilized.
+
+## Gas detection
+
+The generated CO₂ and NH₃ were identified using a gas detector (Dräger X-am 7000). In a glass bottle filled with urea dissolved in PBS (10 mL, 200 mM) or acetate buffer (10 mL, 200 mM), UrNMs (2.5 mg) were added, and the caps were securely fastened to prevent gas release. After 30 min, the bottle caps were removed, and the probe of the gas detector was placed over the solutions to record the generated gases.
+
+## Videos analysis
+
+To investigate the dynamics of swarms over time, the recorded videos were analyzed by pixel intensity distribution and density maps. For pixel intensity distribution, snapshots of videos were captured at 12-second intervals. Subsequently, a region of interest (ROI) measuring 300 pixels by 300 pixels was selected, and the pixel intensity distribution within the ROI was analyzed using ImageJ software. To perform density map analysis, the videos were initially processed to remove the background using ImageJ software. Then 40-second segments were extracted from these videos. The cumulative pixel intensity of these segments, consisting of 1000 frames each, was computed and visualized using the turbo colormap.
+
+## Particle image velocimetry (PIV)
+
+The PIV of recorded videos was conducted by a custom Python code based on the OpenPIV library. The consecutive frames of videos within desired time intervals were extracted and then loaded into the code OpenPIV, with an interrogation window size of 32×32 pixels (width×height), an overlap of 16×16 pixels (horizontal×vertical), and a frame rate of 3.33 fps. The results were then reloaded into the Python code to adjust the arrow size and display particle velocities in color bars.
+
+# References
+
+1. Wang, H. & Pumera, M. Coordinated behaviors of artificial micro/nanomachines: From mutual interactions to interactions with the environment. *Chem. Soc. Rev.* **49**, 3211–3230 (2020).
+
+2. Shim, G., Devenport, D. & Cohen, D. J. Overriding native cell coordination enhances external programming of collective cell migration. *Proc. Natl. Acad. Sci. U.S.A.* **118**, e2101352118 (2021).
+
+3. Peleg, O., Peters, J. M., Salcedo, M. K. & Mahadevan, L. Collective mechanical adaptation of honeybee swarms. *Nat. Phys.* **14**, 1193–1198 (2018).
+
+4. Zitterbart, D. P., Wienecke, B., Butler, J. P. & Fabry, B. Coordinated movements prevent jamming in an emperor penguin huddle. *PLoS One* **6**, e20260 (2011).
+
+5. Yang, M. et al. Swarming magnetic nanorobots bio-interfaced by heparinoid-polymer brushes for in vivo safe synergistic thrombolysis. *Sci. Adv.* **9**, eadk7251 (2023).
+
+6. Yu, J. et al. Active generation and magnetic actuation of microrobotic swarms in bio-fluids. *Nat. Commun.* **10**, 5631–5642 (2019).
+
+7. Xie, H. et al. Reconfigurable magnetic microrobot swarm: multimode transformation, locomotion, and manipulation. *Sci. Robot.* **4**, eaav8006 (2019).
+
+8. Bente, K. et al. Selective actuation and tomographic imaging of swarming magnetite nanoparticles. *ACS Appl. Mater. Interfaces.* **4**, 6752–6759 (2021).
+
+9. Chen, M. et al. Programmable dynamic shapes with a swarm of light-powered colloidal motors. *Angew. Chem. Int. Ed.* **60**, 16674–16679 (2021).
+
+10. Aubret, A., Youssef, M., Sacanna, S. & Palacci, J. Targeted assembly and synchronization of self-spinning microgears. *Nat. Phy.* **14**, 1114–1118 (2018).
+
+11. Xu, T. et al. Reversible swarming and separation of self-propelled chemically powered nanomotors under acoustic fields. *J. Am. Chem. Soc.* **137**, 2163–2166 (2015).
+
+12. Tang, S. et al. Structure-dependent optical modulation of propulsion and collective behavior of acoustic/light-driven hybrid microbowls. *Adv. Funct. Mater.* **29**, 1809003–1809009 (2019).
+
+13. Leunissen, M. E., Vutukuri, H. R. & Van Blaaderen, A. Directing colloidal self-assembly with biaxial electric fields. *Adv. Mater.* **21**, 3116–3120 (2009).
+
+14. Ma, F., Wang, S., Wu, D. T. & Wu, N. Electric-field-induced assembly and propulsion of chiral colloidal clusters. *Proc. Natl. Acad. Sci. U. S. A.* **112**, 6307–6312 (2015).
+
+15. Duan, W., Liu, R. & Sen, A. Transition between collective behaviors of micromotors in response to different stimuli. *J. Am. Chem. Soc.* **135**, 1280–1283 (2013).
+
+16. Gentile, K., Somasundar, A., Bhide, A. & Sen, A. Chemically powered synthetic “living” systems. *Chem* **6**, 2174–2185 (2020).
+
+17. Ziepke, A., Maryshev, I., Aranson, I. S. & Frey, E. Multi-scale organization in communicating active matter. *Nat. Commun.* **13** (2022).
+
+18. Zhang, F. et al. Extremophile-based biohybrid micromotors for biomedical operations in harsh acidic environments. *Sci. Adv.* **8**, eade645 (2022).
+
+19. Yan, X. et al. Multifunctional biohybrid magnetite microrobots for imaging-guided therapy. *Sci. Robot.* **2**, eaaq1155 (2017).
+
+20. Rajabasadi, F. et al. Multifunctional 4D-printed sperm-hybrid microcarriers for assisted reproduction. *Adv. Mater.* **34**, 2204257–2204271 (2022).
+
+21. Akter, M. et al. Cooperative cargo transportation by a swarm of molecular machines. *Sci. Robot.* **7**, eabm0677 (2022).
+
+22. Keya, J. J. et al. DNA-assisted swarm control in a biomolecular motor system. *Nat. Commun.* **9**, 453–460 (2018).
+
+23. Chen, S. et al. Dual-source powered nanomotor with integrated functions for cancer photo-theranostics. *Biomaterials* **288**, 121744–121753 (2022).
+
+24. Somasundar, A. et al. Positive and negative chemotaxis of enzyme-coated liposome motors. *Nat. Nanotechnol.* **14**, 1129–1134 (2019).
+
+25. Kumar, B. V. V. S. P., Patil, A. J. & Mann, S. Enzyme-powered motility in buoyant organoclay/DNA protocells. *Nat. Chem.* **10**, 1154–1163 (2018).
+
+26. Song, J., Shklyaev, O. E., Sapre, A., Balazs, A. C. & Sen, A. Self‐propelling macroscale sheets powered by enzyme pumps. *Angew. Chem. Int. Ed.* e202311556 (2023).
+
+27. Sapre, A. et al. Enzyme catalysis causes fluid flow, motility, and directional transport on supported lipid bilayers. *ACS Appl. Mater. Interfaces* (2024).
+
+28. Ruiz-González, N. et al. Swarms of enzyme-powered nanomotors enhance the diffusion of macromolecules in viscous media. *Small* 2309387–2309403 (2024).
+
+29. Fraire, J. C. et al. Light-triggered mechanical disruption of extracellular barriers by swarms of enzyme-powered nanomotors for enhanced delivery. *ACS Nano* **17**, 7180–7193 (2023).
+
+30. Simó, C. et al. Urease-powered nanobots for radionuclide bladder cancer therapy. *Nat. Nanotechnol.* (2024).
+
+31. Kessler, J. O. & Hill, N. A. The growth of bioconvection patterns in a uniform suspension of gyrotactic micro-organisms. *J. Fluid Mech.* **195**, 223–237 (1988).
+
+32. Bodenschatz, E., Pesch, W. & Ahlers, G. Recent Developments in Rayleigh-Bénard Convection. *Annu. Rev. Fluid Mech.* **32**, 709–778 (2000).
+
+33. Bees, M. A. Advances in bioconvection. *Annu. Rev. Fluid Mech.* **52**, 449–476 (2020).
+
+34. Metcalfe, A. M. & Pedley, T. J. Falling plumes in bacterial bioconvection. *J. Fluid Mech.* **445**, 121–149 (2001).
+
+35. Feng, Y. et al. Self-adaptive enzyme-powered micromotors with switchable propulsion mechanism and motion directionality. *Appl. Phys. Rev.* **8** (2021).
+
+36. Hales, J. M. & Drewes, D. R. Solubility of ammonia in water at low concentrations. *Atmos. Environ.* **13**, 1133–1147 (1979).
+
+37. Gottwald, F. et al. Physical properties of ammonia solutions. Ammonium nitrate-ammonia-water and urea-ammonia-water. *Ind. Eng. Chem.* **44**, 910–913 (1952).
+
+38. Estiu, G. & Merz, K. M. Catalyzed decomposition of urea. Molecular dynamics simulations of the binding of urea to urease. *Biochemistry* **45**, 4429–4443 (2006).
+
+39. M. Ishii & K. Mishima. Two-fluid model and hydrodynamic constitutive relations. *Nucl. Eng. Des.* **82**, 107–126 (1984).
+
+40. Valdez, L., Shum, H., Ortiz-Rivera, I., Balazs, A. C. & Sen, A. Solutal and thermal buoyancy effects in self-powered phosphatase micropumps. *Soft Matter* **13**, 2800–2807 (2017).
+
+41. Aranson, I. S. & Sapozhnikov, M. V. Theory of pattern formation of metallic microparticles in poorly conducting liquids. *Phys. Rev. Lett.* **92**, 234301–234304 (2004).
+
+42. Hortelao, A. C. et al. Swarming behavior and in vivo monitoring of enzymatic nanomotors within the bladder. *Sci. Robot.* **6**, eabd2823 (2021).
+
+# Supplementary Files
+
+- [Supplementaryinfofinalversion.docx](https://assets-eu.researchsquare.com/files/rs-3999734/v1/6981602d85a5f29920f90d4f.docx)  
+  Supplementary info
+
+- [SIVideo1Controlfactorsinfluenceenzymaticswarmingbehavior.mp4](https://assets-eu.researchsquare.com/files/rs-3999734/v1/f6983d3b5c9745a653db9860.mp4)  
+  SI Video1
+
+- [SIVideo2ControlfactorsinfluenceswarmingbehaviorofMSNPs.mp4](https://assets-eu.researchsquare.com/files/rs-3999734/v1/94ec11c67d666b30f1839873.mp4)  
+  SI Video2
+
+- [SIVideo3Bubblesproducedbyenzymaticcatalysisreaction.mp4](https://assets-eu.researchsquare.com/files/rs-3999734/v1/c0647e658dc627a523359818.mp4)  
+  SI Video3
+
+- [SIVideo4Verticalconfinementshapesswarmingbehavior.mp4](https://assets-eu.researchsquare.com/files/rs-3999734/v1/7eed4f900161ad294ce48efe.mp4)  
+  SI Video4
+
+- [SIVideo5Computationalmodelingofthecontrolfactors.mp4](https://assets-eu.researchsquare.com/files/rs-3999734/v1/9a20c2da2008be0eed4e73c0.mp4)  
+  SI Video5
+
+- [SIVideo6Computationalmodelingoftheverticalconfinementeffets.mp4](https://assets-eu.researchsquare.com/files/rs-3999734/v1/53ca96019711f35dc6d9fd27.mp4)  
+  SI Video6

0986baca78f5b40819f004daf2eb76d7f4f4f10f91bb0b965eadad53992e8b0b/preprint/images_list.json

@@ -0,0 +1,34 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Overall structure of ABA3 proteins. (a) The overall structures of apo-BcABA3 (PDB ID, 8ZAC [http://doi.org/10.2210/pdb8ZAC/pdb]), apo-RuABA3 (PDB ID, 8ZAD [http://doi.org/10.2210/pdb8ZAD/pdb]) and apo-SkABA3 (PDB ID, 8ZAF [http://doi.org/10.2210/pdb8ZAF/pdb]) in the homodimeric configuration are shown. The dimeric counterpart in each drawing is shown in gray cartoon. The metal ions are displayed in spheres and helix a11\u2019 in chain A of apo-SkABA3 structure is colored in purple. (b) One polypeptide chain of apo-BcABA3 is displayed with each \u03b1-helix labeled numerically. The either end of the missing fragment between helix \u03b111 and \u03b112 are indicated by dashed circles. (c) The Zn2+ ion-coordination residues in each ABA3 structure are displayed in sticks. The Zn2+ ions are shown in blue spheres. The dash lines measure the distance between residues and the metal ion and the length is shown in parentheses (unit, \u00c5).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Enzyme-ligand interaction networks in ABA3 complex structures. (a) The overall structure of RuABA3/FSPP (PDB ID, 8ZAE [http://doi.org/10.2210/pdb8ZAE/pdb]) and SkABA3/PPi (PDB ID, 8ZAG [http://doi.org/10.2210/pdb8ZAG/pdb]) presented in cartoon models. (b) The enzyme-ligand interaction networks in the complex structures are depicted, with protein residues shown in lines. The 2Fo-Fc electron density maps of FSPP, PPi, Mg2+ ions and coordinating waters contoured at 1.0 s are shown in mesh. Two views relative at the Y-axis by 180\u00b0 are presented. Dashed lines indicate distance < 3.5 \u00c5. The polder omit maps of the bound ligands are shown in Supplementary Fig. 11. In both panels, the bound ligands, Mg2+ ions, Zn2+ ions, and water molecules are shown in sticks, green spheres, blue spheres and red small spheres.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Substrate-binding pocket of ABA3 proteins. Apo-BcABA3 (PDB ID, 8ZAC [http://doi.org/10.2210/pdb8ZAC/pdb]), RuABA3/FSPP (PDB ID, 8ZAE [http://doi.org/10.2210/pdb8ZAE/pdb]) and SkABA3/PPi (PDB ID, 8ZAG [http://doi.org/10.2210/pdb8ZAG/pdb]) are superimposed. The overall protein of RuABA3/FSPP is depicted and shown in cartoon model. The residues that comprise substrate-binding pocket are shown in line (green, BcABA3; magenta, RuABA3/FSPP; cyan, SkABA3/PPi). FSPP and PPi are shown in sticks and metal ions are shown in spheres. Two Mg2+ ions observed in SkABA3/PPi are displayed as cyan spheres and denoted as Mg2+A and Mg2+B. The Mg2+A-corresponding ion in RuABA3/FSPP is presented as a magenta sphere.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "The roles of the active site residues in BcABA3-catalyzed reaction. (a) The PPi release activity of wild type and variant BcABA3. The relative activities of each sample were presented as percentages of the wild type. The individual and average values of each sample in the triplicate assay were presented in dots and bars, respectively. (b) The reaction mixtures of wild type and variant BcABA3were analyzed by GC-MS. The GC-MS chromatograms (m/z 133 and 148) of reaction mixture of each enzyme are displayed. The mass spectrum of (E)-\u03b2-farnesene is shown in Supplementary Fig. 16. (c) Two views depicting the substrate interaction network observed in RuABA3/FSPP related at Y-axis by 180\u00b0 are displayed. Amino acid residues, FSPP and metal ions are shown in lines, sticks and spheres. The hydrophilic interactions measured within 3.5 \u00c5 are shown in purple dashed lines. The Y96-mediated packing force is noted by a role of short strokes. The corresponding residues in BcABA3 are labeled by green letters. Some carbon atoms of FSPP are labeled by italic Arabic numbers. The numbers shown in parentheses indicate distance (unit, \u00c5). (d) The proposed BcABA3-catalyzed cyclization process. The substrate and reaction intermediates are colored in black, and the protein residues and pyrophosphate ion are colored in green. WT, wild type. Source data of Fig. 4a and 4b are provided as a Source Data file.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0986baca78f5b40819f004daf2eb76d7f4f4f10f91bb0b965eadad53992e8b0b/preprint/preprint.md

@@ -0,0 +1,159 @@
+# Abstract
+
+Terpenoid cyclases (TCs) account for the synthesis of the most widespread and diverse natural compounds. A sesquiterpene cyclase termed BcABA3 from an abscisic acid-producing fungus *Botrytis cinerea* that yields (2*Z*,4*E*)-a-ionylideneethane but lacks signature feature of canonical TCs represents a distinct type of TCs. Here, we report the crystal structures of BcABA3, a closely related RuABA3 from *Rutstroemia* sp. and a bacterial SkABA3 from *Shimazuella kribbensis*. These ABA3 proteins adopt an all-α-helix fold and bind pyrophosphate moiety of farnesyl pyrophosphate by Glu-chelated Mg²⁺ ion cluster. We conduct mutagenesis experiments to validate the role of the substrate-binding residues. SkABA3 appears to yield compounds that are distinct from (2*Z*,4*E*)-a-ionylideneethane. These results not only provide the molecular insight into ABA3 proteins that serve as an important basis to the future investigation of this class of TCs, but also reveal the existence of more uncharacterized terpenoids synthesized via dedicated machineries.
+
+[Biological sciences/Structural biology/X-ray crystallography](/browse?subjectArea=Biological%20sciences%2FStructural%20biology%2FX-ray%20crystallography) [Biological sciences/Biochemistry/Biocatalysis](/browse?subjectArea=Biological%20sciences%2FBiochemistry%2FBiocatalysis) [Biological sciences/Biochemistry/Enzyme mechanisms](/browse?subjectArea=Biological%20sciences%2FBiochemistry%2FEnzyme%20mechanisms) [Biological sciences/Biochemistry/Enzymes](/browse?subjectArea=Biological%20sciences%2FBiochemistry%2FEnzymes) [Biological sciences/Chemical biology/Natural products/Natural product synthesis](/browse?subjectArea=Biological%20sciences%2FChemical%20biology%2FNatural%20products%2FNatural%20product%20synthesis)
+
+# Introduction
+
+Terpenoids (or isoprenoids) are the most abundant natural products that more than 180,000 compounds from all kingdoms of life have been reported (http://terokit.qmclab.com/)¹. These structurally and functionally diverse compounds exhibit various physiological activities and many have found application potentials in biotechnological and pharmaceutical industries²,³,⁴. All terpenoids originate from common precursors comprising acyclic C₅ₙ isoprenoid pyrophosphates, which could be elongated or in most cases cyclized to afford various compounds.
+
+The cyclization of isoprenoid pyrophosphates is catalyzed by a large group of enzymes known as terpenoid cyclases (TCs). The action of TCs start from generating the initial carbocation and the subsequent cyclization cascade that afford mono- and polycyclic carbon skeletons that could contain multiple stereocenters. The strategy utilized for carbocation generation is used to classify TCs. Class I TCs utilize a trinuclear metal cluster to ionize isoprenoid pyrophosphates to yield allylic carbocations while class II TCs exploit a general acid (an aspartic acid) to protonate a terminal double bond of the substrate. In spite of the distinct starts of carbocation species and diverse substrate binding cavity that govern the cyclization steps, all known TCs possess the signature aspartate-rich motifs to bind metal ions in type I TCs (DDXXD/E and NSE/DTE) or serve as general acid (DXDD) in type II TCs⁵,⁶. These motifs are located on the upper wall of the substrate-binding pocket, which participate in the coordination of bivalent metal ions, mainly Mg²⁺ ion, and play an essential role in the binding and ionization of pyrophosphate moiety.
+
+Abscisic acid (ABA) is an important sesquiterpene phytohormone that plays a role in growth regulation and stress responses⁷,⁸,⁹,¹⁰. In plant, ABA is derived through an indirect pathway, in which the precursor C40 carotenoids are cleaved and oxidized to afford the final product (Supplementary Fig. 1)¹¹,¹². In phytopathogenic or phytosymbiotic fungi that produce ABA to modulate host-pathogen crosstalk and accelerate disease progression¹³,¹⁴, ABA is synthesized through C15 farnesyl pyrophosphate (FPP) cyclization followed by a series of oxidations (Supplementary Fig. 1)¹⁵,¹⁶. The gene cluster of microbial ABA biosynthetic pathway in an ABA-producing fungus *Botrytis cinerea* that contains four enzymes was illustrated¹⁷,¹⁸. Two putative P450 monooxygenases and a short-chain dehydrogenase/reductase that are respectively denoted as BcABA1, BcABA2 and BcABA4 should consecutively oxidize the FPP cyclization product (2Z,4E)-α-ionylideneethane (Supplementary Fig. 1). An open reading frame without showing significant similarity to known proteins termed BcABA3 was supposed to play a role as the sesquiterpene cyclase that catalyzes the committed step of FPP cyclization. However, BcABA3 shares very low sequence identity to characterized proteins and lacks the signature feature of aspartate-rich metal ion-binding motif used to define TCs.
