Add files using upload-large-folder tool

.gitattributes

@@ -63,3 +63,23 @@ dd9fa622529abdee9160c93e7c3ad7399577e19fb9862692ba9a7f882cc52c8a/preprint/images
 507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_6.tif filter=lfs diff=lfs merge=lfs -text
 507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_7.tif filter=lfs diff=lfs merge=lfs -text
 507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_8.tif filter=lfs diff=lfs merge=lfs -text
+f5c80281108a293c84cfbd7b658f6a42c351103e8bea162bd6e9fa156f8fbb7c/preprint/images/Figure_1.tif filter=lfs diff=lfs merge=lfs -text
+f5c80281108a293c84cfbd7b658f6a42c351103e8bea162bd6e9fa156f8fbb7c/preprint/images/Figure_3.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_4.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_5.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_2.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_3.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_1.tif filter=lfs diff=lfs merge=lfs -text
+e4bad1fa57ec3421099366c905112a01ef7b2d78605745726457e03df0878ad3/preprint/images/Figure_1.tif filter=lfs diff=lfs merge=lfs -text
+e4bad1fa57ec3421099366c905112a01ef7b2d78605745726457e03df0878ad3/preprint/images/Figure_2.tif filter=lfs diff=lfs merge=lfs -text
+e4bad1fa57ec3421099366c905112a01ef7b2d78605745726457e03df0878ad3/preprint/images/Figure_3.tif filter=lfs diff=lfs merge=lfs -text
+e4bad1fa57ec3421099366c905112a01ef7b2d78605745726457e03df0878ad3/preprint/images/Figure_4.tif filter=lfs diff=lfs merge=lfs -text
+e4bad1fa57ec3421099366c905112a01ef7b2d78605745726457e03df0878ad3/preprint/images/Figure_5.tif filter=lfs diff=lfs merge=lfs -text
+e4bad1fa57ec3421099366c905112a01ef7b2d78605745726457e03df0878ad3/preprint/images/Figure_6.tif filter=lfs diff=lfs merge=lfs -text
+e4bad1fa57ec3421099366c905112a01ef7b2d78605745726457e03df0878ad3/preprint/images/Figure_7.tif filter=lfs diff=lfs merge=lfs -text
+71a319bec34be095ccbdc2b64f1e46fe8285018b930a6fcec682fee5e5907994/preprint/images/Figure_6.tif filter=lfs diff=lfs merge=lfs -text
+71a319bec34be095ccbdc2b64f1e46fe8285018b930a6fcec682fee5e5907994/preprint/images/Figure_5.tif filter=lfs diff=lfs merge=lfs -text
+71a319bec34be095ccbdc2b64f1e46fe8285018b930a6fcec682fee5e5907994/preprint/images/Figure_1.tif filter=lfs diff=lfs merge=lfs -text
+71a319bec34be095ccbdc2b64f1e46fe8285018b930a6fcec682fee5e5907994/preprint/images/Figure_3.tif filter=lfs diff=lfs merge=lfs -text
+71a319bec34be095ccbdc2b64f1e46fe8285018b930a6fcec682fee5e5907994/preprint/images/Figure_4.tif filter=lfs diff=lfs merge=lfs -text
+71a319bec34be095ccbdc2b64f1e46fe8285018b930a6fcec682fee5e5907994/preprint/images/Figure_2.tif filter=lfs diff=lfs merge=lfs -text

031e368a469fa7a49d7f45c82ca8e17995cf9776aa7caf1952c8b7077ee3f2da/preprint/images_list.json

@@ -0,0 +1,82 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "(a) POM diagrams of PR-5CB resins at 30 \u00b0C; (b) Relationship of viscosity and shear rate for PB-5CB resins; (c) kinetics of the PR-5CB resins\u2019 curing reaction.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Mechanical and thermal properties of PR-5CB printed in parallel and perpendicular directions. (a) 3D printing directions; (b) stress-strain curves; (c) bending-strain curves; (d) storage modulus and loss factor curves; (e) tensile properties; (f) flexural properties; (g) impact strength and hardness.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "SEM of PR-5CB printed in parallel and perpendicular directions.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Polarizing optical microscopy (POM) results for PR-5CB samples printed in parallel and perpendicular directions at 0\u00b0 and 45\u00b0.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Mechanical properties of 3D printed samples obtained using PR-5CB resins with different printing resolutions. (a) Stress-strain curves and bending-strain curves; (b) tensile strength; (c) flexural strength; (d) impact strength; (e) hardness.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "(a) POM of PR-5CB-3 after extracting 5CB LC; (b) Effect of print resolution and 5CB content on the structure of 3D printed products",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_7.png",
+        "caption": "POM images of PR-5CB resins with different print resolutions",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_8.png",
+        "caption": "SEM images of PR-5CB resins with different print resolutions",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_9.png",
+        "caption": "(a) Storage modulus curves, (b) loss factor curves, (c) calculated crosslinking density, (d) and TGA curves of PR-5CB printed samples",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_10.png",
+        "caption": "(a) Printing badges with different printing directions (b) Printing model of PR-5CB-3 resin (c) Gear model of PR-5CB-3 resin with different printing layer thickness",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

031e368a469fa7a49d7f45c82ca8e17995cf9776aa7caf1952c8b7077ee3f2da/preprint/preprint.md