+
+The role of BcABA3 as a TC had not been confirmed until 2018, when Takino *et al*. demonstrated that BcABA3 can catalyze the FPP cyclization to afford (2Z,4E)-α-ionylideneethane through both *in vitro* biochemical experiments and *in vivo* reconstitution assay. They found that BcABA3 is a metalloenzyme that demands the presence of Mg²⁺ ion to exert its activity. Notably, it is proposed that the cyclization course undergoes in an unprecedented fashion, which involves the formation of two neutral intermediates, β-farnesene and allofarnesene¹⁹. Moreover, BcABA3 homologous gene can be found in the genome of > 100 various microorganisms, including ABA-producing fungi and others that do not possess genes required to construct the microbial ABA biosynthesis and might behave otherwise.
+
+In this work, we solved the crystals structures of BcABA3, a closely related homolog of *Bc* ABA3 from *Rutstroemia* species (RuABA3) that also cyclizes FPP to yield (2Z,4E)-α-ionylideneethane and a bacterial ABA3 from *Shimazuella kribbensis* (SkABA3) to obtain the fundamental understanding about this class of TCs. More importantly, the crystal structures of RuABA3 in complex with a FPP analog farnesyl *S*-thiolodiphosphate (FSPP) and SkABA3 in complex with PPi were also solved. The roles of substrate- and Mg²⁺ ion-binding residues revealed in the complexes were validated via mutagenesis experiments. These results illustrate the overall structure and substrate-binding mode of a distinct class of TCs.
+
+# Results and Discussion
+
+## Characterization and crystallization of ABA3 proteins
+
+The recombinant protein of BcABA3 can be expressed in *E. coli* but no protein crystals were grown. We suspected that the crystallization process could be hampered by the low protein stability as less than 10% protein remained intact after incubated at crystallization temperature (22°C) for two days (Supplementary Fig. 2). To increase the probability of obtaining the structural information of BcABA3, we turned to two homologous genes from *Rutstroemia* sp. and *C. inanum* (Supplementary Table 1). We obtained the recombinant protein of RuABA3, which shares 86.2% sequence identity to BcABA3 and can convert FPP to product that shares identical features of (2*Z*,4*E*)-α-ionylideneethane (Supplementary Fig. 3).
+
+As shown in Supplementary Fig. 2, the recombinant protein of RuABA3 is as labile as BcABA3 and did not grow any crystals. According to the protein sequence alignment shown by Takino *et al.*, the N-terminal portion among BcABA3 and homologs is highly diverse <sup>19</sup>. We thus decided to express N-terminal region-truncated ABA3 proteins, and eventually obtained crystals of BcABA3-Δ59 and RuABA3-Δ65. In the following contents, the BcABA3-Δ59 and RuABA3-Δ65 are designated as BcABA3 and RuABA3, and the full-length proteins are noted as BcABA3-FL and RuABA3-FL for simplicity. Notably, the N-terminus truncation does not influence the enzyme activity as BcABA3 and RuABA3 exhibit similar catalytic rate to their full-length counterparts and can produce (2*Z*,4*E*)-α-ionylideneethane (Supplementary Fig. 3).
+
+## Overall structure of ABA3 proteins
+
+The crystal structures of apo-BcABA3 and apo-RuABA3 (Figs. 1a and 1b, Supplementary Fig. 4a and Supplementary Table 4), which are highly identical to each other (Cα RMSD value of 0.335 Å, Supplementary Fig. 4a), contain two and one polypeptide chains in an asymmetric unit, respectively. The two polypeptides in the apo-BcABA3 structure form a homodimer that is arranged in a compact configuration (Fig. 1a). The same dimerization status was also formed by RuABA3, in which the equivalent dimeric counterpart was identified from a symmetric molecule (Fig. 1a). This was validated by size exclusion chromatographic (SEC) analyses, which indicate that both ABA3 proteins, with their N-terminus truncated or not, exist as a dimer in solution (Supplementary Fig. 5). Both structures adopt an all-α-helix fold and have a missing fragment between helix α11 and α12 (E326-V351 in apo-BcABA3, S327-N353 in apo-RuABA3) (Fig. 1b and Supplementary Fig. 4a).
+
+The searching for homologous structures with the DALI server indicates that all closely related structures of BcABA3 are canonical TCs (Supplementary Table 5). The N-terminal part of apo-BcABA3 folds into helix α1 to α10 in a similar topology as do helix α2 to α12 in canonical TCs such as CotB2 <sup>20</sup>, aristolochene synthase <sup>21</sup>, labdane synthase <sup>22</sup> and selinadiene synthase <sup>23</sup>, while the C-terminal part contains four additional helices (α11 to α14) (Supplementary Fig. 6). Notably, a metal ion coordinated in a tetrahedral geometry comprising four cysteines was identified on the extensive loop α13-α14 (Fig. 1c and Supplementary Fig. 7), which resembles the Zn<sup>2+</sup> ion-coordination in other known structures <sup>24, 25</sup>. The presence of Zn<sup>2+</sup> ion in BcABA3 and RuABA3 has also been confirmed by atomic spectrometric analysis (Supplementary Table 6).
+
+The Zn<sup>2+</sup> ion-binding motif is located distant from the active center (see below), thus was not supposed to involve in the catalytic reaction. Unexpectedly, substituting the Zn<sup>2+</sup> ion-coordinating residues, except for C410, with Ala leads to the reduction in enzyme activity (Supplementary Fig. 8a). All variants lose Zn<sup>2+</sup> ion-binding capacity (Supplementary Fig. 8b) while remain the same composition of secondary structure (Supplementary Fig. 8c), indicating that the loss of the metal ion does not significantly impact the overall protein fold. Notably, the loop that houses the Zn<sup>2+</sup> ion-binding motif is lengthy and extends to the top of the substrate-binding pocket (Supplementary Fig. 8d). Thus, changing the Zn<sup>2+</sup> ion-binding residues might affect the loop localization and influence the integrity of the architecture of the active center. Intriguingly, Takino *et al.* demonstrated that a natural variant termed BcABA3-S that harbors a shorter 3’-terminal region cannot produce (2*Z*,4*E*)-α-ionylideneethane (Supplementary Fig. 8d) <sup>19</sup>, indicating the intact C-terminal region of BcABA3 should be required to the enzyme reaction. In this context, we propose that residues C384, C387 and C407 might form interactions to each other or nearby residues for local stabilization, such that variant C410A can still be fully active. Further structural investigations of these variants are in progress to verify these hypotheses.
+
+## Structure of a bacterial ABA3
+
+As demonstrated in the previous study, the BcABA3 homologous gene is widespread in numerous fungal strains with some documented as abscisic acid-producing strains <sup>19, 26, 27, 28, 29</sup>. In addition, some actinomycete bacteria also possess BcABA3 homologous gene in their genome, which very likely also function as sesquiterpene cyclases considering the degree of identity to BcABA3 (Supplementary Table 7). We thus proceeded to express the recombinant protein of a bacterial homologue of BcABA3 from *Shimazuella kribbensis* termed SkABA3 and solved its crystal structure (Supplementary Table 4). It is noteworthy that the recombinant protein of SkABA3 was much more stable than BcABA3 and RuABA3, such that the recombinant protein remained intact within four days (Supplementary Fig. 2).
+
+SkABA3 adopts an overall fold similar to the other two ABA3 proteins (Cα RMSD value, 0.783 and 0.914 Å), and two out of three polypeptides in an asymmetric unit also form the same homodimeric configuration (Fig. 1a and Supplementary Fig. 9). The SEC analysis also suggests the dimeric configuration of *Sk* ABA3 in solution (Supplementary Fig. 5). Although the full-length recombinant protein was used for crystallization, no electron density maps for the first 20 amino acids on the N-terminus can be observed such that the overall structure of SkABA3 starts from a loop prior to helix α1 as that in apo-BcABA3 and apo-RuABA3 (Supplementary Fig. 4). Compared with apo-BcABA3 and apo-RuABA3, an additional helix following α11 termed α11’ can be observed in apo-SkABA3 structure (Fig. 1a and Supplementary Fig. 4b). As in the other two ABA3 structures, there is also a missing fragment between helix α11’ and α12 (K298-E312) (Supplementary Fig. 4b). A metal ion located in the Zn<sup>2+</sup> ion-site in BcABA3 and RuABA3 was observed in apo-SkABA3 albeit the coordinating residues comprise three cysteines and a serine (Fig. 1c and Supplementary Fig. 7). We also detected the existence of Zn<sup>2+</sup> ion in the recombinant protein of SkABA3 (Supplementary Table 6), thus this metal ion was modeled with a Zn<sup>2+</sup> ion.
+
+Judged from PPi release assay, SkABA3 can catalyze FPP ionization to release inorganic PPi (Supplementary Fig. 3a). The GC-MS analyses indicated that SkABA3 did not produce (2*Z*,4*E*)-α-ionylideneethane but yielded a compound that was eluted at various time (7.65 min for (2*Z*,4*E*)-α-ionylideneethane, 8.36 min for SkABA3 product) (Supplementary Fig. 10). SkABA3 should be able to convert FPP to an isomer of (2*Z*,4*E*)-α-ionylideneethane as a fragment ion at *m*/*z* 204.20 can be identified but no hit can be found from searching the spectral library (Supplementary Fig. 10b). These analyses indicate that the main product of SkABA3 should be an unprecedented sesquiterpenoid compound, and the research on resolving its structure is now in progress in our laboratory.
+
+## Substrate-binding mode of ABA3
+
+Next, we aimed to investigate the substrate-binding mode of ABA3 and conducted crystal-soaking trials to obtain complex structure. After great efforts, we solved the crystal structures of RuABA3 in complex with a FPP analog FSPP and SkABA3 in complex with PPi (Supplementary Fig. 11 and Supplementary Table 4). In both complexes, ligands were bound in a central pocket with their PPi moiety occupying the identical position (Figs. 2a and 3). The upper part of the substrate-binding cavity that comprises several polar residues accounts for the binding of PPi, and the lower part that is lined mainly with hydrophobic residues accommodates the hydrocarbon part of FSPP (Figs. 2b and 3). The PPi-interaction network is strictly conserved while the hydrocarbon-interacting residues are more diverse among ABA3 homologues (Fig. 3 and Supplementary Fig. 12). BcABA3 and RuABA3 might contact to the hydrocarbon chain of FPP with several aromatic residues including Y96, F167 and Y242. These residues might provide cation-π interactions for electrostatic stabilization of multiple transition states in the terpenoid cyclization cascade, as that has been proposed for many TCs.
+
+Two metal ions adjacent to PPi in the SkABA3/PPi complex were observed (Fig. 2). Based on the octahedral six-coordinate geometry and the fact that Mg<sup>2+</sup> ion is required for the reactivity of BcABA3 <sup>19</sup>, these metal ions were modeled with Mg<sup>2+</sup> ions and denoted as Mg<sup>2+</sup><sub>A</sub> and Mg<sup>2+</sup><sub>B</sub> (Fig. 2a). The RuABA3/FSPP complex only contains one Mg<sup>2+</sup> ion, which was assigned as Mg<sup>2+</sup><sub>A</sub> according to its location relative to the Mg<sup>2+</sup> ions in the SkABA3/PPi complex (Figs. 2a and 3). A strictly conserved glutamate (E124, E126 and E80 in BcABA3, RuABA3 and SkABA3, respectively) and several water molecules, instead of the signature aspartic acid-rich motifs in the canonical TCs, account for the Mg<sup>2+</sup> ion coordination (Figs. 2b, 3 and Supplementary Fig. 12). Neither the Mg<sup>2+</sup><sub>C</sub>-corresponding site nor the Mg<sup>2+</sup><sub>C</sub>-binding NSE/DTE motif-comprising residues in the canonical TCs was found in RuABA3/FSPP and SkABA3/PPi (Supplementary Fig. 13). Instead, the location of NSE/DTE motif is occupied by Y200, R161 and H208 (Y242, R202 and H250 in BcABA3) (Supplementary Fig. 13). Notably, the RY pair comprising R271 and Y272 (R312 and Y313 in BcABA3) that also accounts for the PPi interaction in canonical TCs was identified in the similar location (Supplementary Fig. 13) <sup>30</sup>.
+
+While the apo-form and complex structures of RuABA3 and SkABA3 share similar overall fold (Cα RMSD ~ 0.3 Å), we noticed that helix α11’ in SkABA3 was rotated by ~ 90° in the PPi-containing complex (Supplementary Fig. 14). In addition, the loop region following helix α11’ that traverses to cover the substrate-binding pocket was also revealed (Supplementary Fig. 14). The substrate-binding induced conformational change in the active site is common in canonical TCs, which results in the positioning of metal ion-binding motifs <sup>21, 23, 31, 32</sup>. The “capping” is also considered to secure the highly reactive carbocation intermediate from quenching by the bulk solvent prior to the end of the reaction <sup>5</sup>. In SkABA3/PPi complex, no direct contact between the cap region and bound ligands was observed, yet the active site capping might still serve to protect the reactive intermediates.
+
+## The roles of the active site residues of BcABA3
+
+The structural information enables us to conduct mutagenesis experiments to investigate the role of the substrate-binding residues (Fig. 4). In general, substituting PPi moiety- and Mg<sup>2+</sup> ion-binding residues with Ala exhibit very low PPi release activity and generate no product (Figs. 4a and 4b). The only exception is H250A that retains ~ 48% PPi ionization activity and produces significant amounts of (2*Z*,4*E*)-α-ionylideneethane (~ 28% relative to the wild type enzyme). Thus, H250 might make less contribution to the binding of FPP compared with other residues that interact with PPi. CD analysis indicated that all variants shared the same secondary structures composition to the wild type protein, suggesting that the reduced activity is not resulted from the alternation of the overall protein structure (Supplementary Fig. 15).
+
+For the hydrocarbon-interacting residues, Y96A exhibits very low PPi release activity and produces no product (Figs. 4a and 4b). However, Y96F retains ~ 40% PPi release activity and still can produce (2*Z*,4*E*)-α-ionylideneethane (~ 29% relative to the wild type enzyme), suggesting that the aromatic ring but the hydroxyl group is the functional moiety at this position. As the phenolate side chain of Y96 parallels the C4-C5 bond (distant by 3.6 Å), it is possible that this residue might provide packing forces to stabilize the reaction intermediate such as carbocation or allofarnesene as previously proposed (Fig. 4c). I121 is supposed only minimally related to the catalytic reaction as variant I121A can still catalyze PPi ionization and product formation. This is not surprising as this residue is located farther from the substrate than the other residues and is not strictly conserved among ABA3 homologs (Supplementary Fig. 12). Notably, variants Y96A, I117A, F167A, L207A and F302A produce trace amount of (*E*)-β-farnesene, the putative reaction intermediate, and no product (Fig. 4b and Supplementary Fig. 16). Interestingly, these variants are still capable of conducting FPP ionization, suggesting that these variants might play a role after (*E*)-β-farnesene is formed. Since these hydrophobic residues cannot protonate and deprotonate, they should solely participate in the forming of the hydrocarbon-binding pocket. It appears that altering the contour the hydrocarbon-binding pocket poses significant effects to the late-stage cyclization process. This principle also applies to many terpene cyclases, such that modifying the pocket-forming residues could change the product profile <sup>33, 34, 35</sup>.
+
+Based on the above results, we proposed several critical residues that might participate in each step in the previously reported reaction process <sup>19</sup>. Takino *et al.* analyzed the reaction product converted from a series of site-specifically deuterium-labeled FPPs and demonstrated that H<sub>S</sub> at C4, H<sub>R</sub> at C8, an undefined H at C5 and all H at C13 are retained in the product. Notably, the hydrogens on C1 and C12 are partially lost based on H-D exchange occurred in the D<sub>2</sub>O-containing reaction. Together with the fact that (*E*)-β-farnesene and allofarnesene can be converted to (2*Z*,4*E*)-α-ionylideneethane by BcABA3, a cyclization process was proposed (Fig. 4d). The cleavage of pyrophosphate occurs first, which should be catalyzed by E124 that coordinates the Mg<sup>2+</sup> ions. The resulting allylic cation <b>A1</b><sup><b>+</b></sup> is subsequently deprotonated to afford (*E*)-β-farnesene. Y313 or the ionized pyrophosphate anion are within H-bond distance to C12 and are potential deprotonation agents (Figs. 4c and 4d). Next, (*E*)-β-farnesene is supposed to be protonated on C1 and C12 to yield <b>A2</b><sup><b>+</b></sup> and <b>B</b><sup><b>+</b></sup>, respectively. The possible deprotonation agents near C1 and C12 include Y313, pyrophosphoric acid and R202 (Figs. 4c and 4d). Substituting Y313 and R202 with Ala abolished the product generation, suggesting that these two residues are essential for the enzyme activity. <b>A2</b><sup><b>+</b></sup> and <b>B</b><sup><b>+</b></sup> are supposed to be deprotonated at C5 accompanied by the transfer of the hydride ion from C4 to C1 to yield allofarnesene. Here, Y313 is the sole candidate to serve as a base to abstract the H at C5. Next, the allofarnesene is supposed to be protonated on C10 to yield C<sup>+</sup>. No candidate residue that is competent to serve this function can be identified. One possibility is that polarized water molecules that are not revealed in the current structures acts to practice protonation. Otherwise, C10 could undergo a displacement during the reaction and is relocated to alternative location where contains a functional base. Finally, the ring closure between C6 and C11 accompanies by deprotonation of H<sub>R</sub> at C8 as that has been proposed to form the ε-ring in lycopene <sup>36</sup>. R312 that is located within the H-bond distance to C8 is a potential candidate of deprotonation agent, resembling the case that has been reported in other types of enzymes <sup>37</sup>. The importance of Y313, R202 and R312 is illuminated by mutagenesis experiments, as their Ala substituents exhibit very low PPi release activity and produce no product (Figs. 4a and 4b). However, these residues provide interaction for the binding of PPi moiety of FPP, thus their definitive roles in the abovementioned protonation/deprotonation steps should demand more experiments to confirm.
+
+In conclusion, the structural analyses of a distinct type of TCs are demonstrated, which illustrate their overall folding and substrate-binding behaviors. These TCs were tentatively named ABA3 after BcABA3, a fungal enzyme that catalyzes the cyclization of FPP to construct the carbon skeleton of an important phytohormone abscisic acid. As RuABA3 that is highly identical to BcABA3 should catalyze the same reaction and produce (2*Z*,4*E*)-α-ionylideneethane, a bacterial member SkABA3 appears to convert FPP to different products. One feature of ABA3 proteins is the Zn<sup>2+</sup>-binding motif, which is located at a site remote to the substrate-binding site but might stabilize an extensive loop that stretches to the active center. Additionally, ABA3 proteins employ a strictly conserved Glu, instead of the Asp-rich motifs in canonical TCs, to coordinate the Mg<sup>2+</sup> ion cluster to facilitate the FPP ionization. Four basic residues (R202, K249, H250 and R312 in BcABA3) and two Tyr residues (Y242 and Y313) join the Mg<sup>2+</sup> ion cluster to form the molecular recognition motif for the PPi moiety of the substrate. Among them, the RY pair comprising R312 and Y313 is consistently seen in type I TCs. Notably, the RY pair in BcABA3 also form interactions to the hydrocarbon part of FPP and might play a role in the product production process. These results shed lights on the molecular insights into ABA3 proteins and provide important fundamental information to guide the future investigations into this distinct class of TCs. Furthermore, the identification of more uncharacterized terpenoids and their biosynthetic machineries could also be envisioned.
+
+# Methods
+
+## Plasmid construction and mutagenesis
+
+The genes encoding BcABA3 (GenBank accession No., XP_024550392.1), BcABA3 homologs from *Rutstroemia* sp. (RuABA3, GenBank accession No., PQE10596.1; protein sequence identity to BcABA3, 86.2%), *Colletotrichum inanum* (GenBank accession No., KZL82383.1; protein sequence identity to BcABA3, 55.8%), and *Shimazuella kribbensis* (SkABA3, GenBank accession No. WP_051272304.1, protein sequence identity to BcABA3, 36.3%) were synthesized by Sangon Biotech Co., Ltd (Shanghai, China) and cloned into the pET46Ek/LIC plasmid for recombinant protein expression (Supplementary Table 1). The variant ABA3-expressing plasmids were constructed by PCR-based site-directed mutagenesis with oligonucleotides listed in Supplementary Table 2. All plasmids were verified by direct sequencing.
+
+## Recombinant protein expression and purification
+
+The recombinant plasmids were transformed into *E. coli* BL21 (DE3) competent cells and cultured in Luria–Bertani medium containing 100 µg mL⁻¹ ampicillin at 37°C to an optical density at 600 nm (OD600) of ~0.6. The protein expression was induced by 0.2 mM isopropyl β-D-thiogalactopyranoside at 16°C for 20 h. Cells were harvested by centrifugation at 8,000 × g for 10 minutes and suspended in the buffer A containing 25 mM Tris-HCl (pH 7.5), 150 mM NaCl and 20 mM imidazole, followed by disruption with a French press (JuNeng biology & Technology Co., Ltd., Guangzhou, China). Cell lysate was centrifuged at 4°C for 50 min at 17,000 × g to remove cell debris. The supernatant was applied to a fast protein liquid chromatography (FPLC) system (GE Healthcare, Piscataway, NJ) coupled with a Ni-NTA column and eluted with a gradient of 20–500 mM imidazole in buffer A. The target protein-containing fractions were collected and dialyzed against buffer A that contains tobacco etch virus (TEV) protease to remove the His-tag at 4°C overnight. Afterward, the mixture was passed through Ni-NTA column again for collecting the untagged protein. The protein was concentrated by ultrafiltration tubes with a molecular weight cut-off of 10 kDa (Merck Millipore, Billerica, MA) and the purity was analyzed by SDS-PAGE. The protein yields for three ABA3 are similar, ranging from 15 to 20 mg L⁻¹. The N-terminal truncation and point mutation did not reduce the protein yield except for the Ala substituents of the Zn²⁺ ion-coordination residues, whose yields were reduced to 2 to 5 mg L⁻¹.
+
+## Crystallization and structure determination
+
+The crystallization was performed at 22 ℃ using the sitting-drop vapour-diffusion method. The protein solution contains 10 mM DTT and 10 mM MgCl₂ was mixed with the reservoir solution at a ratio of 1:1 in 48-well Cryschem plates, and equilibrated against 100 µL of the reservoir solution. Generally, 1 µL of protein solution and 1 µL of reservoir solution were mixed. The crystallization conditions and cryoprotectants of each dataset are shown in Supplementary Table 3. Before data collection, crystals were dipped with cryoprotectants. For complex crystal preparation, crystals were soaked in cryoprotectants containing 10 mM farnesyl *S*-thiolodiphosphate (FSPP) or pyrophosphate (PPi) for 10 min. X-ray diffraction datasets were collected at beamlines TPS-05A of the National Synchrotron Radiation Research Center (NSRRC) and processed and scaled by HKL-2000³⁸ and on a BRUKER D8 VENTURE X-ray diffractometer at Hubei University and processed with PROTEUM3 v2020.6. A RuABA3 model predicted by AlphaFold2³⁹ was used as a template to solve the crystal structure of apo-RuABA3 by molecular replacement (MR). Structural adjustment and refinement were performed using Refmac5 and Coot in CCP4i suit⁴⁰,⁴¹,⁴². Prior to structural refinement, 5% randomly selected reflections were set aside for calculating *R*<sub>free</sub> as a monitor of model quality. The crystal structures of apo-BcABA3 and apo-SkABA3 were solved with MR by using the apo-RuABA3 structure as the searching model. The obtained apo-form structures were utilized as a searching template to solve respective complex structures. All structures were refined as abovementioned. Structural graphs were depicted by PyMOL program (http://pymol.sourceforge.net/).
+
+## Size exclusion chromatography analysis
+
+The molecular mass of ABA3 proteins in solution was estimated through size exclusion chromatography (SEC) analysis by using FPLC system coupled with a Superdex 200 10/300 GL column. 100 µL protein in a buffer of 25 mM Tris-HCl (pH 7.5) and 150 mM NaCl was loaded onto the column and eluted at a flow rate of 0.2 mL min⁻¹. The protein molecular mass was calculated based on the standard curve built by protein markers with known molecular weight.
+
+## Reaction product analysis by gas chromatography-mass spectrometer
+
+The reaction products of ABA3 proteins were analyzed by using gas chromatography-mass spectrometer (GC-MS). Reaction mixture (500 µL) containing 200 µM FPP, 10 mM DTT, 5 mM MgCl₂ and 20 µM protein in Tris-HCl buffer (pH 7.5) was incubated at 30 ℃ for 5 hr and then quenched by adding 200 µL hexane. The mixture was vortexed and centrifuged at 12,000 × g to remove the pellet. The supernatant was analyzed on a Shimadzu GCMS-TQ8050 NX (Shimadzu Corporation, Japan) equipped with a flame-ionization detector and a HP-5MS column (0.25 mm × 30 m, 0.25 µm). Sample was injected into the column at 100°C in the splitless mode. The column temperature is maintained at 100°C for 3 min and increased to 268°C at the rate of 14°C min⁻¹. The flow rate of the helium carrier gas was 0.66 mL min⁻¹. Reactions without containing enzyme serve as negative control. All the mass spectra were recorded applying the following conditions: ion source, 200°C; scan interval, 0.3 s; scan speed, 2000 amu/s; scan range, 45 to 600 (*m*/*z*). All data were analyzed using GCMSsolution version 4.53SP1. The results of one reaction of each protein was displayed (*n* = 1), while at least three independent trials that yielded similar results have been conducted.
+
+## Atomic absorption spectrometry
+
+2 mg of the purified recombinant protein of ABA3 protein was added to 0.3 mL 69% nitric acid, heated until completely dissolved and diluted to 10 mL by ddH₂O. The samples were analyzed by using ICP-AES optimal 8000 (PerkinElmer, Waltham, MA, USA) and detected for Zn²⁺ (λ = 206.200 nm), Mg²⁺ (λ = 285.213 nm) and Ni²⁺ (λ = 231.604 nm).
+
+## Enzyme activity measurement by PPi release assay
+
+The catalytic efficiency of wild type and variant ABA3 proteins was measured with a pyrophosphate (PPi) release assay as previously described⁴³,⁴⁴,⁴⁵. The reaction mixture contains 140 µM 2-amino-6-mercapto-7-methyl purine ribonucleoside, 4 µg mL⁻¹ purine nucleoside phosphorylase and 4 µg mL⁻¹ inorganic pyrophosphatase, 50 mM Tris-HCl (pH 7.5), 1 mM MgCl₂ and 0.1 mM sodium azide were aliquot into a 96-well plate. 100 µM FPP and 2.4 µM enzyme were added into the reaction mixture, and the released inorganic phosphate formation was measured by the absorbance at 360 nm using the SpectraMax i3x microplate reader (Molecular Devices, Sunnyvale, CA). All assays were performed in triplicate. The change in OD360 within five min was recorded (ΔOD360) to represent the enzyme activity and presented as a percentage of wild type BcABA3 activity as following.
+
+Equation (1)
+
+(ΔOD360<sub>sample</sub> – ΔOD360<sub>no enzyme</sub>) / (ΔOD360<sub>WT</sub> – ΔOD360<sub>no enzyme</sub>) × 100%
+
+## Circular dichroism spectroscopy
+
+0.5 mg mL⁻¹ wild type or variant BcABA3 in PBS buffer was placed in a quartz cuvette and scanned from 200 to 260 nm with a JASCO J-810 spectrometer (JASCO Co., Ltd., Japan). Scans were performed immediately at 25°C and collected at a scanning speed of 200 nm min⁻¹ with a 1 s response time.
+
+# References
+
+1. Zeng T et al (2020) TeroKit: A database-driven web server for terpenome research. J Chem Inf Model 60:2082–2090
+2. Pichersky E, Raguso RA (2018) Why do plants produce so many terpenoid compounds? New Phytol 220:692–702
+3. Rudolf JD, Alsup TA, Xu B, Li Z (2021) Bacterial terpenome. Nat Prod Rep 38:905–980
+4. Yang W, Chen X, Li Y, Guo S, Wang Z (2020) Yu, X. Advances in pharmacological activities of terpenoids. Nat Prod Commun 15:1–13
+5. Christianson DW (2017) Structural and chemical biology of terpenoid cyclases. Chem Rev 117:11570–11648
+6. Oldfield E, Lin FY (2012) Terpene biosynthesis: modularity rules. Angew Chem Int Ed Engl 51:1124–1137
+7. Ding Z, De, Smet I (2013) Localised ABA signalling mediates root growth plasticity. Trends Plant Sci 18:533–535
+8. Dörffling K (2015) The discovery of abscisic acid: A retrospect. J Plant Growth Regul 34:795–808
+9. Shu K, Liu X, Xie Q, He Z (2016) Two faces of one seed: Hormonal regulation of dormancy and germination. Mol Plant 9:34–45
+10. Ton J, Flors V, Mauch-Mani B (2009) The multifaceted role of ABA in disease resistance. Trends Plant Sci 14:310–317
+11. Nambara E, Marion-Poll A (2005) Abscisic acid biosynthesis and catabolism. Annu Rev Plant Biol 56:165–185
+12. Wasilewska A et al (2008) An update on abscisic acid signaling in plants and more. Mol Plant 1:198–217
+13. Cao FY, Yoshioka K, Desveaux D (2011) The roles of ABA in plant-pathogen interactions. J Plant Res 124:489–499
+14. Lievens L, Pollier J, Goossens A, Beyaert R, Staal J (2017) Abscisic acid as pathogen effector and immune regulator. Front Plant Sci 8:587
+15. Inomata M, Hirai N, Yoshida R, Ohigashi H (2004) The biosynthetic pathway to abscisic acid via ionylideneethane in the fungus *Botrytis cinerea*. Phytochemistry 65:2667–2678
+16. Inomata M, Hirai N, Yoshida R, Ohigashi H (2004) Biosynthesis of abscisic acid by the direct pathway via ionylideneethane in a fungus, *Cercospora cruenta*. Biosci Biotechnol Biochem 68:2571–2580
+17. Siewers V, Kokkelink L, Smedsgaard J, Tudzynski P (2006) Identification of an abscisic acid gene cluster in the grey mold *Botrytis cinerea*. Appl Environ Microbiol 72:4619–4626
+18. Siewers V, Smedsgaard J, Tudzynski P (2004) The P450 monooxygenase BcABA1 is essential for abscisic acid biosynthesis in *Botrytis cinerea*. Appl Environ Microbiol 70:3868–3876
+19. Takino J et al (2018) Unveiling biosynthesis of the phytohormone abscisic acid in fungi: Unprecedented mechanism of core scaffold formation catalyzed by an unusual sesquiterpene synthase. J Am Chem Soc 140:12392–12395
+20. Tomita T et al (2017) Structural insights into the CotB2-catalyzed cyclization of geranylgeranyl diphosphate to the diterpene cyclooctat-9-en-7-ol. ACS Chem Biol 12:1621–1628
+21. Chen M et al (2013) Mechanistic insights from the binding of substrate and carbocation intermediate analogues to aristolochene synthase. Biochemistry 52:5441–5453
+22. Centeno-Leija S et al (2019) The structure of (*E*)-biformene synthase provides insights into the biosynthesis of bacterial bicyclic labdane-related diterpenoids. J Struct Biol 207:29–39
+23. Baer P et al (2014) Induced-fit mechanism in class I terpene cyclases. Angew Chem Int Ed Engl 53:7652–7656
+24. Li J, Qian X, Sha B (2003) The crystal structure of the yeast Hsp40 Ydj1 complexed with its peptide substrate. Structure 11:1475–1483
+25. Tagliani A et al (2021) Structural and functional insights into nitrosoglutathione reductase from *Chlamydomonas reinhardtii*. Redox Biol 38:101806
+26. Assante G, Merlini L, Nasini G (1977) (+)-Abscisic acid, a metabolite of the fungus *Cercospora rosicola*. Experientia 33:1556–1557
+27. Dörffling K, Petersen W, Sprecher E, Urbasch I, Hanssen H-P (1984) Abscisic acid in phytopathogenic fungi of the genera *Botrytis*, *Ceratocystis*, *Fusarium*, and *Rhizoctonia*. *Z. Naturforsch. C* 39, 683–684
+28. Marumo S, Katayama M, Komori E, Ozaki Y, Natsume M, Kondo S (1982) Microbial production of abscisic acid by *Botrytis cinerea*. Agric Biol Chem 46:1967–1968
+29. Masi M, Meyer S, Cimmino A, Clement S, Black B, Evidente A (2014) Pyrenophoric acids B and C, two new phytotoxic sesquiterpenoids produced by *Pyrenophora semeniperda*. J Agric Food Chem 62:10304–10311
+30. Dickschat JS (2016) Bacterial terpene cyclases. Nat Prod Rep 33:87–110
+31. Aaron JA, Lin X, Cane DE, Christianson DW (2010) Structure of epi-isozizaene synthase from *Streptomyces coelicolor* A3(2), a platform for new terpenoid cyclization templates. Biochemistry 49:1787–1797
+32. Driller R et al (2018) Towards a comprehensive understanding of the structural dynamics of a bacterial diterpene synthase during catalysis. Nat Commun 9:3971
+33. Li J-X et al (2013) Rational engineering of plasticity residues of sesquiterpene synthases from *Artemisia annua*: Product specificity and catalytic efficiency. Biochem J 451:417–426
+34. Li R et al (2014) Reprogramming the chemodiversity of terpenoid cyclization by remolding the active site contour of *epi*-isozizaene synthase. Biochemistry 53:1155–1168
+35. Zhang F, An T, Tang X, Zi J, Luo H-B, Wu R (2020) Enzyme promiscuity versus fidelity in two sesquiterpene cyclases (TEAS versus ATAS). ACS Catal 10:1470–1484
+36. Cunningham FX Jr., Pogson B, Sun Z, McDonald KA, DellaPenna D, Gantt E (1996) Functional analysis of the beta and epsilon lycopene cyclase enzymes of *Arabidopsis* reveals a mechanism for control of cyclic carotenoid formation. Plant Cell 8:1613–1626
+37. Guillén SYV, Hedstrom L (2005) A twisted base? The role of arginine in enzyme-catalyzed proton abstractions. Arch Biochem Biophys 433:266–278
+38. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in oscillation mode. Meth Enzymol 276:307–326
+39. Jumper J et al (2021) Highly accurate protein structure prediction with AlphaFold. Nature 596:583–589
+40. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr D: Struct Biol 53:240–255
+41. Emsley P, Cowtan K (2004) Coot: model-building tools for molecular graphics. Acta Crystallogr D: Struct Biol 60:2126–2132
+42. Potterton E, Briggs P, Turkenburg M, Dodson E (2003) A graphical user interface to the CCP4 program suite. Acta Crystallogr D: Struct Biol 59:1131–1137
+43. Chen C-C et al (2021) Terpene cyclases and prenyltransferases: Structures and mechanisms of action. ACS Catal 11:290–303
+44. Gao J et al (2018) Head-to-middle and head-to-tail *cis*-prenyl transferases: structure of isosesquilavandulyl diphosphate synthase. Angew Chem Int Ed Engl 57:683–687
+45. Webb MR (1992) A continuous spectrophotometric assay for inorganic phosphate and for measuring phosphate release kinetics in biological systems. *Proc. Natl. Acad. Sci. U. S. A.* 89, 4884–4887
+
+# Supplementary Files
+
+- [20241127AuthorChecklist.docx](https://assets-eu.researchsquare.com/files/rs-4382230/v1/7e3a6ed9da3ae055e80c6b98.docx)  
+  Author checklist
+
+- [SourceData.xlsx](https://assets-eu.researchsquare.com/files/rs-4382230/v1/0c452b440f8da3f50101584d.xlsx)  
+  Dataset Fig. 4a, Dataset Fig. 4b, Dataset Fig. s2, Dataset Fig. s3a, Dataset Fig. s3b, Dataset Fig. s8a, Dataset Fig. s10a
+
+- [rs.pdf](https://assets-eu.researchsquare.com/files/rs-4382230/v1/f600788b55e907eaae3a1037.pdf)  
+  Reporting Summary
+
+- [SI.pdf](https://assets-eu.researchsquare.com/files/rs-4382230/v1/fe691be5f60034898a7fa9b9.pdf)  
+  Supplementary Information

0a6b8025fe63ede42bbc75764a31f72c0e5e57a292ccb22e0161a2c8d90351c2/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Structure of the NavAb channel and HS-AFM sample, and the voltage-dependencies of the NavAb constructs.\n(A) Schematic diagram of the NavAb channel. (B) Crystallographic structure of the NavAb tetramer viewed from extracellular side (PDB: 4EKW). The black and white arrowheads in (A) and (B) indicate the S1-S2 and S3-S4 loops, respectively. (C) Schematic illustration of the HS-AFM imaging of the NavAb channel in a lipid bilayer. The NavAb channel with a His-tag at the C-terminal is attached to the Ni2+-coated mica surface and reconstituted into a lipid bilayer. The AFM tip was used to scan the extracellular surface of the NavAb channel. (D) I-V curves of the NavAb constructs used in this study.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "HS-AFM imaging of the NavAb channel in a lipid bilayer.\n(A, E, I and L) HS-AFM snapshots of the NavAb channel in a lipid bilayer. The black and white arrowheads in (A, E and L) indicate particles derived from the pairs of helices of S1 and S2, and S3 and S4, respectively. Frame rates: 10 frame/sec for A and E, 5 frame/sec for I and L. (B, F, J and M) Time-averaged images of HS-AFM movies A, E, I, and L, respectively. (C, G, K and N) LAFM images of HS-AFM movies A, E, I, and L, respectively. The total frame numbers used for B and C, F and G, J and K, and M and N are 458, 599, 65, and 97, respectively. The used constructs are the WT for (A to C), the N49K mutant for (E to G) and the E32Q/N49K mutant for (I to N). (D, H and O) Schematic illustration of the observed structure of the NavAb channels.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Coupled motions of the PDs and VSDs.\n(A, B and C) Trajectories of NavAb channels. The dark blue, dark green, red and orange lines correspond to the trajectories of the PDs. The light blue, light green, pink and light orange lines correspond to the trajectories of the S1-S2 particles. The trajectories of the particles derived from S3-S4 are not shown for simplicity of presentation. (D, E and F) Time series of the distance between the PD and neighboring VSD (S1 and S2) indicated as colored arrows in A, B and C. (G, H and I) Histograms of the distance between the PD and neighboring VSD (S1-S2 particle). The numbers of pairs (channels) are 24 (6), 25 (7) and 4 (1) for G, H and I, respectively. The total data numbers are 5727, 8231 and 747, respectively. (J, K and L) Primary components of coupled motion calculated by PCA. The dots and lines indicate eigenvectors of coupled motion, with the dots indicating the origin of vectors. The colors of the vectors are the same as their corresponding trajectories (A, B and C). The analyzed constructs are the WT for A, D, G and J, the N49K mutant for B, E, H and K, and the E32Q/N49K mutant for C, F, I and L.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Interchannel dimerization of VSDs.\n(A) HS-AFM snapshots of sets of four particles arranged in parallelogram. All panels show different sets of four particles. The black dots indicate the centers of particles. The molecules are aligned at a similar angle. (B) Structure of a VSD dimer after 1 ms MD simulation. The S4 helices are colored in cyan and yellow. (C) Averaged-simulated AFM image of the extracellular surface of the modeled dimer. The virtual membrane surfaces were set at 4 nm height from the intracellular surface, and the assumed tip radius was 0.5 nm for the simulated AFM image. The time-averaged image was calculated by averaging 100 simulated AFM images, in which they were calculated by using 100 structures extracted from 1 ms MD simulation. (D) Particles of 20 nm diameter placed randomly in an area of 1 mm2. The diameter roughly corresponds to that of the resting NavAb channel. The center coordinates of the particles were randomly generated without considering the exclusive area of other particles, so the particles were allowed to overlap each other. The density is 1500 channels/mm2. (E) Contact probability at different densities of particles with diameters of 10 and 20 nm. Probabilities were calculated using 10000 sets of randomly generated snapshots, including 100-2000 particles in area of 1 mm2. The proportion of particles in contact with at least one other particle was plotted.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Interchannel network between resting channels and cooperative gating.\nSuggested structural changes associated with the voltage-gating of Nav channels; resting Nav channels form a network via VSD dimers, and the VSD dimers dissociate and attach to the PDs upon activation.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0a6b8025fe63ede42bbc75764a31f72c0e5e57a292ccb22e0161a2c8d90351c2/preprint/preprint.md

@@ -0,0 +1,225 @@
+# Abstract
+
+Understanding voltage-gated sodium (Na<sub>v</sub>) channels is significant since they generate action potential. Na<sub>v</sub> channels consist of a pore domain (PD) and a voltage-sensor domain (VSD). All resolved Na<sub>v</sub> structures in different gating states have VSDs that tightly interact with PDs; however, it is unclear whether VSDs attach to PDs during gating under physiological conditions. Another problem is that the key structure facilitating positive cooperativity in the rising phase of the action potential is unknown. Here, we reconstituted three different voltage-dependent Na<sub>v</sub> Ab channels into a lipid membrane and observed their structural dynamics by high-speed atomic force microscopy. Interestingly, VSDs dissociated from PDs in the mutant in the resting state and further dimerized to form cross-links between channels. This dimerization would occur at a realistic channel density, offering a potential explanation for the facilitation of positive cooperativity.