@@ -0,0 +1,303 @@
+# Abstract
+
+A new printing resin with excellent performance (PR-5CB) was prepared by mixing 4'-pentyl-4-cyanobiphenyl (5CB) with acrylic photosensitive resin. The combination of the stereo lithography appearance (SLA) technique and the PR-5CB photosensitive resin allows precise adjustment of the existing morphology of liquid crystals in the resin to further control the mechanical properties of the printed product. Upon the addition of 5CB, the light-driven orientation of 5CB drives other acrylate prepolymers to orient along the orientation direction of 5CB, so that the entire fixed 3D printing polymer becomes anisotropic as observed by polarized optical microscopy. By controlling the 3D printing lamination method, printing resolution, and 5CB content, the mechanical properties of the 3D printed products can be effectively improved. The rheological properties, mechanical properties, and heat resistance of the PR-5CB resins were systematically investigated. The tensile strength, elongation at break, flexural strength, impact strength, and storage strength of the PC-5CB-3//(25 µm) printed products were 121.2 MPa, 25.5%, 222.0 MPa, 11.09 kJ/m², and 1702.3 MPa respectively; these values are 281%, 241%, 270%, 275%, and 186% of those of the commercial inks. The initial decomposition temperature of the printed sample of PR-5CB-3// (25 µm) was 298.5 °C and the maximum decomposition temperature was 423.5 °C, which were also higher than those of the commercial resins. The results of this study are significant for the development of light-cured 3D printing. The developed approach offers unlimited potential for achieving autonomous design of structures that cannot be achieved by current additive manufacturing processes.
+
+Physical sciences/Materials science/Structural materials/Mechanical properties  
+Physical sciences/Materials science/Techniques and instrumentation/Characterization and analytical techniques
+
+# Introduction
+
+The fast and easy molding method and diversified molding design make 3D printing technology different from traditional equal and reduced material manufacturing <sup>1-3</sup>. Therefore, 3D printing has become one of the main approaches for the development of future product processing methods <sup>4-7</sup>. This promising technology has developed rapidly in the last few years, and has found numerous applications in various fields, ranging from personalized consumer products to biomedical engineering, and automotive and aerospace industries <sup>5–9</sup>. This has led 3D printing to be considered as a key component and a paradigmatic example of the next industrial revolution <sup>10</sup>. Many additive manufacturing (AM) technologies currently exist, such as stereo lithography appearance (SLA), selective laser sintering (SLS), and fused deposition modeling (FDM) <sup>11,12</sup>. Among these, SLA is one of the most popular and widely applied techniques because of its high precision, and because in SLA, only a light source is necessary to induce photopolymerization, leading to low energy consumption and almost total absence of environmental pollution <sup>13–15</sup>. Furthermore, Tumbleston et al. reported a novel 3D printing approach called continuous liquid interface production (CLIP) that can significantly improve the printing speed <sup>16</sup>.
+
+However, some inevitable defects restrict the further development of SLA <sup>17-19</sup>. For example, the performance of the photosensitive resin directly affects the mechanical properties, heat resistance, and the accuracy of the light-cured parts, and the imported photosensitive resin is expensive, greatly increasing the cost of the manufactured products. Furthermore, the formed parts prepared by the development of the photosensitive resin have poor surface quality, and show high shrinkage, poor heat resistance, poor toughness, and poor mechanical properties, among other shortcomings. Finally, the shortage of photosensitive resins has significantly hindered the further development and application of 3D printing technology. Therefore, to enable further progress of the 3D printing industry, it is crucial to develop new high-performance photosensitive resin materials for 3D printing <sup>20-26</sup>. Several studies have aimed to solve this problem, and three strategies have been proposed: the use of different monomers <sup>18–21</sup>, addition of inert fillers or additives <sup>22–24</sup>, and adoption of an epoxy acrylate hybrid system <sup>25, 26</sup>. Among these three approaches, the addition of inert fillers or additives is the most convenient and common method, because it can not only enhance the mechanical performance of the resulting object but also reduces volume shrinkage <sup>17</sup>.
+
+Liquid crystal (LC) is an intermediate state of matter between solid and liquid. LCs have the molecular orientation properties of a solid crystal, but change shape as a fluid <sup>27-30</sup>. Liquid crystals are increasingly prepared for use in applications by enclosing the LCs in polymers. There are clear benefits of this approach because the LCs within the polymer are tightly and stably enclosed, providing a compact, reliable, and easy-to-use system <sup>31</sup>. Phase separation procedure is a useful technique in which the LCs and the polymer are first blended together into a common solution. Then, when the obtained mixture is sprayed, the LC molecules detach from the polymer, staying in the medium but still producing distinct LC droplets <sup>32-34</sup>. Polymer-stabilized liquid crystals (PSLC) and polymer-dispersed liquid crystals (PDLC) systems with specific properties can be produced by manipulating the microstructure of the liquid crystal/polymer complexes <sup>35-44</sup>. PDLC films are obtained by dispersing LC droplets in a polymer matrix. Polymerization is induced by methods such as thermal curing and UV irradiation. When these methods are applied, the solubility of the LCs in the polymer matrix decreases and causes phase separation and the formation of microphase-separated structures <sup>45-51</sup>.
+
+Electronically controlled LC composites respond only to electric signals, while thermally-controlled composites respond to both electric power and the temperature changes in their environment <sup>52-54</sup>. Inspired by this idea, we considered that in the light-curing 3D printing process, the photosensitive resin is polymerized under UV light irradiation and the polymerization process produces some heat. If a certain amount of LC is added to the photosensitive resin, and the heat generated by the polymerization of the photosensitive resin causes the added liquid crystal to undergo orientation, this effect will be equivalent to that of the addition of the particles with a certain orientation to the polymer and will enhance the performance of the polymer to a certain extent.
+
+One of the most famous LC series of calamitic molecules is 4-cyano- 4′-alkylbiphenyl (nCB), where “n” represents the number of the carbon atoms present within the alkyl chain. The nCB series displayed an odd–even trend of CN⋅⋅⋅CN and CN⋅⋅⋅phenyl interactions in all crystal structures, apart from 5CB and 6CB. The odd–even effect in the packing modes for higher homologues (n ≤ 7) is known to be a particular case of CN⋅⋅⋅CN or CN⋅⋅⋅phenyl interactions. 4′-Pentyl-4-cyanobiphenyl (5CB) has been one of the most investigated LC materials due to its readily accessible nematic range around room temperature that enables its use as a suitable model compound for studying the physical behavior of simple nematics <sup>55-57</sup>. LCs are an intermediate state between solid crystals and isotropic liquids, and show strong nonlinear optical effects. The 5CB LCs studied in our experiments are linear LCs at room temperature (20-35 °C). In the absence of photon absorption, the laser power required for the photoelectric field to sense the molecular orientation change is 10<sup>2</sup>–10<sup>3</sup> W/cm<sup>2</sup> <sup>58</sup>. The energy required for photoelectric field sensing of the change in the molecular orientation originates from a small shift in the optical frequency <sup>59</sup>. However, retention of the excellent performance of a liquid crystal while using only a simple operation for 3D printing of photosensitive resin for the LC preparation will open a broad range of new possible applications for industry <sup>60-63</sup> and makes this approach more competitive compared to the traditional use of liquid crystal photosensitive materials in 3D printing <sup>64-65</sup>. For example, self-assembly methods containing materials such as copolymers, colloids, and liquid crystals (LC) have created microstructures with tunable optical properties, providing a large number of unique material systems for AM methods to develop smart optical devices <sup>66</sup>. Additionally, combining LCs with advanced 3D printing processes, smart materials with sensing capabilities can be fabricated based on the nematic and anisotropic phase transitions of LCs <sup>67-69</sup>.
+
+Given the above, this study proposes a 3D printing strategy using a novel photosensitive resin with controllable liquid crystal morphology. The strategy includes three design steps: first, 4′-pentyl-4-cyanobiphenyl <strong></strong> (5CB) is used for blending with an acrylic acid-based photosensitive resin to prepare 5CB/photosensitive resin (PR-5CB) and the dispersion and rheological properties of PR-5CB resins are characterized. Second, SLA is selected to print PR-5CB. During the 3D printing process, the lamination direction and printing resolution of the stack are regulated by controlling the printing parameters so that the liquid crystal orientation in the resin is fixed, and when the single-layer ink is cured, the rod-like structure of the liquid crystal is fixed when the entire network system is cross-linked. The combination of 3D printing technology and precise modulation of the liquid crystal morphology to meet the needs of the desired product is used for the first time in this work. Compared to the conventional reinforcing materials, the use of liquid crystals with controlled orientation is highly suitable for commercial use, offering wide range of new applications for the development of AM. Finally, the thermal and mechanical properties of the printed samples of each formulation of PR-5CB were studied to further investigate the enhancement mechanism of the printed samples obtained by the SLA printing technology.
+
+# Results
+
+## Dispersibility and rheological properties of PR-5CB resins
+
+Good dispersion is important for obtaining high performance of 3D printed samples of PR-5CB resins. For the PR-5CB resins, a well-mixed resin was observed under a polarizing microscope, and as shown in Fig. 1 (a), a large number of rod-shaped LC molecules were distributed throughout the field of view. This shows that 5CB is well-dispersed in the resin substrate, demonstrating the feasibility for obtaining the desired properties of the PR-5CB resin. The sizes of the rod LCs obtained from the POM calculation are between 13 and 20 µm.
+
+In 3D printing molding, the resin must be sufficiently leveled or scraped before each layer of raw material is cured and set, which usually requires good flowability. As shown in Fig. 1 (b), the addition of PR-5CB increases the system viscosity, and the viscosity of PR-5CB-5 resin is maximum when the shear rate is 1 s⁻¹ (5.7 pa·s). As the shear rate increases, the overall viscosity of the system tends to decrease significantly. This phenomenon is consistent with the fluid "shear thinning" behavior. Under high-speed shear, the viscosity of PR-5CB (0.2 pa·s) is lower than that of commercial inks, and is only 25% of that of commercial inks (0.8 pa·s). This offers the possibility of mass production for 3D printing.
+
+To further study the dynamic mechanical changes in the curing process of the PR-5CB resins shown in Fig. 1 (c), the UV light was turned on, and the PR-5CB resins were gradually cured. After 20 s, the energy storage modulus gradually increased, and after the curing was completed, the energy storage modulus remained the same, and the maximum energy storage modulus was obtained for the 3D printed sample based on the PR-5CB resins. In particular, the light-curing speed of PR-5CB-3 resin was slightly improved compared to that of commercial resin. This indicates that the prepared PR-5CB resins has fast light-curing ability.
+
+## Investigating the Light Orientation Mechanism of PB-5CB Resins by SLA 3D printing
+
+Since 5CB LCs show strong UV light absorption (absorption coefficient of ~10⁶ /cm), we exploited this property of 5CB by incorporating it into a photosensitive resin so that the UV laser irradiation used to cure the photosensitive resin triggers the LC molecular orientation. Based on this idea, we performed experiments to verify the orientation of the 5CB LC under UV laser irradiation. In the 3D printing process, we print samples with the lamination direction parallel to the tensile test direction (PC-5CB//) and samples with the lamination direction perpendicular to the tensile test direction (PC-5CB⊥), as shown in Fig. 2 (a). Then, we verify the orientation of the LCs by verifying the difference between the mechanical properties of the two different samples.
+
+Figures 2 (b),(c) show the stress-strain curves and bending-strain curves of PC-5CB-0 and PC-5CB-3 in both directions, whereas Fig. 2 (e),(f) show the corresponding results for the tensile strength, elongation at break, flexural modulus, and flexural strength, respectively. It is observed from Fig. 2 (b),(c) that the tensile and bending properties of the 3D printed products of the resin with a small amount of added 3% 5CB LC are improved relative to those of the samples without the added LC. At the same time, the properties of different stacking forms also differed greatly after adding 5CB LC, while the mechanical properties of the resin samples without the LC were almost unaffected by their stacking forms. Figure 2 (e) shows that the tensile strengths for PC-5CB-0//, PC-5CB-0⊥, PC-5CB-3// and PC-5CB-0⊥ are 45.6, 43.2, 121.2 and 94.7 MPa, respectively. Meanwhile, the elongation at break values are 10.2%, 10.6%, 25.5% and 16.9%, respectively. These results show that for the isotropic PC-5CB-0, the mechanical properties of the conventional photosensitive resins do not change significantly for printed products obtained with different printing directions. However, for PC-5CB-3, the tensile strength and elongation at break of PC-5CB-3// were 1.3 and 1.5 times higher than those of PC-5CB-0⊥, respectively. The above results mean that the mechanical properties of the 3D printed products along the direction of the printed layer stacking are better, indicating that LCs are ordered along the layer stacking direction, and the long-range ordered LCs improve the strength of the material. To further confirm our conjecture, the samples were subjected to additional bending tests, impact strength, and hardness tests. The test results showed that for the PC-5CB-0 sample, changing the 3D printing molding orientation does not affect its mechanical properties. However, the flexural strength, flexural modulus and impact strength of the PC-5CB-3// sample were found to be 156.5 MPa, 2836.8 MPa and 11.09 kJ/m², which were not only higher than those of the PC-5CB-0 sample, but also were 1.5, 1.4 and 1.1 times higher than those of the PC-5CB-3⊥ sample. Dynamic mechanical thermal analysis (DMA) is an important tool for the characterization of the compatibility of photosensitive resins with 5CB LCs and the curing and crosslinking of polymers. Therefore, the variation of the energy storage modulus and loss factor of the PR-5CB resin printed products with the temperature was investigated using the DMA technique, and the results are shown in Fig. 2 (d). The storage modulus values of the products are 904.1, 765.2, 1634.2 and 1524.1 MPa for PC-5CB-0//, PC-5CB-0⊥, PC-5CB-3//, and PC-5CB-0⊥, respectively, while the Tg values are 71.7, 76.2, 91.6, and 85.5°C, respectively. This indicates that despite the usual trade-off between strength and toughness, the addition of 5CB can improve not only the strength but also the toughness of 3D printed products. This is mainly due to the orientation of 5CB during the curing of the sample that occurs in the same direction as the light irradiation.
+
+According to the theory of material fracture mechanics, the fracture damage process can be divided into two steps of crack initiation and crack extension, where the crack initiation includes microcrack initiation (microhole formation), microcrack extension, and microcrack termination. For most thermosetting resins with a high cross-link density, when the material is subjected to stress, the molecular chains can only slide in a small area due to the chemical crosslinking, and the material mainly shows elastic deformation, while some molecular chains break, and less energy is consumed by crack extension and material failure. These materials are brittle. Therefore, in the study of the cross-sectional morphology of the cured products of thermosetting resins, we mainly observe the crack initiation, i.e., the microcrack initiation source, the microcrack extension area and the microcrack termination area for which the morphology is generally centered on the initiation source, and the microcrack extends radially outward, accompanied by the microcrack termination phenomenon.
+
+To further show that the 5CB LC orientation also improves the toughness of the printed products, the impact fractures of the printed products obtained with different print directions of PR-5CB-0 and PR-5CB-3 were observed by scanning electron microscopy, with the results shown in Fig. 2. It is observed that the PR-5CB-0 system has a smooth cross-section and a single direction of the microcrack due to the poor molecular movement of the cured resin. This makes it difficult to produce yield deformation, and the microcrack is essentially unhindered during the expansion, which is typical of brittle damage. With the addition of 3% content of 5CB, tough nested and river-like cross-sections appear in the cross-section of PR-5CB-3. In PR-5CB-3//, a clear branching extension of microcracks showing whitening is present, and the microcrack extension ends in a typical river-like section. For the PR-5CB-3⊥ cross-section, in addition to the appearance of a river-like cross-section, the surface granulation-like phenomenon is also more pronounced. This result indicates that in the absence of 5CB addition, the difference in the printing direction has almost no effect on the morphology of the cross-sectional fracture, whereas the addition of 5CB effectively enhances the toughness fracture of 3D printed products, and the change in the printing direction also affects the fracture surface morphology. The above analysis is consistent with the results of the mechanical properties tests.
+
+To verify that the 5CB LC is oriented under UV laser irradiation, we lay the sample strips with different printing directions flat on the table and use a cutter to intercept the upper surface to prepare a surface film with a layer thickness of 1 mm. A film with a sample cross-section of approximately 1 mm was also cut for polarized light microscopy observation. The observation was carried out by placing the cut film between two coverslips, rotating the circular platform, and observing the light and dark changes in the photographic field of view. Figure 2 shows the POM images of the 3D printed sample surface and cross-section at the 0° and 45° directions. The surface and cross-section of the 3D printed products without 5CB are always fully dark in POM observation when the rotary stage is turned 360° (Video 1, 2 in Supplementary Information (SI)), indicating that the sample is isotropic, regardless of whether it is printed along the parallel or perpendicular direction. However, we were surprised to find that when the 5CB LC was added to PR-5CB-3//, the surface under POM observation showed clear 4 light and 4 dark images (Video 3 in SI), while the cross-sectional POM was always fully dark (Video 4 in SI), indicating that the PR-5CB-3// polymer was anisotropic. To further verify this result, PR-5CB-3⊥ was observed by POM, and its surface under POM was fully dark (Video 5 in SI), while its cross-sectional POM showed 4 light and 4 dark images (Video 6 in SI). According to our original concept that only 5CB is oriented during the printing process, and the other photosensitive resin fixes 5CB by anchoring, the POM Fig. should present only bright lines with a certain direction. However, the above results indicate that in contrast to our expectation, upon adding 5CB, the process of 5CB orientation also drives other acrylate prepolymers to orient along the orientation direction of 5CB, so that the whole fixed 3D printing polymer is anisotropic. It is also observed from the POM diagrams and schematics of the PR-5CB-3// surfaces and the PR-5CB-3⊥ cross-sections that the polymers are oriented along the direction of light and 5CB is anchored between the 3D printed layers. This is an important finding regarding LCs used for 3D printing. Typically, for 3D printing, LC prepolymers with photosensitive groups (e.g., RM257) are used that are important for the preparation of LC elastomers. However, this study uses small and unreactive LCs 5CB that can be oriented under UV laser irradiation to drive the orientation of the entire photosensitive resin. Thus, the results of this study are important for the development of light-cured 3D printing.
+
+## Investigating the influence of printing resolution and LC content
+
+The above study showed that different printing directions strongly affect the performance of 3D printed products, and it was found that the performance of 3D printed products is improved when the printing lamination direction is parallel to the stretching direction. Therefore, only parallel-direction printing was used in our subsequent studies.
+
+The LC content associated with the material is another key factor affecting the oriented structure. Samples printed with different contents of 5CB LCs (3%, 5%, and 7%) were used to study the effect of 5CB content on the orientation state. Here, 3D printing was used for the first time to study the light-oriented photocurable resins. The length of 5CB LCs was found to be roughly between 13 and 20 µm based on the POM images discussed above. Therefore, we consider that 3D printing serves to enhance the polymer by anchoring the 5CB LCs in each printed layer through the polymerization process of the photosensitive resin into the polymer during the layer printing process. Additionally, to investigate how the orientation of the photosensitive resin is induced by 5CB, we investigated the change in the polymerization by varying the printing precision of 3D printing (25, 50, and 100 µm) to determine the optimal content of added 5CB and printing precision.
+
+First, mechanical properties of the 3D printed samples obtained using PR-5CB resins with different printing resolutions were investigated, with the results shown in Fig. 5. Comparison of the results shows that the mechanical properties of the products printed using the resin without LC (PR-5CB-0) were almost independent of the printed layer thicknesses; for example, tensile strengths of 43.1, 40.1 and 41.9 MPa were obtained for the thicknesses of 25, 50, and 100 µm, respectively. This indicates that the mechanical properties of the products are only weakly affected by the changes in the printing resolution in the absence of the 5CB LC addition. However, it was found that surprisingly, the addition of 3% of 5CB LC together with the control of the printing to a certain print layer thickness can effectively improve the mechanical properties of the 3D printed products. In particular, the increase in the mechanical properties is pronounced for the print resolution of 25 µm. The tensile strength, elongation at break, flexural strength, flexural modulus, hardness and impact strength for PB-5CB-3 (25 µm) reached 121.2 MPa, 25.5%, 222.0 MPa, 2836.8 MPa, 76.2 HD, and 11.09 kJ/m² respectively, which were 210.2%, 144.9%, 136.6%, 185.7%, 103.8%, and 409.0% of the corresponding values for PB-5CB-0 (25 µm), respectively. Thus, despite the usual trade-off between strength and toughness, not only the strength, but also the toughness of the product was effectively increased. However, it was also found that the mechanical properties became worse when the print resolution was increased to 50 µm and 100 µm. In particular for the print resolution of 100 µm, the mechanical properties decreased significantly and were even lower than those of the products printed with the PB-5CB-0 resin. The tensile strength, elongation at break, flexural strength, flexural modulus, hardness, and impact strength were 100.0 MPa, 23.4%, 156.5 MPa, 2556.9 MPa, 73.5 HD and 8.62 kJ/m² for PB-5CB-3 (50 µm) and were 30.7 MPa, 26.1%, 104.7 MPa, 2012.9 MPa, 71.6 HD and 6.22 kJ/m² for PB-5CB-3 (100 µm), respectively. This shows that simply adding the 5CB LC does not necessarily enhance the performance of 3D printed products; rather, the performance can be effectively enhanced only by controlling the print resolution because of the different effects of the different print resolutions on the LCs in the polymer.
+
+In the printing process, the polymerization of the photosensitive resin under UV laser irradiation is exothermic, and the generated temperature field gives rise to the orientation of the 5CB LC molecules. When the printing accuracy is 25 µm, the thickness of the printed layer is approximately the same as the length of the rod LC and the molding time is short, so that the LC also drives the surrounding photosensitive resin to orient together in the light direction. Then, when the photosensitive resin polymerizes to obtain a polymer, it anchors the 5CB in one direction, giving rise to the anisotropy of the polymer. To demonstrate that the polymer also undergoes orientation, we removed the small LC 5CB from the PR-5CB-3 (25 µm) printed product using dichloromethane while leaving the polymer, and examined the surface of the product by POM. The test results (Video 7 in SI) show that under rotation of the carrier table by 360°, the phenomenon of alternating 4 bright and 4 dark images is observed, with the bright and dark images shown in Fig. 6 (a). This indicates that the photosensitive resin also undergoes orientation during the polymerization process, forming a structure with optical anisotropy similar to that of LCs. In other words, the photosensitive resin polymerizes along the orientation direction of the small LCs during the polymerization process, and this orientation is retained after polymerization, as shown in Fig. 6 (b). When the printing resolution is increased to 50 µm, the LC is also oriented together with the surrounding photosensitive resin in the light orientation process because the thickness of the polymer layer is greater than the length of the LC rod. However, because too much photosensitive resin is present, its orientation is not complete. When the printing accuracy is further increased to 100 µm, the 5CB LC is completely surrounded by the photosensitive resin that cannot be oriented and dispersed during the illumination. Therefore, the orientation of the resin is relatively random, so that the polymer cannot follow the orientation of the LC.
+
+To further verify our conclusion, the 5CB content was increased to 5% and 7% to prepare the PR-5CB-5 and PR-5CB-7 resins, and 3D printing was performed at different printing resolutions, with the results also shown in Fig. 5. First, as the LC content increases, the mechanical properties of the printed product gradually decrease. The tensile strength, elongation at break, flexural strength, flexural modulus, and impact strength decrease from 121.2 MPa, 222.0 MPa, 2836.8 MPa and 11.09 kJ/m² for PB-5CB-3 (25 µm) to 110.2 MPa, 106.7 MPa, 2034.1 MPa, 10.9611.09 kJ/m² for PB-5CB-5 (25 µm) and to 88.0 MPa, 81.6 MPa, 1763.4 MPa, 5.79 kJ/m² for PB-5CB-7 (25 µm), respectively. As mentioned above, the degree of LC orientation of PR-5CB-3 is quite strong, but due to the increase in the LC content, the "anchoring point" provided by the photosensitive resin is not sufficient to induce partial LC orientation, and the 5CB is retained in the droplet form. For the PR-5CB-7 printed sample, the lower degree of orientation resulted in a significant reduction in the tensile strength and modulus, which is unfavorable for the improvement of the product strength. Second, it was found that the mechanical properties decreased when increasing the print resolutions to 50 µm and 100 µm. In particular for the print resolution of 100 µm, the mechanical properties decreased significantly, showing performance that was inferior even to that of the products printed with the PB-5CB-0 resin. Because of the long printing time and high polymerization heat, the 5CB liquid crystals become random and agglomerate into spheres above the orientation temperature, and float on the surface of the printed sample. However, after the photosensitive resin is converted to the polymer, these spherical liquid crystals are fixed on the surface of the sample and become microcrystalline structures upon cooling, as shown in Fig. 6 (b).
+
+To verify our conjecture regarding the effect of the 5CB content and printing resolution on the mechanical properties of the 3D printed products, the POM images for each sample are shown in Fig. 7. The POM image of the surface of the 3D printed products without 5CB (PR-5CB-0) is always fully dark when the rotary stage is turned by 360°, indicating that the surface is isotropic for different print resolutions. For the 25 µm and 50 µm print resolutions, the PR-5CB-3 and PR-5CB-5 products showed optical anisotropy similar to that of liquid crystals, indicating that the photosensitive resin polymerizes along the orientation direction of small liquid crystals during the polymerization process. The POM images show that rotating the carrier table by 360°, the phenomenon of alternating 4 bright and 4 dark images is observed, with the bright and dark images shown in Fig. 7. For 100 µm print resolution, the POM images show under the rotation of the carrier table by 360°, the images of the PR-5CB-3 and PR-5CB-5 samples do not change between light and dark and are always fully bright, indicating the presence of cryptocrystalline or microcrystalline aggregates on the sample surface. This supports the conclusion that due to the long printing time and high polymerization heat, the 5CB liquid crystals become random and agglomerate into spheres above the orientation temperature, and float on the surface of the printed sample. However, after the photosensitive resin is converted into a polymer, these spherical liquid crystals are fixed on the surface of the sample and transform into microcrystalline structures upon cooling. For the PR-5CB-7 samples, even when the print resolution is changed from 25 µm to 100 µm, the POM images are always bright, and some spherical aggregates can be observed on the surface. Due to the increase in the liquid crystal content, the "anchor point" provided by the photosensitive resin was not sufficient to induce partial liquid crystal orientation, and 5CB was retained in the form of liquid droplets, as shown in Fig. 6 (b). For the PR-5CB-73D printed samples, the lower orientation and the presence of liquid crystal aggregates on the surface lead to a significant decrease in the tensile strength and modulus, while the droplet form of 5CB is unfavorable for product strength improvement.
+
+To further show that the 5CB liquid crystal orientation is also favorable for improving the toughness of the printed products, the impact fracture of the printed products with different print resolutions of PR-5CB resins was observed by scanning electron microscopy, with the results shown in Fig. 8. For the PR-5CB-0 system with different print resolutions, smooth cross-sections and a single direction of microcrack are observed due to the poor molecular movement of the cured resin. This makes it difficult to produce yield deformation, and the microcrack is essentially unhindered during the expansion, which is typical of brittle damage. The tearing stripes on the fracture surface are caused by the energy absorbed during the fracture of the product. In particular, the fracture surfaces of the 25 µm printed samples showed more irregular tears with evident wrinkles, indicating the excellent toughness of the printed products. The roughness of the fracture surface increased with increasing printed layer thickness, which is consistent with the impact strength results described above.
+
+To further demonstrate that the strategy of light-oriented 3D printing of 4'-pentyl-4-cyanobiphenyl (5CB) liquid crystal/photocurable resins is not only applicable to the above-mentioned formulations, but rather is effective for all acrylic photocurable resins, we selected different commercial printing formulations (Table S1) with 3% content of 5CB and printed sample strips at 25 µm print resolution to test their tensile properties (Fig. S1) and examine their POM images (Fig. S2). The POM images show that upon rotating the carrier table by 360°, the phenomenon of alternating 4 bright and 4 dark images is observed for all other printing formulations (Videos 8–10 in SI). These results demonstrate that all photosensitive resins with added 5CB have good orientation that in turn improves their mechanical properties.
+
+## Thermal mechanical properties of PR-5CB printing samples
+
+Dynamic mechanical thermal analysis (DMA) is an important tool for the characterization of the compatibility of photosensitive resins with 5CB liquid crystals and the curing and crosslinking of polymers. Therefore, the temperature variations in the energy storage modulus and loss factor of the PR-5CB resins printed products at 25 µm printing resolution at the printing lamination direction parallel to the stretching direction were further investigated using the DMA technique, and the results are shown in Fig.s 9(a),(b). It is observed that the introduction of the 5CB liquid crystal significantly improved the energy storage modulus of the product. The storage modulus values for PR-5CB-3, PR-5CB-5, and PR-5CB-7 were 1702.3, 1629.8 and 1555.6 MPa, corresponding to 186%, 183% and 178% of the value for PR-5CB-0 (875.4 MPa), respectively. The Tg values for PR-5CB-3, PR-5CB-5, and PR-5CB-7 were 91.6, 89.2, and 88.4°C, which are all higher than that for PR-5CB-0 (71.7°C). However, the gradual decrease in the energy storage modulus is particularly noteworthy because it indicates that the stiffness of the printed product decreases with higher 5CB content. This result is attributed to the reduction in the cross-linkage of the polymer, for which the cross-link density is generally calculated by the theory of elastic dynamics of rubber:
+
+$$\mathcal{V}\mathcal{e}={E}^{{\prime }}/3RT$$
+
+where $E^{\prime}$ is the storage modulus in the rubbery state ($E^{\prime}$ at Tg + 50°C), $R$ is the gas constant (8.314 J/mol·K), and $T$ is the absolute temperature at Tg + 50°C. According to the calculated results (Fig. 9 (c)), the crosslinking density of PR-5CB-3 is the highest at 22.5×10³ mol/m³ and decreases with increasing liquid crystal content. Therefore, an addition of more than 3% of 5CB will reduce the crosslinking density of the system from 22.5 × 10³ to 18.5 × 10³ mol/m³, reducing the energy storage modulus of the resin. The increase in the crosslinking density of PR-5CB-3 compared to PR-5CB-0 is due to the presence of rigid structures in 5CB and the anchoring force between the well-ordered liquid crystals and the polymer that makes the system more compact. When the 5CB liquid crystal content is further increased, the unoriented liquid crystals become agglomerated and the molecular network interactions become poor.
+
+The thermal stability performance of the printed samples is demonstrated by the TGA curves (Fig. 9 (d)). The thermal decomposition curves based on PR-5CB resins show the same trends, with no significant differences in the weight loss rates, and are mainly divided into three weight loss stages. For example, for the initial decomposition temperature (T−5%) and maximum decomposition temperature (Tmax), the values for PR-5CB-3 (312.0 ℃, 423.5℃) are approximately 7 ℃ and 5 ℃ higher than those for PR-5CB-0 (306.3 ℃, 418.4℃), indicating that an addition of the appropriate amount of 5CB enhances the thermal stability of the printed samples. This result can be explained based on two effects: (1) the rigid benzene ring structure in 5CB restricts the motion of the chain segments in the system; (2) the short rod-like liquid crystal structure accounts for a small fraction of degrees of freedom in the system and can be tightly interspersed in the molecular network at high temperatures. When the 5CB content increased further, T−5℃ and Tmax decreased slightly, from 312.0℃ to 296.2℃ and from 423.5℃ to 413.1℃, respectively. In this case, 5CB mostly exists in droplet form, reducing the degree of crosslinking between the polymer molecules and facilitating the movement of the molecular chain segments.
+
+## Analysis of visual differences and accuracy of printed products
+
+Since uniform orientation arrangement of liquid crystal molecules is a prerequisite for their use in displays and optics and such an orientation is obtained in PR-5CB printed samples, complex models were printed in different printing orientations to investigate the possible use of PR-5CB printed samples in optical devices (Fig. 10 (a)). We observed a visual difference between the printed models. When the printing direction is parallel to the layer stacking direction, the liquid crystal orientation direction is perpendicular to the printing direction, and the model shows a higher haze with a milky white color. When the printing direction is perpendicular to the printing layer stacking direction, the liquid crystal is oriented in the same direction as the printing direction, changing the refractive index and enabling visible light to pass through. Therefore, the surface haze of the model decreased and the model became more transparent. The anisotropic orientation leads to visual differences. The innovative PR-5CB 3D printing material requires only simple processing, enables microscopic control of molecular orientation, and is not limited to thin-film products, thus significantly reducing the cost of the product. Thus, it is promising for applications in areas such as liquid crystal displays and information recording.
+
+As shown in Fig. 10 (b), models with high precision requirements were printed using the PR-5CB-3 resin, and the details of each model are clearly visible, illustrating the feasibility of the use of this photosensitive resin in fields such as advanced precision instruments. In addition to the accuracy limitations, the poor shrinkage of the samples printed by commercial inks is a major bottleneck limiting the application of 3D printing. Figure 10 (c) shows the gear model of the PR-5CB-3 resin with different printing layer thicknesses and printing direction parallel to the direction of the UV polarized light orientation, respectively. The results showed that the relative error of printing accuracy of the teeth printed with the PR-5CB-3 resin was 0.14–0.28%. Overall, these results demonstrate that the PR-5CB resin prepared in this study has high printing resolution and can be used for printing various products with high accuracy requirements.
+
+# Discussion
+
+In this study, PR-5CB resins were successfully prepared by blending 4'-pentyl-4-cyanobiphenyl with an acrylic photosensitive resin, and were applied to SLA 3D printing. The rheological properties of PR-5CB resins were investigated and showed fast light-curing ability. By controlling the 3D printing lamination method, printing resolution and 5CB content, the mechanical properties of 3D printed products can be effectively improved. It was found that upon the addition of 5CB, the light-driven orientation process of 5CB itself also drives other acrylate prepolymers to orient along the orientation direction of 5CB, so that the entire fixed 3D printing polymer was observed to be anisotropic by polarized optical microscopy. The tensile strength, elongation at break, flexural strength, flexural modulus and impact strength of PR-5CB-3// (25 µm) for 25 µm printing resolution were 121.2 MPa, 25.5%, 222.0 MPa, 2836.8 MPa and 11.09 kJ/m², respectively. These values are 2.81, 2.41, 2.70, 2.12 and 2.75 times higher than the corresponding values of commercial inks. The method introduced in this study is simple and inexpensive and presents the first innovative solution for 3D printing-assisted liquid crystal light-driven orientation. The excellent mechanical properties of the obtained printed products are of great significance for broadening the application of AM to high-end fine displays and optical devices. The use of 3D printing technology for fabrication of device systems is a new approach that requires further exploration, and the use of PR-5CB resins for 3D printing is highly significant for the development of self-reinforced composite materials meeting the needs of different fields.
+
+# Methods
+
+## Materials
+
+Hydroxyethyl methacrylate (HEMA) and 2,4,6-trimethylbenzoyl diphenylphosphine oxide (TPO) were obtained from Rohn Chemical Reagents (Shanghai, China). Aliphatic urethane acrylate (CN9010), polyurethane acrylate (CN991) ethoxylated pentaerythritol tetraacrylate (SR494) were purchased from Juncai Material Technology Co., Ltd. (Shanghai, China). 4′-Pentyl-4-cyanobiphenyl (5CB) was purchased from Shanghai Biogenic Leaf Co., Ltd. (Shanghai, China).
+
+## Preparation of PR-5CB resins
+
+Photosensitive resins containing different contents of 5CB were prepared using solution blending. The PR-5CB resins were prepared by mixing different amounts of CN9010, SR494, active diluent HEMA, and photoinitiator TPO, emulsifying and stirring at 80°C for 30 min, and then cooling to room temperature. Finally, a certain amount of 5CB was added into the mixture. These resins were labeled according to their compositions as described in Table 1.
+
+**Table 1**  
+Formulations of PR-5CB resins.
+
+| Materials | PR-5CB -0 | PR-5CB–3 | PR-5CB -5 | PR-5CB -7 |
+|---------|----------|---------|---------|---------|
+| 5CB | 0 | 3.0 | 5.0 | 7.0 |
+| CN9010 | 30.0 | 30.0 | 30.0 | 30.0 |
+| CN991 | 30.0 | 30.0 | 30.0 | 30.0 |
+| SR494 | 15.0 | 15.0 | 15.0 | 15.0 |
+| HEMA | 20.0 | 20.0 | 20.0 | 20.0 |
+| TPO | 5.0 | 5.0 | 5.0 | 5.0 |
+
+## Stereo Lithography Appearance (SLA) Printing
+
+An SLA printer (Form 2, Formlabs Inc., Somerville, MA, USA) was used to produce the 3D molded parts. The model was first designed using the SolidWorks software, and then was layered using the printing software to control the UV light-curing time and the exposure state of the liquid resin; this allowed the photographic resin to be printed layer-by-layer according to the pre-designed modeling structure. The printing strategy focuses on printing molded parts by controlling the thickness of the print monolayer. The obtained samples were washed with ethanol and dried at 30°C under airflow.
+
+Tensile, bending, and notched specimens were printed using a Form 2 3D printer according to the Chinese GB/T 1040.1e2018, GB/T 9341e2008, and GB/T 18658e2018 national testing standards. These samples were post-treated with UV light for 10 min and left to stand for 24 h prior to measurements and performance tests.
+
+## Polarizing optical microscopy (POM)
+
+A polarized optical microscope (POM, 59XA-2) was used to observe the dispersion state of the photosensitive resins and the orientation of the 3D printed films. An appropriate amount of dispersed liquid was dropped between two cover glasses to form a thick layer of solution, and its state was observed. Using a cutter to cut the spline surface in different printing directions, a film with a thickness of 1 mm was prepared and placed between two cover glasses. Then, the circular platform was rotated and the light and dark changes of the shooting field of vision were observed. The sample was placed between two overlays, and the two overlays were slightly cut to form a thin film of the solution and were photographed for observation.
+
+## Rheological behavior measurement
+
+The rheological behavior of the resin was measured using a TA Instruments Discovery Hybrid Rheometer (DHR-2). The steady-state shear rates ranged from 1 to 1000 s⁻¹. The kinetics of the light-curing process were evaluated using a rheometer equipped with a UV light-emitting diode (LED). The gap between the two geometries was set to 0.1 mm. The upper and lower plates were composed of aluminum and transparent polymethyl methacrylate, respectively. The duration of the experiment was 80 s, and the samples were irradiated at an intensity of 80 MW/cm².
+
+## Tensile test
+
+The tensile properties were evaluated using a universal material testing machine (LD24, Labsans, China). The tensile test speed was 10 mm min⁻¹. The test results reported for each sample are the average values of six replicate tests.
+
+## Notched impact performance
+
+A WH-8050 digital pendulum impact tester (Ningbo Zhenhai Weiheng Testing Instrument Co., Ltd.) was used for the notched impact performance test. According to the GB/T 1043e2008 standard, the test sample dimensions were 80 ± 2, 10 ± 0.2, and 4 ± 0.2 mm, and the V-shaped notch depth was 2 ±���0.1 mm.
+
+## Hardness
+
+The hardness was tested using a D-type digital rubber hardness tester according to the GB/T 531e99, GB/T 2411e80, HG/T 2489e93, and JJG 304-003 standards.
+
+## Dynamic mechanical analysis
+
+Dynamic mechanical analysis (DMA) was performed using a DMA Q800 instrument from TA Instruments (USA). DMA tests were performed in the double cantilever in the temperature range of 30–300°C at a heating rate of 3°C/min and frequency of 1 Hz, using a sample with the dimensions of 35.0 × 10.0 × 3.0 mm³.
+
+## Scanning electron microscopy
+
+Scanning electron microscopy (SEM; JEOL JEM-2010, Japan) was used to observe the surface fracture morphology of the 3D printed parts. The gold film was sputtered on the sample surface with a magnetron sputterer and the scanning acceleration voltage was 10 kV.
+
+## Thermogravimetric analysis
+
+Thermogravimetric analysis (TGA) was performed using a STA449C instrument from TA Instruments (USA) at a temperature range of 30–800°C and a weight of 5–10 mg under nitrogen atmosphere with a flow rate of 100 ml min⁻¹ and a ramp rate of 10°C/min.
+
+# References
+
+1. Liu, X., Shang, Y., Liu, J., Shao, Z. & Zhang C. 3D printing-enabled in-situ orientation of BaTi<sub>2</sub>O<sub>5</sub> nanorods in beta-PVDF for high-efficiency piezoelectric energy harvesters, *ACS Appl. Mater. Interfaces*. **14(11)**, 13361–13368 (2022).
+
+2. Thiam, B.G., Magri, A. El, Vanaei, H.R. & Vaudreuil, S. 3D printed and conventional membranes—A review. *Polymers (Basel)*. **14(5)**, 1023 (2022).
+
+3. Tsai, S. C., Chen, L. H., Chu, C.P., Chao, W.C. & Liao, Y. C. Photo curable resin for 3D printed conductive structures. *Addit. Manuf.* **51**, 102590 (2022).
+
+4. He, Y. et al. A photocurable hybrid chitosan/acrylamide bioink for DLP based 3D bioprinting. *Mater. Des.* **202**, 109588 (2021).
+
+5. Y. Lu. et al. The molecular design of photo-curable and high-strength benzoxazine for 3D printing. *Chem. Commun. (Camb)*. **57(27)**, 3375–3378 (2021).
+
+6. Wilson, S.A., Cross, L.M., Peak, C.W. & Gaharwar, A.K. Shear-thinning and thermo-reversible nanoengineered inks for 3D bioprinting. *ACS Appl. Mater. Interfaces.* **9(50)**, 43449–43458 (2017).
+
+7. Wu, Y. et al. Photocurable 3D printing of high toughness and self‐healing hydrogels for customized wearable flexible sensors *Adv. Funct. Mater.* **31(52)**, 2107202 (2021).
+
+8. Barraza B. et al. Superhydrophobic SLA 3D printed materials modified with nanoparticles biomimicking the hierarchical structure of a rice leaf. *Sci. Technol. Adv. Mater.* **23(1)**, 300–321 (2022).
+
+9. Choo, S. Jin, S. & Jung J. Fabricating high-resolution and high-dimensional microneedle mold through the resolution improvement of stereolithography 3D printing. *Pharmaceutics* **14(4)**, 766 (2022).
+
+10. Zhu, Y. et al. Recent advancements and applications in 3D printing of functional optics. *Addit. Manuf.* **52**, 102682 (2022).
+
+11. Shahzadi, L., Maya, F., Breadmore, M.C. & Thickett, S.C. Functional materials for DLP-SLA 3D printing using thiol–acrylate chemistry: Resin design and postprint applications. *ACS Appl. Polym. Mater.* **4(5)**, 3896–3907 (2022).
+
+12. Xu, X. et al. Stereolithography (SLA) 3D printing of a bladder device for intravesical drug delivery. *Mater. Sci. Eng. C Mater. Biol. Appl.* **120**, 111773 (2021).
+
+13. Kafle, A. et al. 3D/4D printing of polymers: Fused deposition modelling (FDM), selective laser sintering (SLS), and stereolithography (SLA). *Polymers (Basel)* **13(18)**, 3101 (2021).
+
+14. Karakurt, I., Aydoğdu, A., Çıkrıkcı, S., Orozco, J. & Lin, L.. Stereolithography (SLA) 3D printing of ascorbic acid loaded hydrogels: A controlled release study. *Int. J. Pharm.* **584**, 119428 (2020).
+
+15. Xu, X. et al. Stereolithography (SLA) 3D printing of an antihypertensive polyprintlet: Case study of an unexpected photopolymer-drug reaction. *Addit. Manuf.* **33**, 101071 (2020).
+
+16. Wangpraseurt, D. et al. Bionic 3D printed corals. *Nat. Commun.* **11(1)**, 1748 (2020).
+
+17. Zhang, J. et al. A comparative study of the osteogenic performance between the hierarchical micro/submicro-textured 3D-printed Ti<sub>6</sub>Al<sub>4</sub>V surface and the SLA surface. *Bioact. Mater.* 5(1), 9–16 (2020).
+
+18. Zhou, T. et al. SLA 3D printing of high quality spine shaped β-TCP bioceramics for the hard tissue repair applications. *Ceram. Int.* **46(6)**, 7609–7614 (2020).
+
+19. Jalili, A.R. et al. Liquid crystal-mediated 3D printing process to fabricate nano-ordered layered structures. *ACS Appl. Mater. Interfaces* **13(24)**, 28627–28638 (2021).
+
+20. Kotikian, A., Truby, R.L., Boley, J.W., White, T.J. & Lewis, J.A.. 3D printing of liquid crystal elastomeric actuators with spatially programed nematic order. *Adv. Mater.* **30(10)**, 1706164 (2018).
+
+21. Mistry, D. et al. Soft elasticity optimises dissipation in 3D-printed liquid crystal elastomers. *Nat. Commun.* **12(1)**, 6677 (2021).
+
+22. Gantenbein, S. et al. Three-dimensional printing of hierarchical liquid-crystal-polymer structures. *Nature* **561(7722)**, 226–230 (2018).
+
+23. Liu, F. et al. Process-driven microstructure control in melt-extrusion-based 3D printing for tailorable mechanical properties in a polycaprolactone filament. *Macromol. Mater. Eng.* **303(8)**, 1800173 (2018).
+
+24. Chen, S. et al. Synthesis of novel acrylic liquid-crystal resin and its in-situ enhancement in light-curing 3D printing performance. *J. Mater. Res. Technol.* **17**, 2158–2174 (2022).
+
+25. Kim, K. et al. 4D printing of hygroscopic liquid crystal elastomer actuators. *Small* **17(23)**, e2100910 (2021).
+
+26. Liu, X. et al. 3D printing of living responsive materials and devices. *Adv. Mater.* **30(4)**, (2018) 1704821.
+
+27. Devadiga, D. & Ahipa, T.N. Heterodimeric hydrogen-bonded mesogens comprising pyridine moiety: A review. *Liq. Cryst. Rev.* **8(1)**, 5–28 (2020).
+
+28. Devadiga, D. & Ahipa, T.N. An Up-to-date review on halogen-bonded liquid crystals. *J. Mol. Liq.* **333**, 115961 (2021).
+
+29. Devadiga, D. & Ahipa, T.N. Recent synthetic advances in pyridine-based thermotropic mesogens. *RSC Adv.* **9(40)**, 23161–23228 (2019).
+
+30. Devadiga, D. & Ahipa, T.N. Recent advancements in the mesogens comprising of 1,3,5-triazine core moiety. *Liq. Cryst. Rev.* **7(2)**, 107–141 (2019).
+
+31. Popov, N. et al. Thermotropic liquid crystal-assisted chemical and biological sensors, *Materials (Basel)* **11(1)**, 20 (2017).
+
+32. Buyuktanir, E.A., Frey, M.W. & West, J.L. Self-assembled, optically responsive nematic liquid crystal/polymer core-shell fibers: Formation and characterization. *Polymer* **51(21)**, 4823–4830 (2010).
+
+33. West, J.L., Wang, J.R. & Jákli, A. Airbrushed liquid crystal/polymer fibers for responsive textiles. *Adv, Sci. Technol.* **100**, 43–49 (2016).
+
+34. Wang, J., Jákli, A. & West, J.L. Morphology tuning of electrospun liquid crystal/ polymer fibers. *Chem.Phys.Chem.* **17(19)**, 3080–3085 (2016).
+
+35. Cheng, Z., Wang, T., Li, X., Zhang, Y. & Yu, H. NIR-vis-UV light-responsive actuator films of polymer-dispersed liquid crystal/graphene oxide nanocomposites. *ACS Appl. Mater. Interfaces* **7(49)**, 27494–27501 (2015).
+
+36. Kumano, N. et al. Multicolor polymer-dispersed liquid crystal. *Adv. Mater.* **23(7)**, 884–888 (2011).
+
+37. Kim, M. et al. Fabrication of microcapsules for dye-doped polymer-dispersed liquid crystal-based smart windows. *ACS Appl. Mater. Interfaces* **7(32)**, 17904–17909 (2015).
+
+38. Seo, J. et al. Broadband PH-sensing organic transistors with polymeric sensing layers featuring liquid crystal microdomains encapsulated by di-block copolymer chains. *ACS Appl. Mater. Interfaces* **8(36)**, 23862–23867 (2016).
+
+39. Yu, L., Cheng, Z., Dong, Z., Zhang, Y. & Yu, H.. Photomechanical response of polymer-dispersed liquid crystals/graphene oxide nanocomposites. *J. Mater. Chem. C* **2(40)**, 8501–8506 (2014).
+
+40. Kikuchi, H., Yokota, M., Hisakado, Y., Yang, H. & Kajiyama, T. Polymer-stabilized liquid crystal blue phases. *Nat. Mater.* **1(1)**, 64–68 (2002).
+
+41. Broer, D.J., Lub, J. & Mol, G.N. Wide-band reflective polarizers from cholesteric polymer networks with a pitch gradient. *Nature* **378(6556)**, 467–469 (1995).
+
+42. Yu, L. & Yu, H. Light-powered tumbler movement of graphene oxide/polymer nanocomposites. *ACS Appl. Mater. Interfaces* **7(6)**, 3834–3839 (2015).
+
+43. Zheng, Z., Wang, C. & Shen, D. Dichroic-dye-doped polymer stabilized optically isotropic chiral liquid crystals. *J. Mater. Chem. C* **1(39)**, 6471 (2013).
+
+44. Zheng, Z.G. et al. Low-temperature-applicable polymer-stabilized blue-phase liquid crystal and its kerr effect. *J. Soc. Inf. Disp.* **20(6)**, 326 (2012).
+
+45. Allouchery, V., Roussel, F. & Buisine, J.M. Thermodynamic and electro-optic characteristics of UV-cured monofunctional acrylate/nematic liquid crystal mixtures. *Mol. Cryst. Liq. Cryst. Sci. Technol. A Mol. Cryst. Liq. Cryst.* **329(1)**, 227–237 (1999).
+
+46. Nastł,E., Żurańska,E. & Mucha, M. Effect of curing progress on the electrooptical and switching properties of PDLC system. *J. Appl. Polym. Sci.* **71(3)**, 455–463 (1999).
+
+47. Vaz, N.A. & Montgomery, G.P. Refractive indices of polymer-dispersed liquid-crystal film materials: Epoxy-based systems. *J. Appl. Phys.* **62(8)**, 3161–3172 (1987).
+
+48. Smith, G.W. & Vaz, N.A. The relationship between formation kinetics and microdroplet size of epoxy-based polymer-dispersed liquid crystals. *Liq. Cryst.* **3(5)**, 543–571 (1988).
+
+49. Souĺe, E.R., Abukhdeir, N.M. & Rey, A.D. Thermodynamics, transition dynamics, and texturing in polymer-dispersed liquid crystals with mesogens exhibiting a direct isotropic/smectic-a transition. *Macromolecules* **42(24)**, 9486–9497 (2009).
+
+50. Kuschel, F. Hartmann, L. & Bauer, M. Simplified preparation of polymer-encapsulated LC layers by thermally induced phase separation. *Liq. Cryst.* **38(3)**, 325–331 (2011).
+
+51. Srivastava, J.K., Singh, R.K., Dhar, R. & S. Singh. Thermal and morphological studies of liquid crystalline materials dispersed in a polymer matrix. *Liq. Cryst.* **38(7)**, 849–859 (2011).
+
+52. Yang, H., Kikuchi, H. & Kajiyama, T. Study on thermal-addressing characteristics of (polymer network/liquid crystals/chiral dopant) composite. *Chem. Lett.* **32(3)**, 256–257 (2003).
+
+53. Yang, H., Kikuchi, H. & Kajiyama, T. (Homeotropically Oriented Polymer Network/Side Chain Type Liquid Crystalline Polymer/Liquid Crystals/Chiral Dopant) Composite with Thermally Addressed and Electrically Erasable Properties. *Mol. Cryst. Liq. Cryst.* **381(1)**, 85–99 (2002).
+
+54. Pan, G. et al. Effects on thermo-optical properties of the composition of a polymer-stabilised liquid crystal with a smectic A-chiral nematic phase transition. *Liq. Cryst.* **35(9)**, 1151–1160 (2008).
+
+55. Seo, D.S., Matsuda, H., Oh-ide, T. & Kobayashi, S. Alignment of Nematic Liquid Crystal(5CB) on the Treated Substrates: Characterization of Orientation Films, Generation of Pretilt Angles, and Surface Anchoring Strength. *Mol.Cryst.liq. Cryst.* **224**, 13-31 (1993).
+
+56. Bauman, D. Optical and thermodynamic properties of mixtures of pleochroic dye with some alkyl-cyanobiphenyl liquid crystals. *Mol.Cryst., Liq. Cryst.* **172(1)**, 41–50 (1989).
+
+57. Pan, R.P., Chen, S.M. & Yang, H.Y. Free-surface effects on the response of nematic liquid crystal films to a laser pulse. *Liq. Cryst.* **14(5)**, 1561–1571 (1993).
+
+58. Eichler, H.J., Macdonald, R. & Dettmann, C. Nonlinear diffraction of CW-laserbeams by spatial selfphase-modulation in nematic liquid crystals. *Mol.Cryst., Liq. Cryst.* **174**, 153–168 (1989).
+
+59. Tabiryan, N.V., Sukhov, A.V., Ya, B. & Zel’dovich. Orientational optical nonlinearity of liquid crystals. *Mol.Cryst., Liq. Cryst.* **136**, 1–139 (1986).
+
+60. Joralmon, D. et al. Three-dimensional printing of liquid crystals with thermal sensing capability via multimaterial vat photopolymerization. *ACS Appl. Polym. Mater.* **4(4)**, 2951–2959 (2022).
+
+61. Prévôt, M.E. et al. Synchrotron microbeam diffraction studies on the alignment within 3D-printed smectic-A liquid crystal elastomer filaments during extrusion. *Crystals* **11(5)**, 523 (2021).
+
+62. Rodriguez-Palomo, A. et al. Nanostructure and anisotropy of 3D printed lyotropic liquid crystals studied by scattering and birefringence imaging. *Addit. Manuf.* **47**, 102289 (2021).
+
+63. Yang, C. et al. 3D-printed biomimetic systems with synergetic color and shape responses based on oblate cholesteric liquid crystal droplets. *Adv. Mater.* **33(10)**, e2006361 (2021).
+
+64. Rodriguez-Palomo, A. et al. Nanostructure and anisotropy of 3D printed lyotropic liquid crystals studied by scattering and birefringence imaging. *Addit. Manuf.* **47**, 102289 (2021).
+
+65. Chen, S. et al. Synthesis of novel acrylic liquid-crystal resin and its in-situ enhancement in light-curing 3D printing performance. *J Mater. Res. Technol.* **17**, 2158-2174 (2022).
+
+66. Cecil, F., Guijt, R. M., Henderson, A. D., Macka, M. & Breadmore, M. C. One step multi-material 3D printing for thefabrication of a photometric detector flow cell. *Anal. Chim. Acta* **1097**, 127−134 (2020).
+
+67. Dumanli, A. G. & Savin. T. Recent advances in the biomimicry of structural colours. *Chem. Soc. Rev.* **45**, 6698−6724 (2016).
+
+68. White, T. J. & Broer, D. J. Programmable and adaptive mechanics with liquid crystal polymer networks and elastomers. *Nat. Mater.* **14**, 1087−1098 (2015).
+
+69. Nocentini, S., Martella, D., Parmeggiani, C. & Wiersma, D. S. 3D printed photoresponsive materials for photonics. *Adv. Opt. Mater.* **7**, 1900156 (2019).
+
+# Supplementary Files
+
+- [Supplementaryinformation20230214NC.docx](https://assets-eu.researchsquare.com/files/rs-2589056/v1/0d18e17b5c3551e282562da3.docx)
+- [Video1PR5CB0parallel.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/d592d3e0bd0d97c863afcb5e.mp4)  
+  Video 1 PR-5CB-0 parrallel
+- [Video2PR5CB0perpendicular.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/7d3fcdd62134dd5f3bff37e2.mp4)
+- [Video3PR5CB3parallelsurface.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/65bc8f0b0243947b41d7ab80.mp4)  
+  Video 3 PR-5CB-3 parrallel surface
+- [Video4PR5CB3parallelcrosssection.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/ba852926ac26ac759c8ef890.mp4)  
+  Video 4 PR-5CB-3 parrallel cross section
+- [Video5PR5CB3perpendicularsurface.mp4.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/11b41dfdaac655056267c10b.mp4)
+- [Video6PR5CB3perpendicularcrosssection.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/05e4c1bc20f6088182318566.mp4)
+- [Video7removed5CBofPR5CB3.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/4b29415c4fa9f17bcaa79b73.mp4)  
+  Video 7 removing 5CB of PR-5CB-3
+- [Video8F1parallel.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/051a298b0a4f47f2139ff4a5.mp4)  
+  Video 8 F1 parrallel-SI
+- [Video9F2parallel.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/50a917f471224860c55a6de2.mp4)  
+  Video 9 F2 parrallel-SI
+- [Video10F3parallel.mp4](https://assets-eu.researchsquare.com/files/rs-2589056/v1/bfacf52642a4416d96ce933a.mp4)  
+  Video 10 F3 parrallel-SI
+- [GA.png](https://assets-eu.researchsquare.com/files/rs-2589056/v1/c29718b47f69e7217f4092ed.png)  
+  Graphical Abstract

0a6e6a3a08da886347f9cea32930b301d9ee96c0ebea0c3acd080f10c7740148/preprint/images_list.json