+
+[Biological sciences/Biophysics/Single-molecule biophysics](/browse?subjectArea=Biological%20sciences%2FBiophysics%2FSingle-molecule%20biophysics)  
+[Biological sciences/Physiology/Neurophysiology](/browse?subjectArea=Biological%20sciences%2FPhysiology%2FNeurophysiology)  
+[Biological sciences/Biological techniques/Nanobiotechnology](/browse?subjectArea=Biological%20sciences%2FBiological%20techniques%2FNanobiotechnology)
+
+# INTRODUCTION
+
+Since voltage-gated cation channels (VGCCs) play essential roles in neural transmission, it is important to understand the molecular mechanism of their gating.¹ VGCCs have two structural modules: the voltage-sensor domain (VSD) and pore domain (PD). Previous studies have revealed the basic structure of VGCCs; four PDs arranged in a square to form ionic pores at the tetrameric center with four VSDs directly attached to the PDs.²–⁵ There are two types of arrangements: a VSD on an adjacent subunit associating with a PD (swap type) or a VSD on the same subunit associating with a PD (nonswap type).⁶ In any case, VSDs are tightly associated with PDs. Therefore, it is supposed that voltage-dependent structural changes in VSDs are transmitted to PDs via physical contact and that the formed pore is gated in a voltage-dependent manner.⁷–¹¹ The basic structure of these domains and their conformation are similar to those of other voltage-gated Na⁺ channels, K⁺ channels, and Ca²⁺ channels, implying that the molecular arrangement of VSDs and PDs plays a key role in voltage gating.
+
+To consider structural changes due to changes in membrane potential, the structure in the resting state is important because it is the initial structure of the structural change process. However, there is little structural information about this state because it is difficult to perform structural analysis under a membrane potential. It has only recently been reported that structure of a nonswap-type VGCC under an applied electric field.¹² Regarding the swap-type VGCC, the VSD and PD structures of cross-linked NaᵥAb in the resting state were determined.⁹ In both structures, VSD was tightly associated with PD. On the other hand, molecular dynamics (MD) simulation of the voltage-gated potassium channel predicted that VSDs dissociate from PDs when the channel closes.¹³ This implies that there is a possibility of VSDs generally dissociating from the PDs of VGCCs in the resting state, which has not been experimentally confirmed.
+
+All of the resting state structures have been obtained from single-particle analysis,⁹,¹² which allows three-dimensional structures to be determined by averaging a large number of monodisperse protein particles observed by cryo-electron microscopy.¹⁴ This may be the reason why cross-linking was necessary for structural analysis. The VSDs that move freely away from PDs are not suitable for the averaging performed by single-particle analysis. In the case of MD simulations, a single individual channel is investigated, albeit with computational reproduction. Thus, further experimental analysis of a single channel without any restriction is needed to completely understand the molecular mechanism of voltage gating.
+
+High-speed atomic force microscopy (HS-AFM) can be used to image the nanostructure and subsecond motion of biological molecules, including membrane proteins, in solution.¹⁵–²¹ HS-AFM can be used to obtain structural information on a single molecule. Therefore, to confirm whether VSDs are always associated with PDs during gating and how Naᵥ channels interact with each other, we studied NaᵥAb channels reconstituted in a planar membrane by HS-AFM. We imaged the nanostructure of the reconstituted NaᵥAb channels in different gating states using HS-AFM, which revealed the dissociation of VSDs from PDs in the resting state. Furthermore, dimerized VSDs between channels were observed, and this dimerization was theoretically predicted to be a plausible factor for the positive cooperability known to be crucial for the rapid onset of action potentials.
+
+# RESULTS
+
+HS-AFM of Na<sub>v</sub>Ab channels in different gating states.
+
+We used the Na<sub>v</sub>Ab channel as a model of the voltage-gated Na<sub>v</sub> channel. The Na<sub>v</sub>Ab channel is the Na<sub>v</sub> channel of *Arcobacter butzleri*, which has a relatively simple structure consisting of 268 amino acids per subunit with conserved important parts, such as a PD and VSD, and has been used in many structural studies as a model of the Na<sub>v</sub> channel.<sup>3,4,9,22</sup> The Na<sub>v</sub>Ab channel has six transmembrane helices denoted as S1 through S6 (Fig. 1A), which form homotetrameric channels. S5 and S6 compose the PD. When the activation gate at the cytoplasmic side of the PD opens, Na<sup>+</sup> ions permeate through the pore formed at the center of the tetrameric PDs. Helices S1 to S4 compose the VSD. Between the PD and the VSD, there is an S4-S5 linker. The VSDs of the tetrameric Na<sub>v</sub>Ab channel surround the PDs by domain swapping (Fig. 1B). On the extracellular side of the VSD, there are two loops linking helices S1 to S2 and S3 to S4 (indicated by the black and white arrows shown in Fig. 1A and B). To reliably observe the single-molecule structural dynamics of the Na<sub>v</sub>Ab channel by HS-AFM, we truncated the C-terminal cytoplasmic domain (230–268) and added a His-tag on the C-terminus (for readability, this paper will omit describing the notation of the deletion of the C-terminus and the addition of a His-tag). We attached the His-tagged Na<sub>v</sub>Ab channel onto a Ni<sup>2+</sup>-coated mica substrate, reconstituted it into a lipid bilayer with its extracellular surface facing upward and observed it by HS-AFM (Fig. 1C). To observe the structure of the Na<sub>v</sub>Ab channel in different gating states in HS-AFM, two mutants, N49K and E32Q/N49K, were used in addition to the wild type (WT). These three constructs have different voltage dependencies; for example, activation starts at -100 mV in the WT, at -50 mV in the N49K mutant, and at 0 mV in the E32Q/N49K mutant, meaning that at 0 mV, the WT channel is fully activated, and the E32Q/N49K channel is mostly in the resting state (Fig. 1D). We also attempted to observe the KAV mutant,<sup>9</sup> which is in a complete resting state at 0 mV, but were unable to perform imaging because the tetramer was too unstable and difficult to reconstitute properly in the lipid membrane (Figure S1, Movies S1 and S2).
+
+The HS-AFM movies of all Na<sub>v</sub>Ab channels (WT, N49K, E32Q/N49K) showed tetrameric PDs, whose four particles aligned as squares (Fig. 2A, E, I, L). This shape is similar to that of the K<sup>+</sup> channel without VSDs observed in previous work<sup>23</sup> and is consistent with the fact that both Na<sup>+</sup> and K<sup>+</sup> channels share the same basic structure of tetrameric PDs.
+
+VSDs were observed around the PDs. In the WT channel, four particles surrounded the PDs (black arrows in Fig. 2A, Movie S3). These particles can be seen more clearly in the time-averaged and localization AFM (LAFM)<sup>24</sup> images (Fig. 2B and C). The positions of these particles corresponded to the position of the S1-S2 linkers when compared to the crystal structure (black arrows in Fig. 1B). Thus, the surrounding particles originated from the S1 and S2 of the VSDs (Fig. 2D). Here, particles corresponding to S3-S4 were not seen in the WT channel. In the crystal structure (4EKW)<sup>25</sup>, the S3-S4 linker was more embedded in the transmembrane region than the S1-S2 linker, which may explain why no S3-S4-derived particles were observed in the HS-AFM image of the WT channel.
+
+Particles surrounding the tetrameric PDs were also observed for the N49K mutant (Movie S4), which is less likely to open the gate than the WT (Fig. 1D). There were two differences observed when comparing this mutant to the WT. One was that some VSDs were not visible in N49K (e.g., upper right in Fig. 2G), while VSDs were almost always associated with PDs in WT. The S1-S2-derived particles at the upper right were clearly observed in frames at 0.0, 3.3, and 4.8 s (black arrowheads) but were barely visible most of the time, suggesting the intermittent binding of the VSDs to the PDs in the N49K mutant. On the other hand, the S1-S2-derived particle in the lower right was observed around the PDs at most moments (e.g., black arrowhead at 19.6 s). This is why the VSDs are not clearly visible relative to the PDs in the time-averaged and LAFM images (Fig. 2F and G). The probability of VSD (S1-S2) detection is summarized in Figure S2. The other difference was that additional small particles (white arrowheads in Fig. 2E) were sometimes observed in a position of S3-S4 (Fig. 1B, white arrowhead); it is not clear why these particles were observed in the N49K mutant while they were not visible in the WT. One possible explanation is that some structural change occurred in the VSD, since N49K is a mutation that weakens the interaction between S2 and S4. These observations suggested that the interaction between the VSDs and PDs was weaker in the N49K mutant than in the WT (Fig. 2H).
+
+AFM images of the resting mutant E32Q/N49K were completely different from those of the WT and N49K mutant (Movie S5). In the E32Q/N49K mutant, while VSDs surrounded PDs when without lipid bilayer (Figure S3), VSDs were not visible in the vicinity of PDs in reconstituted membrane (Fig. 2I, J and K; Figure S2). When the scanned area was enlarged, four two-set particles were observed at a distance away from the PDs (Fig. 2L, M and N). We concluded that these two-set particles were VSDs for the following reasons. First, no VSDs were observed in the vicinity of the PDs. Second, the distance between the two particles in the set was close to the distance between the S3-S4-derived particle and the S1-S2-derived particle observed occasionally for the N49K mutant (Figure S4). Third, the simulated AFM image of the extracellular surface of a single VSD (4ekw) also showed two particles (Figure S5). The VSDs observed in the E32Q/N49K were clearly visible in the time-averaged and LAFM images, but they were sometimes obscured in the movie (e.g., upper right in snapshot of 17.2 s in Fig. 2L) because they moved around in the lipid bilayer. These were unquestionable data indicating that the VSDs were truly dissociated from the PDs in the resting state of the Na<sub>v</sub>Ab channel (Fig. 2O).
+
+Molecular motion of the Na<sub>v</sub>Ab channels in different gating states.
+
+To clarify how the gating states affect the channel motion, we extracted trajectories of each particle from the HS-AFM movies and analyzed the fluctuation of the PD–VSD distance and their coupled motion (Fig. 3). From the trajectories, we calculated the time variation of the distance from each VSD to the PD (Fig. 3D, E and F). According to the histograms analyzed from the distances between all pairs, the distances in the N49K mutant and WT were similar (Fig. 3G and H). In contrast, although the E32Q/N49K mutant showed a multipeaked distribution because of an insufficient sample number (Fig. 3I), even the nearest peak of the E32Q/N49K mutant was much farther away than those of the WT and N49K mutant. Using the measured distances and two assumptions, the maximum distance between the PD and S1-S2 was found to be approximately 7.5 nm (Figure S6, and see explanations in the caption). Although it was not possible to determine from the images which particles were connected to each other intracellularly since the extracellular side was observed, the S4-S5 linker was considered to be unfolded and disordered because the observed distance between the PDs and S1-S2 particles shown in Fig. 3I (3–11 nm) was much longer than the estimated maximum distance (7.6 nm).
+
+Principal component analysis (PCA) was performed to investigate how the motions of the VSDs were coupled to those of the PDs. In this analysis, eigenvectors appeared if the motions were strongly coupled; otherwise, no vectors appeared. In the WT channels, vectors appeared along the rotational direction for most of the PDs and VSDs (Figs. 3J and S7). This indicated that the rotational motions of PDs were tightly coupled to those of VSDs, which is expected since they were seen in the vicinity in the HS-AFM images. On the other hand, for the N49K mutant, a few large vectors appeared for the VSDs, but almost no vectors appeared for the PDs (Figs. 3K and S8). This was also the case for the second and third principal components (Figure S8). This indicated that the motions of VSDs were not coupled to those of PDs and supported the intermittent binding of VSDs to PDs observed in the AFM images. Accordingly, it strongly suggested that the interaction between VSDs and PDs was almost lost in the N49K mutant, even though the VSDs were observed to be in the vicinity of the PDs in the HS-AFM images. For the E32Q/N49K mutant, VSDs were observed to be dissociated from PDs, resulting in a minor degree of coupling (Figs. 3L and S9).
+
+Dimerization of dissociated VSDs in the resting state.
+
+In the reconstituted membrane of the E32Q/N49K mutant, there were sets of four particles arranged not in a square but in a parallelogram (Fig. 4A). This structure can also be seen clearly in the lower lefthand corner of Fig. 2N. The dissociated VSDs showed two particles derived from S1-S2 and S3-S4 (Fig. 2L, M, and N); thus, we assumed that these sets of four particles were arranged in parallelograms as dimers of two VSDs. To confirm this clearly, the equilibrium structure of the VSD dimer in the membrane was obtained by MD simulation using a homology model based on a dimer of the H<sub>v</sub> channel as the initial channel structure (Fig. 4B). The H<sub>v</sub> channel is a proton channel composed solely of VSD-like domains that form a dimer and cooperatively gates.<sup>26–30</sup> In the simulated structure, S4 helices interacted to form the channel interface in the half of the membrane facing the intracellular solution (Movie S6). The S4 and S3 helices interacted with the PDs in the activated conformation,<sup>31</sup> meaning that this dimeric structure was formed only when the VSDs were detached from the PDs. Using the MD simulation data, we computed the time-averaged image of the simulated AFM images<sup>32</sup> of the VSD dimer (Fig. 4C). The simulated AFM images showed four particles arranged in a parallelogram similar to the experimental AFM images, confirming that the observed structures were dimers of VSDs.
+
+Estimation of dimerization probability in the cell membrane.
+
+To estimate whether the channels could interact with each other with a realistic density of channels in the cell membrane, we calculated the contact probability with densities in the range of 100–2000 µm<sup>2</sup>. The density of the Na<sub>v</sub> channels at the Ranvier node is ~1500 channels/µm<sup>2</sup>.<sup>1,33,34</sup> The coordinates of particles were randomly generated, and contact probabilities were calculated with different particle sizes and densities (Fig. 4D, E and S10). Note that we generated coordinates without considering the exclusive area occupied by other particles to estimate the lowest contact probability; the actual contact probability should be slightly higher than this estimation. Assuming a particle diameter of 20 nm, which was the size of the resting channel seen in Fig. 2L, 84% of channels could interact with at least one other channel at a density of 1500 channels/µm<sup>2</sup> (Fig. 4E). Therefore, it is plausible that when VSDs dissociate from PDs, VSDs can cross-link between at least one other channel in the node of Ranvier. If we assume the size of the Na<sub>v</sub> channel to be 10 nm (10 nm roughly corresponds to the diameter of the activating channel), the contact probability becomes 37% (Fig. 4E). This simple estimation suggests that, in addition to the well-known structural changes of individual VSDs, collective structural changes, such as the formation and dissociation of interchannel networks, occur when the activation gate of the Na<sub>v</sub> channel opens and closes (Fig. 5).
+
+# DISCUSSION
+
+## Dissociation of VSDs from PDs in the resting state
+
+In this study, we performed HS-AFM on Na<sub>v</sub> Ab channels and found that VSDs dissociate from PDs in the resting state. By various structural analyses, the conformational changes of VSDs have been investigated during membrane potential transitions, <sup>9,35</sup> but the effects of such conformational changes on PDs and the whole channel have not been well analyzed. The dissociation of VSDs from PDs has not yet been evidenced by structural analysis. <sup>9</sup> However, the capability of HS-AFM to obtain structural information from a single molecule allowed this dissociated VSD structure to be observed.
+
+The VSDs were found to be tightly bound to the PDs in the WT, which corresponds to a fully activated structure in the depolarization state. In the E32Q/N49K mutant, in which some VSDs were not fully activated without a membrane potential, all VSDs were dissociated from the PDs. The N49K mutant, whose voltage dependency is moderate, appeared to have VSDs that were tightly attached to the PDs, as in the WT, but the PCA results of the coupled motion (Fig. <span class="InternalRef" refid="Fig2">3</span> J, K and L) revealed that the interaction between VSDs and PDs was diminished. During an MD simulation, which lasted for hundreds of µs, of the K<sub>v</sub> channel, a channel of the VGCC family, it was observed that the S4-S5 linker unraveled, and the VSDs detached from the PDs and moved far away. <sup>36</sup> Our results support this MD simulation result. Even in the N49K crystal structure, VSDs have been observed to be slightly farther from the PDs compared to those in the WT crystal structure. <sup>4</sup> Accordingly, it is expected that the tight association of VSDs with PDs is strengthened by structural changes in the VSDs associated with the depolarization of the membrane potential. A consideration of the association and dissociation of VSDs from PDs would be beneficial for elucidating the gating mechanism of VGCCs.
+
+From the viewpoint of VGCC diversity, the dissociation of VSDs is also suggested. VGCCs include a swap-type VGCC in which the VSD contacts the PD of the adjacent subunit and a nonswap-type VGCC in which the VSD contacts the PD of the same subunit. <sup>6</sup> In a recent study, swap-type and nonswap-type structures were observed in the same K<sub>v</sub> AP channel, the most well-understood VGCC. <sup>37</sup> This indicates that VSDs can move drastically when VGCCs are functionally expressed on cells. Furthermore, even in structures considered to be in a resting state in which VSDs and PDs were cross-linked by disulfide bonds, the cytosolic-side transition of the S4 helix was observed. <sup>9</sup> Under natural conditions, without disulfide bonds, it is quite possible that the S4-S5 linker is elongated. The structure of dissociated VSDs has not been observed thus far because it is difficult to detect them. Considering that the VSD is shared in all VGCC families, it is quite possible that VGCCs have such characteristics. Therefore, it will be interesting to see what kind of information will be discovered by methods for obtaining structural information from a single molecule, such as HS-AFM.