@@ -0,0 +1,66 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpeg",
+        "caption": "Schematic representation of the different testing and tracing strategies and which parts of the chain of transmission they can uncover. On the left, a transmission chain is shown where COVID-19 spreads from a parent case to an index case and their sibling case at a source event (attendees circled with dotted line). The index, sibling and child cases all spread their infection further. Black arrows show transmission events, while green diagonals show the infectious period of each case. The index case develops symptoms on day 0 and gets tested as soon as possible. Double full vertical lines highlight when each case is detected as a contact, considering a combined testing and tracing delay of 1 day and testing of identified contacts as soon as possible. The standard and extended contact tracing windows are shown above the timeline. The testing and contact tracing strategies and which additional case they identify is shown on the right. As especially the parent case demonstrates, a possible drawback of backward contact tracing is that some infected contacts are detected at a later stage of their infection, decreasing the effectiveness of testing and quarantine measures. It must be noted that the directionality of transmission and thus the position of an infected individual in the transmission tree is usually difficult to ascertain in practice.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Schematic representation of contact tracing strategies. Thin black and thick green arrows indicate the directions of transmission and contact tracing respectively. I: index case. C: child case. P: parent case. S: sibling case. White circle: undetected case. Grey circle: case detected through symptomatic screening. Green circle: case detected through contact tracing (a) When an index case is diagnosed, the child case at event D-1 is identified through standard forward tracing. A source investigation would fail at this stage, because there is no indication of further infections at the source event. (b) Source investigation does succeed when a second index case I2 is diagnosed independently of the initial index case I1. As the source event becomes clear due to identification of multiple infections, all attendants are traced. (c) An extended tracing window quickly identifies parent, sibling and child cases as direct contacts of the first index case.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpeg",
+        "caption": "Exclusion flow chart for the number of cases and contacts included and excluded in the main alpha-dominant cohort.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Outcomes, positivity rates and risk ratios for contacts of index cases. The dotted line indicates the positivity rate in the control group. The error bars indicate 95% confidence intervals. * indicates a statistically significant difference in comparison to the control group (p<0.05). Section (a) tests the main hypothesis by comparing the extended tracing window to the symptomatic control group. Subgroups by the numbers of days from onset or test of the index case to the last interaction with the index case are shown in section (b) and (c) for the extended and standard tracing windows respectively. Section (d) shows subgroups according to presence at suspected source events, and subgroups by relationship type are shown in panel (e).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.jpeg",
+        "caption": "Symptom onset in infected contacts relative to symptom onset (panel a) or sampling (panel b) of the index case. Asympt: asymptomatic. Case-contact pairs which were excluded from the mean calculation because either or both were asymptomatic, are shown on the right. Panel (a) shows the delay between symptom onset of an index case and their infected contact. Based on published serial intervals for the Alpha strain of 4-5 days38, the arrows roughly indicate the ranges where we expect to find most parent, sibling and child cases. The results suggest that the backward traced group consisted of a high proportion of sibling cases and few parent or child cases. Panel (b) shows the delay between detection of an index case and symptom onset of their infected contact. \u00a0Forward traced symptomatic contacts were detected on average 1.77 days earlier in their infectious cycle than their backward traced counterparts, assuming equal testing and tracing delays.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.jpeg",
+        "caption": "Timing of RT-qPCR testing in contacts as performed in the study period and the diagnostic accuracy of such tests by day since last exposure. Panel (a) shows the number of contacts who underwent either one or two tests at our test centre after their exposure. This demonstrates how testing immediately after exposure (\u201ctest to trace\u201d) was most often complemented with testing after a latent period (\u201ctest to release\u201d). While the former mainly supports iterative tracing and a shortened isolation period for asymptomatic infected contacts, the latter allows shortening of quarantine for non-infected contacts. As the delay between last exposure and symptom onset or testing of the index case increased, the percentage of contacts requiring two tests decreased. Panel (b) shows the timing of first and seconds tests at our centre for contacts, relative to their last exposure. The difference in timing of the first and second tests is reduced as the contact tracing window is extended further back in time. \u00a0Panel (c) shows the test results of infected contacts by day after last exposure, demonstrating how the sensitivity of RT-qPCR testing increased rapidly in the first days after exposure.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_7.png",
+        "caption": "Outcomes, positivity rates and risk ratios for contacts of index cases in selected periods, differing with regards to the dominant variants of concern, immunity, level of viral circulation, social contact restrictions and government testing/quarantine strategy. The error bars indicate 95% confidence intervals. Panel (a) repeats the main study outcomes from Figure 4 panel (a), while the results from subsequent periods are shown in panels (b) and (c).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_8.jpeg",
+        "caption": "Schematic representation of two possible strategies for backward contact tracing, based on our results. Panel (a) shows a hybrid strategy, which avoids testing contacts in the extended tracing window who were not present at the suspected source event. Panel (b) shows an extended tracing window strategy with systematic testing of all contacts.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0a6e6a3a08da886347f9cea32930b301d9ee96c0ebea0c3acd080f10c7740148/preprint/preprint.md

@@ -0,0 +1,283 @@
+# Abstract
+
+Standard contact tracing practice for COVID-19 is to identify persons exposed to an infected person during the contagious period, assumed to start two days before symptom onset or diagnosis. In the first large cohort study on backward contact tracing for COVID-19, we extended the contact tracing window by 5 days, aiming to identify the source of the infection and persons infected by the same source. The risk of infection amongst these additional contacts was similar to contacts exposed during the standard tracing window and significantly higher than symptomatic individuals in a control group, leading to 42% more cases identified through contact tracing. Compared to standard practice, backward traced contacts required fewer tests and shorter quarantine, but if infected they were identified later in their infectious cycle. Our results support implementing backward contact tracing when rigorous suppression of viral transmission is warranted.
+
+[Health sciences/Diseases/Infectious diseases/Viral infection](/browse?subjectArea=Health%20sciences%2FDiseases%2FInfectious%20diseases%2FViral%20infection)
+
+[Health sciences/Health care/Public health/Epidemiology](/browse?subjectArea=Health%20sciences%2FHealth%20care%2FPublic%20health%2FEpidemiology)
+
+# Introduction
+
+## The role of contact tracing in COVID-19
+
+Case-based interventions such as case isolation or contact tracing with quarantine have been crucial in controlling the ongoing COVID-19 pandemic, while reducing the need for indiscriminate contact reductions with high economic cost<sup>1,2</sup>. Contact tracing aims to identify and interrupt transmission chains by isolating infected patients and quarantining those at risk from infection. More infections are prevented, and epidemic control is improved, if the identification of patients and contacts at risk is rapid and comprehensive<sup>3–6</sup>. It has been a staple public health intervention in a variety of infectious diseases, notably sexually transmitted diseases and tuberculosis<sup>7,8</sup>. Worldwide investments in contact tracing programs and research on the topic have not prevented repeated resurgence of community transmission of COVID-19, underscoring the urgent need for improved knowledge on the effective implementation of this key public health measure<sup>6,9</sup>.
+
+## Forward contact tracing
+
+Forward contact tracing of an index case (the person diagnosed with COVID-19 undergoing contact tracing) intends to interrupt onward transmission from child cases (persons infected by the index case) by quarantining and/or testing contacts the index case has encountered during their infectious period<sup>10–12</sup>. In the light of substantial asymptomatic and pre-symptomatic transmission, the infectious period is generally assumed to start 2 days prior to onset of symptoms or diagnosis, whichever came first<sup>13–18</sup>. In addition to child cases, any practical forward tracing strategy probably identifies the parent case (the infector of the index case) and sibling cases (infected by the same parent case) some of the time, for example if the index case had repeated contact with their parent or sibling case during their own infectious period, or if the time from the index case's infection to their symptom onset or diagnosis was less than two days<sup>12</sup>. Forward contact tracing is the focus in most jurisdictions and has shown its ability to decrease COVID-19 transmission (Fig. <span class="InternalRef" refid="Fig1">1</span>)<sup>13,14,19</sup>.
+
+## Backward contact tracing
+
+Backward contact tracing, or bidirectional contact tracing, which combines both approaches, specifically aims to identify the parent case and sibling cases by going back further in time<sup>5,10−12</sup>. In any practical implementation, additional child cases may also be identified through backward contact tracing, for example if the index case’s infectiousness started more than two days before symptom onset<sup>12</sup>. Backward contact tracing is particularly promising in COVID-19 because a small proportion of index cases, the so-called superspreaders, generate the majority of secondary infections<sup>11,20−27</sup>. This phenomenon favours allocating resources to the identification of source cases and events, as a high rate of infection can be expected amongst individuals exposed to the same source. Endo et al estimate bidirectional contact tracing to result in 2–3 times the number of subsequent cases averted compared to forward contact tracing alone in a simple branching model for COVID-19<sup>10</sup>. Kojaku et al show backward contact tracing to be highly effective in terms of the number of prevented cases per quarantine when running an SEIR (Susceptible-Exposed-Infectious-Removed) model on synthetic and empirical contact networks, even if contact tracing comprehensiveness is low<sup>11</sup>. One potential difficulty of backward contact tracing lies in the inherent delays involved in testing, tracing and quarantine – where infected contacts who are sibling or parent cases risk being detected after or near to the end of their infectious period<sup>3,18</sup>. This could reduce efficiency and increase the relative cost of testing and quarantine (Fig. <span class="InternalRef" refid="Fig1">1</span>). Due to these delays, immediate testing of identified contacts in support of iterative tracing may be especially relevant in backward contact tracing.
+
+## Types of backward contact tracing
+
+The real-world implementation of backward contact tracing can be broadly subdivided into a source event approach and an extended contact tracing window approach (Fig. <span class="InternalRef" refid="Fig2">2</span>). Several countries have rolled out an approach focusing on source events, which are events where the index case is suspected to have contracted COVID-19. The identification of such an event leads to the screening of attendants at risk, which usually includes more individuals than the direct contacts of the index case under investigation<sup>28–32</sup>. This is because the risk at these events is not related to the index case, but to an unknown parent case. High positivity rates have been reported for attendants of some source events<sup>33</sup>. In practice, this approach is usually reliant on the identification of multiple confirmed or probable infected cases at the same event, for example by pooling of contact tracing data from different index cases or asking the index case about other cases in their environment. As a result, the approach can fail to identify the source event at the time of identification of the initial index case. Another approach is to extend the contact tracing window back in time and to systematically refer all close contacts for quarantine and/or testing (Fig. <span class="InternalRef" refid="Fig1">1</span>, <span class="InternalRef" refid="Fig2">2</span>). This assumes that, if the tracing window is extended backward by at least the incubation period of the index case, the parent case can be identified, as well as sibling cases present at a shared source event. To this end, the contact tracing window should be extended far enough to include most of the variability in incubation periods<sup>34</sup>. Several modelling studies underscore the benefits of extending the contact tracing window for COVID-19. Bradshaw et al show in a stochastic branching-process model that extending the contact tracing window from 2 to 6 days before onset or diagnoses improves the reduction in the effective reproduction number by 85%-275% when using manual contact tracing only (performed by humans rather than through digital means)<sup>12</sup>. Their findings are robust to contextual factors such as case ascertainment rate, test sensitivity, basic reproduction number and the percentages of asymptomatic, pre-symptomatic and environmental transmission. Fyles et al also show in a branching process model that an extended contact tracing window results in a linear decrease in the growth rate up until around 8 to 10 days prior to symptom onset or diagnosis, although additional gains are highly reliant on recall decay<sup>5</sup>.
+
+## Hypothesis and research question
+
+Whilst there is evidence from modelling studies pointing at the potential benefits of backward contact tracing, no study has evaluated the efficiency in practice. The positivity rate of screened contacts has been proposed as an indicator for efficient allocation of testing and quarantine<sup>35,36</sup>. In this cohort study we thus determined the positivity rate of additional close contacts (for the purpose of this article this includes co-attendants of high risk events of up to 20 persons) identified in an extended contact tracing window, starting 7 days before onset of symptoms or diagnosis, whichever was earlier. This window was chosen to include the source event most of the time<sup>32–34</sup>. We tested the hypothesis that the positivity rate amongst additional contacts in the extended tracing window would be at least as high as amongst a control group of patients attending the test centre for symptoms suggestive of COVID-19. In a first subgroup analysis, we explored how far back the contact tracing window should extend, by calculating the positivity rate of identified contacts grouped by day of last exposure. Our second hypothesis was that the risk would not be limited to possible source events identified at the time of the tracing interview. Therefore, the second subgroup analysis compared our strategy to a source investigation approach, by subgrouping contacts last exposed in the extended contact tracing window according to presence at suspected source events.
+
+# Results
+
+## Study cases and contacts
+
+Our test and trace program started in September 2020 and is still active in April 2022. Due to gradual improvements in organisation and data collection, there was a marked increase in the ratio of contacts with outcome data after the initial months of the program (Supplementary Fig. 3). The study period for the main analyses was chosen from 1st February 2021 to 31st May 2021, which was after the initial set-up phase of the program and included both an upward and a downward trend in country-wide infection rates.
+
+14,917 students underwent RT-qPCR testing at our centre in this period (3.8 tests per 1,000 persons daily), resulting in 498 students with a new diagnosis of COVID-19. A further 231 positive RT-qPCR test results of students in the study population were reported to us from external sources, resulting in a total of 729 cases. 36 (4.9%) of these were interpreted as a past infection or false positive by the treating physician, leaving 693 actual cases (14-day incidence of 245 per 100,000). Six cases (0.9%) were considered lost to follow-up, because they could never be contacted by the contact tracing team, and 28 (4.1%) were excluded because data on presence of symptoms was missing. Therefore, 659 index cases remained in the analysis (Fig. 3).
+
+72.5% of index cases self-reported being symptomatic at the time of testing, which was similar to the national average<sup>37</sup>. Index cases had a mean age of 21.4 years (SD: 3.60 years, missing data 15.0%) and were 51.1% male (missing data 12.1%).
+
+Contact tracing of the index cases resulted in 3,971 case-contact pairs (mean 6.0 contacts per case, 2.2 times the national average<sup>37</sup>), of which 956 (24.1%) were excluded because the contact person already had a positive test result 0 to 60 days before the positive test of the index case. Another 331 (11.0%) contacts were excluded because they already had a known exposure to a different infected individual within 7 days before the tracing interview. Finally, 288 contacts (10.7%) were lost to follow-up.
+
+The resulting 2,403 contacts were divided into two groups. The standard tracing window group, which would have been identified through standard practice, consisted of 1610 individuals in close contact with the index case in the period from 2 days before onset or test until the contact tracing interview. The study group consisted of 793 additional contacts in the extended tracing window, i.e. their last close interaction with the index case was 3 to 7 days before onset or test. For the main analysis, we did not make assumptions on the directionality of transmission. Therefore, both the forward and backward traced group likely included parent, sibling and child cases.
+
+We did not collect demographic data on contacts of index cases.
+
+The control group consisted of all 1,461 students who attended our test centre for the first time with self-reported symptoms suggestive of COVID-19 as the main reason for their test.
+
+There was a slightly higher percentage of women in the control group (56.5%, missing data 3.0%) compared to the index cases, while the mean age was similar (22.0 years; SD 3.84 years, missing data 3.0%). The temporal distribution of individuals in the backward traced contact and symptomatic control groups is shown in Supplementary Fig. 1, panel (a).
+
+## High risk of infection in the extended tracing window
+
+By extending the contact tracing window, 49% more contacts at risk and 42% more cases were identified, compared to standard contact tracing practice alone.
+
+The risk of infection in the standard and extended tracing window groups was similar, namely 17.1% in the former (CI 15.3–19.1%) and 14.6% in the latter (CI 12.2–17.3%). The risk in the extended tracing window group was significantly higher (risk ratio 2.22, CI 1.72–2.88, p < 0.0001) than the risk of 6.5% (CI 5.3–7.9%) in the control group, demonstrating the relative efficiency of extending the contact tracing window to 7 days prior to symptom onset or test (Fig. 4).
+
+Contacts in the standard and extended tracing window groups were subgrouped by their last day of contact with the index case, relative to symptom onset or test. The results show that the number of additional identified close contacts per day decreased markedly as the tracing window was extended backward. The risk of infection varied from 8.5–19.2%, and the confidence interval lower bound did not drop below 3.5% for any of these subgroups in the extended tracing window. For day 3, 4 and 5 before onset or test, the risk was significantly higher than the control group (p < 0.05).
+
+## The risk is not limited to suspected source events
+
+An important consideration when deciding between a source investigation approach and an extended tracing window is the risk of infection for contacts not present at suspected source events. A suspected source event was identified for 80.6% of index cases. If the contact tracing interview failed to suggest a source event, the risk of infection for extended tracing window contacts was 17.1% (CI 11.9–23.6%). If a source event was identified, the risk was around 4 times higher for contacts who attended the event (absolute risk 27.5%, CI 21.6–34.2%) compared to those who did not. The latter group still had a risk of 6.9% (CI 4.6–9.8%), which was similar to the symptomatic control group but not significantly higher (Fig. 4, panel a).
+
+## Risk by relationship type
+
+In an explorative subgroup analysis, extended tracing window contacts were grouped according to relationship type with the index case. The majority of identified contacts were either family (28.6%), fellow residents in student housing (12.3%), or friends (48.2%). Each of these three groups had a significantly increased infection risk as compared to the symptomatic control group. The other subgroups lacked sufficient numbers for statistical power. (Fig. 4, panel e)
+
+## Backward contact tracing identifies mostly sibling cases
+
+Effective contact tracing requires the detection of infected contacts as soon as possible, before they reach the end of their contagious period. The sibling and especially parent cases targeted by backward contact tracing can be expected to be in a later stage of infection compared to forward traced contacts, potentially leading to lower efficiency of tracing, testing and quarantine measures.
+
+The differences in onset dates between index cases and their infected contacts give an indication of their relative position in the transmission tree (Fig. 5, panel a). Our data suggest that the relative share of sibling cases was indeed higher amongst infected backward traced contacts, compared to contacts in the standard tracing window, while parent cases formed a minority in both groups.
+
+As a result, when comparing the date of detection of an index case with the onset date of their infected contact, the infected contacts in the extended tracing window were on average 1.8 days later in their infectious cycle compared to those in the standard tracing window (Fig. 5, panel b). The difference could be interpreted as a reduction in contact tracing efficiency equal to an additional testing or tracing delay of the same period.
+
+## Less tests and shorter quarantine in the backward traced group
+
+The value of contact testing depends not only on test specific diagnostic performance, but also on timing. Immediate testing after contact identification can accelerate iterative tracing (“test to trace”) and – if an asymptomatic contact tests positive – reduce to total duration spent in quarantine and isolation. Tests after a latent period are more sensitive and can thus be used to allow shortening of quarantine for non-infected contacts (“test to release”) (Fig. 6, panel c)<sup>39</sup>.
+
+During the study period, contacts were requested to undergo RT-qPCR tests both as soon as possible after identification and 7 days after last exposure, which is reflected in the timing of contact testing in our dataset (Fig. 6, panel b).
+
+As backward traced contacts were detected a minimum of three days after their last exposure by definition and an average of 4.0 days longer after last exposure than forward traced contacts in our dataset, a single test at identification was more likely to serve both a “test to trace” and “test to release” strategy concurrently.
+
+Another consequence of this inherent difference in last exposure date is that, in our dataset, the duration of quarantine is up to 4.0 days shorter for contacts in the backward traced group compared to the forward traced group. The average duration of quarantine depends also on the distribution of exposure dates relative to index case detection, tracing delays and the set quarantine duration (Supplementary Fig. 6).
+
+## Impact of changing viral variants
+
+Consecutive SARS-CoV-2 variants of concern (VOC) may have challenged the effectiveness of contact tracing in several ways. First, their increased intrinsic transmissibility more rapidly overwhelms the public health system<sup>40,41</sup>. Second, their shortened incubation time and serial interval risk outpacing the delays inherent in testing and tracing<sup>34,38,42</sup>. To assess the influence of these altered transmission dynamics, the main analysis was repeated for periods when the Delta and Omicrons VOCs were dominant in the population (Fig. 7). These periods differed from the main study period not only in terms of the dominant circulating VOC, but also in the immune status of the target population, the general contact restrictions in place, the COVID-19 incidence rate and the government requirements concerning testing and quarantine (Supplementary Figs. 3,7–9)<sup>37,45,46</sup>.
+
+Unfortunately, follow-up rates dropped markedly after the main study period, especially for contacts in the extended tracing window. During the periods characterized by Delta dominance, backward traced contacts had similar positivity rates (PR) to both forward traced contacts and symptomatic controls, further supporting our main hypothesis (Fig. 7, panel b). During the periods characterized by Omicron dominance and an almost fully vaccinated population, backward traced contacts retained a very high PR (mean 13.3%, CI 8.5–19.5)<sup>43,44</sup>. It was however significantly lower than the much increased PR in symptomatic controls and forward traced contacts (Fig. 7, panel c).
+
+# Discussion
+
+This study lays out a strategy for backward contact tracing which markedly improves the effectiveness of contact tracing in the setting of COVID-19. It identified an additional 42% of cases not detected through the contact tracing protocol used in most jurisdictions, gains which are likely to have a major impact on epidemic control<sup>12</sup>. The main trade-off was that infected backward traced contacts were identified on average 1.8 days later in their infectious cycle than forward traced contacts. However, the burden of testing and quarantine was lower in backward traced contacts due to inherent differences in the timing of their last exposure to the index case. Our results contradict perceptions on cost efficiency, which continue to hamper the broader introduction of backward contact tracing as a standard mitigation strategy.
+
+Our approach was to extend the contact tracing window back in time from 2 to 7 days before symptom onset or test, and to systematically refer all identified close contacts in this period for testing, as well as co-attendees of small high-risk events. This simple change in standard protocol, which could be implemented both in manual and digital contact tracing, allowed mostly sibling cases to be identified quickly as direct contacts of the index case.
+
+Our data show that only 49% more contacts at risk are identified by extending the contact tracing window backward by 5 additional days. This could be explained by recall decay or by recurring contacts with the same individuals.
+
+Crucially, contacts last encountered during the extended tracing window had a higher risk of testing positive compared to symptomatic patients in the same population. These results were independent of whether they were friends, family or fellow residents of the index case<sup>34,38,42</sup>.
+
+Positivity rates amongst symptomatic individuals are dependent on many factors, such as the level of community transmission of SARS-CoV-2 and other respiratory viruses. Still, this group was chosen as a control group, because it represents a high bar and testing of symptomatic patients is standard in most protocols globally<sup>37,45,46</sup>.
+
+Unfortunately, we were unable to replicate the high follow-up rates of the main study period in subsequent periods with different dominant variants of concern. We attribute this mainly to gradual loosening of government-mandated testing protocols and higher viral circulation, forcing the contact tracing team to prioritise contact notification over follow-up<sup>45,46</sup> (<em>Supplementary Fig. 3</em>). The control group probably also suffered a further reduction in reliability after the main study period, due to the rollout of alternative testing methods such as pharmacy-based and self-administered rapid antigen tests and the progressive scaling back of RT-qPCR testing in general<sup>37,45,46</sup>. Based on follow-up rates, we chose four subsequent periods of interest, characterised by Delta and Omicron VOC dominance, for analysis (<em>Supplementary Figs. 3 and 6</em>). These periods also differed from the main study period with regards to several other factors, such as general contact restrictions, population immunity and government test and quarantine strategy (<em>Supplementary Figs. 8 and 9</em>)<sup>37,45,46</sup>. In the Delta periods, the positivity rate of backward traced contacts was similar to the symptomatic control and forward traced groups. In the Omicron periods, it was significantly lower than that of symptomatic and forward traced reference groups (Fig.<span class="InternalRef" refid="Fig7">7</span>). However, the lower bound of the positivity rate of backward traced contacts remained above the threshold positivity rates of 5% and 4% that the World Health Organisation (WHO) and European Centre for Disease Control (ECDC) recommended as target indicator for comprehensive testing, when considering all tests performed in a population<sup>43,44</sup>. It should be noted that we did not adjust the range of the extended tracing window to accommodate shorter incubation period and serial interval reported for the Delta and Omicron VOCs<sup>34,38,42</sup>.
+
+The high positivity rate observed in contacts last seen before the contagious period can be explained by several mechanisms. First, the index case may have become contagious more than 2 days before symptom onset or test. However, the number of child cases observed in the backward traced group was low. Second, the source case is likely to be amongst earlier contacts, but parent cases also formed a minority in this group. Third, due to a proven individual propensity to shed live virus and an above average number of social interactions, the source case is likely to have initiated other infections among the index’s contacts<sup>11,47</sup>. This explanation is supported by the fact that secondary infections linked to a particular index case followed a binomial distribution with a value for the dispersion parameter<em>k</em> of 0.407 (<em>Supplementary Fig. 4</em>), which implies a degree of overdispersion corresponding with literature<sup>20,22,23</sup>. The finding that backward traced contacts seem to consist mostly of sibling rather than parent cases, also supports this mechanism (Fig.<span class="InternalRef" refid="Fig5">5</span>). Fourth, more distant relatives in the transmission tree could be detected due to wider circulation in an index case’s social circle. Fifth, recall decay may cause index cases to forget contacts with whom they had shorter, fewer and less close interactions. There may also be a more intentional tendency to mention only those contacts who the index case considers at risk.
+
+A question that arises is whether it is worthwhile to quarantine and test a contact in the extended tracing window, if a source event was identified which the contact at hand did not attend. Our results show that the risk of infection for such a contact (6.9%, CI 4.7–9.7%) was only a quarter of that of source event attendees, but still similar to the symptomatic control group. It was also higher than the WHO and ECDC targets of 5% and 4% mentioned above<sup>43,44</sup>.
+
+These results speak in favour of simply referring all close contacts in the extended tracing window for testing and quarantine, even if they were not present at the suspected source event. Additionally, we show that jurisdictions favouring the implementation of a source investigation strategy would do well to switch to an extended contact tracing window approach when no clear source event is identified at the time of the contact tracing interview (hybrid strategy, Fig.<span class="InternalRef" refid="Fig8">8</span>).
+
+Previous studies have emphasized that the benefits of backward contact tracing hinge on the ability to identify first the parent case and then sibling cases in a two-step process, which is likely to be highly susceptible to testing and contact tracing delays<sup>5,10−12</sup>. However, the distribution of differences in symptom onset dates between index cases and their backward traced contacts suggests that most backward traced contacts were sibling cases identified as direct contacts of the index case, without the need to first identify the parent case (Fig.<span class="InternalRef" refid="Fig5">5</span>, <em>panel a</em>). As a result, backward traced contacts were detected only 1.8 days later in their infectious cycle, compared to forward traced contacts consisting mostly of child cases. Still, the later detection of infected backward traced contacts probably has a negative effect on efficiency, because they are isolated for a smaller fraction of their contagious period<sup>3</sup>.
+
+We would argue that this effect is compensated for by a lower testing and quarantine burden for backward traced contacts.
+
+Compared to forward traced contacts, the last exposure of backward traced contacts to the index case was 4.0 days earlier. This reduces the duration of their quarantine by up to the same amount, depending on exact exposure date, testing and tracing delay and policy set quarantine duration, and often eliminates the need for two tests (Fig.<span class="InternalRef" refid="Fig6">6</span> and <em>Supplementary Fig. 5</em>).
+
+In both backward and forward traced contacts, the rapidly increasing test sensitivity in the first days after exposure supports the implementation of an initial “test to trace” immediately after identification, which accelerates iterative tracing and shortens the total duration of quarantine and isolation for asymptomatic infected contacts (Fig.<span class="InternalRef" refid="Fig6">6</span> and <em>Supplementary Fig. 5</em>). A “test to release" after a latent period can have sufficient sensitivity to end the quarantine of uninfected contacts. As backward traced contacts are, by definition, identified late after exposure to the index case, a “test to trace” and “test to release” can be combined into a single test more often than in their forward traced counterparts (<em>Supplementary Fig. 5</em>).
+
+Overall, our results show that, on an individual level, the immediate cost and burden of backward contact tracing are proportional to the additional cases identified. Our data do not allow inferences about the impact on a population scale. However, several modelling studies have suggested that the improved epidemic control offered by backward contact tracing has the potential for dramatically lower costs to society, in the form of reduced testing, quarantine and illness<sup>5,10−12</sup>.
+
+The study has several limitations. First, the main analyses took place in the setting of moderate general contact restrictions, which altered social patterns significantly and likely increased the efficiency of identifying source individuals by decreasing the number of contacts in general and casual contacts in particular, which are harder to identify through manual contact tracing. Second, index cases were young adults in tertiary education, whose socio-economic status and contact patterns may differ significantly from other age and social groups, limiting generalisability<sup>48</sup>. Third, the population was almost entirely unvaccinated during the main study period. Fourth, the main variant circulating in the population at the time was the Alpha strain, with lower transmissibility than the subsequent Delta and Omicron VOCs (<em>Supplementary Fig. 9</em>)<sup>40,41</sup>. Our analyses of periods dominated by Delta and Omicron strains do not allow the same strong conclusions due to reduced data quality. Fifth, a testing and contact tracing program is a complex public health intervention, and the particular methods of implementation and contextual factors have a major impact on its overall effectiveness. The influence of host-, pathogen- and environment-related factors on the comparative efficiency of backward contact tracing strategies merits further study.
+
+Our results indicate that in the context of significant community transmission of COVID-19 and in the presence of moderate contact restrictions, there can be a marked added benefit, at low relative cost, to extending the contact tracing window backward beyond the infectious period of the index case.
+
+# Main References
+
+1. Davis, E. L. et al. Contact tracing is an imperfect tool for controlling COVID-19 transmission and relies on population adherence. *Nat. Commun.* **12**, (2021).
+
+2. Yalaman, A., Basbug, G., Elgin, C. & Galvani, A. P. Cross-country evidence on the association between contact tracing and COVID-19 case fatality rates. *Sci. Rep.* **11**, (2021).
+
+3. World Health Organization and others. *Contact tracing in the context of COVID-19: interim guidance, 10 May 2020*. (2020).
+
+4. European Centre for Disease Prevention and Control. Contact tracing: Public health management of persons, including healthcare workers, having had contact with COVID-19 cases in the European Union-second update. *ECDC, Stock.* (2020).
+
+5. Endo, A. et al. Implication of backward contact tracing in the presence of overdispersed transmission in COVID-19 outbreaks. *Wellcome Open Res.* **5**, (2021).
+
+6. Kojaku, S., Hébert-Dufresne, L., Mones, E., Lehmann, S. & Ahn, Y.-Y. The effectiveness of backward contact tracing in networks. *Nat. Phys.* **17**, (2021).
+
+7. Bradshaw, W. J., Alley, E. C., Huggins, J. H., Lloyd, A. L. & Esvelt, K. M. Bidirectional contact tracing could dramatically improve COVID-19 control. *Nat. Commun.* **12**, (2021).
+
+8. Demirgüç-Kunt, A., Lokshin, M. & Torre, I. The sooner, the better: The economic impact of non‐pharmaceutical interventions during the early stage of the COVID‐19 pandemic. *Econ. Transit. Institutional Chang.* **29**, (2021).
+
+9. Kretzschmar, M. E. et al. Impact of delays on effectiveness of contact tracing strategies for COVID-19: a modelling study. *Lancet Public Heal.* **5**, (2020).
+
+10. Scarabel, F., Pellis, L., Ogden, N. H. & Wu, J. A renewal equation model to assess roles and limitations of contact tracing for disease outbreak control. *R. Soc. Open Sci.* **8**, (2021).
+
+11. Fyles, M. et al. Using a household-structured branching process to analyse contact tracing in the SARS-CoV-2 pandemic. *Philos. Trans. R. Soc. B Biol. Sci.* **376**, (2021).
+
+12. Karlinsky, A. & Kobak, D. Tracking excess mortality across countries during the COVID-19 pandemic with the World Mortality Dataset. *Elife* **10**, (2021).
+
+13. Ge, Y. et al. COVID-19 Transmission Dynamics Among Close Contacts of Index Patients With COVID-19. *JAMA Intern. Med.* (2021) doi: 10.1001/jamainternmed.2021.4686.
+
+14. Ren, X. et al. Evidence for pre-symptomatic transmission of coronavirus disease 2019 (COVID‐19) in China. *Influenza Other Respi. Viruses* **15**, (2021).
+
+15. Casey-Bryars, M. et al. Presymptomatic transmission of SARS-CoV-2 infection: a secondary analysis using published data. *BMJ Open* **11**, (2021).
+
+16. Hu, S. et al. Infectivity, susceptibility, and risk factors associated with SARS-CoV-2 transmission under intensive contact tracing in Hunan, China. *Nat. Commun.* **12**, (2021).
+
+17. Adam, D. C. et al. Clustering and superspreading potential of SARS-CoV-2 infections in Hong Kong. *Nat. Med.* **26**, (2020).
+
+18. Endo, A., Abbott, S., Kucharski, A. J. & Funk, S. Estimating the overdispersion in COVID-19 transmission using outbreak sizes outside China. *Wellcome Open Res.* **5**, (2020).
+
+19. R, L. et al. Epidemiology and transmission dynamics of COVID-19 in two Indian states. *Science* **370**, (2020).
+
+20. Bi, Q. et al. Epidemiology and transmission of COVID-19 in 391 cases and 1286 of their close contacts in Shenzhen, China: a retrospective cohort study. *Lancet. Infect. Dis.* **20**, 911–919 (2020).
+
+21. Edwards, D. A. et al. Exhaled aerosol increases with COVID-19 infection, age, and obesity. *Proc. Natl. Acad. Sci.* **118**, (2021).
+
+22. Chen, P. Z. et al. Heterogeneity in transmissibility and shedding SARS-CoV-2 via droplets and aerosols. *Elife* **10**, (2021).
+
+23. Sneppen, K., Nielsen, B. F., Taylor, R. J. & Simonsen, L. Overdispersion in COVID-19 increases the effectiveness of limiting nonrepetitive contacts for transmission control. *Proc. Natl. Acad. Sci.* **118**, (2021).
+
+24. Whaley, C. M., Cantor, J., Pera, M. & Jena, A. B. Assessing the Association Between Social Gatherings and COVID-19 Risk Using Birthdays. *JAMA Intern. Med.* **181**, (2021).
+
+25. Shimizu, K. & Negita, M. Lessons Learned from Japan’s Response to the First Wave of COVID-19: A Content Analysis. *Healthcare* **8**, (2020).
+
+26. Yong, S. E. F. et al. Connecting clusters of COVID-19: an epidemiological and serological investigation. *Lancet Infect. Dis.* **20**, (2020).
+
+27. Finkel, A., Jasper, L. & Weeramanthri, T. *National Contact Tracing Review: A report for Australia’s National Cabinet*. (2020).
+
+28. Leclerc, Q. J., Fuller, N. M., Knight, L. E., Funk, S. & Knight, G. M. What settings have been linked to SARS-CoV-2 transmission clusters? *Wellcome Open Res.* **5**, (2020).
+
+29. Kasy, M. & Teytelboym, A. Adaptive targeted infectious disease testing. *Oxford Rev. Econ. Policy* **36**, (2020).
+
+30. Pezzutto, M., Bono Rosselló, N., Schenato, L. & Garone, E. Smart testing and selective quarantine for the control of epidemics. *Annu. Rev. Control* **51**, (2021).
+
+31. Wang, C. C. et al. Airborne transmission of respiratory viruses. *Science (80-. )* **373**, eabd9149 (2021).
+
+32. World Health Orgnisation. Public health criteria to adjust public health and social measures in the context of COVID-19 Annex to Considerations in adjusting public health and social measures in the context of COVID-19. https://apps.who.int/iris/bitstream/handle/10665/332073/WHO-2019-nCoV-Adjusting_PH_measures-Criteria-2020.1-eng.pdf?sequence=1&isAllowed=y (2020).
+
+33. ECDC. Maps in support of the Council Recommendation on a coordinated approach to travel measures in the EU. https://www.ecdc.europa.eu/en/covid-19/situation-updates/weekly-maps-coordinated-restriction-free-movement.
+
+34. Mossong, J. et al. Social Contacts and Mixing Patterns Relevant to the Spread of Infectious Diseases. *PLoS Med.* **5**, (2008).
+
+35. Shah, A. S. V. et al. Effect of Vaccination on Transmission of SARS-CoV-2. *N. Engl. J. Med.* (2021) doi: 10.1056/NEJMc2106757.
+
+36. Ke, R., Martinez, P., Smith, R., Gibson, L. & Achenbach, C. Longitudinal analysis of SARS-CoV-2 vaccine breakthrough infections reveal limited infectious virus shedding and restricted tissue distribution. *medRxiv* (2021) doi: https://doi.org/10.1101/2021.08.30.21262701.
+
+37. Raymenants, J. et al. Integrated PCR testing and extended window contact tracing system for COVID-19 to improve comprehensiveness and speed.
+
+# Methods
+
+## Study design and context
+
+In this cohort study we investigated the risk of contracting COVID-19 for contacts traced in an extended contact tracing window. Their risk was compared to a control group of patients from the target population, who were tested for self-reported symptoms of COVID-19 in the same period (Supplementary Fig. 1).
+
+A second reference group consisted of contacts exposed to an index case during the standard “forward” contact tracing window. The main outcome measure was a positive test in the 14 days after the last contact with the index case, or - for the control group - after the onset of symptoms.
+
+The study was performed in the context of a dedicated test and trace system for a target population of an estimated 32,965 higher education students residing in the city of Leuven, Belgium. A low-threshold test centre offered free RT-qPCR tests upon self-referral, while a team of contact tracers performed manual bidirectional contact tracing. The program relied heavily on community involvement and benefited from maximum integration of testing and tracing from a human process and information technology (IT) point of view. We elaborate on the operational aspects in a published testing and contact tracing protocol and show the delays involved in each step in this cascade during the study period in Supplementary Fig. 2<sup>49</sup>.
+
+The study protocol was approved by the Ethics Committee Research UZ / KU Leuven. Informed consent was waived as the data gathered did not exceed what was required for the purpose of safeguarding public health.
+
+We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines<sup>50</sup>.
+
+## Study participants
+
+Students attached to one of Leuven’s tertiary education facilities were included in the study if they either had a positive RT-qPCR test result at the KU Leuven test centre or if they were reported to the tracing team as having had a positive RT-qPCR test result elsewhere and had recently resided in or had come into contact with others in the city of Leuven. The main analysis included cases testing positive from 1st February until 31st May 2021 and their contacts.
+
+Cases were excluded if the treating physician interpreted the result as falsely positive, or as a past infection with COVID-19. Cases who could not be contacted by the tracing team after repeated attempts were also excluded, as well as cases where information on symptom onset was missing.
+
+Cases were asked about all their close interactions with contact persons in the period from 7 days before symptom onset or test until the time of the contact tracing interview.
+
+Contacts were included as a close contact if they were reported by the index case as having had either direct physical contact, an interaction at less than 1.5 meters without face masks, an interaction at less than 1.5 meters for more than 15 minutes, or an interaction without face masks for more than 15 minutes. Also included as close contacts were co-attendants at a “high risk event” of up to 20 attendees, defined as fitting at least 2 of the following 3 criteria: crowding (at least 5 individuals belonging to at least two households), close contact (< 1.5 meters without masks) and closed environment (indoor).
+
+Individuals who were already identified as contacts exposed to a previously diagnosed index case within 7 days before the contact tracing interview were excluded as contacts from the second identified index case, while still being considered as contacts for the first. While this approach introduces ambiguity as to the exact day of last exposure, it is reflective of our focus on decision making at the time of first identification of a contact.
+
+Contacts who had already tested positive on the same day as the index case or up to 60 days before, were also excluded. All other contacts were advised to quarantine while undergoing RT-qPCR testing as soon as possible and, if the test was negative, seven days after the last exposure to a positive case.
+
+Contacts were assigned to either the standard tracing window group, a reference group mirroring standard practice, or to the extended tracing window group, based on when their last close contact with the index case took place.
+
+As a control group, we selected all students who attended the test centre for the first time during the study period, and who self-reported symptoms suggestive for Covid-19 as the reason for their test. Only the first test was included, to reduce selection bias towards students with a lower threshold for testing.
+
+When comparing the symptom onset date of contacts to the sampling or onset date of their index case, the analysis was restricted to case-contact pairs where the contact was also included as a case in the main analysis.
+
+When computing the timing of testing after last exposure, the analysis was restricted to pairs where the contact was tested in the university testing centre and thus a student, as testing of other contacts didn’t fall under the responsibility of the university contact tracing team and therefore was not subjected to similarly rigorous follow-up.
+
+In the analysis assessing the sensitivity of RT-qPCR testing depending on the day after exposure, a contact was only labelled as “not infected” if they had a negative test between 7 and 14 days after last exposure. Test sensitivity on a particular day post-exposure was calculated for infected contacts who had not yet been diagnosed and was defined as the number of positive tests divided by the total number of tests in this group.
+
+For the analyses of more recent cohorts, time periods were chosen according to the main circulating VOC and the lost to follow-up rates of contacts (Supplementary Fig. 3). All index cases and their respective contacts were included by means of the same inclusion criteria as for the main Alpha dominant period. Inclusion and exclusion flowcharts are shown in Supplementary Fig. 6. In the last period, with the Omicron strain dominant, contacts who did not develop symptoms or undergo testing in the 7 days after exposure were considered “not infected”.
+
+## Data sources
+
+For cases and contacts tested in our test centre, RT-qPCR test results were reported directly by the laboratory. Students who tested positive elsewhere were reported by the government contact tracing teams, by the infected students themselves or by their contacts attending the test centre. The date of onset of symptoms was reported by the index case when attending the test centre and confirmed when being called by the contact tracing team.
+
+For each of their listed close contacts, we asked the index case about the dates and nature of their interactions, and the type of their relationship. Cases could supply this information using an online web form, and were contacted by telephone for confirmation and clarification during a thorough interview. Contacts were grouped into events if multiple people were present at the same time. These contact data were coded into a customized version of Go.Data, an outbreak investigation tool developed by the WHO and GOARN (Global Outbreak Alert and Response Network) partners<sup>51</sup>.
+
+Test dates and results of contacts who were tested outside of our test centre were obtained by telephone. This information was coded into Go.Data in a similar fashion.
+
+## Variables
+
+Contacts were assigned one of three possible outcomes. “Infected” includes those contacts who were diagnosed with COVID-19 1 to 14 days after the diagnosis of the index case. “Not infected” denotes other contacts who underwent an RT-qPCR or antigen test with a negative result 1 to 7 days after their last contact with the index case. All other contacts were considered “lost to follow-up”.
+
+The day of last contact was defined as the difference in days between the last date of interaction with the index case on the one hand, and on the other hand either the date of the positive test or the date of onset of symptoms, whichever was earlier.
+
+Each contact of an index case was assigned a relationship type from the following list: partner, family, friend, fellow resident, acquaintance, fellow student or other.
+
+Suspected source events were defined as events which, at the time of the contact tracing interview with the index case, were identified as the likely source of the infection, because the index case knew that an individual was present with a confirmed infection or suggestive symptoms. If the index case had been in quarantine since travelling from abroad, travel was considered the source event and travel companions were considered present. Multiple suspected source events were taken into account per index case if applicable. Suspected source events were required to fall within the backward tracing window at least partly to be labelled as such.
+
+## Study size
+
+The data feeding into this study were gathered in the light of the ongoing public health response for COVID-19. The exact study period was chosen from February onwards since gradual improvements in data gathering - through updates of the IT infrastructure and human capacity building - allowed for follow-ups of all contacts to be consistently recorded from February onwards. The end of the main study period marks the end of the academic semester, at which point testing and case numbers fell precipitously. The resulting number of cases and contacts is a consequence of the epidemiological trajectory within the study period.
+
+## Statistical methods
+
+Positivity rates were calculated with two-sided 95% confidence intervals according to the Clopper-Pearson method. Small-sample adjusted risk ratios were determined with two-sided normal approximation 95% confidence intervals.
+
+Missing demographic data was ignored in the calculations and the amount of missing data reported. Contacts with missing outcome data were considered lost to follow-up.
+
+Cases and contacts lost to follow-up were not included in the analysis.
+
+### Methods references
+
+47. von Elm, E. et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for reporting observational studies. Int. J. Surg. 12, (2014).
+
+48. World Health Organization. Go.Data. https://www.who.int/tools/godata (2021).
+
+# Supplementary Files
+
+- [sourcedatamainanalyses.csv](https://assets-eu.researchsquare.com/files/rs-952839/v2/9e2630147cbe40992411f233.csv)  
+  Dataset 1
+
+- [revisedsupplementaryinformation.docx](https://assets-eu.researchsquare.com/files/rs-952839/v2/ae7a83d2f66994b7a12bbfa7.docx)  
+  Supplementary Information

0c6e2f5cba0e7af6c6cfb672d733f15ac527ee25f7ea358f37e9e4f51d58e9db/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.jpg",
+        "caption": "Spatial distributions of wind and solar power prediction errors and the error-induced carbon emissions by province in China. a, Wind energy.b, Solar energy. The larger bubbles indicate the provincial wind and solar energy installations, and the smaller ones indicate the average wind and solar energy generation (8760 hours) by province. The provinces are divided into four groups according to the provincial prediction error (average value of 8760 hours) and marked with four colors. The province abbreviations are listed in the Supplemental Information. The thick red line marks the boundaries of the four areas of China, I. North China, II. East China, III. Central China, and IV. Southwest China. Individual provinces are indicated with lighter white lines. c, Carbon emissions caused by solar and wind energy prediction error. A darker color indicates a higher emission level of the province. d, Trajectory of the prediction error-induced carbon emissions from 2021 to 2030. The stacked bars show installed capacity of different generation technologies. The purple curve shows the proportion of the prediction error-induced carbon emissions in the entire power sector. Radii of bubbles on the curve indicate the amount of the prediction error-induced carbon emissions.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.jpg",
+        "caption": "Impacts of installed capacity, power generation and first-order difference of time series. a, wind installed capacity, b, solar installed capacity, c,wind hourly first-order difference, and d, solar daily first-order difference. The radius of each bubble indicates wind or solar generation, wind daily first-order difference, and solar hourly first-order difference, respectively. The number of bubbles is 30, representing the 30 provinces of China, excluding Tibet, Hong Kong, Macao, and Taiwan. The black linear regression line fits the center of the bubbles. The color of each bubble indicates the different categories: red\u2014category with the largest prediction error; yellow\u2014category with the second-largest prediction error; blue\u2014category with the third-largest prediction error; green\u2014category with the smallest prediction error.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.jpg",
+        "caption": "Peaks distribution and the impact on the wind and solar power prediction errors. a,Influence of the wind hourly peaks. The radius of each bubble indicates the ratio of the wind daily peaks. b, Wind hourly peak distribution in 10 power generation intervals for Beijing (BJ), Guizhou (GZ), Fujian (FJ), and Inner Mongolia (IM). c, Influence of the solar daily peaks. The radius of each bubble represents the ratio of the solar hourly peaks. d, Solar daily peak distribution in 10 power generation intervals for BJ, Jilin (JL), Chongqing (CQ), IM. In aand c, the number of bubbles is 30, representing the 30 provinces of China, excluding Tibet, Hong Kong, Macao, and Taiwan. The black line indicates the fit through the centers of the bubbles. The color of each bubble indicates the different categories: red\u2014category with the largest prediction error; yellow\u2014category with the second-largest prediction error; blue\u2014category with the third-largest prediction error; green\u2014category with the smallest prediction error.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.jpg",
+        "caption": "Provincial probability distribution function (PDF) and prediction errors in each interval. a-d, The upper figures show the PDFs of wind prediction in Beijing (BJ), Guizhou (GZ), Fujian (FJ), and Inner Mongolia (IM), and the lower figures show the wind prediction error in each interval. e-h, The upper figure shows the PDFs of solar prediction in BJ, Jilin (JL), Chongqing (CQ), and IM, and the lower figure shows the solar prediction error in each interval. The color corresponds to the prediction in each interval: pale turquoise: interval 1; cornflower blue: interval 2; dark salmon: interval 3; burlywood: interval 4; purple: interval 5; pale green: interval 6; light sky blue: interval 7; yellow: interval 8; deep sky blue: interval 9; light coral: interval 10. Each box shows the distribution of the prediction errors. The lower/upper end of each box indicates the minimal/maximal value, the lower and upper percentiles indicate 25% and 75%, respectively. The short red line indicates the median, and the bubble line indicates the average prediction error of each box.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.jpg",
+        "caption": "Temporal analysis of wind and solar prediction errors. a, Wind, c, solar prediction error in the 30 provinces in spring, summer, autumn, and winter. Each chord and arc represent the prediction error (%) between a province and the season, where the thickness is proportional to the level of prediction error. Regarding province arcs, each segment corresponds to the prediction error in each season; regarding season arcs, each segment corresponds to the prediction error in each province. The number next to the arc indicates the cumulative prediction error. b,Hourly prediction error of wind power in Beijing (BJ), Guizhou (GZ), Fujian (FJ), and Inner Mongolia (IM). d, Hourly prediction error of solar in BJ, Jilin (JL), Chongqing (CQ), and IM. The line indicates the hourly prediction error, and the bar indicates the average prediction error in the four seasons: Green\u2014spring; red\u2014summer; yellow\u2014autumn; and blue\u2014winter.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0c6e2f5cba0e7af6c6cfb672d733f15ac527ee25f7ea358f37e9e4f51d58e9db/preprint/preprint.md