+
+## Possible gating mechanism due to VSD dissociation in the resting state
+
+The phenomenon that the VSD is distant from the PD in the resting state does not align with the model of the gating mechanism in which the S4-S5 linker helix closes the channel due to the sinking of the S4 helix in the resting state, and the S4-S5 linker helix loosens because of the rising of the S4-S5 linker helix upon depolarization and opens the channel. <sup>9</sup> The fact that dissociated VSDs cause PDs to close raises the question of what causes the closing of the channel. Lipids can be considered as candidates to study how closing occurs without VSDs. That is, when the VSD is dissociated from the PD, lipids associate tightly with the PD. Recently, the single-particle analysis of large pore channels has suggested that lipids may be directly involved in channel gating. <sup>38–40</sup> In these cases, it is predicted that lipids directly fill the pores of the channels. There may be a mechanism by which tightly adhering lipids cause PDs to close the channel, and then activated VSDs approach the PDs and replace the lipids around the PDs to reopen the channel. Again, note that our observation aligns with the gating mechanism observed in the MD simulation, wherein the VSD dissociates when the channel closes. <sup>35</sup> Detailed analyses of such MD simulation trajectories are necessary to draw conclusions.
+
+## Dimerization of dissociated VSDs
+
+While the interchannel dimerization of dissociated VSDs in the resting state of the Na<sub>v</sub> channel is a new finding, the observed dimer structure is similar to that in the structural model of an H<sub>v</sub> dimer <sup>28</sup> that is an H<sup>+</sup> channel composed of only a VSD-like domain <sup>26–30</sup>. It has also been reported that VSDs interact with other VSDs during early S4 transitions. <sup>41</sup> While it was assumed that the coupled activation between VSDs occurred in a single channel, this observation could also be considered an interchannel interaction based on our results. Voltage-gated phosphatase, another example of a protein with a VSD, has also been reported to be a dimer. <sup>42</sup> S4 is usually attached to PDs in the activated state, so it is natural that they dimerize when dissociated from PDs. Thus, it seems possible that the VSD itself has a dimerization ability, and dimerization is a common phenomenon in the VGCC family.
+
+In the case of VGCCs, it is assumed that the interaction of VSDs occurs between other channels because of the steric hindrance of PDs. In our previous study, the gradual inactivation of the channel current was observed in NaChBac, the VSD of which had a single cysteine mutation. <sup>43</sup> This inactivation occurred when the concatemer of the NaChBac tetramer was mutated by the introduction of cysteine residues into not adjacent VSDs but into diagonal VSDs. The HS-AFM results provide a clear answer to the gradual inactivation of a single cysteine mutant. By applying a deep negative potential for the first time during whole-cell patch clamp measurement, NaChBac enters a resting state, causing VSD to dissociate from PD and cross-link with VSD on other channels. When VSD cross links with that of other channels, it is possible that the channel is gradually deactivated because the channel cannot be activated.
+
+Recently, Clatot <em>et al</em>. reported that Na<sub>v</sub> 1.5 forms a dimer by interacting with a 14-3-3 protein <sup>44,45</sup> and shows coupled gating. <sup>46</sup> By combining experimental data <sup>45,46</sup> and the structures of Na<sub>v</sub> 1.5 <sup>2</sup> and the 14-3-3 protein <sup>44</sup>, the dimeric structure of Na<sub>v</sub> 1.5 has been modeled. <sup>47</sup> In this model, two VSDs from different channels associate with each other with the supportive binding of the 14-3-3 protein. We propose another possibility: the VSDs of Nav1.5 form a dimer prior to the binding of the 14-3-3 protein, and then the 14-3-3 protein binds and stabilizes it. To our knowledge, there are no conflicting reports of the dimerization of dissociated VSDs.
+
+## Implication for clustering and functional cooperativity
+
+Historically, the Hodgkin-Huxley equation has been used to model action potential generation. <sup>48,49</sup> Naundorf <em>et al</em>. have reported that the cooperative activation of sodium channels is important for reproducing the experimental dynamics of action potential initiation. <sup>50</sup> However, the molecular basis of functional cooperativity is still unclear, even though the functional cooperativity of Na<sup>+</sup> channels has been extensively investigated. <sup>47,50–58</sup> Since it has been reported that dimers of H<sub>v</sub> channels are cooperatively activated <sup>29</sup>, it is not surprising that the dimerization of VSDs leads to the cooperative activation of Na<sub>v</sub> channels. Na<sub>v</sub> channels are densely distributed, especially at the node of Ranvier in myelinated nerves <sup>59,60</sup>, with a density of ~1500 channels/µm<sup>2</sup>. <sup>1,33,34</sup> Recent super resolution microscopic imaging of the node of Ranvier showed the nanoscale localization of Na<sub>v</sub> channels with 190-nm periodicity in the node, <sup>60</sup> meaning that the local density around Na<sub>v</sub> channels is much higher than 1500 channels/µm<sup>2</sup>. Even though the channels are randomly arranged at a density of 1500 channels/µm<sup>2</sup>, 84% of channels can interact with other channels when their diameter is 20 nm, which corresponds to the diameter of the resting channel observed in this study (Fig. <span class="InternalRef" refid="Fig3">4</span> E). Thus, in the actual node of Ranvier, it is plausible that a network of interchannel cross-linking via dissociated VSDs is formed. Interestingly, Huan <em>et al</em>. reported that a small fraction, 5–15%, of strongly cooperative channels generate action potentials with the most rapid onset dynamics. <sup>53</sup> Based on our simple estimation, ~12% of channels with a 20 nm diameter interact with the other four channels at a density of 1500 channels/µm<sup>2</sup> (Figure S10I), implying that these channels are strongly cooperative channels. Integrating all this information, we expect that the dimerization of dissociated VSDs stabilizes the resting state of channels and is a strong candidate for the molecular basis of the cooperative activation of Na<sub>v</sub> channels (Fig. <span class="InternalRef" refid="Fig4">5</span>).
+
+Many other voltage-gated K<sup>+</sup> and Ca<sup>2+</sup> channels in the VGCC family also form clusters and cooperatively gate. <sup>47,61–65</sup> In addition to the VGCC family, the KcsA channel, sharing the structure of PDs, showed clustering-dispersion upon gating. <sup>66</sup> Since the basic structure of PDs and VSDs in voltage-gated channels is shared among VGCCs <sup>1</sup>, it would be very interesting if the interchannel cross-linking by dissociated VSDs in the resting state is widely shared in other voltage-gated channels and is the molecular basis for their cooperative activation.
+
+# METHODS
+
+## Expression and purification of Na<sub>v</sub> Ab
+
+The Na<sub>v</sub> Ab mutated DNAs were subcloned into the modified pBiEX-1 vector (Novagen) that was modified by replacing the fragment from the NdeI site to the SalI site in a multicloning site with that of the previously described modified pET 15b (Novagen) vector between BamHI and SalI. Therefore, all Na<sub>v</sub> Ab channels have an N-terminal hexa-histidine tag with a thrombin digestion site. The polymerase chain reaction accomplished site-directed mutagenesis using PrimeSTAR® Max DNA Polymerase (Takara bio.). For AFM measurement, all Na<sub>v</sub> Ab channels were truncated at the C-terminal 37 amino acids from His231 to C-terminal Asn267, corresponding to the cytosolic helix, and a 7 amino acid linker (ENLYFQG) and C-terminal hexa-histidine tag without a thrombin digestion site were added. For brevity, the notation of the deletion of the C-terminus and the addition of the His-tag were omitted in the main text. All clones were confirmed by DNA sequencing.
+
+Proteins were expressed in the *E. coli* KRX strain (Promega). Cells were grown at 37°C to an OD<sub>600</sub> of 0.6, induced with 0.2% rhamnose (Wako), and grown for 16 h at 25°C. The cells were suspended in TBS buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl) and lysed using a high-pressure homogenizer LAB1000 (SMT Co. Ltd.) at 12,000 psi. Cell debris was removed by low-speed centrifugation (12,000×g, 30 min, 4°C). Membranes were collected by centrifugation (100,000×g, 1 h, 4°C) and solubilized by homogenization in TBS buffer containing 30 mM n-dodecyl-β-D-maltoside (DDM, Anatrace). After centrifugation (40,000×g, 30 min, 4°C), the supernatant was loaded onto a HIS-Select® Cobalt Affinity Gel column (Sigma). The protein bound to the cobalt affinity column was washed with 10 mM imidazole in TBS buffer containing 0.05% lauryl maltose neopentyl glycol (LMNG, Anatrace) instead of DDM. After washing, the protein was eluted with 300 mM imidazole, and the N-terminal His-tag was removed by thrombin digestion (overnight, 4°C). Eluted protein was purified on a Superdex-200 column (Cytiva) in TBS buffer containing 0.05% LMNG. Excessive surfactants were removed from the gel-filtration fraction of Na<sub>v</sub> Ab by GraDeR<sup>67</sup>. The top layer buffer contained 20 mM Tris (pH 8.0), 150 mM NaCl, 5% glycerol, and 0.003% lauryl maltose neopentyl glycol. The bottom layer buffer contained 20 mM Tris (pH 8.0), 150 mM NaCl, and 25% glycerol. After stacking the buffers, the gradient was generated with Gradient Master 108 (BioComp Instruments). The gel-filtration fractions containing the Na<sub>v</sub> Ab proteins were loaded on top of the gradient and centrifuged at 35,000 rpm (209,678 g) for 18 h at 4°C with an SW41Ti rotor. The solution was fractionated by a peristaltic pump from bottom to top. The fractions containing Na<sub>v</sub> Ab protein were detected by tryptophan fluorescence-detection size exclusion chromatography. The protein fractions were dialyzed against 20 mM Tris (pH 8.0) and 150 mM NaCl buffer overnight at 4°C and concentrated to a protein concentration of 0.1 mg/ml.
+
+## Electrophysiological measurement of the Na<sub>v</sub> Ab
+
+The recordings were performed using SF-9 cells. SF-9 cells (ATCC catalog number CRL-1711) were grown in Sf-900™ II medium (Gibco) supplemented with 0.5% 100× Antibiotic-Antimycotic (WAKO) at 27°C. Cells were transfected with target channel-cloned pBiEX vectors and enhanced green fluorescent protein (EGFP)-cloned pBiEX vectors using FuGENE HD transfection reagent (Promega). First, the channel-cloned vector (1.5 µg) was mixed with 0.5 µg of the EGFP-cloned vector in 100 µL of the culture medium. Next, 3 µL of FuGENE HD reagent was added, and the mixture was incubated for 10 min before the transfection mixture was gently dropped onto cultured cells. After incubation for 16–48 h, the cells were used for electrophysiological measurements.
+
+For measurement, a pipette solution contained 135 mM NaF, 15 mM NaCl, 10 mM EGTA, and 10 mM HEPES (pH 7.4 adjusted by NaOH), and a bath solution contained 150 mM NaCl, 1.5 mM CaCl<sub>2</sub>, 1.0 mM MgCl<sub>2</sub>, 10 mM HEPES (pH 7.4 adjusted by NaOH) and 10 mM glucose. Cancellation of the capacitance transients and leak subtraction were performed using a programmed P/10 protocol delivered at -140 mV. The bath solution was changed using the Dynaflow® Resolve system. All experiments were conducted at 25 ± 2°C. All sample numbers represent the number of individual cells used for each measurement. Cells that had a leak current smaller than 0.5 nA were used for measurement. When any outliers were encountered, these outliers were excluded if any abnormalities were found in other measurement environments and were included if no abnormalities were found. All results are presented as the mean ± standard error.
+
+## Preparation of destabilized liposomes
+
+We dissolved the phospholipid 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) in chloroform in a glass tube and dried for at least 2 h in vacuo. The lipid film at the glass bottom was hydrated by buffer (10 mM HEPES [pH 7.5], 200 mM KCl) to form a liposome solution (10 mg/mL). n-Dodecyl-<em>β</em>-D-maltopyranoside (DDM) was added to the liposome solution to 0.06% to solubilize the liposomes. This comicellar solution of DDM and DMPC was diluted with DDM-free buffer (20 mM Tris, [pH 7.4], 135 mM NaCl, 20 mM KCl) to a final DDM concentration of 0.006%, resulting in DDM-destabilized liposomes. All procedures were performed at room temperature.
+
+## Reconstitution of Na<sub>v</sub> Ab into lipid bilayers
+
+We reconstituted the Na<sub>v</sub> Ab channel into a lipid bilayer by a previously reported method<sup>68,69</sup>. To suppress diffusion in the bilayer, channels were immobilized on the mica surface through a Ni<sup>2+</sup>–His-tag interaction. First, solubilized Na<sub>v</sub> Ab channels with His-tag at cytoplasmic C-termini (1–10 µg/mL in 20 mM Tris, [pH 7.4], 150 mM NaCl) were applied onto Ni<sup>2+</sup>-coated mica surface for 5 min, then excess channels floating in solution were washed out by buffer (20 mM Tris, [pH 7.4], 150 mM NaCl). Next, DDM-destabilized DMPC liposomes (50–200 µg/mL DMPC) were applied, incubated for 5–10 min, and then rinsed with buffer (20 mM Tris, [pH 7.4], 150 mM NaCl). All procedures were performed at 25°C. Since the His-tag was on the cytoplasmic side, the extracellular side of the Na<sub>v</sub> Ab channel faced upward in the resulting reconstituted membrane and was imaged by HS-AFM.
+
+## HS-AFM observation and image processing of reconstituted Na<sub>v</sub> Ab
+
+We used a laboratory-built HS-AFM<sup>15,70</sup> and the cantilever AC7 (Olympus Co., Tokyo, Japan) with an electron beam deposition tip. The typical resonant frequency and quality factor in water of AC7 are 700 kHz and 2, respectively. Imaging buffers were 20 mM Tris, (pH 7.4), 150 mM NaCl for WT and E32Q/N49K, and 50 mM NaCl for N49K. The data from the 50 mM NaCl condition were used in the analysis because the resolution was higher in N49K, but the overview of the channel structure did not change at either 50 mM or 150 mM. Time-averaged images and LAFM images are calculated by the z-project and LAFM plugin<sup>24</sup> on ImageJ after removing noise and tilt as described in the next section. To compensate for changes in molecular size due to piezo nonlinearity and other factors, the movies were scaled so that the average distance between adjacent PDs was 3.3 nm. We used BioAFMViewer<sup>32</sup> for simulation of AFM images.
+
+## Particle tracking in HS-AFM movies
+
+To track particles correctly, we removed background, high-frequency noise and lateral drift of HS-AFM movies using subtract background, FFT filter and template matching plugin on ImageJ, respectively. We detected and tracked particles by using the TrackMate plugin<sup>71</sup> on ImageJ. The estimated particle size was set to 2.5 nm, which corresponds to a longer diameter of two α-helices when they are close together. Obvious particle misrecognitions and trajectory swaps were corrected manually. Only for the PCA described below were coordinates generated by linear interpolation for frames where no particles were recognized. In E32Q/N49K, we assigned the two VSD-derived particles as follows: the closest particle clockwise from the PD with the center of the tetramer as the origin as S3-S4 particles and the next closest particles as S1-S2 particles.
+
+### PCA
+
+PCA was applied using the conventional method<sup>72</sup>, that is, by diagonalizing the covariance matrix (<em>C</em>) defined as follows:
+
+$$
+{C}_{ij}=⟨({\overrightarrow{r}}_{i}-⟨{\overrightarrow{r}}_{i}⟩)({\overrightarrow{r}}_{j}-⟨{\overrightarrow{r}}_{j}⟩)⟩
+$$
+
+where $\overrightarrow{r}$ represents the x and y positions of the center pixel of particles detected by TrackMate in ImageJ and angular brackets represent the time average. By diagonalizing <em>C</em>, eigenvectors were computed.
+
+## Modeling of VSD dimer and MD simulation
+
+In the homology modeling, the H<sub>v</sub> dimer was obtained by superimposing the H<sub>v</sub> monomer (pdb code: 3wkv) on each helix of the cytosolic coiled-coil helices structure of H<sub>v</sub> (pdb code: 3vmx). The VSD of Na<sub>v</sub> Ab was extracted from the Na<sub>v</sub> Ab crystal structure (PDB code: 5yuc) and superimposed on the Hv dimer based on the S1-S4 helix to form the activated VSD dimer of Na<sub>v</sub> Ab.
+
+The homology model of the VDS dimer was embedded in the preequilibrated POPC membrane with approximately 150 mM NaCl solution. Amber ff19SB<sup>73</sup>, Lipid21<sup>74</sup>, TIP3P<sup>75</sup>, and the Joung-Cheatham<sup>76</sup> models were employed for the Na<sub>v</sub> Ab channel, the POPC lipid molecule, water, and ions, respectively. The system was composed of 1 VSD dimer, 249 POPC molecules, 21,206 water molecules, 58 Na<sup>+</sup> ions, and 68 Cl<sup>−</sup> ions. The total number of atoms in the system was 101,040.
+
+The bad contact made in the embedding process was removed by 1,000-step minimization. The MD simulation was performed for 600 ps with constant temperature (300 K) and pressure conditions (1 bar), where harmonic restraints were imposed on all atoms in the Na<sub>v</sub> Ab channel with a force constant of 2 kcal/mol/ Å<sup>2</sup>. The Langevin thermostat with a collision frequency of 2 ps was used to control the temperature, and the Monte Carlo barostat with anisotropic scaling was used to control the pressure. The SHAKE algorithm<sup>77</sup> was used to keep the bond length having H atoms constant, enabling the use of a time step of 2 fs. The periodic boundary condition was imposed, and long-range interactions were calculated by the particle mesh Ewald method<sup>78</sup> with a 10 Å cutoff in real space. Then, the MD simulation was performed for 100 ns with constant temperature (300 K) and volume conditions, where harmonic restraints were imposed on all atoms in the Na<sub>v</sub> Ab channel with a force constant of 2 kcal/mol/Å<sup>2</sup>. Finally, the constraints on the channels were removed, and the MD simulation was performed for 1 µs with constant temperature (300 K) and volume conditions, and 100 VSD-dimer structures every 10 ns were saved for AFM simulation.
+
+# References
+
+1. Hille, B. *Ion channels of excitable membranes*. (Sinauer, 2001).
+2. Jiang, D. *et al.* Structure of the Cardiac Sodium Channel. *Cell* **180**, 122–134.e10 (2020).
+3. Payandeh, J., Scheuer, T., Zheng, N. & Catterall, W. A. The crystal structure of a voltage-gated sodium channel. *Nature* **475**, 353–358 (2011).
+4. Irie, K., Haga, Y., Shimomura, T. & Fujiyoshi, Y. Optimized expression and purification of NavAb provide the structural insight into the voltage dependence. *FEBS Lett.* **592**, 274–283 (2018).
+5. De Lera Ruiz, M. & Kraus, R. L. Voltage-Gated Sodium Channels: Structure, Function, Pharmacology, and Clinical Indications. *J. Med. Chem.* **58**, 7093–7118 (2015).
+6. Barros, F., de la Peña, P., Domínguez, P., Sierra, L. M. & Pardo, L. A. The EAG Voltage-Dependent K+ Channel Subfamily: Similarities and Differences in Structural Organization and Gating. *Front. Pharmacol.* **11**, 1–19 (2020).
+7. Catterall, W. A. Ion channel voltage sensors: Structure, function, and pathophysiology. *Neuron* **67**, 915–928 (2010).
+8. Catterall, W. A., Wisedchaisri, G. & Zheng, N. The chemical basis for electrical signaling. *Nat. Chem. Biol.* **13**, 455–463 (2017).
+9. Wisedchaisri, G. *et al.* Resting-State Structure and Gating Mechanism of a Voltage-Gated Sodium Channel. *Cell* **178**, 993–1003.e12 (2019).
+10. Huang, G. *et al.* Unwinding and spiral sliding of S4 and domain rotation of VSD during the electromechanical coupling in Na<sub>v</sub>1.7. *Proc. Natl. Acad. Sci.* **119**, 1–9 (2022).
+11. Amaral, C., Carnevale, V., Klein, M. L. & Treptow, W. Exploring conformational states of the bacterial voltage-gated sodium channel NavAb via molecular dynamics simulations. *Proc. Natl. Acad. Sci.* **109**, 21336–21341 (2012).
+12. Mandala, V. S. & MacKinnon, R. Voltage-sensor movements in the Eag Kv channel under an applied electric field. *Proc. Natl. Acad. Sci. U. S. A.* **119**, 1–12 (2022).
+13. Jensen, M. Ø. *et al.* Mechanism of voltage gating in potassium channels. *Science (80-. ).* **336**, 229–233 (2012).
+14. R.A. Crowther. *The Resolution Revolution: Recent Advances In cryoEM*. (Elsevier Inc., 2016).
+15. Uchihashi, T., Kodera, N. & Ando, T. Guide to video recording of structure dynamics and dynamic processes of proteins by high-speed atomic force microscopy. *Nat. Protoc.* **7**, 1193–1206 (2012).
+16. Kodera, N., Yamamoto, D., Ishikawa, R. & Ando, T. Video imaging of walking myosin V by high-speed atomic force microscopy. *Nature* **468**, 72–6 (2010).
+17. Uchihashi, T., Iino, R., Ando, T. & Noji, H. High-Speed Atomic Force Microscopy Reveals Rotary Catalysis of Rotorless F1-ATPase. *Science (80-. ).* **333**, 755–758 (2011).
+18. Shibata, M., Yamashita, H., Uchihashi, T., Kandori, H. & Ando, T. High-speed atomic force microscopy shows dynamic molecular processes in photoactivated bacteriorhodopsin. *Nat. Nanotechnol.* **5**, 208–12 (2010).
+19. Shibata, M. *et al.* Real-space and real-time dynamics of CRISPR-Cas9 visualized by high-speed atomic force microscopy. *Nat. Commun.* **8**, 1430 (2017).
+20. Ando, T., Uchihashi, T. & Scheuring, S. Filming biomolecular processes by high-speed atomic force microscopy. *Chem. Rev.* **114**, 3120–88 (2014).
+21. Heath, G. R. & Scheuring, S. Advances in high-speed atomic force microscopy (HS-AFM) reveal dynamics of transmembrane channels and transporters. *Curr. Opin. Struct. Biol.* **57**, 93–102 (2019).
+22. Lenaeus, M. J. *et al.* Structures of closed and open states of a voltage-gated sodium channel. *Proc. Natl. Acad. Sci.* **114**, E3051–E3060 (2017).
+23. Sumino, A., Sumikama, T., Uchihashi, T. & Oiki, S. High-speed AFM reveals accelerated binding of agitoxin-2 to a K<sup>+</sup> channel by induced fit. *Sci. Adv.* **5**, eaax0495 (2019).
+24. Heath, G. R. *et al.* Localization atomic force microscopy. *Nature* **594** (2021).
+25. Payandeh, J., Gamal El-Din, T. M., Scheuer, T., Zheng, N. & Catterall, W. A. Crystal structure of a voltage-gated sodium channel in two potentially inactivated states. *Nature* **486**, 135–139 (2012).
+26. Sasaki, M., Takagi, M. & Okamura, Y. A voltage sensor-domain protein is a voltage-gated proton channel. *Science (80-. ).* **312**, 589–592 (2006).
+27. Okamura, Y., Fujiwara, Y. & Sakata, S. Gating mechanisms of voltage-gated proton channels. *Annu. Rev. Biochem.* **84**, 685–709 (2015).
+28. Mony, L., Stroebel, D. & Isacoff, E. Y. Dimer interaction in the Hv1 proton channel. *Proc. Natl. Acad. Sci. U. S. A.* **117**, 20898–20907 (2020).
+29. Fujiwara, Y. *et al.* The cytoplasmic coiled-coil mediates cooperative gating temperature sensitivity in the voltage-gated H<sup>+</sup> channel Hv1. *Nat. Commun.* **3** (2012).
+30. Lee, S. Y., Letts, J. A. & MacKinnon, R. Dimeric subunit stoichiometry of the human voltage-dependent proton channel Hv1. *Proc. Natl. Acad. Sci. U. S. A.* **105**, 7692–7695 (2008).
+31. Payandeh, J., Scheuer, T., Zheng, N. & Catterall, W. A. The crystal structure of a voltage-gated sodium channel. *Nature* **475**, 353–8 (2011).
+32. Amyot, R. & Flechsig, H. BioAFMviewer: An interactive interface for simulated AFM scanning of biomolecular structures and dynamics. *PLoS Comput. Biol.* **16**, 1–12 (2020).
+33. Waxman, S. G. & Murdoch Ritchie, J. Organization of ion channels in the myelinated nerve fiber. *Science (80-. ).* **228**, 1502–1507 (1985).
+34. Salzer, J. L. Clustering sodium channels at the node of Ranvier: Close encounters of the axon-glia kind. *Neuron* **18**, 843–846 (1997).
+35. Schwaiger, C. S., Liin, S. I., Elinder, F. & Lindahl, E. The conserved phenylalanine in the K<sup>+</sup> channel voltage-sensor domain creates a barrier with unidirectional effects. *Biophys. J.* **104**, 75–84 (2013).
+36. Jensen, M. Ø. *et al.* Mechanism of voltage gating in potassium channels. *Science* **336**, 229–33 (2012).
+37. Tao, X. & Mackinnon, R. Cryo-EM structure of the KvAP channel reveals a non-domain-swapped voltage sensor topology. *Elife* **8**, 1–15 (2019).
+38. Kuzuya, M. *et al.* Structures of human pannexin-1 in nanodiscs reveal gating mediated by dynamic movement of the N terminus and phospholipids. *Sci. Signal.* **15**, 1–11 (2022).
+39. Syrjanen, J., Michalski, K., Kawate, T. & Furukawa, H. On the molecular nature of large-pore channels. *J. Mol. Biol.* **433**, 166994 (2021).
+40. Burendei, B. *et al.* Cryo-EM structures of undocked innexin-6 hemichannels in phospholipids. *Sci. Adv.* **6** (2020).
+41. Chanda, B., Asamoah, O. K. & Bezanilla, F. Coupling Interactions between Voltage Sensors of the Sodium Channel as Revealed by Site-specific Measurements. *J. Gen. Physiol.* **123**, 217–230 (2004).
+42. Rayaprolu, V., Royal, P., Stengel, K., Sandoz, G. & Kohout, S. C. Dimerization of the voltage-sensing phosphatase controls its voltage-sensing and catalytic activity. *J. Gen. Physiol.* **150**, 683–696 (2018).
+43. Nagura, H. *et al.* Evidence for lateral mobility of voltage sensors in prokaryotic voltage-gated sodium channels. *Biochem. Biophys. Res. Commun.* **399**, 341–346 (2010).
+44. Yang, X. *et al.* Structural basis for protein-protein interactions in the 14-3-3 protein family. *Proc. Natl. Acad. Sci. U. S. A.* **103**, 17237–17242 (2006).
+45. Allouis, M. *et al.* 14-3-3 Is a regulator of the cardiac voltage-gated sodium channel Nav1.5. *Circ. Res.* **98**, 1538–1546 (2006).
+46. Clatot, J. *et al.* Voltage-gated sodium channels assemble and gate as dimers. *Nat. Commun.* **8**, 1–14 (2017).
+47. Dixon, R. E., Navedo, M. F., Binder, M. D. & Fernando Santana, L. Mechanisms and Physiological Implications of Cooperative Gating of Clustered Ion Channels. *Physiol. Rev.* **102**, 1159–1210 (2022).
+48. A. L. Hodgkin and A. F. Huxley. A quantitative description of membrane current and its application to conduction and excitation in nerve. *J. Physiol.* **117**, 500–544 (1952).
+49. Häusser, M. The hodgkin-huxley theory of the action potential. *Nat. Neurosci.* **3**, 1165 (2000).
+50. Naundorf, B., Wolf, F. & Volgushev, M. Unique features of action potential initiation in cortical neurons. *Nature* **440**, 1060–1063 (2006).
+51. Undrovinas, A. I., Fleidervish, I. A. & Makielski, J. C. Inward sodium current at resting potentials in single cardiac myocytes induced by the ischemic metabolite lysophosphatidylcholine. *Circ. Res.* **71**, 1231–1241 (1992).
+52. Pfeiffer, P. *et al.* Clusters of cooperative ion channels enable a membrane-potential-based mechanism for short-term memory. *Elife* **9**, 1–27 (2020).
+53. Huang, M., Volgushev, M. & Wolf, F. A small fraction of strongly cooperative sodium channels boosts neuronal encoding of high frequencies. *PLoS One* **7** (2012).
+54. Sigworth, F. J. Covariance of nonstationary sodium current fluctuations at the node of Ranvier. *Biophys. J.* **34**, 111–133 (1981).
+55. Armstrong, C. M. & Wm Gilly, F. Fast and slow steps in the activation of sodium channels. *J. Gen. Physiol.* **74**, 691–711 (1979).
+56. Elender, F. & Århem, P. Mechanisms of the tetrahydroaminoacridine effect on action potential and ion currents in myelinated axons. *Eur. J. Pharmacol. Mol. Pharmacol.* **208**, 1–8 (1991).
+57. Elinder, F., Frankenhaeuser, B. & Århem, P. Non-stationary fluctuation analysis of the Na current in myelinated nerve fibers of Xenopus laevis: Experiments and stochastic simulations. *BioSystems* **62**, 13–28 (2001).
+58. Vandenberg, C. A. & Bezanilla, F. A sodium channel gating model based on single channel, macroscopic ionic, and gating currents in the squid giant axon. *Biophys. J.* **60**, 1511–1533 (1991).
+59. Rasband, M. N. & Shrager, P. Ion channel sequestration in central nervous system axons. *J. Physiol.* **525**, 63–73 (2000).
+60. D’Este, E., Kamin, D., Balzarotti, F. & Hell, S. W. Ultrastructural anatomy of nodes of Ranvier in the peripheral nervous system as revealed by STED microscopy. *Proc. Natl. Acad. Sci. U. S. A.* **114**, E191–E199 (2017).
+61. Nakamura, Y. *et al.* Nanoscale Distribution of Presynaptic Ca<sup>2+</sup> Channels and Its Impact on Vesicular Release during Development. *Neuron* **85**, 145–158 (2015).
+62. Zhang, J. *et al.* Clustering and Functional Coupling of Diverse Ion Channels and Signaling Proteins Revealed by Super-resolution STORM Microscopy in Neurons Article Clustering and Functional Coupling of Diverse Ion Channels and Signaling Proteins Revealed by Super-resolut. *Neuron* **92**, 461–478 (2016).
+63. Bhargava, A. *et al.* Super-resolution scanning patch clamp reveals clustering of functional ion channels in adult ventricular myocyte. *Circ. Res.* **112**, 1112–1120 (2013).
+64. Tateishi, Y. *et al.* Cluster formation of inositol 1,4,5-trisphosphate receptor requires its transition to open state. *J. Biol. Chem.* **280**, 6816–6822 (2005).
+65. Marx, S. O., Ondrias, K. & Marks, A. R. Coupled gating between individual skeletal muscle Ca<sup>2+</sup> release channels (ryanodine receptors). *Science (80-. ).* **281**, 818–821 (1998).
+66. Sumino, A., Yamamoto, D., Iwamoto, M., Dewa, T. & Oiki, S. Gating-Associated Clustering-Dispersion Dynamics of the KcsA Potassium Channel in a Lipid Membrane. *J. Phys. Chem. Lett.* **5**, 578–584 (2014).
+67. Hauer, F. *et al.* GraDeR: Membrane Protein Complex Preparation for Single-Particle Cryo-EM. *Structure* **23**, 1769–1775 (2015).
+68. Sumino, A., Uchihashi, T. & Oiki, S. Oriented Reconstitution of the Full-Length KcsA Potassium Channel in a Lipid Bilayer for AFM Imaging. *J. Phys. Chem. Lett.* **8**, 785–793 (2017).
+69. Sumino, A., Sumikama, T., Uchihashi, T. & Oiki, S. High-speed AFM reveals accelerated binding of agitoxin-2 to a K<sup>+</sup> channel by induced fit. *Sci. Adv.* **5**, 1–11 (2019).
+70. Ando, T. *et al.* A high-speed atomic force microscope for studying biological macromolecules. *Proc. Natl. Acad. Sci. U. S. A.* **98**, 12468–72 (2001).
+71. Tinevez, J. Y. *et al.* TrackMate: An open and extensible platform for single-particle tracking. *Methods* **115**, 80–90 (2017).
+72. García, A. E. Large-amplitude nonlinear motions in proteins. *Phys. Rev. Lett.* **68**, 2696–2699 (1992).
+73. Tian, C. *et al.* Ff19SB: Amino-Acid-Specific Protein Backbone Parameters Trained against Quantum Mechanics Energy Surfaces in Solution. *J. Chem. Theory Comput.* **16**, 528–552 (2020).
+74. Dickson, C. J., Walker, R. C. & Gould, I. R. Lipid21: Complex Lipid Membrane Simulations with AMBER. *J. Chem. Theory Comput.* **18**, 1726–1736 (2022).
+75. Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W. & Klein, M. L. Comparison of simple potential functions for simulating liquid water. *J. Chem. Phys.* **79**, 926–935 (1983).
+76. Joung, I. S. & Cheatham, T. E. Determination of alkali and halide monovalent ion parameters for use in explicitly solvated biomolecular simulations. *J. Phys. Chem. B* **112**, 9020–9041 (2008).
+77. Ryckaert, J. P., Ciccotti, G. & Berendsen, H. J. C. Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. *J. Comput. Phys.* **23**, 327–341 (1977).
+78. Essmann, U. *et al.* A smooth particle mesh Ewald method. *J. Chem. Phys.* **103**, 8577–8593 (1995).
+
+# Supplementary Files
+
+- [Supportinginformation.docx](https://assets-eu.researchsquare.com/files/rs-2817693/v1/1fc68f9eb2affba25ea38c3c.docx)  
+  Supplemental material
+
+- [MovieS1clusteringKAV.avi](https://assets-eu.researchsquare.com/files/rs-2817693/v1/49492c344eef2045a15f531c.avi)  
+  Movie S1
+
+- [MovieS2deformingKAV.avi](https://assets-eu.researchsquare.com/files/rs-2817693/v1/6e97a64d1e7b2b267c9b91e8.avi)  
+  Movie S2
+
+- [MovieS3WT0017.avi](https://assets-eu.researchsquare.com/files/rs-2817693/v1/0842e5dd4e7a501998eed531.avi)  
+  Movie S3
+
+- [MovieS4N49K0041.avi](https://assets-eu.researchsquare.com/files/rs-2817693/v1/d68d7d4caf0a2e48f9260ce6.avi)  
+  Movie S4
+
+- [MovieS5E32QN49K0018.avi](https://assets-eu.researchsquare.com/files/rs-2817693/v1/23638aa2b7908867b3bc59ed.avi)  
+  Movie S5
+
+- [MovieS6MDsimulationofVSDdimer.mp4](https://assets-eu.researchsquare.com/files/rs-2817693/v1/4a87c831dcba5bfdebe4c722.mp4)  
+  Movie S6

0bb09d0a804159970c9d0eff9ced942ed2c788723b84cdeef3100fa1c8fa3d32/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpeg",
+        "caption": "Illustration of the fabrication process and microstructure characterizations. a Specific \u201cSolvent exchange-reprotonation\u201d processing strategy. b FESEM image of Mica@TiO2 microplatelet with lateral size of about 20 \u03bcm. c FESEM image of TiO2 nanograins with diameters of about 30 nm uniformly distributed on Mica surface. d,e OM (d) and FESEM (e) images of ANFs network with fibrillar joints, showing hyperbranched morphology. f Photograph of AMTA (with 50 wt% Mica@TiO2 loading) displaying its flexibility. g,h Top-view (g) and section-view (h) FESEM images of AMTA, showing a typical nacreous structure with lamellar micropores.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpeg",
+        "caption": "Mechanical properties and multiscale extrinsic toughening mechanisms. a Tensile stress-strain curves of ANFs/Mica@TiO2 composites via different preparation methods. b,c Ultimate strength, maximum Young\u2019s modulus (b) and strain at break, work of toughness (c) obtained by uniaxial tensile tests. d Crack propagation behaviors of AMTA observed from FESEM image and EDS elemental maps of Si, Ti and C, showing long-range crack deflection, branching and bridging.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpeg",
+        "caption": "Optical properties comparison among Mica@TiO2, Mica and TiO2. a-c Simulated scattering efficiencies (a), near-field electric field distributions at 0.5 \u03bcm wavelength (b) and far-field scattering phase function at 0.5 \u03bcm wavelength (c) for different monodisperse scatterers, where the size consistent with real situation. E and k represent electric field and wave vector of the incident light, respectively. 90\u00b0 denotes as incident direction. d Schematic illustration of multiple scattering behaviors of AMTA caused by dielectric contrast, including Mica@TiO2 and interlaminar micropores. e Solar reflectivity spectra of PDRC films respectively correspond to Mica@TiO2, Mica and TiO2 scatterers in the UV-VIS-NIR region, AM 1.5 global solar spectrum was shaded as reference. Inset highlights the reflectivity spectra of PDRC films with Mica@TiO2 scatterers.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpeg",
+        "caption": "Comprehensive optical properties and theoretical cooling performance. a Thickness optimization of AMTA based on the solar reflectivity. b Comparison of solar as a function of thickness with previous reports. c Spectral reflectivity and emissivity of a 25-\u03bcm-thick AMTA across 0.3-25 \u03bcm wavelength range against AM 1.5 solar spectrum and realistic atmospheric window. d Blackbody emissivity spectra from 0 to 50 \u2103, demonstrating that the emissivity of about 10 \u03bcm wavelength is more effective for improving the radiative cooling capacity of PDRC films. e Calculated cooling power of AMTA under different solar irradiance and atmospheric transmittance. f,g Calculated net cooling power of AMTA at daytime (f) and nighttime (g), respectively, with different q values.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.jpeg",
+        "caption": "Subambient radiative cooling performance and potential applications. a Schematic of simulated solar heating test. b,c Thermal infrared imaging (b) and temperature tracking (c) of bare, commercial white paint-covered and AMTA-covered Al plate, respectively. d,e Digital picture (d) and corresponding temperature variation (e) of commercial car cover and AMTA on the surface of identical car models at high noon in summer. f,g Digital picture (f) and corresponding temperature variation (g) of mobile phone covered with AMTA tailored for its shape and size, producing a comfortable state for outdoor running. h,i Photograph (h) and schematic (i) of home-built setup for subambient radiative cooling measurements. j 24 h real-time temperature recordings of ambient air, commercial white paint-covered and AMTA-covered Al plate (August 13, 2022). Detailed solar irradiance (Isolar), relative humidity (RH) and subambient temperature drop (\u0394T) were also shown in the corresponding graph. k Comparison of tensile strength and solar reflectivity of AMTA with various reported PDRC films. l Comprehensive performance comparison of AMTA with other universal cable insulation sheath.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0bb09d0a804159970c9d0eff9ced942ed2c788723b84cdeef3100fa1c8fa3d32/preprint/preprint.md

@@ -0,0 +1,155 @@
+# Abstract
+
+Passive daytime radiative cooling (PDRC) is a promising path to tackle energy, environment and security issues originated from global warming. However, the contradiction between optical properties (especially desired high solar reflectivity) and other applicable performance (e.g. strength, modulus, durability and thickness) limits the practical applications of PDRC. Herein, we demonstrate a nacreous PDRC film integrating aramid nanofibers (ANFs) network with core-shell TiO<sub>2</sub>-coated mica microplatelet (Mica@TiO<sub>2</sub>) scatterers via a “Solvent exchange-Reprotonation” processing strategy for enhancing mechanical strength and durability without compromising optical performance. The slow but complete two-step protonation transition regulates the three-dimensional dendritic ANFs network with strong fibrillar joints, where overloaded scatterers (> 50 wt%) are stably grasped and anchored in alignment, thereby resulting in a high strength of ~ 112 MPa. Meanwhile, the strong backward scattering excited by multiple interfaces of core-shell Mica@TiO<sub>2</sub> and interlamellar micropores guarantees a balanced reflectivity (~ 92%) and thickness (~ 25 µm). Notably, such design renders excellent environmental durability, including high temperature, UV radiation, water rinsing and scratch damage, to meet the realistic requirements. The practical PDRC cooling capability is further revealed by outdoor tests where attainable subambient temperature drops are ~ 3.35 ℃ for daytime and ~ 6.11 ℃ for nighttime, thus providing possibility for thermal protection of vehicles, mobile phones and cables exposed to direct sunlight. Consequently, both the cooling capacity which equals those of state-of-the-art PDRC designs and comprehensive outdoor-services performance, greatly push PDRC towards practical applications.