@@ -0,0 +1,269 @@
+# Abstract
+
+Solar and wind resources are vital for the sustainable and cleaner transition of the energy supply. Although renewable energy potentials are assessed in the literature, few studies examine the statistical characteristics of the inherent uncertainties of renewable generation arising from natural randomness, which is inevitable in stochastic-aware research and applications. Here we develop a rule-of-thumb statistical learning model for wind and solar power prediction and generate an hourly and year-long dataset of prediction errors in 30 provinces of China. The results reveal diversified spatial and temporal distribution patterns of prediction errors, indicating that more than 70% of wind prediction errors and 50% of solar prediction errors arise from scenarios with high utilization rates. We discover that the first-order difference and peak ratio of generation series are two primary indicators explaining the distribution characteristics of prediction errors. Furthermore, the prediction errors could result in additional CO<sub>2</sub> emissions from coal-fired thermal plants. We estimate that such emission would potentially reach 319.7 megatons in 2030, accounting for 7.7% of China’s power sector. Finally, improvements in investment incentives and interprovincial scheduling could be suggested.
+
+Earth and environmental sciences/Environmental social sciences/Energy and society  
+Physical sciences/Energy science and technology/Energy modelling
+
+# Introduction
+
+To realize China’s dual carbon goals proposed in 2020<sup>1</sup>, the installed capacity of renewable energy resources should be significantly increased. As China mentioned in the 2020 Climate Ambition Summit, the installation of wind and solar energy should reach no less than 1.2 TW in 2030, almost 3 times more than that in 2019<sup>2</sup>, becoming the dominant electricity generation resource. However, due to the salient intermittency and volatility, wind and solar energy operation and modeling face the critical challenges of a high degree of uncertainty, which must be considered in energy research<sup><span additionalcitationids="CR4" citationid="CR3" class="CitationRef">3</span>–<span citationid="CR5" class="CitationRef">5</span></sup>.
+
+Various studies have investigated the generalized spatial and temporal characteristics of renewable energy resources in regional areas and compiled standardized test datasets, including statistical analysis studies of current wind and solar resources<sup><span additionalcitationids="CR7 CR8 CR9" citationid="CR6" class="CitationRef">6</span>–<span citationid="CR10" class="CitationRef">10</span></sup> and important impact factors of renewable energy generation<sup><span citationid="CR11" class="CitationRef">11</span></sup>, current wind and solar energy resource estimation studies using meteorological data and prediction methods<sup><span additionalcitationids="CR13" citationid="CR12" class="CitationRef">12</span>–<span citationid="CR14" class="CitationRef">14</span></sup>, and future wind and solar energy resource assessment studies based on wind speed and solar irradiation data<sup><span additionalcitationids="CR16 CR17 CR18" citationid="CR15" class="CitationRef">15</span>–<span citationid="CR19" class="CitationRef">19</span></sup>. However, in contrast to traditional energy resources, renewable energy resources rely on weather conditions and thus are highly unstable, posing great challenges in accurate and reliable prediction. Thus, in addition to resource assessment of renewable energy, it is important to proactively consider the risk caused by uncertainties in renewables during real-time operation of energy systems and design of energy policies<sup><span citationid="CR20" class="CitationRef">20</span>–<span citationid="CR21" class="CitationRef">21</span></sup>. However, little research focuses on analysis of the spatiotemporal uncertainty distributions of renewable energy. Research gaps exist in error-analysis benchmarks with the long-term, high-granularity, and nationwide scales of wind and solar output prediction, especially within the context of China.
+
+Error-analysis benchmarks for wind and solar output forecasting are of great value in academic research and industry. First, a prediction error database of the wind and solar output should be obtained via benchmark prediction methods, e.g., neural network-based<sup><span citationid="CR22" class="CitationRef">22</span></sup>, data mining<sup><span citationid="CR23" class="CitationRef">23</span></sup>, and regression methods<sup><span citationid="CR24" class="CitationRef">24</span></sup>. Second, a wide variety of studies, e.g., power system planning and operation<sup><span additionalcitationids="CR26 CR27" citationid="CR25" class="CitationRef">25</span>–<span citationid="CR28" class="CitationRef">28</span></sup>, energy scheduling<sup><span additionalcitationids="CR30" citationid="CR29" class="CitationRef">29</span>–<span citationid="CR31" class="CitationRef">31</span></sup>, and market operation and mechanism design studies<sup><span citationid="CR32" class="CitationRef">32</span>–<span citationid="CR33" class="CitationRef">33</span></sup>, must consider the intermittency and volatility of renewable energy resources via robust optimization<sup><span citationid="CR34" class="CitationRef">34</span>–<span citationid="CR35" class="CitationRef">35</span></sup>, stochastic programming<sup><span citationid="CR36" class="CitationRef">36</span>–<span citationid="CR37" class="CitationRef">37</span></sup>, and statistical analysis methods<sup><span citationid="CR38" class="CitationRef">38</span>–<span citationid="CR39" class="CitationRef">39</span></sup>. Third, the prediction error of renewable power determines the revenue risk of power generation companies, especially in markets with deviation punishment. In this regard, prediction error analysis can provide an important reference for decision-making of intermittent renewables.
+
+The motivation of this work is to develop a year-long error-analysis benchmark for hourly wind and solar generation forecasting in 30 provinces of China, which is expected to constitute a valuable resource and toolkit for market operators or planners. To this end, we use a one-year standard dataset including hourly wind and solar output data for 30 provinces of China<sup><span citationid="CR11" class="CitationRef">11</span></sup>. Here, we establish a rule-of-thumb prediction model to conduct hourly prediction of the wind and solar output in a rolling fashion and to obtain basic prediction datasets. The results reveal the nationwide spatial distribution of the wind and solar energy uncertainty through the prediction error. The first-order difference and peak ratio of output data are determined as primary factors of the prediction error. To further analyze provincial forecasting characteristics, we provide the provincial probability distribution function (PDF) of prediction errors and distribution regularities, the influence of power generation intervals on prediction in each province, and the temporal features of uncertainty via seasonal analysis.
+
+# Nationwide Analysis Of The Uncertainty Of Wind And Solar Generation
+
+We obtain an error-analysis benchmark for the hourly wind and solar output in 30 provinces of China in 2018 from a newly developed rule-of-thumb prediction model based on installation and hourly generation data retrieved from our previous study<sup>11</sup>. The spatial distributions of the wind and solar uncertainty across China are analyzed through the prediction error, as shown in Fig. 1a and Fig. 1b, respectively, excluding Tibet, Taiwan, Hong Kong, and Macao Provinces (unsuitable for wind energy construction<sup>10</sup> or data limitations). The prediction error is calculated as the predicted value minus the actual value (please refer to Methods). The wind prediction error ranges from 5–13.5%, with the largest error in Chongqing (CQ) and the smallest error in Inner Mongolia (IM). The overall prediction error of solar energy is smaller than that of wind energy, ranging from 3.9%~9.8%, and the largest provincial prediction error is observed in Beijing (BJ), while the smallest provincial prediction error comes from Xinjiang (XJ). Detailed error analysis of wind and solar power for each province are shown in Supplemental Fig. 1 and Fig. 2, respectively. We divide the 30 provinces into four groups according to the wind prediction error: i) > 9.5%, ii) 8%~9.5%, iii) 6.5%~8%, iv) < 6.5%. Four groups can also be distinguished in terms of solar energy according to the prediction error: i) > 8%, ii) 7%~8%, iii) 6%~7%, iv) < 6%. The details of each group are provided in the Supplemental Files.
+
+The results demonstrate that, except for Northwest China, the wind prediction error in the other regions is relatively large, especially large in the southwestern area, i.e., CQ, Sichuan (SC), Yunnan (YN), Guangxi (GX), and Guizhou (GZ), and central area including Ningxia (NX), Shaanxi (SN), Henan (HA), HB, Jiangxi (JX), and Hunan (HN), ranging from 8.3%~13.5% and 6.5%~8.9%, respectively. These two areas account for 21.8% and 20.5%, respectively, of the total prediction error in China. Regarding solar energy, the prediction error is concentrated in the areas of North China covering BJ, Tianjin (TJ), Liaoning (LN), Shandong (SD), Jilin (JL), Shanxi (SX), and Hebei (HE), Central China, and East China including Shanghai (SH), Jiangsu (JS), Anhui (AH), and Zhejiang (ZJ), ranging from 6.8%~9.8%, 6%~8.7%, and 6.8%~9.3%, respectively, accounting for 28.5%, 21.1%, and 15.4%, respectively, of the total prediction error in China.
+
+Furthermore, the prediction errors may result in additional CO<sub>2</sub> emission as generators with higher emission intensity have to be scheduled to provide spinning reserve capacity (Supplementary Fig. 1). Here we discover that the prediction erros of wind and solar power would potentially cause 54.5 megatons and 45.0 megatons of CO<sub>2</sub> in 2018, respectively, in total accounting for approximately 30.8% of France’s annual emission<sup>40</sup>. As can be seen in Fig. 1c, the provincial prediction errors of wind and solar and the provincial CO<sub>2</sub> emission caused by spinning reserves are not monotonous in space. In some provinces, such as Hubei (HB), IM, and XJ, a lower prediction error results in a higher CO<sub>2</sub> emission. This is because the CO<sub>2</sub> emission is determined by both the amount of spinning reserve requirement and the heterogeneity of thermal coal consumption rates. A more diversified coal consumption rate leads to a larger value of marginal coal consumption for providing additional reserve capacity, which will increase the prediction error-related CO<sub>2</sub> emission.
+
+Here the carbon emissions caused by prediction errors are also projected for ten years from 2021 to 2030 (Fig. 1d). Assuming the identical statistical characteristics of wind and solar power prediction errors, the additional CO<sub>2</sub> emissions would rise to 319.7 megatons, accounting for 7.7% of China’s power sector’s C0<sub>2</sub> emissions.
+
+# Key Factors Affecting Prediction Errors
+
+Two statistical indicators are proposed to explore the factors impacting prediction errors. Due to the irregular distribution of the wind output and the daily periodicity of the solar output, we use hourly output data and daily output data to analyze the wind and solar prediction errors, respectively (please refer to Methods and Supplemental Fig. 3). We use the coefficient of determination (CoD) $R^{2}$, which measures the linear correlation, to quantify the relationship between the prediction error and various factors. The installed capacity is independent of the prediction error, with $R^{2}=0.2$ for wind energy (Fig. 3a) and $R^{2}=0.005$ for solar energy (Fig. 3b). In addition, the power generation reflected by the bubble size exhibited no correlation with the prediction error (Fig. 2a and Fig. 2b).
+
+As shown in Fig. 2c and Fig. 2d, the results indicate that the first-order difference is a major influencing factor of the prediction error, which comprises a series of changes from one period to the next. The relationship between the prediction error and first-order difference is approximately linear. In regard to wind power, the relationship between the prediction error and hourly first-order difference yields $R^{2}=0.922$ (Fig. 2c), while the daily first-order difference does not impact the wind prediction error (please refer to the bubble size in Fig. 2c). Regarding solar power, the CoD between the prediction error and the daily first-order difference is $R^{2}=0.765$ (Fig. 2d). The hourly first-order difference, however, could not reflect the prediction error, as indicated by the bubble size in Fig. 2d. The reason is that wind power prediction is conducted hour-by-hour, and the daily wind power generation is irregular and cannot reflect the hourly wind generation pattern. Regarding solar power, power generation varies periodically on a daily basis, and the characteristics of the hourly first-order difference could be masked by this daily periodicity.
+
+Another significant factor influencing the prediction error is the peak ratio, which reflects the frequency of the tendency changes in the power output series, with CoD $R^{2}=0.827$ for the hourly wind output (Fig. 3a) and $R^{2}=0.648$ for the daily solar output (Fig. 3c). Similar to the first-order difference, wind and solar energy differ in their hourly and daily features. Specifically, in regard to solar energy, the hourly peak ratio of the entire output series is similar across all the provinces (the bubble size in Fig. 3c) and varies by approximately 5.3%. This also occurs because there are periodic solar energy peaks every day.
+
+To further explore the impact of different power generation levels on the prediction error, we evenly divided the installed generation capacity into 10 intervals. We also select a representative province in each wind and solar energy category for detailed analysis. The representative wind energy provinces are BJ, GZ, FJ, and IM; the representative solar energy provinces are BJ, JL, CQ, and IM. We express the peak distribution in each power generation interval as a frequency (Fig. 3b for wind energy and Fig. 3d for solar energy). Regarding wind energy, peaks in provinces with a large prediction error, e.g., BJ: 10.7%; GZ: 8.3%; and FJ: 7.9%, occur in both higher and lower power intervals, and the frequency fluctuates at 10%. However, in provinces with a small prediction error (IM: 5%), peaks are concentrated in lower power intervals from 1 ~ 4, at 94.6%. In contrast, solar energy peaks are mainly located in higher power intervals, with the peaks in intervals above 5 accounting for 70.5%, 69.6%, 64.9%, and 89.6% in BJ, JL, CQ, and IM, respectively.
+
+# Provincial And Temporal Analysis Of The Prediction Errors
+
+We examine the PDF and prediction error in each province within the above 10 power generation intervals to further analyze the spatial characteristics of the prediction error (Fig. 4). The results reveal that the more concentrated the PDF is within a certain interval, the smaller the prediction error within this interval. In terms of wind generation, the average prediction error within interval 1 in BJ is small (only 7.4%), and the PDFs within this interval are concentrated from intervals 1–4; in contrast, the prediction error within interval 10 reaches 22.7%, and the PDF within this interval is distributed across almost all intervals. The prediction error within each interval also reflects the variance and fluctuation magnitude within the interval. As shown in Fig. 4a, the average prediction error within interval 10 in BJ is larger than that within interval 1, and the fluctuation range within these two intervals is 0 ~ 74% with a variance of $5.6\times {10}^{-2}$, and 0 ~ 20.5% with a variance of $8.5\times {10}^{-3}$, respectively.
+
+As illustrated in Fig. 4, we also discover that most of the provinces with large prediction errors achieve wind and solar prediction errors in high power intervals. The proportions of intervals above 5 in BJ for wind energy, GZ for wind energy, FJ for wind energy, BJ for solar energy, JL for solar energy, and CQ for solar energy are 72.8%, 76.6%, 70.4%, 59%, 57.1%, and 58.9%, respectively. This phenomenon is more obvious for wind energy because solar power never occurs at full generation, and there is almost no solar power generation within intervals 9–10. Instead, the prediction errors in provinces with a small prediction error are distributed almost equally among all intervals, e.g., the wind prediction error within each interval in IM ranges from 7.0%~10.0%. This occurs because high power generation generally exhibits peak or inflection points, which fluctuate wildly and are difficult to predict. The proportion of peaks within each interval is provided in Supplemental Table 3. Thus, the uncertainty of power generation can be intuitively assessed based on power generation.
+
+We also analyze the seasonal characteristics of the generation uncertainty of solar and wind power on a provincial level. Here, we compare the provincial prediction error in spring, summer, autumn, and winter. Nationally, we determine that spring and summer are dominant seasons for the wind uncertainty, accounting for 55.7% of the total prediction error (Fig. 5a), and spring and winter are dominant seasons for the solar uncertainty, accounting for 57.4% of the total prediction error (Fig. 5c). The provincial characteristics are also similar, as illustrated in Fig. 5bd. The wind uncertainties in BJ and GZ in spring and summer account for 60.7% and 57.8%, respectively, of the total prediction error; the solar uncertainties in BJ, JL, and IM in spring and winter account for 57.7%, 63.9%, and 65.7%, respectively, of the total prediction error. This occurs because the solar irradiation in summer and autumn is sufficient with fewer rainy days, resulting in more stable solar power generation and relatively accurate prediction results.
+
+# Conclusions And Policy Implications
+
+We provide an error-analysis benchmark for hourly wind and solar generation in 30 provinces of China with significance for research, industry, and policy decision-making. The proposed benchmark reveals statistical characteristics of wind and solar uncertainty, which is indispensable for academic research. For example, it can help to build PDF of wind and solar generation, providing scenario basis for stochastic economic dispatch<sup>41</sup>. Energy scheduling may also use renewable generation and consider their prediction errors as a probability distribution<sup>42</sup>. Second, the benchmark is applicable for robust optimization, because the best and worst-case operating condition can be obtained through prediction results. It can also replace the assumed prediction errors to generate reasonable probability distribution and be used as expected form in optimization formulations<sup>43–44</sup>. Third, risk assessment can also benefit from the benchmark, as the security region of power systems can be depicted based on the prediction results and errors<sup>45</sup>. Without our work, most of these research use assumed renewable generation and prediction error. In industry, the benchmark plays a critical role as a guiding reference for intuitive analysis of resource distributions and fluctuations, which could help to evaluate investment revenue and the risk of renewable projects. If prediction errors are large and renewable generation is unstable, the renewable projects will take more risks and the investment should be reduced. In addition, policy-makers and system planners need information contained in the benchmark when determining development strategies for cleaner energy systems. An emergent and valuable issue entails the implementation of energy storage devices to mitigate the power balance stress in power systems with an increasing share of renewable resources<sup>46–47</sup>, and the optimal sizing and setting processes of energy storage devices rely heavily on the spatial and temporal uncertainties of renewable generation.
+
+The analysis indicates that the first-order difference and peak ratio of renewable generation are two primary influencing factors of prediction errors, both reflecting fluctuations in power generation. The wind prediction error is affected by the hourly power generation because the prediction model is employed based on the irregular hourly wind output. In contrast, the solar prediction error is affected by daily fluctuations since solar generation exhibits daily periodicity.
+
+Our results reveal the provincial distribution of the uncertainty of wind and solar generation, indicating different priorities for renewable energy in development different areas. The top 5 provinces with the largest wind prediction error are CQ, SC, TJ, BJ, and YN, with values of 13.5%, 11.4%, 10.9%, 10.7%, and 10%, respectively. In contrast, the solar prediction error in these provinces is 6.7%, 4.9%, 9.3%, 9.8%, and 4.2%, respectively, which indicates that CQ, SC, and YN should prioritize the development of solar energy due to the small prediction errors and fluctuations. BJ and TJ are commercial provinces with small areas and are not suitable for wind and solar energy development. IM, GD, HI, GS, and QH should develop wind energy due to their smallest prediction errors of 5%. 5.2%, 5.3%, 5.5%, and 6%, respectively. BJ, SH, TJ, NX, and LN are the top 5 provinces with larger solar prediction errors, namely, 9.8%, 9.3%, 9.3%, 8.7%, and 7.7%, respectively, while the wind prediction errors in SH and LN reach 7.8% and 6.7%, respectively, and the corresponding potential wind capacity factor is approximately 30<sup>10</sup>. Therefore, wind energy development in these two provinces is a recommended pathway to reduce the adverse impact of renewable generation on power system operation.
+
+Our temporal analyses demonstrate that renewable generation in spring exerts the greatest impact on the power system, requiring proactive deployment of flexible resources. Combined with the spatial distribution, the solar prediction error in Northeast and North China in winter exhibits a large prediction error, ranging from 8.4–11.9%, with an average value of 9.6%, larger than the total prediction error of 5.9%~9.8%, with an average value of 7.7%. As the Chinese government has issued the Electric Heating Policy to provide heat in Northeast and North China in winter, the load demands in the power sector have increased significantly<sup>48</sup>. The flexibility-adjustable resources and volatility on the power source side exhibit inverse distributions, which has become a central problem in the consumption of renewable energy in these regions. In contrast, Southeast China achieves the smallest prediction error in regard to both wind and solar energy in winter, with average values of 6.8% and 4.8%, respectively. Additionally, existing research has suggested abundant offshore wind power resources in the area, with wind capacity factors higher than 50%, almost ranking at the top in China<sup>10–11</sup>. Due to the obvious seasonal distribution of offshore wind power, which dominates in spring and winter<sup>49</sup>, wind power represents a suitable alternative resource to offset the winter load peak in North and Northeast China.
+
+In Southwest China, the wind prediction error in summer ranges from 9.8%~16.2%, with an average value of 12.1%, which is larger than the total prediction error, with a range of 8.3%~13.5% and an average value of 10.4%. Fortunately, Southwest China contains the most abundant water resources in China, accounting for 2/3 of the national reserves, and the technically exploitable capacity reaches 425 GW, accounting for 71% of the national capacity<sup>50</sup>. Due to the characteristics of the monsoon climate in China, the flood season in the southwest occurs in summer, from April to August, which could provide abundant hydropower and could offset the adverse impact of wind fluctuations on the power system.
+
+Here, we summarize two policy suggestions for China. First, the government should provide adequate policy support or incentives to encourage wind energy development in the southeastern and northwestern areas of China and solar energy development in the areas of Northwest, Southwest, and Southeast China. According to our analysis, the wind and solar fluctuations in these areas are limited, reducing the adverse impact on the power system. Second, the government should plan interprovincial energy transmission in the space dimension to reduce the winter load peak in North China and increase renewable energy consumption. As concluded, the wind and solar fluctuations in North China are notable, accounting for 28.5% and 21.3%, respectively, of the total prediction error in China. In addition, winter exhibits the highest solar energy fluctuations with a proportion of 27.7% of the total prediction error. As such, the government should improve the power system infrastructure, systematically evaluate potential network congestion issues, and plan additional power transmission lines according to the resource uncertainty and load distribution.
+
+# Methods
+
+## Wind and solar output data
+
+Hourly wind and solar output data for 2018 pertaining to 30 provinces of China are retrieved from previous work<sup>11</sup>, except Tibet, Taiwan, Hong Kong, and Macao. The dataset contains 8760 hours of wind and solar output data, and wind and solar installed capacity data for these 30 provinces are included. We denote the hourly wind output as $W_{i,t+1,0}$ and the hourly solar output as $S_{i,t+1,0}$, where $i$ and $t$ are province and time slot indices, respectively, for $i\in [1,N],t\in [1,T]$, $N=30$, and $T=8760$. As previously mentioned, daily wind and solar output data are also required for the analysis, which can be calculated as follows:
+
+$$
+{W}_{\text{Day},i,c,0}={max}({W}_{i,t,0},{W}_{i,t+1,0},\cdots {W}_{i,t+23,0}),t=24\cdot (c-1)
+$$
+
+$$
+{S}_{\text{Day},i,c,0}={max}({S}_{i,t,0},{S}_{i,t+1,0},\cdots {S}_{i,t+23,0}),t=24\cdot (c-1)
+$$
+
+where ${S}_{\text{Day},i,c,0}$ and ${W}_{\text{Day},i,c,0}$ are the daily solar and wind output, respectively, of province $i$ in time slot $t$, and $c$ is a day index, for $c\in \left[1, C\right] \text{and} C=365$.
+
+## Autoregressive Integrated Moving Average (Arima)-based Benchmark Prediction Model
+
+Time series prediction is based on historical data, among which the autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) techniques are typical methods to study stationary time series and are suitable for a large number of problems. However, the fluctuations in wind and solar energy indicate that their power generation involves a nonstationary time series with a time-varying mean value and variance, which is difficult to study with these methods. Thus, to predict nonstationary sequences, the ARIMA prediction model is introduced by Box-Jerkins. Considering a certain number of differences in the ARIMA prediction model, wind and solar power generation series can be converted into a stationary series, convenient for prediction analysis. In the literature, the ARIMA model is widely used in short-term renewable forecasting and is validated to yield satisfactory results.
+
+In prediction model construction, it is necessary to first determine whether the series is stationary. If the series is not stationary, it should be differentiated until the series meets the stationarity requirements. Suppose the real wind and solar power generation series are ${Y}_{t}$, the differential order can be denoted by $d$, and the differential process can be expressed as follows:
+
+$${X}_{t}=(1-B{)}^{d}{Y}_{t}, \text{ADFtest}({X}_{t})=1$$
+
+where ${X}_{t}$ is the stationary series of the original real data, $B$ is the lag operator, and $ADFtest=1$ passes the stationarity test. Except for the differential order $d$, the ARIMA model should also determine the autoregressive order $p$ and moving average order $q$, and the autoregressive moving average model for ${X}_{t}$ can be expressed as follows:
+
+$$\left(1-{\sum }_{i=1}^{p}{\phi }_{i}{B}^{i}\right){X}_{t}={\mu }_{0}+(1-{\sum }_{i=1}^{q}{\mu }_{i}{B}^{i}){\alpha }_{t}$$
+
+where ${\phi }_{i}$ and ${\mu }_{i}$ are the autoregressive parameter and moving average parameter, respectively, ${\alpha }_{t}$ is white noise with a mean of 0, ${\mu }_{0}$ is a deterministic trend quantity greater than 0, and ${B}^{i}$ is the $i$th power of $B$. Via the use of the prediction model, we can obtain the predicted series ${X}_{predict,t}$, which is a differential series of the predicted wind and solar power generation. Thus, the predicted power generation can be obtained through the following calculation equation:
+
+$${Y}_{predict,t}=(1-B{)}^{-d}{X}_{predict,t}$$
+
+where ${Y}_{predict,t}$ denotes the predicted results of the ARIMA-based prediction model, and in this paper, this variable indicates the wind and solar output.
+
+There are three major parameters of the ARIMA-based prediction model: differential order $d$, autoregressive order $p$, and moving average order $q$. Parameter $d$ is determined based on the minimum number of differences required to obtain a stationary time series. The $d$ value is generally smaller than 3 because the greater the difference order, the more information would be lost<sup>51</sup>. It should be noted that parameter $d$ is completely determined by the properties of the original sequence, while the selection of $p$ and $q$ should consider the overall prediction effect. In general, $p$ and $q$ should remain within 1/5 of the length of the input data. Due to the large amount of wind and solar power generation data in each province in one year, usually 8760 hours, we separate multiple prediction windows for each province and used the moving window method to predict wind and solar power generation. At present, the methods for $p$ and $q$ determination usually include the Akaike information criterion (AIC) and Bayesian information criterion (BIC), but the optimal parameter configuration can only be provided for a single prediction window. To unify the prediction models with the different prediction windows in the same provinces and minimize the prediction error, we randomly select 5 weeks of data throughout the year as a sample and traverse $p$ and $q$ for each province to obtain the best parameters with the minimum prediction error. The detailed parameters for each province are listed in Supplemental Table 2.
+
+Other parameters, such as the autoregressive parameter ${\phi }_{i}$ and moving average parameter ${\mu }_{i}$, can vary with the input data. These two parameters are determined by the autocorrelation coefficient and autocovariance, respectively, which can be obtained with the Yule–Walker estimation, least squares estimation or maximum likelihood estimation method<sup>52</sup>. In this paper, we build the ARIMA-based prediction model, and all the parameters except $p$, $d$, and $q$ could be automatically generated.
+
+In this paper, we set 6 hours as the prediction time scale and 168 hours as the input data dimension to predict wind and solar power generation. The reason is that power generation is independent of time, i.e., previous hours slightly affect the next hour, and the prediction accuracy can decrease with increasing prediction time scale.
+
+## Prediction Error Calculation
+
+In this paper, the prediction error of wind and solar energy could be calculated as the unit MW prediction error. When using the ARIMA-based benchmark prediction model, we could obtain the predicted wind and solar energy generation, and the prediction error can then calculated as follows:
+
+$$
+{\epsilon }_{i,t}=\frac{{W}_{i,t,*}-{W}_{i,t,0}}{{C}_{\text{W},i}}\cdot 100\%
+$$
+$$
+{\epsilon }_{i,t}=\frac{{S}_{i,t,*}-{S}_{i,t,0}}{{C}_{\text{S},i}}\cdot 100\%
+$$
+
+where ${\epsilon }_{i,t}$ is the prediction error in province $i$ in time slot $t$, ${W}_{i,t,*}$ and ${S}_{i,t,*}$ are the predicted wind and solar output, respectively, of province $i$ in time slot $t$, and ${C}_{\text{W},i}$ and ${C}_{\text{S},i}$ are the wind and solar installed capacities, respectively, in province $i$. When determining the prediction error in a given province, we calculate the average value over 8760 hours.
+
+## Estimation Of Co Emission Caused By Prediction Errors
+
+Additional operating reserve capacity has to be scheduled to accommodate the prediction errors of wind and solar generation, which will push up the CO$_2$ emission from thermal generation units. In this paper, the additional CO$_2$ emission caused by prediction errors of wind and solar generation is estimated as follows:
+
+$${E}_{i,t}={\epsilon }_{i,t}{C}_{\text{W},i}\left({h}_{i}^{max}-{h}_{i}^{avr}\right){e}_{i},$$
+$${E}_{i,t}={\epsilon }_{i,t}{C}_{\text{S},i}\left({h}_{i}^{max}-{h}_{i}^{avr}\right){e}_{i},$$
+
+where ${E}_{i,t}$ is the prediction error-related CO$_2$ emission in province $i$ in time slot $t$. Coefficients ${h}_{i}^{max}$ and ${h}_{i}^{avr}$ denote the maximum and average thermal coal consumption rate for electricity generation in province $i$, respectively. Coefficient ${e}_{i}$ is the average CO$_2$ emission rate of thermal coal, which is set as 2.83 in this paper.
+
+## First-order Difference
+
+The first-order difference can be used to assess the variation in discrete time-series data. With the use of the first-order difference, we can obtain the increment in the original data, which can reflect gradient information. In this paper, prediction is conducted hour-by-hour, and the prediction accuracy is primarily determined by the hourly change in the generation data. Thus, in terms of wind energy, we use the first-order difference of hourly wind generation data to measure the hourly change, which can be calculated as follows:
+
+$${F}_{\text{H},i,t}=\frac{{W}_{i,t+1,0}-{W}_{i,t,0}}{{C}_{\text{W},i}}$$
+
+where ${F}_{\text{H},i,t}$ is the hourly first-order difference in province $i$ in time slot $t$ and ${W}_{i,t+1,0}$ and ${W}_{i,t,0}$ are the real wind energy generation in time slots $t+1$ and $t$, respectively. When evaluating the hourly first-order difference in a province, we calculate the average value over 8760 hours.
+
+Regarding solar energy, power generation exhibits daily periodicity, so we use daily solar energy generation data to measure the fluctuation, which can be expressed as follows:
+
+$${F}_{\text{Day},i,c}=\frac{{S}_{\text{Day},i,c+1,0}-{S}_{\text{Day},i,c,0}}{{C}_{\text{S},i}}$$
+
+where ${F}_{\text{Day},i,c}$ is the daily first-order difference in province $i$ on day $c$. We also calculate the average value over 365 days to evaluate the solar energy fluctuations in a given province.
+
+## Analysis And Calculation Of The Peak Ratio
+
+In this paper, we use the peak ratio to evaluate the prediction error. It should be noted that all the prediction methods learn the variation tendency of a given data series to predict future data. The easier a tendency is to learn, the more accurate the prediction. Thus, we aim to obtain a feature that could indicate the change in tendency to better measure the prediction error. The peaks of series data indicate inflection points, with previous data exhibiting an upward tendency and subsequent data exhibiting a downward tendency, which is a key feature reflecting the tendency change.
+
+In regard to wind energy, we use 4 consecutive time slots to determine hourly peaks and traverse the time series to find all peaks, i.e., $t=t+1$. The power generation in these four time slots should satisfy the following conditions to reach a peak: the first three hours should continuously increase, the first three hours should increase by more than 10% of the installed capacity, and the fourth hour should decrease, which can be expressed as follows:
+
+$${P}_{\text{H},i,t}=1, \text{ for }{W}_{i,t,0}-{W}_{i,t-1,0}<0,{W}_{i,t-1,0}-{W}_{i,t-2,0}\ge 0,\\ {W}_{i,t-2,0}-{W}_{i,t-3,0}\ge 0,{W}_{i,t-1,0}-{W}_{i,t-3,0}\ge 0.1\cdot {C}_{\text{W},i}$$
+
+$$N{P}_{\text{H,}i}={\sum }_{t\in T}{P}_{\text{H,}i,t}$$
+$$R{P}_{\text{H},i}=N{P}_{\text{H},i}/T$$
+
+where ${P}_{\text{H},i,t}$ denotes the hourly peaks in province $i$ in time slot $t$, $N{P}_{\text{H},i}$ is the number of hourly peaks in province $i$, and $R{P}_{\text{H},i}$ is the ratio of hourly peaks in province $i$. We also calculate the average value over 8760 hours to evaluate the wind energy fluctuations in each province.
+
+Regarding solar energy, we use daily power generation data to obtain daily peaks. Similar to the hourly peak calculation, four consecutive days are chosen to determine peaks, and similar conditions should be satisfied, which can be expressed as follows:
+
+$${P}_{\text{Day},i,c}=1, \text{ for }{S}_{\text{Day},i,c,0}-{S}_{\text{Day},i,c-1,0}<0,{S}_{\text{Day},i,c-1,0}-{S}_{\text{Day},i,c-2,0}\ge 0,\\ {S}_{\text{Day},i,c-2,0}-{S}_{\text{Day},i,c-3,0}\ge 0,{S}_{\text{Day},i,c-1,0}-{S}_{\text{Day},i,c-3,0}\ge 0.1\cdot {C}_{\text{S},i}$$
+
+$$N{P}_{\text{Day,}i}={\sum }_{c\in C}{P}_{\text{Day,}i,c}$$
+$$R{P}_{\text{Day,}i}=N{P}_{\text{Day,}i}/C$$
+
+where ${P}_{\text{Day},i,c}$ is the daily peak in province $i$ on day $c$, $N{P}_{\text{Day,}i}$ is the number of daily peaks in province $i$, and $R{P}_{\text{Day,}i}$ is the ratio of daily peaks in province $i$. The average value over 365 days is also calculated to express the solar energy fluctuations in each province.
+
+# References
+
+1. China Xinhua News. Xi's statement at the General Debate of the 75th Session of the United Nations General Assembly. [http://www.qstheory.cn/yaowen/2020-09/22/c_1126527766.htm] (2020)
+
+2. Climate Ambition Summit. Leaders statements of president Xi Jinping. [http://www.gov.cn/xinwen/2020-12/13/content_5569138.htm] (2020)
+
+3. PrakashKumar K. & Saravanan B. Recent techniques to model uncertainties in power generation from renewable energy sources and loads in microgrids–A review. Renewable and Sustainable Energy Reviews. 71, 348–358 (2017).
+
+4. Mahdi S., Helena L. & Nilay S. Integrated renewable electricity generation considering uncertainties: The UK roadmap to 50% power generation from wind and solar energies. Renewable and Sustainable Energy Reviews. 72, 385–398 (2017).
+
+5. Salvador P., Juan M., Trine B. Impact of forecast errors on expansion planning of power systems with a renewables target. European Journal of Operational Research. 248, 1113–1122 (2016).
+
+6. Ayik A., Ijumba N., Kabiri C. & Goffin P. Preliminary wind resource assessment in South Sudan using reanalysis data and statistical methods. Renewable and Sustainable Energy Reviews. 138, 110621 (2021).
+
+7. Kies A., Schyska B., Bilousova M., Sayed O., Jurasz J. & Stoecker H. Critical review of renewable generation datasets and their implications for European power system models. Renewable and Sustainable Energy Reviews. 152, 111614 (2021).
+
+8. Rourke F., Boyle F. & Reynolds A. Ireland’s tidal energy resource; An assessment of a site in the Bulls Mouth and the Shannon Estuary using measured data. Energy Conversion and Management. 87, 726–734 (2014).
+
+9. Han J., Mol A., Lu Y. & Zhang L. Onshore wind power development in China: Challenges behind a successful story. Energy Policy. 37, 2941–2951 (2009).
+
+10. Davidson M., Zhang D., Xiong W., Zhang X. & Karplus V. Modelling the potential for wind energy integration on China’s coal-heavy electricity grid. Nature Energy. 1, 16086 (2016).
+
+11. Lu X., McElroy M., Peng W., Liu S., Nielsen C. & Wang H. Challenges faced by China compared with the US in developing wind power. Nature Energy. 1, 16061 (2016).
+
+12. Gadad S. & Deka P. Offshore wind power resource assessment using Oceansat-2 scatterometer data at a regional scale. Applied Energy. 176, 157–170 (2016).
+
+13. Churio O., Marley S., Chamorro V. & Ochoa G. Wind and solar resource assessment and prediction using Artificial Neural Network and semi-empirical model: case study of the Colombian Caribbean region. Heliyon. 7, e07959 (2021).
+
+14. Pereira S., Abreu E., Iakunin M., Cavaco Af., Salgado R. & Canhoto P. Method for solar resource assessment using numerical weather prediction and artificial neural network models based on typical meteorological data: Application to the south of Portugal. Solar Energy. 236, 225–238 (2022).
+
+15. Weekes S., Tomlin S., Vosper S., Skea A., Gallani M. & Standen J. Long-term wind resource assessment for small and medium-scale turbines using operational forecast data and measure–correlate–predict. Renewable Energy. 81, 760–769 (2015).
+
+16. Joshi S, Mittal S, Holloway P, et al. High resolution global spatiotemporal assessment of rooftop solar photovoltaics potential for renewable electricity generation. Nature communications. 12, 1–15 (2021).
+
+17. Abreu E., Canhoto P., Prior V. & Melicio R. Solar resource assessment through long-term statistical analysis and typical data generation with different time resolutions using GHI measurements. Renewable Energy. 127, 398–411 (2018).
+
+18. Tahir Z. & Asim M. Surface measured solar radiation data and solar energy resource assessment of Pakistan: A review. Renewable and Sustainable Energy Reviews. 81, 2839–2861 (2018).
+
+19. Sweerts B., Pfenninger S., Yang S., Folini D., Zwaan B. & Wild M. Estimation of losses in solar energy production from air pollution in China since 1960 using surface radiation data. Nature Energy. 4, 657–663 (2019).
+
+20. Yin J, Molini A, Porporato A. Impacts of solar intermittency on future photovoltaic reliability. Nature communications. 11, 1–9 (2020).
+
+21. Anadón D. L., Baker E. & Bosetti V. Integrating uncertainty into public energy research and development decisions. Nature Energy. 2, 17071 (2017).
+
+22. Qazi A., Fayaz H., Wadi A., Raj R., Rahim N. & Khan W. The artificial neural network for solar radiation prediction and designing solar systems: a systematic literature review. Journal of Cleaner Production. 104, 1–12 (2015).
+
+23. Colak I., Sagiroglu S. & Yesilbudak M. Data mining and wind power prediction: A literature review. Renewable Energy. 46, 241–247 (2012).
+
+24. Reikard G. Predicting solar radiation at high resolutions: A comparison of time series forecasts. Solar Energy. 83, 342–349 (2009).
+
+25. Lu X., McElroy M. & Kiviluoma J. Global potential for wind-generated electricity. Proc. Natl Acad. Sci. USA. 106, 10933–10938 (2009).
+
+26. Zhang S, Chen W. Assessing the energy transition in China towards carbon neutrality with a probabilistic framework. Nature communications. 13, 1–15 (2022).
+
+27. Schyska B U, Kies A, Schlott M, et al. The sensitivity of power system expansion models. Joule. 5, 2606–2624 (2021).
+
+28. Jeon S. & Choi D. Joint optimization of Volt/VAR control and mobile energy storage system scheduling in active power distribution networks under PV prediction uncertainty. Applied Energy. 310, 118488 (2022).
+
+29. Olauson J. et al. Net load variability in Nordic countries with a highly or fully renewable power system. Nature Energy. 1, 16175 (2016).
+
+30. Morstyn T., Farrell N., Darby S. & McCulloch M. Using peer-to-peer energy-trading platforms to incentivize prosumers to form federated power plants. Nature Energy. 3, 94–101 (2018).
+
+31. Zhou D., Al-Durra A., Zhang K., Ravey A. & Gao F. A Robust Prognostic Indicator for Renewable Energy Technologies: A Novel Error Correction Grey Prediction Model. IEEE Transactions on Industrial Electronics. 66, 9312–9325 (2019).
+
+32. Wu H., Shahidehpour M., Alabdulwahab A. & Abusorrah A. A Game Theoretic Approach to Risk-Based Optimal Bidding Strategies for Electric Vehicle Aggregators in Electricity Markets With Variable Wind Energy Resources. IEEE Transactions on Sustainable Energy. 7, 374–385 (2016).
+
+33. David M., Boland J., Cirocco L., Lauret P. & Voyant C. Value of deterministic day-ahead forecasts of PV generation in PV + Storage operation for the Australian electricity market. Solar Energy. 224, 672–684 (2021).
+
+34. Zhang Y., Gatsis N. & Giannakis G. Robust Energy Management for Microgrids With High-Penetration Renewables. IEEE Transactions on Sustainable Energy. 4, 944–953 (2013).
+
+35. Hosseini S., Carli R. & Dotoli M. Robust Optimal Energy Management of a Residential Microgrid Under Uncertainties on Demand and Renewable Power Generation. IEEE Transactions on Automation Science and Engineering. 18, 618–637 (2021).
+
+36. Liu N., Cheng M., Yu X., Zhong J. & Lei J. Energy-Sharing Provider for PV Prosumer Clusters: A Hybrid Approach Using Stochastic Programming and Stackelberg Game. IEEE Transactions on Industrial Electronics. 65, 6740–6750 (2018).
+
+37. Lu R., Ding T., Qin B., Ma J., Fang X. & Dong Z. Multi-Stage Stochastic Programming to Joint Economic Dispatch for Energy and Reserve With Uncertain Renewable Energy. IEEE Transactions on Sustainable Energy. 11, 1140–1151 (2020).
+
+38. G. E. Constante-Flores and M. S. Illindala, "Data-Driven Probabilistic Power Flow Analysis for a Distribution System With Renewable Energy Sources Using Monte Carlo Simulation," in IEEE Transactions on Industry Applications. 55, 174–181 (2019).
+
+39. M. Fan et al. Uncertainty Evaluation Algorithm in Power System Dynamic Analysis With Correlated Renewable Energy Sources. IEEE Transactions on Power Systems. 36, 5602–5611 (2021).
+
+40. Our World in Data. CO₂ Country Profile. [https://ourworldindata.org/co2/country/united-states?country=~FRA].
+
+41. Wu H., Shahidehpour M., Alabdulwahab A. & Abusorrah A. Demand Response Exchange in the Stochastic Day-Ahead Scheduling With Variable Renewable Generation. IEEE Transactions on Sustainable Energy. 6, 516–525 (2015).
+
+42. Papavasiliou A., Oren S. S. & O'Neill R. P. Reserve Requirements for Wind Power Integration: A Scenario-Based Stochastic Programming Framework. IEEE Transactions on Power Systems. 26, 2197–2206 (2011).
+
+43. Valencia F., Collado J., Sáez D. & Marín L. G. Robust Energy Management System for a Microgrid Based on a Fuzzy Prediction Interval Model. IEEE Transactions on Smart Grid. 7, 1486–1494 (2016).
+
+44. Bouffard F. & Galiana F. D. Stochastic Security for Operations Planning With Significant Wind Power Generation. IEEE Transactions on Power Systems. 23, 306–316 (2008).
+
+45. Lara J. D., Dowson O., Doubleday K., Hodge B.-M. & Callaway D. S. A Multi-Stage Stochastic Risk Assessment With Markovian Representation of Renewable Power. IEEE Transactions on Sustainable Energy. 13, 414–426 (2022).
+
+46. Ziegler M S, Mueller J M, Pereira G D, et al. Storage requirements and costs of shaping renewable energy toward grid decarbonization. Joule. 3, 2134–2153 (2019).
+
+47. Hunt J D, Byers E, Wada Y, et al. Global resource potential of seasonal pumped hydropower storage for energy and water storage. Nature communications. 11, 1–8 (2020).
+
+48. Wang J. et al. Exploring the trade-offs between electric heating policy and carbon mitigation in China. Nature Communications. 11, 6054 (2020).
+
+49. Ren L., Ji J., Lu Z. & Wang K. Spatiotemporal characteristics and abrupt changes of wind speeds in the Guangdong–Hong Kong–Macau Greater Bay Area. Energy Reports. 8, 3465–3482 (2022).
+
+50. China Energy News. The most authoritative and comprehensive interpretation of southwest hydropower development. [http://www.hydropower.org.cn/showNewsDetail.asp?nsId=23181] (2018).
+
+51. Amini M. H., Kargarian A. & Karabasoglu O. ARIMA-based decoupled time series forecasting of electric vehicle charging demand for stochastic power system operation. Electric Power Systems Research. 140, 378–390 (2016).
+
+52. Wei W. S. W. Time Series Analysis: Univariate and Multivariate Methods. CA: Addison-Wesley (USA) (1990).
+
+# Supplementary Files
+
+- [DataSource.xlsx](https://assets-eu.researchsquare.com/files/rs-2284531/v1/52cb7d3c6090368146b4f20c.xlsx)  
+  Dataset
+
+- [SupplementaryInformation.pdf](https://assets-eu.researchsquare.com/files/rs-2284531/v1/b159515b8d70a9182d08cb15.pdf)

0e32d155f6317a62a891efa17fc4b06fe2849be5681a1bd4489fb58b15d0c412/preprint/images_list.json

@@ -0,0 +1,50 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "The schematic diagrams of synthesis and design of RuCu DAs/NGA. a A schematic illustration of the preparation strategy by the pulsed discharge. b A schematic diagram of metal atoms loaded on the porous NGA by the pulsed discharge. c The micropore distribution of NGA and GAs (graphene aerogels), inset is a schematic diagram of N-doped graphene with different micropore types. d The total current curve in the circuit during discharge. e A schematic plot showing the brief formation process of RuCu DAs/NGA changes with energy input. f A research tendency of single atom catalysts to increase the variety of metal elements on the support and form asymmetric coordination structures.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "The characterizations of RuCu DAs/GA. a SEM image of RuCu DAs/NGA. b TEM image of RuCu DAs/NGA. c HADDF-STEM image of RuCu DAs/NGA. d EDS mapping images of RuCu DAs/NGA, C (red), N (blue), Cu (yellow), and Ru (pink). e High magnificationHAADF-STEM image (dark field). fThe local magnified image of RuCu DAs/NGA. g The corresponding 3D intensity profiles for f. h The distance frequency between adjacent Cu and Ru atoms, and the frequency of single sites and Ru-Cu dual sites. i A schematic plot of Ru-Cu pair atoms with different types anchored on the N-doped graphene.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Atomic coordination structure and chemical state of RuCu DAs/NGA. a The Cu K-edge XANES spectra of RuCu DAs/NGA and the references (Cu foil, Cu SAs/NGA and CuO). b Cu K-edge FT k3-weighted EXAFS spectra of RuCu DAs/NGA and references. c The Cu K-edge EXAFS fitting result of RuCu DAs/NGA in the R space. d The Ru K-edge XANES spectra of RuCu DAs/NGA and the references (Ru foil, Ru SAs/NGA, and RuO2). e Ru K-edge FT k3-weighted EXAFS spectra of RuCu DAs/NGA and references. f The Ru K-edge EXAFS fitting result of RuCu DAs/NGA in the R space. g The proposed atomic structure model of RuCu DAs/NGA. h The valence states of Cu and Ru in RuCu DAs/NGA and references based on the first-order derivative of XANES spectra. i The WT-EXAFS results of Cu in RuCu DAs/NGA and references. j The WT-EXAFS results of Ru in RuCu DAs/NGA and references.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "The structural characterizations of expanded MCu DAs/NGA. a TEM image of PtCu DAs/NGA. b EDS mapping images of PtCu DAs/NGA, C (red), N (green), Pt (yellow), and Cu (cyan). c High magnification HAADF-STEM image (dark field) of PtCu DAs/NGA. d The Pt L3-edge XANES spectra of PtCu DAs/NGA and the references (Pt foil and PtO2). e Pt L3-edge FT k3-weighted EXAFS spectra with the fitting in R space of PtCu DAs/NGA and references. f TEM image of Ag-Cu DAs/NGA. g EDS mapping images of AgCu DAs/NGA, C (red), N (green), Ag (cyan), and Cu (yellow). h High magnification HAADF-STEM image (dark field) of AgCu DAs/NGA. i The Ag K-edge XANES spectra of AgCu DAs/NGA and the references (Ag foil and Ag2O). j Ag K-edge FT k3-weighted EXAFS spectra with the fitting in R space of AgCu DAs/NGA and references. k TEM image of PdCu DAs/NGA. l EDS mapping images of PdCu DAs/NGA, C (blue), N (red), Pd (pink) and Cu (yellow). m High magnification HAADF-STEM image (dark field) of PdCu DAs/NGA. n The Pd K-edge XANES spectra of PdCu DAs/NGA and the references (Pd foil and PdO). o Pd K-edge FT k3-weighted EXAFS spectra with the fitting in R space of PdCu DAs/NGA and references. p The oxidation states of metals in PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA respectively. q The frequency of single sites and Cu-M (Pt, Ag, and Pd) dual sites. The optimized structures and differential charge densities of r PtCu DAs/NGA, s AgCu DAs/NGA, and t PdCu DAs/NGA.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "NO3RR performance and in-situ XAFS results of RuCu DAs/NGA. a LSV curves of RuCu DAs/NGA, Ru SAs/NGA, and Cu SAs/NGA measured in 0.1 M KNO3 and 0.1 M KOH electrolyte. b The FEs of NH3 production at varied potentials. c Partial NH3 current densities of three catalysts at different potentials. d Potential-dependent yield rate of NH3 generation over RuCu DAs/NGA. e Cycling tests of RuCu DAs/NGA for NO3RR at -0.4 V vs. RHE. f Comparison of NH3 yield rate and potentials for RuCu DAs/NGA and various catalysts recently reported (See the details in Supplementary Table 4) g Potential-dependent in situ ATR-SEIRAS spectra (1000-2000 cm-1) during the electrochemical NO3RR process. h Cu and i Ru K-edge XANES spectra of RuCu DAs/NGA at different applied potentials (OCP, -0.1 V, -0.2 V, -0.3 V, -0.4 V, and -0.5 V) during the NO3RR process. j The valence states of Cu and Ru in RuCu DAs/NGA and references based on the first-order derivative of XANES spectra. k Cu and l Ru K-edge k3-weighted FT-EXAFS at different potentials during the NO3RR process.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Theoretical NO3RR activity on RuCu DAs/NGA. a The differential charge densities of RuN4/C, CuN4/C, and RuN2CuN3/C. b The projected density of states of RuN2CuN3/C. c Electrochemical NO3RR pathways and relative free energy on RuN4/C, CuN4/C, and RuN2CuN3/C at U=0 vs. RHE. d The energy barriers of the key steps on RuN4/C, CuN4/C, and RuN2CuN3/C.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0e32d155f6317a62a891efa17fc4b06fe2849be5681a1bd4489fb58b15d0c412/preprint/preprint.md

@@ -0,0 +1,126 @@
+# Abstract
+
+Atomically dispersed dual-atom catalysts (DACs) with asymmetric coordination are pivotal for enhancing economic efficiency and sustainable development in the electrochemical nitrate reduction reaction (NO₃ RR) to produce ammonia. However, rational design and rapid synthesis of DACs remain challenging. Here, we demonstrate the pulsed discharge method, which generates microsecond current pulses to inject substantial energy instantaneously into ruthenium (Ru) and copper (Cu) metal salt precursors supported by nitrogen-doped graphene aerogels (NGA). This process results in the atomically dispersed Ru and Cu dual atoms anchoring onto nanopore defects of NGA (RuCu DAs/NGA) through explosive decomposition of the metal salt nanocrystals. X-ray absorption spectroscopy analysis suggests an asymmetric RuN₂-CuN₃ coordination structure on NGA. The RuCu DAs/NGA catalyst exhibits outstanding electrochemical performance in NO₃ RR, achieving a Faraday efficiency of 97.8% and an ammonia yield rate of 3.07 mg h⁻¹ cm⁻² at -0.4 V vs. RHE. *In situ* studies monitor the evolution of RuCu active sites and reaction intermediates during the NO₃ RR process in real time. Density functional theory calculations reveal that the Ru-Cu sites in the asymmetric RuN₂CuN₃/C structure create a synergistic effect, optimizing intermediate adsorption and lowering the energy barrier of key elementary reactions. This pulsed discharge method is simple, ultra-fast, and versatile (e.g., applicable to PtCu, AgCu, and PdCu DAs on NGA), offering a general-purpose strategy for the precise preparation of atomically dispersed dual-atom catalysts, which are traditionally challenging to synthesize.
+
+Physical sciences/Materials science/Nanoscale materials/Synthesis and processing  
+Physical sciences/Nanoscience and technology/Nanoscale materials/Graphene/Synthesis of graphene
+
+# Introduction
+
+The escalating concentration of nitrate (NO₃⁻) in surface and groundwater has increasingly raised environmental and ecological concerns¹⁻³. Addressing NO₃⁻ contamination by converting it into high value-added products, such as ammonia (NH₃), presents a promising solution⁴,⁵. Ammonia is a widely utilized industrial chemical essential for human development, serving as a key component in fertilizers and various industrial processes. Traditionally, NH₃ synthesis has relied primarily on the Haber-Bosch process, which operates under high temperatures and pressures, leading to substantial energy consumption and significant greenhouse gas emissions. In contrast, producing NH₃ via the electrochemical reduction of NO₃⁻, powered by renewable electricity, offers significant potential for economic efficiency and environmental sustainability⁶⁻⁸. Despite these advantages, the process from NO₃⁻ to NH₃ involves multiple complex intermediate conversions and an eight-electron transfer process, posing considerable challenges to the activity and high Faradaic efficiency (FE) of newly developed electrocatalysts. Overcoming these hurdles is crucial for advancing this promising method of NH₃ production and realizing its full potential for both economic and environmental benefits.
+
+Atomically dispersed catalysts have the highest atomic utilization and superior performance over conventional metal nanoparticles⁹⁻¹³. In recent years, some advanced high performance atomically dispersed catalysts have been designed and prepared to produce NH₃ by electrochemically catalyzing NO₃⁻ reduction reaction (NO₃⁻ RR)¹⁴⁻¹⁷. However, the challenges of the activity and selectivity of single metal sites in the electrocatalysis of NO₃⁻ RR remain as well. For example, the atomically dispersed copper (Cu) sites have a high activity and FE_NH3 for NO₃⁻ RR catalysis in alkaline media, but a higher overpotential is required as usual¹⁸,¹⁹. Another problem is the nitrite (NO₂⁻) accumulation at the Cu sites during NO₃⁻ RR, a quasi-stable intermediate carcinogen, ultimately resulting in lower FE for NH₃ production²⁰,²¹. Therefore, some strategies such as modification, doping, and alteration of coordination structure targeting Cu sites have been proposed to enhance the performance of catalysts in the NO₃⁻ RR process²²⁻²⁴. Recent studies demonstrated that ruthenium (Ru) also presents remarkable activity in the electrocatalytic production of NH₃ during the NO₃⁻ RR process²⁵⁻²⁷. More importantly, Ru catalysts can achieve efficient conversion of NO₂⁻ to NH₃ by enhancing the adsorption of intermediates²⁸. Further, it reported that Cu combined with other metal atoms can modulate the local electronic structure and improve the performance of the catalysts²⁹,³⁰. Dual atom catalyst (DAC) is an atomically dispersed catalyst with an active site consisting of two paired metal atoms³¹⁻³⁴. The synergy between pairs of metal atoms in a DAC provides the special ability to further reduce the energy barrier of complex chemical reactions³⁵,³⁶. If an atomically dispersed bimetallic catalyst with Cu and Ru acting together is designed, the synergistic effect of the two metal atoms can not only enhance the adsorption of the intermediates but also reduce the energy barriers of the intermediate reaction steps. Pulsed discharge is a process in which the high density current passes through the medium between electrodes, and huge energy is injected into the medium, resulting in an instant rise in temperature, phase transition, and dramatic volume expansion of the medium³⁷⁻⁴⁰. It is a new environmentally-friendly material preparation method with the advantages of high instantaneous power and good repeatability. This technique shows great potential in the synthesis of novel atomically dispersed materials.
+
+In this work, nitrogen-doped graphene aerogel supported Ru-Cu dual atoms catalyst (RuCu DAs/NGA) is synthesized rapidly by the pulsed discharge method. The metal salts supported by NGA are injected with huge energy in tens of microseconds, resulting in the explosive decomposition of the metal salt nanocrystals and the formation of atomically dispersed RuCu dual sites on the NGA. The asymmetric RuN₂-CuN₃ coordination structure is proposed based on corresponding detection and analysis. Impressively, RuCu DAs/NGA exhibits excellent NH₃ production performance during the electrochemical catalytic NO₃⁻ RR process. Some important hydrogenation intermediates (such as *NH₂OH, *NH₂, etc.) were detected by *in situ* attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS). Furthermore, *in situ* X-ray absorption fine structure (XAFS) was utilized to trace and analyze the active sites of RuCu DAs/NGA during electrocatalytic NO₃⁻ RR. The reaction paths and relative free energies of RuN₄/C, CuN₄/C, and RuN₂CuN₃/C at U = 0 V are calculated through density functional theory (DFT). The modulated asymmetric RuN₂CuN₃/C structure can not only optimize the adsorption of the intermediate but also reduce the energy barrier of the elementary reaction. Additionally, we synthesized other atomically dispersed M-Cu (M = Ag, Pt, Pd) dual atoms sites on NGA and suggested similar asymmetric coordination structures (MN₂-CuN₃).
+
+**Pulsed discharge synthesis of RuCu DAs/NGA.** As a demonstration, we tailored RuCu dual atoms on NGA substrates by pulsed discharge strategy (Fig. 1a). At first, nitrogen-doped graphene hydrogel (NGH) was prepared by the hydrothermal assembly approach, and then soaked in the aqueous solutions of copper chloride (CuCl₂) and ruthenium chloride (RuCl₃) for 6 hours. NGA (Supplementary Fig. 1) supported RuCl₃ and CuCl₂ (RuCl₃-CuCl₂/NGA) could be obtained through a quick freeze-drying method. The RuCl₃-CuCl₂/NGA was packed into a copper discharge tube and compressed by copper plugs. The two ends of the copper tube were connected to the two electrodes of the high-power pulsed discharge system through copper strips (Supplementary Figs. 2 and 3). Once the capacitor was fully charged, the air switch of the discharge system would be triggered to close quickly. As a result of the huge pulse current, RuCl₃-CuCl₂/NGA itself generated an instant thermal shockwave. Furthermore, applying a high-intensity current pulse engenders a potent electromagnetic field, thereby giving rise to several transient regions of elevated temperature on graphene. Metal salt structures explosively break down rapidly owing to huge energy input, even directly into gaseous *Ru, *Cu, and *Cl ions from solid nanocrystals (Supplementary Video 1). These gaseous metal ions are anchored by N atoms on the graphene and form bimetallic atom pairs under the action of the sharp pulsed electromagnetic field (Fig. 1b). The micropore distribution of NGA and graphene aerogels (GAs) by N₂ adsorption-desorption tests are shown in Fig. 1c, with holes with diameters ranging from 0.4–0.8 nm being the main pore defects on N-doped graphene. The inset (Fig. 1c) is a schematic diagram of N-doped graphene with different micropore types, the N atoms provide a large number of sites for the fixing of atomically dispersed melt atoms. These sub-nanopores on two-dimensional graphene are the critical space-confined strategy in our design for anchoring bimetallic atom pairs. The apparent activity of the catalyst could be improved due to its high specific surface area and microporous structure, which is conducive to mass transfer. Figure 1d shows the typical current waveform during pulsed discharge. The voltage amplitude was set at 8 kV (Supplementary Fig. 4), and the value approached 0 after several oscillating attenuations, lasting a total duration of about 600 µs. The first current peak lagged behind the first voltage peak by ~ 25 µs, a result of the combined action of capacitors and inductors in the discharge system. The main peak parameters of the current waveform are listed in Supplementary Table 1 during the pulsed discharge process. The current presented a typical underdamped waveform in this resistance-inductance-capacitance (RLC) circuit, indicating that the resistance in the circuit did not change significantly during the pulsed discharge. The energy barriers of metal salts would blast and decompose when the transient energy input, as illustrated in Fig. 1e. Because the current and the effect of Joule heat are almost synchronized⁴¹,⁴², the temperature rise time from 0 to peak on the metal salt nanocrystals is about tens of microseconds. When metal salt nanocrystals were inputted such a large amount of energy in a short period, they would decompose explosively and the metal ions anchored on NGA during the pulsed discharge (Supplementary Note 1). The magnetic pinching effect generated by the dynamic electromagnetic field inhibits the radial expansion of the formed ions, maintaining the relatively high-density plasma containing Ru and Cu ions in the NGA. The mixed Ru and Cu ions form atomic pairs on the defects of NGA caused by the trapping effect driven by thermal dynamics. Additionally, the air within the porous NGA structure would give rise to localized corona discharge plasmas comprising *O and *N ions (Supplementary Fig. 5). Following multiple discharge events, these conditions facilitate the formation of robust bonds between the metal atom pairs and the NGA carrier. Due to the current peak value being only kept for tens of microseconds and then quickly dropping to 0, the distinctive NGA-supported atomic dispersed metals structure could be retained at high cooling rates and stably exist. During the cooling process, some metal atoms that are very close together inevitably coalesce to form metal nanoclusters. Multiple repetitions of pulsed discharge treatment are necessary, as they induce repeated vaporization and dispersion of the metal atoms. Furthermore, repeated pulsed discharges facilitate the thorough mixing of Ru and Cu metal atoms, promoting their anchoring on NGA to form Ru-Cu atomic pairs. Our experiments have demonstrated that RuCu DAs/NGA can be fully obtained after six cycles of pulsed discharges (Supplementary Fig. 6).
+
+A significant development trend in single-atom catalysts is the incorporation of multiple metal elements on the support to form asymmetric coordination structures (Fig. 1f). Utilizing the rapid thermal shockwave and corona plasma effects of pulsed discharge, atomically dispersed catalysts of multiple metals can be rapidly synthesized. Additionally, the microsecond-scale heating and cooling characteristics create favorable conditions for the formation and stable existence of asymmetric coordination structures between the metal and support atoms. Compared to other synthesis methods for atomically dispersed catalysts, our pulsed discharge synthesis technique (Supplementary Fig. 7) stands out due to its exceptionally high instantaneous temperatures and extremely short duration, on the order of hundreds of microseconds.
+
+After six pulsed discharge treatments, the still intact RuCu DAs/NGA sample was recovered from the discharge tube. Figure 2a displays the profile scanning electron microscope (SEM, Supplementary Note 2) image of RuCu DAs/NGA, the 3D porous structures (tens of microns) are similar to that of the initial NGA. The transmission electron microscopy (TEM) image still exhibits the features of curly porous (tens of nanometers) graphene, as shown in Fig. 2b. Figure 2c shows the high-angle annular dark field-scanning transmission electron microscopy (HAADF-STEM) image of RuCu DAs/NGA, no nanoparticles could be found on graphene. The corresponding energy dispersive spectrum mapping results are illustrated in Fig. 2d, where the elements C (red), N (blue), Cu (yellow), and Ru (pink) are uniformly distributed. The high-resolution HADDF-STEM image reveals a significant number of evenly distributed bright spots on the graphene, as shown in Fig. 2e. These bright spots are thought to be atomically dispersed Ru and Cu atoms because the atomic numbers of metals are much higher than that of C, N, and other nonmetallic elements. Figure 2f is a partial enlargement image of Fig. 2e, due to the atomic number of Ru being much larger than that of Cu, the brightness of Ru atoms in the STEM dark field image is significantly higher than that of Cu atoms. Meanwhile, a large proportion of Ru-Cu metal atoms form pairs. Figure 2g depicts the corresponding two 3D intensity profiles in Fig. 2f, which exhibits the obvious intensity difference between Ru and Cu atoms, as well as the significant difference with C, N, and other atoms in NGA. Furthermore, the frequency histogram of Ru and Cu adjacent atomic distance distribution and the frequency distribution histogram of single and double sites statistically in the HADDF-STEM images are shown in Fig. 2h and Supplementary Fig. 8. The average distance between Ru and Cu atoms is about 0.25 nm, and the two easily interact at this distance. The proportion of Ru-Cu dual sites is about 65.9%, indicating that the RuCu DAs/NGA sample prepared by pulsed discharge is worthy of the name. Notably, lots of N atoms provide rich sites for anchoring metal atoms. The contents of Ru and Cu are 0.39 at% (3.1 wt%) and 0.52 at% (2.6 wt%), respectively (Supplementary Fig. 9), which is consistent with the results (Ru 3.3 wt%, Cu 2.9 wt%) of inductively coupled plasma optical emission spectrometry (ICP-OES) test.
+
+Figure 2i presents a schematic plot of Ru-Cu dual atoms with different types anchored on the N-doped graphene, this is the main reason for the different lengths between Ru and Cu atoms. Furthermore, the varying angles formed between the RuCu and the NGA planes can also influence the interatomic distances within the Ru-Cu atom pairs (Supplementary Fig. 10). The intensity ratio of D (~ 1346 cm⁻¹) to G (~ 1594 cm⁻¹) for RuCu DAs/NGA is slightly higher than that for NGA in the Raman test (Supplementary Fig. 11), demonstrating that more defects were formed in graphene owing to Ru-Cu atoms dopant during pulsed discharge. The X-ray diffraction (XRD) pattern (Supplementary Fig. 12) displays only one distinct wide peak at ~ 25°, attributed to the stacked graphene layers. No peaks of Ru or Cu crystals appear in the RuCu DAs/NGA.
+
+**Analysis of the atomic bond structure.** The atomic bond structure of RuCu DAs/NGA was investigated by X-ray photoelectron spectroscopic (XPS) and XAFS. The XPS spectra of RuCu DAs/NGA are shown in Supplementary Fig. 13, a distinct N 1s peak was identified, which contains a C-N bond and a weak C-Ru/Cu bond. The Cu 2p₃/₂ spectrum has two peaks at 931.5 and 933.4 eV, respectively, assigned to Cu⁺ and Cu²⁺⁴³. Additionally, the weak signals of Ru 3p were detected as well. Figure 3a shows the Cu K-edge X-ray absorption near-edge spectra (XANES) of RuCu DAs/NGA and references (Cu, CuO, and Cu SAs/NGA). The absorption edge of RuCu DAs/NGA is closer to that of CuO than that of Cu foil, indicating that the oxidation state of Cu in RuCu DAs/NGA is closer to CuO. The k³-weighted Fourier transform (FT) from Cu K-edge extended X-ray absorption fine structure (EXAFS) spectra (Fig. 3b) present the peaks of RuCu DAs/NGA and Cu SAs/NGA (Supplementary Note 3) are located at ~ 1.45 Å, which attributed to the Cu-N bond in first shell⁴⁵,⁴⁶. Moreover, the secondary peak at 2.30 Å of RuCu DAs/NGA is close to the first shell location of the Cu foil (2.24 Å), which implies that the existence of the metallic diatomic coordination structure in RuCu DAs/NGA. Figure 3c displays the Cu K-edge EXAFS fitting result of RuCu DAs/NGA in the R space. The fitting and experimental results have a high degree of matching in different spaces (R space, k space, and q space, Supplementary Fig. 14). Correspondingly, the structural parameters were extracted from the Cu K-edge EXAFS fitting results (Supplementary Table 2). The coordination number of Cu was estimated to be 2.9 Cu-N (first peak) and 0.8 Cu-Ru (second peak) in the first shell according to the fitting results, with bond lengths of 1.93 and 2.59 Å, respectively. The Ru K-edge XANES of RuCu DAs/NGA and references (Ru, RuO₂, and Ru SAs/NGA) are shown in Fig. 3d. The absorption edge of RuCu DAs/NGA is located between RuO₂ and Ru. The k³-weighted FT from Ru K-edge EXAFS spectra (Fig. 3e) exhibit the strong peaks of RuCu DAs/NGA and Ru SAs/NGA are located at about 1.50 Å, which correspond to the Cu-N bonds in the first shell. Similarly, the secondary peak at 2.36 Å of RuCu DAs/NGA is close to the peak position of the Ru foil (2.34 Å), indicating the presence of the metallic bond in RuCu DAs/NGA as well. Likewise, a good fitting result in R space is shown in Fig. 3f. Based on the fitting results, the coordination numbers of Ru-N and Ru-Cu are 2.2 and 0.8 in the first shell respectively. The bond lengths of Ru-N and Ru-Cu are 2.07 Å and 2.57 Å, respectively. Ultimately, a proposed asymmetric coordination structure (RuN₂-CuN₃/NGA) is depicted in Fig. 3g. According to the first derivative of the Cu absorption edge of RuCu DAs/NGA and references, the valence states of Cu and Ru in RuCu DAs/NGA were estimated to be 1.32 and 3.0, as shown in Fig. 3h. The Cu and Ru K-edge wavelet transform (WT) EXAFS results of RuCu DAs/NGA and references were utilized to distinguish backscattered atoms, as shown in Figs. 3i and 3j. The maximum intensity position of Cu for RuCu DAs/NGA is ~ 6.7 Å, which is closer to that of CuO (~ 6.2 Å) than that of Cu foil (~ 8.0 Å). Similarly, the maximum intensity position of Ru for RuCu DAs/NGA (~ 6.0 Å) is close to that of RuO₂. The difference in intensity between the RuCu DAs/NGA and the references originates from the combined contribution of Ru-N, Cu-N, and Ru-Cu.
+
+**Generality of synthesis and structural analysis of CuM DAs/NGA (M = Pt, Ag, Pd).** Since the instantaneous temperature generated by the pulsed discharge is super high compared to the thermal decomposition temperature of ordinary metal salts, this method can rapidly prepare the atomical level dispersion for most metals. In addition to the Cu-Ru diatomic structure, the combinations of Cu and several other metals were designed to form asymmetric coordination structures of Cu-M/NC on NGAs in this research. The combination of Cu and other metals atomically dispersed dual catalysts on NGA provides different potential applications for different electrocatalytic scenarios. Here, the general pulsed discharge synthesis method was easily extended to the preparation of other metals dual atoms supported by NGA, such as PtCu DAs/NGA, AgCu DAs/NGA, PdCu DAs/NGA, etc. Figure 4a shows a TEM image of PtCu DAs/NGA, which exhibits 3D pleated graphene characteristics and is not supported with metal nanoparticles or clusters. After washing with water and freeze-drying, PtCu DAs/NGA is the broken lamellar graphene aerogel (inset Fig. 4a). The EDS mapping images of PtCu DAs/NGA are displayed in Fig. 4b, Pt and Cu elements are uniformly distributed on the N-doped graphene sheets. The contents of Pt and Cu are 0.5 at% (7.2 wt%) and 0.55 at% (2.6 wt%), respectively (Supplementary Fig. 15), which is consistent with the results (Pt 7.5 wt%, Cu 3.0 wt%) of ICP-OES test. Figure 4c presents a HADDF-STEM high magnification image of PtCu DAs/NGA, with the red oval dotted line surrounding the Pt-Cu bimetallic pair sites.
+
+Further, the coordination structure of PtCu DAs/NGA was analyzed through XAFS testing. The Pt L₃-edge XANES results of PtCu DAs/NGA and references (Pt foil and PtO₂) are shown in Fig. 4d, the valence state of Pt in PtCu DAs/NGA is between 0 and + 4 based on the intensity of white lines. The valence state of Cu in PtCu DAs/NGA was calculated to be 1.36 (Supplementary Fig. 16). The k³-weighted FT from Pt L₃-edge EXAFS spectrum (Fig. 4e) exhibits the peak of PtCu DAs/NGA is located at ~ 1.63 Å, which corresponds to the Pt-N bond in the first shell. Besides, the second peak at 2.30 Å for PtCu DAs/NGA, is comparable to the first shell of Pt foil (2.60 Å), with proposing the metal-metal bond. The EXAFS fitting result in R space for PtCu DAs/NGA is inserted in Fig. 4e, and it can be seen that the experimental and fitting curves are in agreement. Equally, the characterizations and structures of AgCu DAs/NGA and PdCu DAs/NGA are shown in Figs. 4f-4o and Supplementary Figs. 17–20. Supplementary Table 3 provides the best-fitting structural parameters, suggesting the local coordination structures of PtN₂CuN₃, AgN₂CuN₃, and PdN₂CuN₃ in the three samples. According to the first derivative of the Cu (Ag, Pd) absorption edge (or the white-line peak area of Pt) for PtCu DAs/NGA, AgCu DAs/NGA, PdCu DAs/NGA, and their references, the oxidation states of Cu are estimated to be 1.36, 1.32 and 1.42 in PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA respectively (Fig. 4p). The oxidation states of Pt, Ag, and Pd in three catalysts were calculated to be 1.73, 0.36 and 1.75 respectively. The percentage of Pt-Cu, Ag-Cu, and Pd-Cu dual sites are 64.9%, 61.6%, and 65.4% respectively, demonstrating that interacting metals dual sites dominate as compared to single sites in PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA (Fig. 4q).
+
+The optimized structures and differential charge densities of PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA are shown in Figs. 4r-4s. The bond lengths of Pt-Cu, Ag-Cu, and Pd-Cu pairs are estimated to be 2.22 Å, 2.32 Å, and 2.31 Å respectively. Furthermore, the differential charge densities of PtN₂CuN₃/C, AgN₂CuN₃/C, and PdN₂CuN₃/C were calculated to elucidate the electrical properties of the asymmetric Cu-M sites. The asymmetric deployment of the modulated MN₂-CuN₃ leads to a significantly polarized surface charge distribution. Electron enrichment (cyan) near the MN₂ site and electron deficiency near CuN₃ can be attributed to electron transfer from the CuN₃ site to the PtN₂, AgN₂, and PdN₂ sites, respectively (Supplementary Fig. 21). The three extended studies demonstrate the universal strategy to synthesize unsymmetrical atomic interface structures for the pulsed discharge method. This combination of Cu and noble metals could open more catalytic possibilities for atomically dispersed diatomic catalysts.
+
+**Electrocatalysis and *in situ* study.** The reaction of nitrate reduction to the ammonia was evaluated in 0.1 M KNO₃ and 0.1 M KOH conditions (Supplementary Note 4), and linear sweep voltammetry (LSV) curves are shown in Fig. 5a. RuCu DAs/NGA showed the lowest onset potential and the fastest current density decrease in the three catalysts. The Faradaic efficiency and ammonia evolution rate were studied by chronoamperometry with different operated potentials (Supplementary Fig. 22). Figure 5b presents the FE of the NH₃ product (FE_NH3) at different potentials (-0.1 V to -0.6 V vs. RHE). Impressively, the FE_NH3 of RuCu DAs/NGA reached 97.8% at -0.3 V vs. RHE, the performance is very competitive compared to other catalysts reported recently⁴⁷,⁴⁸. The partial current density of NH₃ (J_NH3) on RuCu DAs/NGA has the optimal performance among the three catalysts from − 0.1 V to -0.6 V vs. RHE (Fig. 5c), suggesting the enhanced NO₃⁻ RR performance through the tailored asymmetric RuN₂-CuN₃ coordination structure strategy. The J_NH3 reached − 18.2 mA cm⁻² on RuCu DAs/NGA at -0.3 V vs. RHE, which is an excellent performance. The NH₃ yield rate results (Fig. 5d) exhibit that RuCu DAs/NGA is more active and selective than Ru SAs/NGA and Cu SAs/NGA. To ascertain the endurance of RuCu DAs/NGA under the conditions of NO₃⁻ RR, a series of 10 continuous electrolysis cycles were executed. Figure 5e illustrates that RuCu DAs/NGA exhibits sustained high Faradaic efficiency and NH₃ yield rates across these cycles, substantiating its exceptional durability. The current density loss was at the negligible operated potential of -0.3 V and − 0.5 V vs. RHE during the 24 h continuous NH₃ testing from NO₃⁻ RR (Supplementary Fig. 23). The NH₃ yield rate of RuCu DAs/NGA reached 3.07 mg h⁻¹ cm⁻² at -0.4 V vs. RHE, which still presents an optimal performance compared to similar catalysts in recent reports (Fig. 5f, Supplementary Table 4). Moreover, the yield rate of RuCu DAs/NGA reached 1.61 mg h⁻¹ cm⁻² and 4.35 mg h⁻¹ cm⁻² at -0.3 V and − 0.5 V vs. RHE respectively, which is better than the references (Ru SAs/NGA and Cu SAs/NGA) and superior to similar catalysts as well.
+
+To further elucidate the intermediates involved in NO₃⁻ RR on various electrocatalysts, *in situ* ATR-SEIRAS measurements were conducted (Supplementary Fig. 24). Figure 5g presents the results for the RuCu DAs/NGA at open circuit potential (OCP), -0.1 V, -0.2 V, -0.3 V, -0.4 V, -0.5 V, and − 0.6 V vs. RHE. A progressively increasing intensity of the NO₃⁻ vibrational band at ~ 1245 cm⁻¹ was observed from OCP to -0.6 V, indicating the ongoing consumption of NO₃⁻ during electrolysis. A broad peak centered around 3400 cm⁻¹ (Supplementary Fig. 25), identified as the *NH₂ species, was consistently observed⁴⁹,⁵⁰. Additionally, the peak intensities of hydrogenation intermediates (*NH₂ at ~ 1168 cm⁻¹, *NH₂OH at ~ 1115 cm⁻¹), NH₄⁺ (NH₃ at ~ 1460 cm⁻¹), and deoxidation intermediates (*NO₂ at ~ 1630 cm⁻¹) progressively increased⁵¹,⁵². These findings suggest that the RuCu DAs/NGA catalyst is highly effective in activating NO₃⁻, subsequently facilitating the formation of substantial amounts of hydrogenation intermediates that are eventually converted into NH₄⁺ (NH₃). The detection of *NH₂OH and *NH₂ species indicates the co-occurrence of both indirect and direct reduction pathways during the NO₃⁻ RR process on RuCu DAs/NGA. Additionally, a significant enhancement of the H-O-H stretching vibrations at approximately 1660 cm⁻¹ and 3484 cm⁻¹ was observed, indicating the dissociation of H₂O into OH⁻ and H⁺ species, which are stabilized on the RuCu DAs/NGA catalyst.
+
+To study the structure-activity relationship of RuCu DAs/NGA on the atomic level, *in situ* XAFS tests were carried out during the electrochemical catalytic NO₃⁻ RR process. Figures 5h and 5i present the *in situ* Cu K-edge and Ru K-edge XANES results of RuCu DAs/NGA at OCP, -0.1 V, -0.2 V, -0.3 V, -0.4 V, and − 0.5 V vs. RHE. The absorption edges of both Cu and Ru in RuCu DAs/NGA tend to gradually move towards lower energy, together with the reduced intensity of their white lines, which implies the decrease in Cu and Ru valence states with the lower applied potentials. According to the results (Supplementary Fig. 26) of the first derivative of the absorption edge for RuCu DAs/NGA, the specific valence states of Cu were decreased from 1.27 to 0.73 when the operated potentials were reduced. Similarly, the valence states of Ru decreased from 2.87 to 2.62 continuously (Fig. 5j). In fact, Cu and Ru disclose similar trends in oxidation states owing to NO₃⁻ being adsorbed on the active site of the catalyst in the electrolyte, resulting in the interaction on the unpaired outermost d orbitals of Cu and Ru with the N 2p orbitals of NO₃⁻. The oxidation states of both Cu and Ru would continue to decrease when lower reaction potentials were applied. The Cu-Ru dual active sites would interact with the adsorbed NO₃⁻ more easily to form Cu/Ru-N/O bonds at lower potentials, which leads to electron redistribution among Cu, Ru, N, and O atoms. The Cu K-edge FT-EXAFS spectra (Fig. 5k) of RuCu DAs/NGA show that the Cu-N peak shifts from 1.50 Å (OCP) to 1.41 Å (-0.3 V), which was considered as the compressing of Cu-N bonds. The Ru K-edge FT-EXAFS spectra (Fig. 5l) of RuCu DAs/NGA exhibit that the Ru-N peak moves from 1.56 Å (OCP) to 1.53 Å (-0.3 V), with the shrinking of Cu-N bonds as well. In addition, the Ru-Cu bonds also tend to shift to the left during the NO₃⁻ RR process. Because the metal atoms are likely not in the same plane as graphene, it is possible to have both metal and metal-N bonds clamped. Besides, the Ru-Cu bonds also tend to move to the left slightly during the NO₃⁻ RR process. The above bond lengths analysis is consistent with the specific fitting results (Supplementary Figs. 27 and 28, and Supplementary Table 5). Due to the metal atoms are likely not on the same plane as graphene, it is possible to be pinched for both the Ru-Cu bonds and the Cu/Ru-N bonds. In brief, the high performance of RuCu DAs/NGA in the electrocatalytic NO₃⁻ RR is due to the joint effect of the RuN₂-CuN₃ coordination moieties. During the NO₃⁻ RR process, the local coordination structure (coordination number and bond length, etc.) near the Ru-Cu sites changed slightly, and a new stable structure would be formed. The detection of the reaction intermediates absorbed by the Ru-Cu active sites might be the main reason.
+
+**Theoretical study of NO₃⁻ RR.** To understand the fundamental mechanism of the NH₃ product reduced from NO₃⁻ for RuCu DAs/NGA, the DFT method was used to investigate the whole process of the NO₃⁻ RR on the asymmetrical RuN₂CuN₃/C moieties. The differential charge densities (Fig. 6a) of RuN₄/C, CuN₄/C, and RuN₂CuN₃/C were calculated to elucidate the electrical properties of Ru and Cu sites and to study the synergistic interactions of the two asymmetric coordination metal atoms as well. The RuN₄ site makes it easier to obtain electrons (electron-rich, green area) than the CuN₄ site, showing a stronger reducing ability. Therefore, it can be predicted that RuN₄ would have a stronger adsorption capacity for intermediates than CuN₄. Through the asymmetric deployment of the modulated RuN₂-CuN₃, the surface charge distribution appears significantly polarized. Electron enrichment near RuN₂ and electron deficiency near CuN₃ are possibly attributed to the electron transfer from the Cu site to the Ru site. To reveal the underlying reason for the interactions between RuN₂-CuN₃ sites and the reactive species, the projected densities of state (PDOS) of CuN₄/C, RuN₄/C, and RuN₂CuN₃/C concentrating on d orbitals of Ru and Cu were simulated (Supplementary Fig. 29 and Fig. 6b). According to the d band center theory⁵³,⁵⁴, the d orbitals of Cu in CuN₄ are further away from the Fermi level compared to Ru in RuN₄, the adsorption of intermediates on CuN₄ is weaker than that on RuN₄. The modulated RuN₂-CuN₃ has a strong synergistic effect, the 4d orbital of Ru is closer to the Fermi level, and it has a stronger adsorption effect for the reaction intermediates, which would be more excellent catalytic activity on RuN₂-CuN₃. Furthermore, another CuN₃ site with a slightly weak adsorption ability may promote the desorption of intermediates, which is conducive to enhancing the reaction rate.
+
+All NO₃⁻ RR pathways and relative free energy on RuN₄/C, CuN₄/C, and RuN₂CuN₃/C at U = 0 vs. RHE are illustrated in Fig. 6c and Supplementary Figs. 30–32. NO₃⁻ is adsorbed first and discharged on the metal sites, forming *NO₃, then transformed into *NO₃H. After that, *NO₃H converts into *NO₂ with the departure of the OH⁻. Subsequently, as protons in H₂O continue to be added to the intermediate and take O atoms away (forming OH⁻), *NH₃ is gradually formed. Finally, *NH₃ is desorbed to yield NH₃ leaving the metal sites. The whole reaction path and the relative free energies of each step are shown in Supplementary Table 6. The NO₃⁻ RR reaction steps can be divided into two parts, namely the reactant or product absorption/desorption process (non-electron gain/loss reaction) and the electron gain/loss reaction steps. From NO₃⁻ to *NO₃H, since the charge transfer number is 0, the process can be considered as the adsorption of the product by the metal sites. According to the energy position of *HNO₃, the adsorption of *HNO₃ by RuN₂CuN₃ is the strongest, followed by RuN₄ and CuN₄ is the worst. A large amount of reactants would accumulate in the RuN₂CuN₃ active site, and the increase of reactant concentration is beneficial to the reaction. From *HNO₃ to *NH₃, the intermediates in the elementary reaction would get an e⁻ at each step. For the elementary reaction of gaining or losing electrons, the smaller the energy barrier is, the easier the reaction is. For RuN₂CuN₃, RuN₄, and CuN₄, the respective maximum energy barrier is 0.252 eV (*NHOH→*NH₂OH), 0.5 eV (*NO→*NOH), and 0.994 eV (*NO→*NOH) in their electron gain/loss steps (Fig. 6d). Note that the energy barrier of *NO→*NOH on RuN₂CuN₃ is 0.242 eV. Compared with Cu, Ru reduces the energy barrier of the key step of *NO→*NOH, while the energy barrier of this step is further decreased in the Ru-Cu dual atoms structure. Thus, *NO→*NOH is no longer the key step in determining the reaction rate on RuN₂CuN₃, while *NHOH→*NH₂OH is the critical step. The most critical intermediate is *NO in NO₃⁻ RR, and reducing the energy barrier of its hydrogenation step is the key to the action of these catalysts. The asymmetric RuN₂CuN₃ structure exhibits the best effect in the three catalysts. Based on the relative free energy positions of *NO and *NOH, the adsorption of the Cu site to intermediates is too weak, and the Ru site could enhance the adsorption of these intermediates, but the adsorption of *NO is too strong and may not be conducive to the reaction. In the RuN₂CuN₃ structure, Ru maintains the strong adsorption of *NOH, but weakens the adsorption of *NO, so the energy barrier of this step is reduced. From the differential charge density of the key intermediates (*NO, Supplementary Fig. 33), *NO gets more charge in the Ru-containing system. The main difference between Ru-Cu and single Ru systems is the adsorption configuration, NO is adsorbed horizontally, and Ru-Cu acts together in the RuN₂CuN₃ system.
+
+For the last step *NH₃→*+NH₃ is the desorption process of the product, the adsorption of the intermediate at the Cu site is weak, and the product is easier to desorption. The desorption ability of Ru containing system is slightly larger, but it is also less than 1eV, which can be carried out at room temperature. The last step is not critical unless desorption is large reach to poison the metal sites, where the barrier has not yet reached the conditions for poison. Under actual reaction conditions, the product NH₃ is highly soluble in water, and the desorption energy of less than 1eV will not poison the metal site. Additionally, the system containing Ru has high selectivity for the NH₃ path, almost only along the path of NH₃ production. While CuN₄/C may produce NO₂⁻ and NO (low energy barrier), NH₃ selectivity is not high enough. This is consistent with the NO₃⁻ RR test results.
+
+# Conclusion
+
+In conclusion, we successfully designed bimetallic atoms utilizing nanopore defects on nitrogen-doped graphene aerogel (NGA) and rapidly synthesized atomically dispersed RuCu diatomic catalysts (RuCu DAs/NGA) through a pulsed discharge method. Our correlation analysis suggests an asymmetric coordination structure of RuN₂CuN₃ on NGA. Notably, this pulsed discharge technique can be adapted to efficiently prepare a variety of diatomic catalysts, including PtCu, AgCu, and PdCu DAs/NGA. The RuCu DAs/NGA catalyst exhibited remarkable electrocatalytic activity and selectivity for ammonia production via NO₃ RR. Both experimental investigations and theoretical calculations demonstrate that the tailored asymmetric RuN₂CuN₃/NGA structure fosters strong cooperativity, optimizing and regulating each elementary reaction effectively. The implications of this research extend beyond academic interest; the scalability and practical applications of our RuCu DAs/NGA catalyst are significant for sustainable NH₃ production through NO₃⁻ reduction. The ability to rapidly prepare bimetallic catalysts with such optimized structures not only enhances catalytic performance but also holds promise for industrial applications where efficient and sustainable processes are crucial. We envision that the pulsed discharge strategy can facilitate the development of advanced catalysts applicable to various energy conversion and catalytic scenarios, thereby contributing to the ongoing efforts in sustainable chemistry and environmental remediation.
+
+# Methods
+
+**Chemicals.** Copper chloride (CuCl₂, 99%, Alfa Aesar), Ruthenium chloride (RuCl₃, 99%, Alfa Aesar), Palladium chloride (PdCl₂, 99%, Alfa Aesar), Hydrogen hexachloroplatinate (H₂PtCl₄•xH₂O, 99.995%, Alfa Aesar), Argentum nitricum (AgNO₃, 98%, Alfa Aesar), mammonia (analytical grade, Alfa Aesar), ammonium hydroxide (NH₃•H₂O, 25–28%, Alfa Aesar), KOH (Sigma Aldrich), Nafion D-521 dispersion (5 wt%, Alfa Aesar).
+
+**Preparation of RuCu DAs/NGA.** The single-layer graphene oxide (GO) was prepared by a modified Hummers method<sup><span citationid="CR55" class="CitationRef">55</span></sup>. GO (30 mg) and NH₃•H₂O (200 mg) were fully dispersed in deionized water (15 mL). The evenly dispersed mixture was poured into a small hydrothermal reactor (20 mL), heated to 180 ℃, and kept there for 6 hours. After cooling to room temperature, the nitrogen-doped graphene hydrogel (NGH) was formed. After cleaning NGA several times with deionized water, it was immersed in an aqueous solution of copper and ruthenium chloride (Cu and Ru 3 at%) for 3 hours. NGA-supported CuCl₂ and RuCl₃ nanocrystals (CuCl₂-RuCl₃/NGA) were obtained after rapid freezing with liquid nitrogen and freeze-drying. The prepared CuCl₂-RuCl₃/NGA was filled into a copper discharge tube using two threaded copper plugs to hold the ends (Supplementary Video 2). Then the discharge tube containing CuCl₂-RuCl₃/NGA was connected in series to the circuit of the high-power pulsed discharge system. The charging voltage was set to 8 kV, and the air switch was automatically triggered after the capacitor finished charging. The pulse current would pass through the copper discharge tube and CuCl₂-RuCl₃/NGA, and CuCl₂-RuCl₃/NGA would be completely transformed into RuCu DAs/NGA after six repeated pulsed discharge processing. The preparation process of PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA was similar to that of RuCu DAs/NGA. The pulsed discharge characteristics are presented in Supplementary Note 1. The preparation of Cu SAs/NGA and Cu SAs/NGA is shown in Supplementary Note 2. The characterizations, the preparation of reference, and the NO₃⁻ RR tests are presented in Supplementary Notes 2–4, the XAFS measurements and data processing are depicted in Supplementary Notes 5–7. The details of DFT calculation methods are shown in Supplementary Note 8.
+
+**In situ ATR-SEIRAS test.** *In situ* ATR-SEIRAS measurements were conducted using a Thermo-Fisher Nicolet iS50 spectrometer. The electrochemical experiments employed a three-electrode cell setup with an electrolyte comprising 0.1 M KNO₃ and 0.1 M KOH. The spectral resolution was set at 4 cm⁻¹, and spectra recorded at open circuit potential (OCP) served as reference points. Measurements were taken across a potential range from 0 V to -0.6 V vs. RHE. To enhance signal sensitivity, a monocrystalline silicon substrate with a gold-plated surface was utilized. The scanning range spanned from 4000 cm⁻¹ to 400 cm⁻¹.
+
+## Data availability
+
+The data supporting the findings of this study are available within the article and its Supplementary Information files. All other relevant source data are available from the corresponding authors upon reasonable request.
+
+# References
+
+1. Suryanto BHR et al (2019) Challenges and prospects in the catalysis of electroreduction of nitrogen to ammonia. Nat Catal 2:290–296
+2. Van Langevelde PH, Katsounaros I, Koper MT (2021) Electrocatalytic nitrate reduction for sustainable ammonia production. Joule 5:290–294
+3. Sun J et al (2021) A hybrid plasma electrocatalytic process for sustainable ammonia production. Energy Environ Sci 14:865–872
+4. Han S, Li H, Li T et al (2023) Ultralow overpotential nitrate reduction to ammonia via a three-step relay mechanism. Nat Catal 6:402–414
+5. Foster SL et al (2018) Catalysts for nitrogen reduction to ammonia. Nat Catal 1:490–500
+6. Christensen CH, Johannessen T, Sørensen RZ, Nørskov JK (2006) Towards an ammonia-mediated hydrogen economy? Catal Today 111:140–144
+7. Chen JG et al (2018) Beyond fossil fuel-driven nitrogen transformations. Science 360:eaar6611
+8. Huang Y, He C, Cheng C et al (2023) Pulsed electroreduction of low-concentration nitrate to ammonia. Nat Commun 14:7368
+9. Li P, Jin Z, Fang Z, Yu G (2021) A single-site iron catalyst with preoccupied active centers that achieves selective ammonia electrosynthesis from nitrate. Energy Environ Sci 14:3522–3531
+10. Shang H, Zhou X, Dong J et al (2020) Engineering unsymmetrically coordinated Cu-S₁N₃ single atom sites with enhanced oxygen reduction activity. Nat Commun 11:3049
+11. Gan T, Wang D (2024) Atomically dispersed materials: Ideal catalysts in atomic era. Nano Res 17:18–38
+12. Suryanto BHR et al (2019) Challenges and prospects in the catalysis of electroreduction of nitrogen to ammonia. Nat Catal 2:290–296
+13. Pan Y et al (2020) Structural Regulation with Atomic-Level Precision: From Single-Atomic Site to Diatomic and Atomic Interface Catalysis. Matter 2:78–110
+14. Li L et al (2021) Efficient nitrogen fixation to ammonia through integration of plasma oxidation with electrocatalytic reduction. Angew Chem Int Ed 60:14131–14137
+15. Bu Y et al (2023) Electrical Pulse-Driven Periodic Self-Repair of Cu-Ni Tandem Catalyst for Efficient Ammonia Synthesis from Nitrate. Angew Chem Int Ed 62:24, e202217337
+16. Wang Y et al (2020) Enhanced nitrate-to-ammonia activity on copper-nickel alloys via tuning of intermediate adsorption. J Am Chem Soc 142:5702–5708
+17. Chen G-F et al (2020) Electrochemical reduction of nitrate to ammonia via direct eight-electron transfer using a copper–molecular solid catalyst. Nat Energy 5:605–613
+18. Guo S et al (2017) Insights into nitrate reduction over indium-decorated palladium nanoparticle catalysts. ACS Catal 8:503–515
+19. Wu Z-Y et al (2021) Electrochemical ammonia synthesis via nitrate reduction on Fe single atom catalyst. Nat Commun 12:2870
+20. Wang Y et al (2020) Unveiling the activity origin of a copper-based electrocatalyst for selective nitrate reduction to ammonia. Angew Chem Int Ed 59:5350–5354
+21. Liu Y, Wei J, Yang Z et al (2024) Efficient tandem electroreduction of nitrate into ammonia through coupling Cu single atoms with adjacent Co₃O₄. Nat Commun 15:3619
+22. Shao J, Jing H, Wei P et al (2023) Electrochemical synthesis of ammonia from nitric oxide using a copper–tin alloy catalyst. Nat Energy 8:1273–1283
+23. Daiyan R et al (2021) Nitrate reduction to ammonium: from CuO defect engineering to waste NOₓ-to-NH₃ economic feasibility. Energy Environ Sci 14:3588–3598
+24. Yin H et al (2023) Electrocatalytic Reduction of Nitrate to Ammonia via a Au/Cu Single Atom Alloy Catalyst. Environ Sci Technol 57:3134–3144
+25. Tao H et al (2019) Nitrogen fixation by Ru single-atom electrocatalytic reduction. Chem 5:204–214
+26. Peng O et al (2022) Swinging hydrogen evolution to nitrate reduction activity in molybdenum carbide by ruthenium doping. ACS Catal 12:15045–15055
+27. Chen F et al (2022) Efficient conversion of low-concentration nitrate sources into ammonia on a Ru-dispersed Cu nanowire electrocatalyst. Nat Nanotechnol 17:759–767
+28. Gao W et al (2023) Alloying of Cu with Ru Enabling the Relay Catalysis for Reduction of Nitrate to Ammonia. Adv Mater 35:19, 2202952
+29. Zhou Y et al (2023) Optimizing Intermediate Adsorption over PdM (M = Fe, Co, Ni, Cu) Bimetallene for Boosted Nitrate Electroreduction to Ammonia. Angew Chem Int Ed 62:24, e202319029
+30. Zhang S, Wu J, Zheng M et al (2023) Fe/Cu diatomic catalysts for electrochemical nitrate reduction to ammonia. Nat Commun 14:3634
+31. Bai J, Lian Y, Deng Y et al (2024) Simultaneous integration of Fe clusters and NiFe dual single atoms in nitrogen-doped carbon for oxygen reduction reaction. Nano Res 17:2291–2297
+32. Pei J et al (2024) Integrating Host Design and Tailored Electronic Effects of Yolk-Shell Zn-Mn Diatomic Sites for Efficient CO₂ Electroreduction. Angew Chem Int Ed. 63, 3, e202316123
+33. Chen C et al (2022) Dual-metal single-atomic catalyst: The challenge in synthesis, characterization, and mechanistic investigation for electrocatalysis. SmartMat 3:533–564
+34. He H, Mai JH, Hu KS et al (2024) Recent advances in electrocatalysts for efficient hydrogen evolution reaction. Rare Met
+35. Zhang W et al (2021) Emerging dual-atomic-site catalysts for efficient energy catalysis. Adv Mater 33:2102576
+36. Guo S et al (2022) Gold boosts nitrate reduction and deactivation resistance to indium-promoted palladium catalysts. Appl Catal B Environ 305:121048
+37. Qiao J et al (2024) Controllable preparation of black phosphorus nanomaterials via liquid-phase pulsed discharge. Mater Today Chem 37:102034
+38. Li C et al (2023) One-step synthesis of structural-controlled metal-graphene nanocomposites via flash atomization and plasma-assisted reactions of electrical explosion. Carbon 213:118296
+39. Liu K et al (2024) Ultra-Fast Pulsed Discharge Preparation of Coordinatively Unsaturated Asymmetric Copper Single‐Atom Catalysts for CO₂ Reduction. Adv Funct Mater 34:2312589
+40. Shao T et al (2018) Atmospheric-pressure pulsed discharges and plasmas: mechanism, characteristics and applications. High Volt 3:14–20
+41. Yao Y, Huang Z, Xie P et al (2019) High temperature shockwave stabilized single atoms. Nat Nanotechnol 14:851–857
+42. Yang C et al (2020) Overcoming immiscibility toward bimetallic catalyst library. Sci Adv 6:eaaz6844
+43. Jia Y et al (2023) Electronic Oxide-Support Strong Interactions in the Graphdiyne-Supported Cuprous Oxide Nanocluster Catalyst. J Am Chem Soc 145:1803–1810
+44. Sarkisov GS, Hamilton A, Sotnikov VI (2018) Inverse polarity effect for electrical explosion of fine metal wires in vacuum. Phys Rev E 98:053203
+45. Chen P et al (2019) Interfacial engineering of cobalt sulfide/graphene hybrids for highly efficient ammonia electrosynthesis. PNAS 116:6635–6640
+46. Cai Y, Fu J, Zhou Y et al (2021) Insights on forming N, O-coordinated Cu single-atom catalysts for electrochemical reduction CO₂ to methane. Nat Commun 12:586
+47. Shang H, Zhou X, Dong J et al (2020) Engineering unsymmetrically coordinated Cu-S₁N₃ single atom sites with enhanced oxygen reduction activity. Nat Commun 11:3049
+48. Tang C, Chen L, Li H et al (2021) Tailoring acidic oxygen reduction selectivity on single-atom catalysts via modification of first and second coordination spheres. J Am Chem Soc 143:20, 7819–7827
+49. Xu Y, Cheng C, Zhu J et al (2024) Sulphur-boosted active hydrogen on copper for enhanced electrocatalytic nitrate-to-ammonia selectivity. Angew Chem Int Ed e202400289
+50. Li X, Shen P, Li X et al (2022) Sub-nm RuOₓ clusters on Pd metallene for synergistically enhanced nitrate electroreduction to ammonia. ACS Nano 17:1081–1090
+51. Luo S et al (2024) Ruthenium-induced hydrolysis effect on Fe₂O₃ nanoarrays for high-performance electrochemical nitrate reduction to ammonia. Appl Catal B Environ 351:123967
+52. Yao Y, Zhu S, Wang H et al (2020) A spectroscopic study of electrochemical nitrogen and nitrate reduction on rhodium surfaces. Angew Chem Int Ed 59:10479–10483
+53. Andersen M (2023) Revelations of the d band. Nat Catal 6:460–461
+54. Shilong J, Xianwei F, Hongwen H et al (2024) Descriptors for the Evaluation of Electrocatalytic Reactions: d-Band Theory and Beyond. Adv Funct Mater 32:2107651
+55. Daniela CM et al (2010) Improved Synthesis of Graphene Oxide. ACS Nano 4(8):4806–4814
+
+# Supplementary Files
+
+- [Video1.mp4](https://assets-eu.researchsquare.com/files/rs-4852122/v1/95680051a12a385a054611b0.mp4)  
+  Animation demonstration of pulsed discharge.
+
+- [Video2.mp4](https://assets-eu.researchsquare.com/files/rs-4852122/v1/e5b95d8de2228e0aa88aa4ee.mp4)  
+  Demonstration of the practical operation of pulsed discharge.
+
+- [SI.docx](https://assets-eu.researchsquare.com/files/rs-4852122/v1/3222603781a0a63412a389b6.docx)