+
+Physical sciences/Materials science/Nanoscale materials/Two-dimensional materials  
+Physical sciences/Materials science/Materials for energy and catalysis/Porous materials  
+Physical sciences/Materials science/Structural materials/Mechanical properties  
+Physical sciences/Energy science and technology  
+Physical sciences/Nanoscience and technology/Nanoscale materials/Organic–inorganic nanostructures
+
+# Introduction
+
+While working to limit global warming to 1.5 ℃ proposed in the Paris Agreement<sup>1</sup>, it is undeniable that climate extremes around the world have been detected frequently, especially in hot summers, which has raised a series of energy, environment and security issues, including electricity consumption (e.g., air conditioners), carbon emissions and self-ignition (e.g., vehicles and cables)<sup>2,3</sup>. Passive daytime radiative cooling (PDRC) as an energy-efficient and eco-friendly cooling technology, with fundamental principles of reflecting most sunlight (0.3–2.5 µm) and emitting long-wave infrared (LWIR) radiation through atmospheric window (8–13 µm), has been highly regarded in recent years<sup>4–7</sup>. Ideally, subambient temperature drop of objects can be passively reached with zero energy input and zero pollution output since there is a sustained cold universe (~ 2.7 K) as power source for PDRC<sup>4,8,9</sup>.
+
+The key-point of PDRC materials is ultra-high reflectivity ($\overline{\text{R}}$)<sub>solar</sub> &gt; 90% across 0.3–2.5 µm wavelengths<sup>10,11</sup>, since just a few percent of solar absorbance would eclipse cooling capacity of LWIR radiation and effectively heat any exposed surfaces (Supplementary Fig. 1). Research in recent decades mainly focuses on two fields to design high-reflectivity PDRC films comprising porous structures<sup>12–18</sup> and polymer-dielectric scatterers composites<sup>11,19−22</sup>. However, the above-mentioned strategies, while achieving high reflectivity, typically mean losses of other applicable performance (e.g., strength, modulus, durability and thickness), which leads PDRC films unable to meet the long-term cooling requirements of outdoor devices in hot summers. Specifically, for porous structures, although air voids in polymer skeleton create strong multiple scattering, thus greatly enhancing solar reflectivity<sup>12,13</sup>. But increased cost and inherent mechanical weakness triggered by elevated thickness with abundant air holes or defects still remain as major problems, especially for all-polymeric PDRC films<sup>10,23</sup>, which were seldom resistant against environmental aging (e.g., UV radiation, water rinsing and scratch damage) and fire<sup>24</sup>. As for the improvement of solar reflectivity, introducing high-content dielectric scatterers to polymer matrix is another common alternative. Nevertheless, with overloaded scatterers (> 50 wt%), both limited-processing and scatterers leakage have resulted in difficult shaping and applicability of PDRC materials<sup>25,26</sup>. More importantly, the severe agglomeration of dielectric particles directly led to the substantial decline of optical and mechanical properties for PDRC films<sup>27</sup>, which was rarely discussed in previous reports. Therefore, to push PDRC towards practical applications, alleviating the agglomeration of scatterers, simultaneously realizing highly efficient PDRC and other applicable performance at relative low cost remain as challenge.
+
+Herein, taking inspiration from robust nacre, we demonstrated a hierarchical-morphology design mechanism, which incorporated dielectric scatterers to the porous structure, directly provides efficient PDRC and excellent comprehensive applicable performance. Two-dimensional (2D) core-shell scatterers composed of exfoliated Mica with surface uniformly distributed TiO<sub>2</sub> nanograins (Mica@TiO<sub>2</sub>) were assembled with aramid nanofibers (ANFs) via a “Solvent exchange-Reprotonation” processing strategy, successfully forming a nacreous “brick-and-mortar” PDRC film (AMTA). Specifically, by regulating the slow but complete two-step protonation transition, a three-dimensional (3D) dendritic ANFs network with strong fibrillar joints is formed, where more than 50 wt% scatterers are stably grasped and orderly embedded in AMTA with the help of “hyperbranched ANFs adhesives”, thereby exhibiting excellent mechanical strength of ~ 112 MPa and Young’s modulus of ~ 4 GPa. Besides, the barely agglomerated TiO<sub>2</sub> nanograins allow the sunlight to fully scattering at core-shell and shell-air interfaces, while the intense group vibrations in Mica greatly enhance the infrared absorption of AMTA within atmospheric window, so as to achieve a high $\overline{\text{R}}$<sub>solar</sub> of 92% and acceptable $\overline{\text{ε}}$<sub>LWIR</sub> of 87% at a fairly low thickness of 25 µm. We experimentally demonstrate high-performance daytime cooling of AMTA with an average subambient temperature drop of ~ 3.35 ℃ under direct sunlight, potentially applied to the mobile phones, vehicles and buildings for effective thermal management. Notably, our AMTA exhibited outstanding environmental durability with its optical and mechanical properties retained, even after challenged against 180 ℃ thermal treatment, 96 h UV radiation, 8 h water rinsing and scratch damage, promising for real-world applications of PDRC.
+
+# Results
+
+## Design, fabrication and characterization
+
+To meet the long-term cooling requirements of outdoor devices in harsh environments, radiative coolers should have the characteristics of high solar reflectivity, strong thermal emissivity, excellent mechanical properties and outstanding environmental durability. The solar reflectivity mainly depends on the extent of multiple sunlight scattering, which is caused by the refractive index contrast at the interface between two phases, that is, impedance mismatch<sup>28</sup>. Thus, a series of interface creation methods are derived, such as introducing pores or dielectric scatterers. As pores with a certain thickness significantly decrease the mechanical properties and thermal conductivity of radiative coolers, dielectric scatterers like TiO<sub>2</sub>, have become an ideal candidate due to its high refractive index (<em>k</em> > 2.5) and low cost. Inspired by the hierarchically ordered microstructure of nacre, ANFs with ultra-high strength (~ 3.6 GPa) and heat-resistance are appropriately used as an organic matrix to fabricate high mechanical performance materials. Mica@TiO<sub>2</sub>, a commercial pearlescent pigment, creatively combines exfoliated Mica microplatelet (Core) with uniformly distributed TiO<sub>2</sub> nanograins (Shell) to achieve both high refractive index and environmental durability, making it a desirable 2D inorganic building block (Fig. <span class="InternalRef" refid="Fig1">1</span> b,c). Besides, functional groups such as Si-C, Si-O and Ti-O can greatly improve the emissivity of radiative coolers within the atmospheric window.
+
+The nacreous PDRC films were assembled in a “Solvent exchange-Reprotonation” processing strategy schematically illustrated in Fig. <span class="InternalRef" refid="Fig1">1</span> a. Aramid microfibers (AMFs) were first dissociated into negatively charged ANFs by deprotonation in a mixture of dimethyl sulfoxide (DMSO) and potassium hydroxide (KOH). Mica@TiO<sub>2</sub> was pretreated with (3-aminopropyl)triethoxysilane (APTES) to make it positively charged (Supplementary Fig. 3). Afterward, ANFs was mixed with APTES treated Mica@TiO<sub>2</sub>, forming a homogeneous flaxen sol via synergetic cross-linking such as electrostatic interaction and hydrogen bonding. Subsequently, the uniform sol was slowly dropped into the high velocity flow field of isopropanol (IPA), involving solvent exchange and turbulent shear<sup>29</sup>. During the process of preliminary protonation, we note that ANFs flocculated into hyperbranched morphology, which was conducive to assemble highly connected networks with fibrillar joints (Fig. <span class="InternalRef" refid="Fig1">1</span> d,e), so that high-content Mica@TiO<sub>2</sub> could be stably dispersed for several hours (Supplementary Fig. 5). Moreover, the vacuum-assisted filtration (VAF) method was used to prepare nacreous films with lamellar micropores (Fig. <span class="InternalRef" refid="Fig1">1</span> h) based on the above dendritic colloid suspension. These films were further immersed into water bath to reprotonate the ANFs network completely, resulting in the high mechanical performance radiative coolers (Fig. <span class="InternalRef" refid="Fig1">1</span> f).
+
+## Mechanical performance and toughening mechanisms analysis
+
+Mechanical performance is particularly considerable for devices that work outdoors for long period of time. To demonstrate the superiority of above “Solvent exchange-Reprotonation” processing strategy, the mechanical properties of nacreous films were systematically studied and compared to those composites fabricated through conventional methods<sup>30, 31</sup>. As shown in Fig. <span class="InternalRef" refid="Fig2">2</span> a-c, the tensile strength, Young’s modulus as well as toughness of H<sub>2</sub>O-stirring composites with the optimal Mica@TiO<sub>2</sub> loading (30 wt%) can barely reach ~ 81.5 MPa, ~ 3.6 GPa and ~ 4.9 MJ/m<sup>3</sup>, respectively. Because the ANFs overquick reprotonation provided by the strong proton donor (H<sub>2</sub>O) is uneven and incomplete. Instead by solvent exchange and turbulent shear in IPA, slow protonation process and the formation of ANFs network with fibrillar joints can significantly improve the mechanical properties of the composites. Furthermore, through the complete reprotonation in water, the robust highly ordered brick-and-mortar microstructure are achieved with the strong electrostatic interaction and multiple hydrogen bonding between microfibers, demonstrating a tensile strength of ~ 193.2 MPa, a Young’s modulus of 4.9 GPa and a toughness of 16.2 MJ/m<sup>3</sup>, which are far superior to the previously reported PDRC films.
+
+To further analyze the hierarchical toughening mechanisms of these ANFs/Mica@TiO<sub>2</sub> nacreous films, the crack propagation behaviors during tensile test are observed by SEM and EDS, illustrated in Fig. <span class="InternalRef" refid="Fig2">2</span> d. The major crack starts from the notch and propagates along a circuitous path in the nacreous films. Specifically, whenever the embedded Mica@TiO<sub>2</sub> microplatelet is encountered, the crack will deflect and branch in advance, resulting in abundant tiny cracks uniformly distributed inside the materials. Besides the typical crack deflection, the nacre-mimetic structure with lamellar micropores also leads crack bridging and delamination, which effectively relieve the local high stress. Meanwhile, the TiO<sub>2</sub> nanograins wrapped on the Mica can achieve extrinsic toughening through inelastic shear resistance caused by the roughness of 19.8 nm (Supplementary Fig. 2d-f), in accordance with those of aragonite whiskers in nacre<sup>32</sup>.
+
+In addition, the effects of APTES modification, dielectric scatterer content and type on mechanical properties of these composites were also studied (Supplementary Figs. 7 and 8). After being pretreated by APTES, the hydroxyl groups on the surface of Mica@TiO<sub>2</sub> react with the silane groups of APTES, which can facilitate the interfacial interaction between Mica@TiO<sub>2</sub> and ANFs network. For different types of dielectric scatterers, such as Mica@TiO<sub>2</sub>, Mica and TiO<sub>2</sub>, their intrinsic geometry has significant impact on the mechanical performance of these PDRC films. Generally speaking, 2D platelet scatterers with inherent ultra-high aspect-ratio, like Mica@TiO<sub>2</sub> and Mica, exhibit outstanding mechanical properties, especially in strength and modulus. At the same time, the nacreous microstructure assembled with 2D platelets and ANFs network can further strengthen and toughen these PDRC films<sup>33</sup>. However, for TiO<sub>2</sub> scatterers, the agglomeration phenomenon caused by high specific surface energy of zero-dimensional nanoparticles leads to poor mechanical properties and weak light scattering effect<sup>34</sup>.
+
+## Optical properties and theoretical radiative cooling performance
+
+Ultra-high solar reflectivity is one of the necessary requirements for these PDRC films, which is mainly determined by the multiple scattering behaviors of sunlight inside materials. Regarding the light scattering, it can be divided into three predominantly defined scattering regions according to the size of scatterers, namely, Rayleigh scattering, Mie scattering and geometric scattering<sup>35</sup>. Because of the larger or comparable size of micro-sized Mica and nano-sized TiO<sub>2</sub> to the wavelength of sunlight, Mica@TiO<sub>2</sub> mainly follows the Mie scattering and geometric scattering mechanisms. Based on the scattering theory, the scattering efficiency (<em>Q</em><sub>sca</sub>) and scattering phase function play crucial roles in the final reflectivity. The <em>Q</em><sub>sca</sub> is derived from the scattering cross-section (<em>C</em><sub>sca</sub>), which is determined by the ratio of scattering light to incident light, while the scattering phase function denotes the intensity of scattering light in various directions. Since backscattering makes greater contributions to the final reflectivity, a parameter (<em>σ</em>) is redefined to describe the ratio of backscattering to total scattering.
+
+$$\begin{array}{c}\text{σ}\text{ }\\text{=}\\text{ }\\frac{{\\text{S}}_{\\text{back}}}{{\\text{S}}_{\\text{total}}}\\text{ }\\text{×}\\text{ }\\text{100}\\text{\\%}\\#\\text{(}\\text{1}\\text{)}\\end{array}$$
+
+where <em>S</em><sub>back</sub> is the integrated area of backscattering, and <em>S</em><sub>total</sub> is the total integrated area of scattering phase function.
+
+Finite-Difference Time-Domain (FDTD) simulations were employed to investigate the optical properties of different monodisperse scatterers. As shown in Fig. <span class="InternalRef" refid="Fig3">3</span> a, compared with Mica microplatelet or TiO<sub>2</sub> nanograin alone, core-shell Mica@TiO<sub>2</sub> exhibits higher <em>Q</em><sub>sca</sub> in the VIS-NIR band, and reaches the peak at ~ 500 nm where the solar irradiation is the strongest. That means that sunlight would diffuse intricate pathways, leading to an enormous increase in the proportion of reflected light. To further verify the backscattering behavior of Mica@TiO<sub>2</sub>, near-field electric field distribution and far-field scattering phase function were established (Fig. <span class="InternalRef" refid="Fig3">3</span> b,c). It can be seen that the scattering direction of TiO<sub>2</sub> has no obvious deviation owing to the Rayleigh scattering, while for the Mica and Mica@TiO<sub>2</sub>, they represent the overall forward scattering, which is one of the typical characteristics of Mie scattering. Nevertheless, in-depth analysis shows that Mica@TiO<sub>2</sub> is provided with much stronger backward electric field distribution, which is different from Mica. Meanwhile, in the far field scattering phase function, Mica@TiO<sub>2</sub> is also consistent with expectation, i.e., the deflection angle of the scattering light to be as large as possible, preferably backward (larger <em>σ</em>).
+
+Since just a few percent of solar absorbance would effectively heat the surface, the spectral reflectivity of the PDRC films across the UV-VIS-NIR band is further quantitatively characterized (Fig. <span class="InternalRef" refid="Fig3">3</span> e). A 25 µm thick AMTA reflects more than 92% solar irradiation (0.3–2.5 µm), and increases rapidly with the elevation of Mica@TiO<sub>2</sub> loading. This is significantly superior to the TiO<sub>2</sub> based composite films at the same mass fraction, which also corresponds to the previous optical simulation results. The distribution of scatterers in AMTA interiorly is a key factor affecting its optical properties. Specifically, regular, dense and non-agglomerated TiO<sub>2</sub> nanograins wrapped on the Mica are conducive to create abundant dielectric contrast interfaces for light scattering (Fig. <span class="InternalRef" refid="Fig3">3</span> d) owing to their high refractive index. So that AMTA is capable to achieve high reflectivity at fairly low thickness (≤ 25 µm), which is much thinner than the conventional polymeric PDRC films usually > 300 µm (Fig. <span class="InternalRef" refid="Fig4">4</span> b). On the other hand, similar to the porous polymers reported in the past<sup>13</sup>, the interlaminar micropores caused by the solvent evaporation contribute to the high reflectivity of AMTA as well through lamellar interface scattering (Supplementary Figs. 12 and 13).
+
+It is worth noting that AMTA with internal Si-C, Si-O and Ti-O vibrations (Supplementary Fig. 6a) shows acceptable atmospheric window emissivity (<span class="InlineEquation"><span class="mathinline">\\(\\overline{\\text{ε}}\\)</span></span><sub>LWIR</sub> ~87%) at such a low thickness, which directly determines the theoretical upper limit of the cooling power. According to the heat transfer mode and the law of conversation of energy, combined with the measured full-wave band spectrum (Fig. <span class="InternalRef" refid="Fig4">4</span> c), the cooling power of AMTA can be theoretically calculated by the relevant equations (detailed calculation methods are shown in the Supplementary Information). Considering the diversity of solar irradiance (<em>I</em><sub>solar</sub>) and atmospheric transmittance (<em>t</em><sub>LWIR</sub>) at different longitude and latitude in the world, we first calculated the cooling power of AMTA under various environments, as shown in Fig. <span class="InternalRef" refid="Fig4">4</span> e. As expected, lower <em>I</em><sub>solar</sub> and higher <em>t</em><sub>LWIR</sub> are beneficial for higher cooling power. And even in a relatively harsh environment (high <em>I</em><sub>solar</sub> and low <em>t</em><sub>LWIR</sub>), ~ 100 W/m<sup>2</sup> cooling power can be obtained, demonstrating the universal and efficient radiative cooling capacity of AMTA. Moreover, the calculated net cooling power (<em>P</em><sub>net</sub>) and achievable cooling temperature (<em>T</em><sub>a</sub> - <em>T</em><sub>c</sub>) of AMTA for different non-radiative heat coefficient (<em>q</em>) are referred in Fig. <span class="InternalRef" refid="Fig4">4</span> f-g. There are two key points should be focus on: one is the <em>P</em><sub>net</sub> at <em>T</em><sub>a</sub> - <em>T</em><sub>c</sub> = 0 (<em>P</em><sub>net</sub>(<em>T</em><sub>a</sub> = <em>T</em><sub>c</sub>)), which means that the cooling power totally derived from thermal radiation, and the other is the maximum cooling temperature at <em>P</em><sub>net</sub> = 0 ((<em>T</em><sub>a</sub> - <em>T</em><sub>c</sub>)<sub>max</sub>). Comparing Fig. <span class="InternalRef" refid="Fig4">4</span> f with Fig. <span class="InternalRef" refid="Fig4">4</span> g, we can notice that <em>P</em><sub>net</sub>(<em>T</em><sub>a</sub> = <em>T</em><sub>c</sub>) and (<em>T</em><sub>a</sub> - <em>T</em><sub>c</sub>)<sub>max</sub> are both lower during daytime. This is mainly resulted from the solar absorbance under direct sunlight since the <span class="InlineEquation"><span class="mathinline">\\(\\overline{\\text{R}}\\)</span></span><sub>solar</sub> < 100%. Notably, <em>P</em><sub>net</sub>(<em>T</em><sub>a</sub> = <em>T</em><sub>c</sub>) of AMTA reaches 74 W/m<sup>2</sup> and 106 W/m<sup><span citationid="CR2" class="CitationRef">2</span></sup> during daytime and nighttime, respectively, both resulting more than 15 ℃ (<em>T</em><sub>a</sub> - <em>T</em><sub>c</sub>)<sub>max</sub> when there is zero non-radiative heat. More significantly, a large temperature drop of ~ 6 ℃ can be still achieved during daytime even with a <em>q</em> = 8 W/m<sup>2</sup> K.