0f15a7bd2d76f5d2bb944609afd0241a2619b021e00ef177a72e85197b771b85/preprint/images_list.json

@@ -0,0 +1,74 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Elevated NAT10 expression is associated with decreased survival and reduced infiltration of immune cells in cancer.\n(A) The expression of NAT10 across different stages of LUAD was assessed using the GEPIA website. (B) Kaplan-Meier survival curve comparison between high- and low-NAT10 expression groups (optimal cut-off) in the TCGA-LUAD cohort. (C) Kaplan-Meier survival curve comparison between high- and low-NAT10 expression groups in 37 lung cancer patients. NAT10 expression was quantified via immunohistochemistry using Image Pro Plus. (D) ROC curves for survival prediction with corresponding AUC values. (E and F) Tumor weight and growth curves for C57BL/6N mice inoculated with TC1 (E) or MCA205 cancer cells (F). 2\u00d710^6 WT cells were subcutaneously inoculated into the backs of C57BL/6N mice (n = 5). Mice received Remodelin or saline via oral gavage for first 5 days at a dose of 100 mg/kg. Tumor size was measured daily using calipers to generate growth curves. (G and H) Tumor weight and growth curves for Nude/Nude mice inoculated with TC1 (G) or MCA205 cancer cells (H). 2\u00d710^6 WT cells were subcutaneously inoculated into the backs of C57BL/6N mice (n = 5). Mice received Remodelin or saline via oral gavage for first 5 days at a dose of 100 mg/kg. Tumor size was measured daily using calipers to generate growth curves. (I) Analysis of immune cell infiltration using the CIBERSORT algorithm between high and low NAT10 expression groups in TCGA-LUAD cohort. (J) Immunohistochemical staining of NAT10 and CD8+ T cells in lung cancer patient samples (n=37). CD8+ T cells counts in the high- and low-NAT10 expression groups are presented on the right. *p<0.05, as determined by unpaired Student\u2019s t-test.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "NAT10 deficiency suppresses tumor growth via immune-dependent mechanisms\n(A and B) Tumor growth curves for C57BL/6N mice bearing TC1 (A) or MCA205 (B) cancer cells transplants. 2*106 WT or sgNAT10 cancer cells were subcutaneously transplanted into back flanks of C57BL/6N mice (n=5), and tumor growth was monitored with calipers at the indicated time. Data are presented as mean \u00b1 SD. ***p < 0.001, Statistical significance was determined by Mann-Whitney U test. (C and D) Tumor weight and growth curves for nude mice inoculated with TC1 (C) or MCA205 cancer cells (D). 2*106 WT or sgNAT10 cacner cells were subcutaneously transplanted into Nude mice that lack mature T lymphocytes. Tumor growth was monitored at the indicated times. Data are presented as mean \u00b1 SD. *p < 0.05, Statistical significance was determined by Mann-Whitney U test. (E-F) Kaplan-Meier survival curves for C57BL/6N mice injected with TC1 (E) and MCA205 (F) cancer cells (n=6 mice for each group). 2*106 WT or sgNAT10 cancer cells were injected intravenously into C57BL/6N mice, and the number of dead mice was recorded every day. ***p < 0.001, Log-rank test. (G-H) Kaplan-Meier survival curves for Nude/Nude mice injected with TC1 (G) and MCA205 (H) cancer cells (n=6 mice for each group). 2*106 WT or sgNAT10 cancer cells were injected intravenously into C57BL/6N mice, and the number of dead mice was recorded every day. (I) C57BL/6N mice were subcutaneously immunized at the left flank with equal numbers of sgNAT10 cancer cells, freeze-thawed WT cancer cells, or PBS (control). Freezing and thawing were performed three times. Fourteen days post-immunization, an equivalent number of live WT cancer cells were subcutaneously transplanted into the right flank of the immunized mice. A schematic representation of the vaccination experiment with sgNAT10 cancer cells is provided in the left panel. Tumor growth was monitored at the specified time points. Data are presented as the percentage of tumor-free mice; n=5 tumors in each group, **p < 0.01, ***p < 0.001.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "NAT10 deficiency triggers immune-response signaling and induces cellular immune responses in vivo.\n(A) Gene Set Enrichment Analysis (GSEA) was conducted on the differentially expressed genes between WT and sgNAT10 TC1 tumor tissue. Three positively regulated 'hallmark' signatures were identified: interferon-alpha response, interferon-gamma response, and inflammatory response (left panel). The gene list was ranked based on the signed likelihood ratio (from log2 fold change [log2FC]) comparing sgNAT10 versus WT TC1 tumors (right panel). (B) Heatmaps illustrating core biological pathways, such as the Antigen Presentation Machinery (APM) and CD8+ T effector cells (Teff), depict gene expression (color-coded by log2FC) in columns. (C) Heatmaps depicting the cell cycle biological pathways illustrate gene expression (color-coded by log2FC) in columns. (D) Multichannel imaging and image analysis were employed to investigate immune cell infiltration in the tumor microenvironment. C57BL/6N mice were subcutaneously transplanted with either WT or sgNAT10 TC1 cancer cells. On day 8, tumor tissues were subjected to six-color immunofluorescence analysis. (E) C57BL/6N mice (n=5/group) were subcutaneously inoculated with 2*10^6 WT or sgNAT10 TC1 cancer cells. They were intravenously administered with 200 \u00b5g of anti-CD8 antibodies per mouse on days -1, 3, and 5. Red arrows indicate the time points of anti-CD8 antibody injections. Tumor growth was monitored at specified time points, starting on day 0. Data are presented as mean \u00b1 SD. **p < 0.01, ***p < 0.001. Statistical significance was determined using the Mann-Whitney U test. (F and G) Flow cytometry was used to analyze the proportions of major immune cell populations in TC1 (F) and MCA205 (G) tumor tissues. Tumor tissues from C57BL/6N mice, transplanted as described in (E), underwent flow-cytometric analysis, focusing on IFN-\u03b3+CD8+ T and GZMB+CD8+ T cells. Data are presented as mean \u00b1 SD, with statistical significance determined using an unpaired Student\u2019s t-test, *p < 0.05 and **p < 0.01. (H) mRNA expression levels of CD8a, IFN-\u03b3, GZMA, GZMB, Cxcl19, and Cxcl10 genes were analyzed using RT-qPCR in TC1 (left panel) and MCA205 (right panel) tumor tissue. Tumor tissues from C57BL/6N mice transplanted as described in (E) underwent RT-qPCR analysis. Data are presented as fold changes relative to WT tumor, with mean \u00b1 SD indicated. Statistical significance was determined using an unpaired Student\u2019s t-test, with *p < 0.05, **p < 0.01, and ***p < 0.001. (I) ELISpot assay was conducted to measure IFN-\u03b3 secretion in TC1 (left panel) and MCA205 (right panel) tumors. Tumor tissues from C57BL/6N mice transplanted as described in (E) underwent ELISpot analysis. The number of spots was quantified using an ELISpot reader, and the results were expressed as spot forming units (SFU). Statistical significance was determined using an unpaired Student\u2019s t-test, with **p < 0.01 and ***p < 0.001. (J and K) FACS analysis was performed to assess the proliferation of CD8 (J) and CD4 (K) T cells in coculture with TC1 and MCA205 cancer cells, with or without NAT10 deficiency. The percentage of proliferating (CFSE-low) cells among all labeled CD4 or CD8 T cells is shown to the right of (J) and (K). Data are expressed as mean \u00b1 SD (n = 3 samples per group).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "NAT10 deficiency triggers IFN-I responses in cancer cells.\n(A and B) GSEA was performed on the DEGs between WT and sgNAT10 group of TC1 (A) and MCA205 (B) cancer cells. Two positively regulated 'hallmark' signatures were identified: interferon-alpha response and interferon-gamma response. Additionally, the heatmaps of the gene list of 'hallmark' signatures were shown on the right. (C and D) The mRNA expression levels of Ifnb, Stat1, Tlr3, Ddx58, Ccl5, Ccl7, Tap1, Mx2 in WT and sgNAT10 of TC1 (C) and MCA205 (D) cancer cells were detected by RT-qPCR, with normalization to GAPDH; **p < 0.01, ***p < 0.001, unpaired Student\u2019s t-test. (E and F) WT or sgNAT10 TC1 (E) and MCA205 (F) cancer cells were subcutaneously transplanted into background of ifnar-/- C57BL/6N mice (n=5). The tumor weight and growth was monitored in the indicated time. Data are presented as mean \u00b1 SD. ***p < 0.001, Statistical significance was determined by Mann-Whitney U test. The weight of the tumors was weighed using an analytical balance, and data are presented as mean \u00b1 SD. **p < 0.01, Statistical significance was determined by an unpaired Student\u2019s t-test.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "NAT10 regulates the stability and translation efficiency of MYC mRNA.\n(A) The highly enriched motif within ac4C peaks was analyzed in acRIPseq. (B) Proportion of ac4C peak distribution in the TSS, 5\u2019 UTR, start codon, stop codon and 3\u2019 UTR region across the entire set of mRNA transcripts. (C) Density distribution of ac4C peaks across mRNA transcripts. (D) Seven candidate genes (Phf2, Myc, Wwc2, Kmt2a, Gigyf1, Timeless, and Nufip2) were identified by acRIP-seq and Label-free quantitative proteomics. (E) Expressions of MYC in sgNAT10 TC1 and MCA205 cancer cells were analyzed by Western blotting. (F) IGV software was used to visualize the peaks with ac4C enrichment in WT and sgNAT10 TC1 cancer cells. Square marked decreased ac4C peaks in sgNAT10 TC1 cacner cells. (G) Schematic representation of positions of ac4C motifs within Myc mRNA (upper panel). The ac4C sites in the 3'UTR of Myc mRNA were mutated to eliminate as many ac4C sites as possible. The lower panel shows the schematic representation of the mutated 3'UTR of the pEZX-MT06 vector for studying the roles of ac4C in Myc mRNA stability. (H) Effect of NAT10 on pEZX-MT06-Myc reporter. TC1 cancer cells were cultured in 24-well plates and transfected with Lipofectamine 3000 reagent according to the manufacturer\u2019s instructions. Specifically, 100 ng/well of pEZX-MT06-Myc and either 0, 150, or 300 ng/well of VP64-NAT10 or empty vector were cotransfected. Additionally, Renilla luciferase plasmids (30 ng/well) were cotransfected as a normalization control for transcription efficiency. Luciferase activity was measured 24 h post-transfection, and the results were presented as relative luciferase activity (luciferase activity normalized to Renilla activity). Data are expressed as mean \u00b1SD. ***P < 0.001. (I) NAT10 RIP-qPCR analysis of Myc mRNA in TC1 cells. (J) The mRNA levels of MYC were detected in sgNAT10 TC1 cells after treatment with Act-D for the indicated times.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_6.png",
+        "caption": "Depletion of NAT10 induces dsRNA-mediated RIG-I-dependent signaling through the Myc/CDK2/DNMT1 pathway.\n(A) The FPKM of individual genes of CDKs family from RNAseq data originating from the WT and sgNAT10 TC1 cancer cells. **p < 0.01, unpaired Student\u2019s t-test. Heatmap depicting the CDKs illustrates gene expression (color-coded by log2FC). (B) The FPKM of individual genes of DNMTs family from RNAseq data originating from the WT and sgNAT10 TC1 cancer cells. **p < 0.01, unpaired Student\u2019s t-test. Heatmap depicting the DNMTs illustrates gene expression (color-coded by log2FC). (C) Western blot analysis of DNMT1, CDK2 and NAT10 in matched WT and sgNAT10 TC1 (upper panel) and MCA205 (lower panel) cancer cells. \u03b2-actin was used as a loading control. (D) Representative immunofluorescence staining of dsRNA in WT, sgNAT10, sgCDK2, sgNAT10 Myc-rescued, and sgNAT10 Cdk2-rescued TC1 cancer cells was detected by using confocal microscopy. Antibody J2 (1:250 dilution, 4 mg/ml) revealed the dsRNA (labelled in red). And the corresponding statistical diagrams were on the right. Statistical analysis was conducted using One-way ANOVA, **p < 0.01, ***p < 0.001. (E) Representative immunofluorescence staining of dsRNA in WT, sgNAT10, sgCDK2, sgNAT10 Myc-rescued, and sgNAT10 Cdk2-rescued MCA205 cancer cells was detected by using confocal microscopy. And the corresponding statistical diagrams were on the right. Statistical analysis was conducted using One-way ANOVA, **p < 0.01, ***p < 0.001. (F) Protein expressions of NAT10 and RIG-I in vector, sgNAT10 and sgNAT10/RIG-I TC1 cancer cells determined by Western blot assay (upper panel). mRNA expressions of Ifnb1, Stat1, Tlr3, Ddx58, Ccl5, Ccl7, Tap1, and Mx2 by RT-PCR in vector, sgNAT10 and sgNAT10/RIG-I TC1 cancer cells.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_7.png",
+        "caption": "Remodelin enhances response to ICIs therapy.\n(A) Schematic diagram showing the combining Remodelin with ICIs therapy in C57BL/6N.\n(B and C) Tumor weight for C57BL/6N mice inoculated with TC1 (B) or MCA205 (C) cancer cells treated with Remodelin and/or anti-PD-1 antibodies. TC1 or MCA205 cancer cells were inoculated subcutaneously into C57BL/6N mice. Mice were received Remodelin or saline via oral gavage for first 7 days at a dose of 100 mg/kg. On day 8, the corresponding group mice were treated with IgG control or anti-PD-1 antibodies. After sacrificing the mice, tumor tissues were excised and the representative images were on the left panel. Additionally, the weight of the tumor tissues was weighed using an analytical balance, and data are presented as mean \u00b1 SD. ***p < 0.001, Statistical significance was determined by unpaired Student\u2019s t-test. (D and E) Representative immunofluorescence staining of CD8+ T cells in TC1 (D) and MCA205 (E) tumor tissues. Tumor tissues from C57BL/6N mice transplanted as in (B) were subjected to immunostaining analysis for CD8+ T cells (red) and nucleus (blue). CD8+ T cells were quantified by counting positive signals in 3 randomly selected fields (20\u00d7) per tumor section using Image J (n=5). Statistical analysis was conducted using One-way ANOVA, **p < 0.01, ***p < 0.001. Scale bar, 100 \u03bcm. (F and G) FACS analysis of the proportions of IFN-\u03b3+CD8+ immune cell populations in TC1 (F) and MCA205 (G) tumor tissues. Tumor tissues from C57BL/6N mice transplanted as in (B) were subjected to FACS analysis for IFN-\u03b3+CD8+ immune cell populations. **p < 0.01, ***p < 0.001. (H and I) ELISpot assay was performed to measure IFN-\u03b3 secretion in TC1 (H) and MCA205 (J) tumor tissues with different treatments. The number of spots and the results were quantified as described above. Statistical significance was determined using an unpaired Student\u2019s t-test, with **p < 0.01, ***p < 0.001.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_8.png",
+        "caption": "Intratumoral delivery of PEI/PC7A/siNAT10 nanoparticles for cancer immunotherapy.\n(A) The diagram of PEI/PC7A and its size distribution. (B) Confocal image of uptake of PEI/PC7A. TC1 cancer cells were cultured in chamber slides overnight, and then added with 20 nM FAM-labeled siRNA for 4 h. Cells were stained with 50 nM Lyso-Tracker Red (Beyotime, Cat#C1046) and 10 \u03bcg/mL Hoechst (Beyotime, Cat#C1022) for 30 min. Immunofluorescence images were acquired on a Nikon A1 fluorescence microscope. (C) mRNA expression levels of NAT10 by RT-PCR in TC1 cancer cells with or without siRNA. 2X10^5 TC1 cancer cells were seeded in 12-well plates overnight. The medium was then replaced with Opti-MEM, and PEI/PC7A was added with a final siRNA concentration of 20 nM. (D) Protein expressions of NAT10 determined by Western blot in TC1 (left panel) and MCA205 (right panel) cancer cells treated with or without PEI/PC7A/siNAT10 nanoparticles. (E) Tumor weight for C57BL/6N mice inoculated with TC1 cancer cells treated with Remodelin or PEI/PC7A/siNAT10 nanoparticles. TC1 cancer cells were inoculated subcutaneously into C57BL/6N mice. Mice were received Remodelin via oral gavage for first 7 days at a dose of 100 mg/kg. PEI/PC7A containing siRNA (5 nmol/kg) dissolved in PBS were intratumorally injected on day 4, 7 and 9. After sacrificing the mice, tumor tissues were excised and the representative images were on the left panel. Additionally, the weight of the tumor tissues was weighed using an analytical balance, and data are presented as mean \u00b1 SD. **p < 0.01, ***p < 0.001. (F) Tumor weight for C57BL/6N mice inoculated with TC1 cancer cells treated with PEI/PC7A/siRNA nanoparticles and/or anti-PD-1 antibodies. TC1 cancer cells were inoculated subcutaneously into C57BL/6N mice. Mice were received intratumorally PEI/PC7A/siNAT10 nanoparticles or saline via on day 4, 7 and 9. On day 8, the corresponding group mice were treated with IgG control or anti-PD-1 antibodies. Data are presented as mean \u00b1 SD. **p < 0.001, ***p < 0.001. (G) Representative immunofluorescence staining of CD8+ T cells in TC1 tumor tissues. Tumor tissues from C57BL/6N mice transplanted as in (F) were subjected to immunostaining analysis for CD8+ T cells (red) and nucleus (blue). CD8+ T cells were quantified by counting positive signals in 3 randomly selected fields (20\u00d7) per tumor section using Image J (n=5). Statistical analysis was conducted using One-way ANOVA, **p < 0.01, ***p < 0.001. Scale bar, 100 \u03bcm. (H) FACS analysis of the proportions of IFN-\u03b3+CD8+ immune cell populations in TC1 tumor tissues. Tumor tissues from C57BL/6N mice transplanted as in (F) were subjected to FACS analysis for IFN-\u03b3+CD8+ immune cell populations. **p < 0.01, ***p < 0.001.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_9.png",
+        "caption": "Schematic of proposed NAT10 blockade-induced antitumor immune responses.\nNAT10 directly acetylates Myc mRNA and promotes Myc transcription, which induces CDK2 expression and promotes cell proliferation. However, inhibition of NAT10 down-regulated MYC-CDK2-DNMT1 expression, which enhanced dsRNA formation to trigger IFN-I responses and antitumor immunity.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