+
+## Subambient radiative cooling performance and potential applications
+
+In order to preliminarily evaluate the capability of sunlight reflection and thermal management of AMTA, simulated sunlight of 1000 W/m<sup>2</sup> was carried out to heat the aluminium (Al) plate, while infrared images and temperatures were captured simultaneously by an IR camera (Fig. <span class="InternalRef" refid="Fig5">5</span> a-c). Within 12 min, the temperature of AMTA-covered Al plate only rose from 32 to 46 ℃ and gradually leveled off, much lower than the bare Al plate (61 ℃) and the one coated with commercial white paint (54 ℃). This is mainly due to the effective solar heat shielding and considerable heat dissipation by AMTA (Supplementary Fig. 15). The actual subambient radiative cooling performance of AMTA was characterized in Chengdu, China (104°2′35″E, 30°38′32″N, 570-m altitude) in early August employing the home-built contraption shown in Fig. <span class="InternalRef" refid="Fig5">5</span> h-i. For better steady and accuracy, insulated polystyrene foams and transparent polyethylene films were applied to eliminate the influence of thermal conduction and convection, respectively. The AMTA tightly covered on the Al plate was subjected to direct sunlight under a clear day, and its temperature variation throughout the day was monitored by a thermocouple. Additionally, temperatures of both the environment and the Al plate coated with commercial white paint were also registered for comparison. The corresponding subambient temperature drop (ΔT) can be obtained by subtracting the ambient temperature curves, as shown in Fig. <span class="InternalRef" refid="Fig5">5</span> j. Promisingly, the AMTA exhibited continuous high-performance subambient radiative cooling. Under the highest <em>I</em><sub>solar</sub> of ~ 600W/m<sup>2</sup> and a relative humidity (RH) of ~ 20% at noon, a critical average ΔT of 3 ~ 4 ℃ was achieved. And at night, it can easily reach a high cooling effect of more than 6 ℃ even under a RH of ~ 60% (low <em>t</em><sub>LWIR</sub>). Compared with commercial white paint used in buildings and vehicles, our AMTA demonstrated obvious advantages in outdoor cooling.
+
+The AMTA with a fairly low thickness presents a combination of high tensile strength and excellent optical properties, substantially superior to a variety of previously reported PDRC films (Fig. <span class="InternalRef" refid="Fig5">5</span> k)<sup>36</sup>, offering a momentous candidate for the cooling requirements of outdoor devices. Besides, electrical insulation and thermal stability are common features of ANFs and Mica (Supplementary Figs. 16 and 18), allowing AMTA potentially helpful in the protection of high-temperature cables to avoid security problems like spontaneous combustion. To mimic real-world operating conditions, several simulated field tests were performed, such as thermal treatment, water rinsing, scratch treatment and UV radiation (Supplementary Fig. 19). The <span class="InlineEquation"><span class="mathinline">\\(\\overline{\\text{R}}\\)</span></span><sub>solar</sub> and tensile strength of AMTA could be maintained stably regardless of different harsh environments, revealing its great environmental durability. Utilizing the remarkable comprehensive performance, AMTA can be widely employed in various thermal management systems in our daily activities. For instance, the application scenarios of AMTA in car cover and mobile phone case are demonstrated as shown in Fig. <span class="InternalRef" refid="Fig5">5</span> d-g. After running a 2 hours experiment exposed to the sunlight, average temperature drop ΔT of 10 ℃ and 3 ℃ were observed for the covered devices, respectively. Such strong cooling effects established a safe and comfortable state for outdoor devices, leaving AMTA a promising PDRC material.
+
+# Discussion
+
+To summarize, we have proposed and demonstrated nacreous ANFs/Mica@TiO₂ radiative coolers with excellent comprehensive performance via an effective “Solvent exchange-Reprotonation” processing strategy, towards real-world applications of PDRC. The two-step protonation transition not only promotes the transformation of nanofibers to strong microfibers, but also shapes a dendritic network with fibrillar joints as cores, enabling overloaded scatterers to be stably grasped and orderly embedded into nacre-like “brick-and-mortar” microstructure with outstanding mechanical strength of ~112 MPa and Young’s modulus of ~4 GPa. Meanwhile, by simultaneously introducing core-shell Mica@TiO₂ scatterers and interlamellar micropores into AMTA, strong multiple scattering at core-shell and shell-air interfaces yields a high $\overline{\text{R}}$<sub>solar</sub> of 92% at a quite low thickness of 25 µm. Further combined with a $\overline{\text{ε}}$<sub>LWIR</sub> of 87%, such design can generate an average subambient temperature drop of ~3.35 ℃ under direct sunlight. Besides, considering the harsh outdoor environment, long-term durability including high temperature, UV radiation, water rinsing and scratch damage have been integrated into AMTA. With respect to its overall performance combinations, AMTA possesses great superiority compared with commercial pearlescent coatings and currently existing polymeric PDRC materials, expecting it a prospective material for meeting long-term cooling requirements of outdoor devices.
+
+# Methods
+
+## Raw materials
+Raw AMFs (Kevlar 49) were purchased from DuPont, USA. Mica, TiO₂ and Mica@TiO₂ were offered by Fujian Kuncai Material Technology Co. Ltd., China. Chemicals including DMSO, IPA, KOH and APTES were all purchased from Aladdin Reagent Limited Corporation, China and used without any further purifications.
+
+## Fabrication of Nacreous PDRC Films
+According to a previous report<sup>37</sup>, 2 g raw AMFs and 2 g KOH were first added into 100 mL DMSO. After mechanically stirring for 7 days at room temperature, a uniform, viscous and crimson ANFs dispersion was obtained and further diluted to 0.8 wt% by DMSO for subsequent use. As pretreatment, 4.5 g Mica@TiO₂ was dispersed in the mixture of ethanol and deionized water (70 mL, 9:1 w/w), added 3 mL APTES, and refluxed at 70 ℃ for 4 h. After that, Mica@TiO₂ dispersion was filtered and washed five times with deionized water. APTES treated Mica@TiO₂ was dispersed into DMSO (10 mg/mL) via ultrasonic 20 min and mixed with the obtained ANFs dispersion under mechanical stirring to form homogeneous flaxen ANFs/Mica@TiO₂ sols with different mass ratio. Subsequently, the uniform sol was slowly dropped into the highly turbulent flow of IPA, which was formed by high-speed shearing of IPA in IKA Magic Lab device (IKA Works Inc.) operating at 10,000 rpm (Solvent exchange)<sup>38</sup>. Based on the obtained stable dendritic colloid suspension, free-standing nacreous PDRC films were directly prepared via a layer-by-layer VAF method followed by soaking in a water bath for 24 h (Reprotonation). A series of nacreous PDRC films with different Mica@TiO₂ loading had been fabricated, in which 50 wt% sample was denoted as AMTA. For comparison, ANFs/Mica and ANFs/TiO₂ composite films were prepared by the similar method described above.
+
+## Characterization
+The digital pictures were photographed by iPhone SE (12MP Main) and the microscopic morphology was revealed with optical microscope (OM, Olympus BX51, Japan), field emission scanning electron microscope (FESEM, Apero S HiVac, FEI, USA) and atomic force microscope (AFM, Bruker, USA). Energy dispersive X-ray spectroscopy (EDS, Octane Elect Super, EDAX, USA) mapping was used to check the distribution of elements on surface. Zeta potential of Mica@TiO₂ was characterized with Brookhaven Zeta PALS 190 Plus. X-ray photoelectron spectroscopy (XPS) spectra were characterized on an ESCALab Xi+ (ThermoScientific, USA) using a monochromatic Al-Kα X-ray source. The component and structure of Mica@TiO₂ as well as nacreous PDRC films were characterized by Fourier-transform infrared spectroscopy (FTIR, Nicolet 6700, Thermo Scientific, USA) and X-ray diffraction (XRD, DY1291, Philips, Holland) with a Cu-Kα radiation source. Thermal stability of nacreous PDRC films was performed on a thermogravimetric analyzer (TGA, Q500, TA, USA) from 25 to 800 ℃ at a heating rate of 10 ℃ min⁻¹ under nitrogen condition. The DC breakdown strength and volume resistivity of the prepared insulating PDRC films were carried out on a voltage-withstanding tester (DDJ-50 kV, Guance Electronics Co. Ltd., China) at room temperature with ~45% relative humidity and an ultra-high resistance micro-current tester (ZST-121, Zhonghang Times Instrument Co. Ltd., China), respectively. Thermal conductivity (λ) of nacreous PDRC films was calculated from λ = α × ρ × Cₚ, where α, ρ and Cₚ respectively correspond to thermal diffusivity, mass density and specific heat. Laser flash analysis (LFA 467 Hyper Flash, Netzsch, Germany) was used to measure α as the voltage and pulse width were set to be 250 V and 200 µs, respectively.
+
+## Mechanical Testing
+Tensile stress-strain curves were recorded using a universal test instrument (Instron 5967, USA) with a 500 N load cell at room temperature. The sample length and width for all the PDRC films were 30 and 5 mm, respectively. The thickness of each tested sample strip was obtained by averaging thickness values at 3 to 5 different positions and at least five specimens were measured for all different groups. The gauge distance and loading rate were 15 mm and 5 mm/min, respectively. For single-edge notched tensile tests, the samples with the size of 30 mm × 10 mm were slightly notched to approximately one-third of their widths by a razor blade. Further tensile tests were carried out to expand notches without fracture of the splines, while the crack propagation behaviors were observed by FESEM.
+
+## Optical Spectrum Characterization
+Optical characterization was conducted across UV to MIR wavelengths. The reflectivity (R) and transmissivity (T) were measured using a UV-VIS-NIR spectrometer (Lambda 1050, PerkinElmer, USA) in the range of 0.25–2.5 µm with an integrating sphere, while in MIR range (2.5–25 µm) using a FTIR spectrometer (Nicolet iS50, Thermo Scientific, USA) equipped with a diffuse gold integrating sphere (Pike Technologies, USA). The emissivity (ε) is equal to absorptivity, which is obtained from the equation ε = 1 – R – T.
+
+Definition of average solar reflectivity ($\overline{R}$)ₐₒₗₐᵣ is given by:
+
+$$
+\begin{array}{c}
+{\overline{R}}_{\text{solar}}\text{ }=\text{ }\frac{{\int }_{\text{0.3 μm}}^{\text{2.5 μm}}{I}_{\text{solar}}\text{(λ)}\text{ }×\text{ }{R}_{\text{solar}}\text{(λ)dλ}}{{\int }_{\text{0.3 μm}}^{\text{2.5 μm}}{I}_{\text{solar}}\text{(λ)dλ}}\#(2)
+\end{array}
+$$
+
+where λ is the wavelength, Iₛₒₗₐᵣ(λ) is the reference direct normal spectral irradiance ASTM G173 under air-mass 1.5, representing the global solar intensity, and Rₛₒₗₐᵣ(λ) is the spectral reflectivity of the surface.
+
+Definition of average atmospheric window emissivity ($\overline{ε}$)ₗᵥᵢᵣ is given by:
+
+$$
+\begin{array}{c}
+{\overline{ε}}_{\text{LWIR}}\text{ }=\text{ }\frac{{\int }_{\text{8 μm}}^{\text{13 μm}}{I}_{\text{BB}}\text{(T, λ) × }ε\text{(T, λ)dλ}}{{\int }_{\text{8 μm}}^{\text{13 μm}}{I}_{\text{BB}}\text{(T, λ)dλ}}\#(3)
+\end{array}
+$$
+
+where I₆₆(T, λ) is the spectral irradiance emitted by a blackbody at temperature T (here assumed to be 25 ℃) shown in Fig. 4d, and ε(T, λ) is the spectral thermal emissivity of the surface.
+
+## Stability and Durability Tests
+(1) To evaluate the thermal stability in long-term direct exposure to the sunlight, AMTA was sandwiched between two Al plates and placed on a heating stage at 180 ℃ for 8 h. (2) To mimic the unexpected rainstorm environment, AMTA was rinsed under the high-speed (~6 m/s) water jet for 8 h. (3) Scratch resistance test was orthogonally applied onto the AMTA (8×8 cm) surface under a normal load, over a distance of 6 cm at a constant scratch rate of 2 cm/s. (4) Accelerated UV aging tests were conducted in a homemade weathering chamber equipped with an iodine lamp (365 nm maximum intensity, UV irradiance of 10.0±0.5 W/m²) at 35 ℃ for 24, 48, 72 and 96 h, respectively. The UV radiation dosage of a 96 h test is equivalent to 1 year of Florida sunshine exposure (annual UV dosage of about 280 MJ/m²), which is an international benchmark for UV durability tests. The $\overline{R}$ₐₒₗₐᵣ and tensile strength of AMTA before and after treatment were shown in Supplementary Fig. 16e.
+
+## Data Availability
+The data that support the findings of this study are available from the corresponding authors upon reasonable request.
+
+# References
+
+1. Ipcc. *Global Warming of 1.5°C: IPCC Special Report on Impacts of Global Warming of 1.5°C above Pre-industrial Levels in Context of Strengthening Response to Climate Change, Sustainable Development, and Efforts to Eradicate Poverty*, (Cambridge University Press, Cambridge, 2022).
+2. The Cold Road to Paris Mapping Pathways Toward Sustainable Cooling for Resilient People and Economies by 2050. (World Bank, 2021).
+3. Xu, C., Kohler, T.A., Lenton, T.M., Svenning, J.-C. & Scheffer, M. Future of the human climate niche. *Proc. Natl. Acad. Sci. U.S.A.* 117, 11350–11355 (2020).
+4. Raman, A.P., Anoma, M.A., Zhu, L., Rephaeli, E. & Fan, S. Passive radiative cooling below ambient air temperature under direct sunlight. Nature 515, 540–544 (2014).
+5. Liang, J., *et al.* Radiative cooling for passive thermal management towards sustainable carbon neutrality. Nat. Sci. Rev. 10, nwac208 (2023).
+6. Fan, S. & Li, W. Photonics and thermodynamics concepts in radiative cooling. Nat. Photonics 16, 182–190 (2022).
+7. Zhao, D., *et al.* Radiative sky cooling: Fundamental principles, materials, and applications. Appl. Phys. Rev. 6, 021306 (2019).
+8. Bhatia, B., *et al.* Passive directional sub-ambient daytime radiative cooling. Nat. Commun. 9, 5001 (2018).
+9. Zhang, Q., *et al.* Recent Progress in Daytime Radiative Cooling: Advanced Material Designs and Applications. Small Methods 6, 2101379 (2022).
+10. Yao, P., *et al.* Spider-Silk-Inspired Nanocomposite Polymers for Durable Daytime Radiative Cooling. Adv. Mater. 34, 2208236 (2022).
+11. Song, J., *et al.* Durable radiative cooling against environmental aging. Nat. Commun. 13, 4805 (2022).
+12. Liu, J., *et al.* Micro-Nano Porous Structure for Efficient Daytime Radiative Sky Cooling. Adv. Funct. Mater. 32, 2206962 (2022).
+13. Mandal, J., *et al.* Hierarchically porous polymer coatings for highly efficient passive daytime radiative cooling. Science 362, 315–319 (2018).
+14. Liu, X., *et al.* Hierarchically Superhydrophobic Stereo-Complex Poly (Lactic Acid) Aerogel for Daytime Radiative Cooling. Adv. Funct. Mater. 32, 2207414 (2022).
+15. Zhong, H., *et al.* Hierarchically Hollow Microfibers as a Scalable and Effective Thermal Insulating Cooler for Buildings. ACS Nano 15, 10076–10083 (2021).
+16. Cai, L., *et al.* Spectrally Selective Nanocomposite Textile for Outdoor Personal Cooling. Adv. Mater. 30, 1802152 (2018).
+17. Xiang, B., *et al.* 3D porous polymer film with designed pore architecture and auto-deposited SiO2 for highly efficient passive radiative cooling. Nano Energy 81, 105600 (2021).
+18. Chen, M., *et al.* Designing Mesoporous Photonic Structures for High-Performance Passive Daytime Radiative Cooling. Nano Lett. 21, 1412–1418 (2021).
+19. Hu, D., *et al.* Hollow Core-Shell Particle-Containing Coating for Passive Daytime Radiative Cooling. Composites Part A 158, 106949 (2022).
+20. Li, P., *et al.* Thermo-Optically Designed Scalable Photonic Films with High Thermal Conductivity for Subambient and Above-Ambient Radiative Cooling. Adv. Funct. Mater. 32, 2109542 (2022).
+21. Zhang, H., *et al.* Biologically inspired flexible photonic films for efficient passive radiative cooling. *Proc. Natl. Acad. Sci. U.S.A.* 117, 14657–14666 (2020).
+22. Mandal, J., Yang, Y., Yu, N. & Raman, A.P. Paints as a Scalable and Effective Radiative Cooling Technology for Buildings. Joule 4, 1350–1356 (2020).
+23. Xu, Y., *et al.* Multiscale porous elastomer substrates for multifunctional on-skin electronics with passive-cooling capabilities. *Proc. Natl. Acad. Sci. U.S.A.* 117, 205–213 (2020).
+24. Li, T., *et al.* A radiative cooling structural material. Science 364, 760–763 (2019).
+25. Li, X., *et al.* Full Daytime Sub-ambient Radiative Cooling in Commercial-like Paints with High Figure of Merit. Cell Rep. Phys. Sci. 1, 100221 (2020).
+26. Huang, W., *et al.* Scalable Aqueous Processing-Based Passive Daytime Radiative Cooling Coatings. Adv. Funct. Mater. 31, 2010334 (2021).
+27. Wang, C.-H., Liu, M.-X. & Jiang, Z.-Y. TiO2 particle agglomeration impacts on radiative cooling films with a thickness of 50 µm. Appl. Phys. Lett. 121, 202204 (2022).
+28. Wang, B. & Cao, Z. A Review of Impedance Matching Techniques in Power Line Communications. Electronics 8, 1022 (2019).
+29. Zhang, X., *et al.* Biomimetic Aerogel for Moisture-Induced Energy Harvesting and Self-Powered Electronic Skin. Adv. Funct. Mater. 33, 2210027 (2023).
+30. Wei, Y., *et al.* “Toolbox” for the Processing of Functional Polymer Composites. Nano-Micro Lett. 14, 35 (2021).
+31. Yang, B., *et al.* Fabrication, Applications, and Prospects of Aramid Nanofiber. Adv. Funct. Mater. 30, 2000186 (2020).
+32. Ritchie, R.O. The conflicts between strength and toughness. Nat. Mater. 10, 817–822 (2011).
+33. Han, Z.-M., *et al.* Nacre-Inspired Nanocomposite Films with Enhanced Mechanical and Barrier Properties by Self-Assembly of Poly(Lactic Acid) Coated Mica Nanosheets. Adv. Funct. Mater. 32, 2202221 (2022).
+34. Zare, Y. Study of nanoparticles aggregation/agglomeration in polymer particulate nanocomposites by mechanical properties. Composites Part A 84, 158–164 (2016).
+35. Xu, R. Light scattering: A review of particle characterization applications. Particuology 18, 11–21 (2015).
+36. Wu, S., Cao, Y., Li, Y. & Sun, W. Recent Advances in Material Engineering and Applications for Passive Daytime Radiative Cooling. Adv. Opt. Mater. 11, 2202163 (2023).
+37. Yang, M., *et al.* Dispersions of Aramid Nanofibers: A New Nanoscale Building Block. ACS Nano 5, 6945–6954 (2011).
+38. Zhang, X., *et al.* Fabricating high performance multi-functional hygroelectric generator through a biomimic approach. Nano Energy 98, 107241 (2022).
+
+# Supplementary Files
+
+- [SupportingInformation.docx](https://assets-eu.researchsquare.com/files/rs-2776901/v1/a7236af9c68efb1a2927dae7.docx)  
+  Supplementary Information
+
+- [SupplementaryMovie1.mov](https://assets-eu.researchsquare.com/files/rs-2776901/v1/1b3cfde7405e78e9c6aff040.mov)  
+  Video of Scattering behavior of Mica@TiO2
+
+- [SupplementaryMovie2.mov](https://assets-eu.researchsquare.com/files/rs-2776901/v1/4c6ccb760d4222fba8ee7926.mov)  
+  Video of Scattering behavior of Mica
+
+- [SupplementaryMovie3.mov](https://assets-eu.researchsquare.com/files/rs-2776901/v1/ef9d483664b7215145cd4624.mov)  
+  Video of Scattering behavior of TiO2