0f15a7bd2d76f5d2bb944609afd0241a2619b021e00ef177a72e85197b771b85/preprint/preprint.md

@@ -0,0 +1,246 @@
+# Abstract
+
+Posttranslational modifications add tremendous complexity to cancer progression; however, gaps remain in knowledge regarding the function and immune regulatory mechanism of newly discovered mRNA acetylation modification. Here, we discovered an unexpected role of N4-acetylcytidine (ac4C) RNA acetyltransferase-NAT10 on reshaping tumor immune microenvironment. Based on analyses of patient datasets, we found that NAT10 was upregulated in tumor tissues, and negatively correlated with overall survival and immune cells infiltration. Inhibition of NAT10 significantly suppressed tumor growth *in vivo* and *in vitro*. NAT10 deficiency in cancer cells significantly upregulated immune cells infiltration and stimulated tumor-specific cellular immune responses, leading to the establishment of robust anti-tumor immunity. Mechanistically, we identified MYC as a key downstream target of NAT10, and then induced CDK2-DNMT1 expression. Meanwhile, inhibition of NAT10 down-regulated MYC-CDK2-DNMT1 expression, which enhanced double-stranded RNAs (dsRNA) formation to induce type I IFN (IFN-I) and trigger immune responses of CD8<sup>+</sup> T cells. In terms of clinical significance, we demonstrated that inhibition of NAT10 using Remodelin or PEI/PC7A/siRNA nanoparticles combined with anti-PD1 treatment synergistically improved tumor immune microenvironment and repressed tumor progression *in vivo*. Therefore, inhibition of NAT10 in cancer cells improve tumor immunogenicity, resulting in tumor suppression by enhancing anti-tumor immune responses. Our study uncovers a crucial role of NAT10 in re-modulating tumor immunogenicity and demonstrates a novel concept for targeting NAT10 in cancer immunotherapy.
+
+**Biological sciences/Cancer/Tumour immunology**  
+**Biological sciences/Cancer/Cancer therapy**
+
+# Background
+
+Cancer represents a considerable global public health challenge. Compared to traditional treatment modalities like chemotherapy, radiotherapy, and surgery, cancer immunotherapy has significantly improved both patient survival and quality of life. The success of immunotherapies has spurred interest in the development of effective antitumor medications, including immune checkpoint inhibitors (ICIs) [1]. However, its effectiveness is primarily hindered by the limited infiltration and activation of immune cells within the tumor microenvironment [2].
+
+Interferons (IFNs) are crucial components of the immune response against infections and malignancies. IFNs play a significant role in promoting the anti-tumor response through the activation and functioning of immune cells, facilitating the elimination of malignant cells [3]. IFNs are primarily released by immune and stromal cells, with cancer cells also contributing to the release of IFNs. Various genotoxic anticancer treatments, such as radiation and chemotherapy, small molecule kinase inhibitors, function by inducing DNA damage or, producing double-stranded RNAs (dsRNA), and then activating IFN pathways to stimulate an immune response against cancer [4–6].
+
+Recent advanced transcriptomic studies have revealed that metazoan cells express diverse types of endogenous dsRNA, including endogenous retroviral elements (ERV), repetitive RNA elements, mitochondrial dsRNAs, mRNAs with inverted Alu-containing 3′ UTRs, and structural dsRNAs with long dsRNA stems [7, 8]. Malignant cells may exhibit increased levels of dsRNAs due to the loss of suppressive epigenetic modifications in repetitive elements, genomic instability, or mitochondrial damage induced by oxidative stress, thereby increasing the dsRNA load in cancer cells [9]. For instance, ERVs make up more than 8% of the human genome, with the majority being silenced in normal somatic cells through promoter DNA methylation. In some cancers, loss of ERV DNA methylation leads to aberrant overexpression of ERVs, and bidirectional transcription of ERVs has been demonstrated to enhance dsRNA formation [10, 11].
+
+Growing evidence indicates that the accumulation of intracellular dsRNA produced by cancer cells stimulates the production of type I IFN, thereby enhancing antitumor immunity [12]. Abnormal accumulation of endogenous dsRNAs triggers an innate antiviral and antitumor immune response by activating dsRNA-sensing pathways, including retinoic acid-inducible gene I (RIG-I), which may lead to chronic inflammation and associated human diseases [13]. Further investigations into these gene signatures and associated changes in the anti-tumor immune response are likely to significantly contribute to predicting patient outcomes and their responses or resistance to chemotherapy or immunotherapy.
+
+Some modifying enzymes, such as DNA modifying enzymes, and mRNA regulators, can influence tumor immunity [14]. In a recent study, Arango and colleagues identify N4-acetylcytidine (ac4C) as a new mRNA acetylation modification catalyzed by N-acetyltransferase 10 (NAT10) that increases the stability and translation efficiency of transcripts [15]. NAT10 has recently been shown to regulate tumor progression, however its impact on tumor immunity remains understudied. In this study, we demonstrate that inhibiting NAT10 effectively enhances the type I IFN response, stimulates adaptive immune responses, and inhibits tumor progression. Combining NAT10 inhibitors with a novel delivery system or ICIs therapy could offer effective means to boost immunity and halt tumor progression, providing targets for clinical treatment.
+
+# Results
+
+NAT10, the only writer for ac4C modification on mRNAs, has been reported to have many important functions, such as affecting stem cells differentiation, promoting glycolysis addiction. Additionally, the potential role of NAT10 in pro-tumor effect has been uncovered, but the anti-tumor immunity of cancer-intrinsic NAT10 has not been reported.
+
+To determine the role of NAT10, we utilized GEPIA database to explore the influence of NAT10 in lung adenocarcinoma (LUAD) [16]. The expression of NAT10 was increased in TCGA-LUAD cohort with the stage increase of the disease (Fig. 1A), which indicated that NAT10 was related with the progression of cancer. Moreover, we download the raw data and survival information from TCGA-LUAD cohort to obtain a Kaplan-Meier survival curve. The results showed that higher NAT10 expression was correlated significantly with shorter overall survival (Fig. 1B). Concurrently, we collected 38 pathology slides of lung cancer patients and performed immuno-histochemical staining to obtain immuno-histochemistry stain score for survival analysis. The results was in accordance with TCGA database, as shown by higher NAT10 expression with lower survival (Fig. 1C). Moreover, ROC curve was utilized to validate the ability of the prognostic efficiency of NAT10, which indicated that the NAT10 has a good predictive value in lung cancer (Fig. 1D). Altogether, these results suggest that NAT10 is a cancer-promoting gene, which may be an important risk factor for the development of lung cancer.
+
+Given our findings above, we further explored the oncogenic effect of NAT10 in vivo. Remodelin, a well-established inhibitor for NAT10 [17], was used for the further study. The immunocompetent (C57BL/6N) mice were used to establish syngeneic tumor models. Mice were injected subcutaneously with murine lung cancer cells (TC1) or murine fibrosarcoma cells (MCA205). The tumor masses on the flank of the mice indicated the successful establishment of the tumor after 14 days. Importantly, the tumors in mice exposed to low-dose Remodelin were significantly reduced in weight and size compared to the saline group (Fig. 1E, 1F). On the other hand, Immunodeficient (Nude/Nude) mice injected subcutaneously with TC1 or MCA205 cells were established. Our results showed that there was no difference in tumor size and weight whether exposed to saline or low-dose Remodelin in the nude mice (Fig. 1G, 1H). Additionally, we further explored the effect of NAT10 on cancer cell in vitro. Colony formation assay was conducted to detect the role of NAT10 in the cell proliferation. The results showed that Remodelin significantly suppressed the proliferation of cancer cells (Fig. S1A, S1B). The above results indicated that the effect of NAT10 on tumor growth was partly related to host immunity.
+
+Analysis of immune cell infiltration was performed using the CIBERSORT algorithm between high and low NAT10 expression groups in TCGA-LUAD cohort [18]. The results showed that LUAD patients with lower expression of NAT10 appeared to have higher proportions of immune cells, include CD8+ T cells and DCs (Fig. 1I). The expression analysis of NAT10 within individual patients was conducted based on the scRNAseq data (GSE148071) [19], whereby the summation of NAT10 expression levels across all cells within each patient was followed by division by the total number of sequenced cells within the same patient. The results also indicated a positive relationship between NAT10 expression and malignant cells percentage while a negative correlation with the proportion of T cells or DCs (Fig. S1C). Furthermore, our immunohistochemical staining results in lung cancer samples (n = 37) showed the same negative association between NAT10 and CD8+ T cells (Fig. 1J). Collectively, these results suggest that NAT10 is a proto-oncogene and maybe affect tumor growth in an immune-dependent manner.
+
+Given the observed phenomenon, it is imperative to further investigate the mechanisms underlying NAT10 in anti-tumor immunity. CRISPR/Cas9 technology utilizing Nat10-specific sgRNA (sgNAT10) pairs was employed to knockout NAT10 in TC1 and MCA205 cells (Figs. 2A, S2A). Wild-type (WT) cells transfected with an empty vector served as controls. To evaluate whether NAT10 inhibition within cancer cells triggers immune responses, we established syngeneic tumor models in immunocompetent (C57BL/6N) mice and immunodeficient (Nude/Nude) mice transplanted with either WT or sgNAT10 cancer cells. All C57BL/6N mice bearing with WT TC1 or MCA205 cells had developed substantial tumor masses post-transplantation, whereas the tumor masses disappeared quickly in those bearing with sgNAT10 TC1 or MCA205 cells (Fig. 2A, 2B), indicating that NAT10 deficiency might impede subcutaneous tumor growth in immunocompetent mice. Subsequently, we established transplant tumor models in immunocompetent Nude mice. The results demonstrated that Nude mice in both WT and sgNAT10 groups developed apparently substantial tumors though with a significant difference (Fig. 2C, 2D), further suggesting that NAT10 deficiency may suppress tumors in an immune-dependent manner. As the mice receiving transplants of NAT10-deficient TC1 and MCA205 cells exhibited slightly smaller tumors compared to those with WT cells in Nude/Nude mice, we speculated that NAT10 might have additional suppressive effects on tumor cell proliferation besides of eliciting adaptive immune responses. Consequently, we performed colony formation assays and CCK8 assays. Our findings demonstrated that NAT10 deficiency significantly impeded colony formation (Figs. S2B, S2C) and proliferation (Figs. S2D, S2E) in both TC1 and MCA205 cells.
+
+Moreover, we further investigated whether NAT10 could influence the survival via the host immune system. C57BL/6N mice bearing NAT10-deficient TC1 or MCA205 cell transplants exhibited significantly prolonged survival compared to mice with WT cell transplants; however, they succumbed to the tumor within 60 days. In contrast, all C57BL/6N mice receiving NAT10-deficient cancer cell transplants survived until day 60 (Fig. 2E, 2F). In the xenograft model using Nude/Nude mice, no significant difference of survival between WT and sgNAT10 groups was found and all mice died within 40 days (Fig. 2G, 2H). These differing outcomes between immunocompetent and immunodeficient mice further support the notion that NAT10 inhibition may enhance mouse survival by activating immunity.
+
+Considering the possible nonspecific effects of the CRISPR/Cas9 system, we restored NAT10 expression in the NAT10-deficient TC1 cells. We first constructed a plasmid VP64-NAT10-GFP, in which the base sequence corresponding to the NAT10 sgRNA position in the VP64-NAT10-GFP plasmid was modified, with the encoded amino acids unchanged to avoid cleavage by the CRISPR CAS9 enzyme. Our findings demonstrated that NAT10 restoration enabled NAT10-deficient TC1 cells to successfully develop tumors (Fig. S2F), providing clear evidence of the pro-oncogene effect of NAT10.
+
+For the assessment of possible involved immune memory, we subcutaneously immunized C57BL/6N mice with either sgNAT10 cancer cells or freeze-thawed WT cancer cells on the left side, followed by re-challenge with comparable numbers of live WT cancer cells on the right side after 2 weeks (Fig. 2I, left panel). Intriguingly, each mouse immunized with sgNAT10 cancer cells completely inhibited WT tumor growth on the right side, resulting in tumor-free mice, whereas those immunized with freeze-thawed WT cancer cells showed a significantly weaker effect (Fig. 2I). These results suggested that NAT10 deficiency elicited immunological protection.
+
+NAT10 deficiency triggers immune responses of CD8+ T cells in vivo
+
+Based on our previous findings suggesting that NAT10 deficiency may impede tumor growth by activating anti-immune mechanisms, we conducted transcriptomic RNA-sequencing (RNA-seq) analysis to comprehensively investigate whether NAT10 has impacts on immune-response signaling in vivo. Tumor tissues inoculated with WT or sgNAT10 cancer cells were harvested on day 8, and total RNA was extracted for RNA sequencing. Gene Set Enrichment Analysis (GSEA) revealed upregulation of "hallmark" signatures including "Interferon-gamma (IFN-γ) response", "Interferon-alpha (IFN-α) response", and "Inflammatory response" in sgNAT10 TC1 tumor tissues [20]. Heatmaps depicting differentially regulated genes from the GSEA analysis in WT and sgNAT10 TC1 tumor tissues showed increased expression of numerous cytokines and chemokines, such as C-X-C motif chemokine ligands 9, 10 and 11 (CXCL9, 10, 11) (Fig. 3A), which contribute to robust anti-tumor immunity. CXCL9/10/11 are responsible for recruiting and activating T cells via binding with CXCR3 [21]. Moreover, our results found that genes associated with the antigen presentation machinery (APM) and CD8+ Teff were upregulated in sgNAT10 tumor tissues (Fig. 3B), while cell cycle-related genes linked to proliferation were downregulated (Fig. 3C). Collectively, these findings indicate that NAT10 deficiency plays a crucial role in anti-tumor immunity through regulating immunological response factors especially genes associated with CD8+ Teff cells.
+
+To further clarify the detailed immune cells involved in NAT10 deficiency-induced anti-tumor immunity, we applied multicolor immunofluorescence experiments. Our results demonstrated a notable elevation in CD8+ T cells and DCs in sgNAT10 group compared to the WT group, while no noticeable difference was initially observed in Tregs (Fig. 3D). Considering that CD8+ T cells play a crucial role in anti-tumor immunity, we conducted antibody-based depletion of CD8+ T cells prior to in vivo transplantation of sgNAT10 TC1 cancer cells. The results showed that depletion of CD8+ T cells markedly impeded NAT10-deficient-induced tumor regression, suggesting deletion of NAT10 primarily exerts anti-tumor immune effects via CD8+ T cells (Fig. 3E).
+
+Given that CD8 antibodies reversed the protective effect of NAT10 deficiency, we further investigated CD8+ T cell infiltration and functionality. Our results of immunofluorescence assay revealed increased tumor-infiltrating CD8+ T cells in NAT10-deficient tumor tissues (Fig. S3A). Moreover, flow cytometry results quantified higher CD8+ T cell frequencies in NAT10-deficient tumor tissues, consistent with immunofluorescence results (Fig. S3B). For the functional assessment, we further demonstrated the elevated IFN-γ and Granzyme B (GZMB) levels in tumor-infiltrating CD8+ T cells in the sgNAT10 tumor group (Fig. 3F, 3G). Importantly, inguinal lymph nodes, critical for anti-tumor immunity, were also used in our study to investigate the CD8+ T cells. Our results also showed the upregulation of IFN-γ+ CD8+ T cells in the NAT10-deficient group (Fig. S3C, S3D). Additionally, gene expression analysis by real-time PCR method confirmed the upregulation of CD8a, IFN-γ, Granzyme A (GZMA), GZMB, CXCL9, and CXCL10 in sgNAT10 group (Fig. 3H). IFN-γ secretion was increased in the sgNAT10 group, as observed in IFN-γ ELISpot assay (Fig. 3I). Moreover, T-cell proliferation assay indicated enhanced proliferation of both CD4+ and CD8+ T cells in NAT10-deficient cancer cells (Fig. 3J, 3K and S3E). These findings collectively suggest adaptive immune responses, particularly CD8+ T cell-mediated antitumor immunity, have been activated induced by NAT10 deficiency in vivo.
+
+NAT10 deficiency induces IFN-I responses in cancer cells
+
+The above results showing an enhanced IFN response and an increased infiltration of tumor-infiltrating lymphocytes (TIL) in the NAT10 deficient tumor microenvironment suggest a possible link between IFN-mediated tumor cell chemokine expression and increased TIL infiltration, which may be responsible for the enhanced antitumor immune responses. To test this hypothesis, RNA-seq was performed with total mRNA extracted from WT and sgNAT10 TC1 or MCA205 cells. GSEA analysis showed that these pathways were mainly involved in the “IFN-I” signaling pathway (Fig. S4A, S4B). Compared to WT cancer cells, NAT10 deletion induced the expression of genes related to IFN-I response (Fig. 4A, 4B). By RT-qPCR analysis, we further confirmed the increased expression of some of these genes in sgNAT10 cancer cells, including the type I IFN gene Ifnb1 itself, the transcription factor Stat1, the antiviral gene Mx2, the pattern recognition receptor genes Tlr3 and Ddx58, the antigen presentation related gene Tap1, as well as the chemokine-encoding genes Ccl5 and Ccl7 (Fig. 4C, 4D).
+
+NAT10 deficiency could induce IFN-I responses in cancer cells, which can play key roles in the activation of cellular components of the immune response, such as dendritic cells and T cells. To verify that IFN-I responses underlined the outcomes, sgNAT10 cancer cells were transplanted into type I IFN receptor KO (Ifnar1 KO) mice. The results showed that both WT and sgNAT10 cancer cells developed apparently substantial tumors in Ifnar1 KO mice, suggesting the effects favoring anti-tumor immune responses triggered by NAT10 deficiency were significantly abolished on an Ifnar1 KO background (Fig. 4E, 4F). These data indicated that NAT10 deficiency in cancer cells may drive IFN-I responses to promote protective anti-tumor CD8+ T cell immunity.
+
+NAT10 increases MYC expression through regulating mRNA acetylation
+
+Next, we explore the mechanism by which NAT10 deletion induces interferon production. To identify whether the acetyltransferase NAT10 directly mediated antitumor immune response, acRIP-seq analysis was performed. The sequential analysis of ac4C peaks showed that typical GAGGAGA motifs were highly enriched within ac4C sites of mRNA (Fig. 5A). Further analytic results showed that the ac4C peaks predominantly occurred within coding sequences (CDS) and 3’untranslated regions (3’UTR) (Fig. 5B, 5C). As reported, the acetyltransferase NAT10 can confer enhanced mRNA stability, and ac4C peaks within wobble sites can stimulate translation efficiency [22]. We therefore investigated potential targets using a combination of acRIP-seq and Label-free quantitative proteomics. We identified 7 candidate genes (Phf2, Myc, Wwc2, Kmt2a, Gigyf1, Timeless, and Nufip2) that showed concomitant decreased mRNA acetylation and reduced protein levels in sgNAT10 cancer cells (Fig. 5D, S5A).
+
+Among the 7 candidate genes, MYC has been reported to be related to both cell proliferation and antitumor immunity [23]. We then performed Western blot, and our results showed that NAT10 deficiency resulted in decreased protein expression of MYC in cancer cells (Fig. 5E). To identify the key ac4C sites that regulate mRNA stability, we further analyzed the acetylation peaks of MYC mRNA. AcRIP-seq data showed that the ac4C peaks were distributed in the CDS and 3’/5’UTR region of MYC mRNA (Fig. 5F). Interestingly, the 3’UTR region of MYC mRNA contains a nucleic acid sequence consistent with the typical GAGGAGA motifs (Fig. 5A), suggesting that this ac4C site may be more dynamic in regulating MYC mRNA stability. Subsequently, we constructed 3’UTR reporters containing wild type or mutant MYC 3’UTR after the firefly luciferase reporter gene (Fig. 5G). The dual-luciferase assay showed significantly attenuated fluorescence activity in the mut-3’UTR groups compared to WT-3’UTR groups, mirroring reduced mRNA stability due to the loss of acetylated position (Fig. 5H). Moreover, the acRIP-PCR results confirmed that NAT10 may bind to the 3\'UTR of MYC (Fig. 5I). Furthermore, our results also showed that the half-life of MYC mRNA was ≈ 16 hours for WT cells and significantly decreased in sgNAT10 cells, meaning reduced ac4C enrichment was accompanied by increased decay of MYC mRNA (Fig. 5J). Overall, NAT10 promoted MYC mRNA stability and translation efficiency via ac4C modification, and the ac4C peak within the 3’UTR region was responsible for mRNA stability.
+
+Considering the important expression-regulating role of NAT10 on MYC, we investigate whether NAT10 modulates anti-tumor immunity via MYC. Firstly, CRISPR/Cas9 technology utilizing MYC-specific sgRNA (sgMYC) pairs was employed to knockout MYC in TC1 (Fig. S5B). To evaluate whether intrinsic MYC deficiency inhibits tumor growth by triggering an immune response, we established syngeneic tumor models in C57BL/6N mice transplanted with either WT or sgMYC cancer cells. The results showed that sgMYC TC-1 tumors exhibited a significant reduction in tumor growth as compared with their WT parental cells (Fig. S5C). At day 10 after subcutaneous transplantation, a considerably higher percentage of CD8+ T cells were observed in sgMYC tumors than in WT tumors (Fig. S5D), indicating that adaptive immunity might be involved in MYC-deficient induced tumor reduction. The MYC protein restored in sgNAT10 TC1 cells was significantly abolished NAT10-deficient induced tumor regression (Fig. S5E, S5F). The elevated IFN-γ secretion induced by NAT10 deficiency in vivo were also significantly abolished in Myc-overexpressed sgNAT10 cells (Fig. S5G). These data suggest that NAT10 might modulate anti-tumor immunity via regulating MYC expression.
+
+NAT10 depletion induces dsRNA-mediated RIG-I-dependent IFN-I signaling via Myc/CDK2/DNMT1 pathway
+
+As shown above that NAT10 enhanced mRNA stability and translation efficiency of MYC, we then aimed to elucidate IFN-I signaling induced by NAT10 inhibition. Considering the ability of NAT10 to promote cell proliferation, we reanalyzed the RNA-seq data and revealed several differentially-expressed genes associated with proliferation, in which CDK2, a member of the cyclin-dependent kinases family [24], was the most significantly down-regulated in NAT10 deficient cells (Fig. 6A, S6A). It has been reported that MYC could directly regulate CDK2 expression [25], and our western blot results also showed that MYC deletion significantly inhibited the expression of CDK2 in cancer cells (Fig. S6B). Moreover, our results showed that knocking out CDK2 (sgCDK2) led to the inhibition of tumor growth (Fig. S6D), consistent with the effect of siNAT10. Next, we explored the effects of CDK2 on antitumor immune response. The results showed that CDK2 deficiency in cancer cell elevated CD8+ T cells infiltration and IFN-γ expression, which is consistent with the phenomenon caused by NAT10 deletion (Fig. S6E, S6F). These findings suggest that NAT10 deficiency might enhance anti-tumor immunity via Myc-mediated regulation of CDK2 expression.
+
+How does CDK2 deletion induce IFN-I responses? CDK2-deficient cells have been proven to inhibit the activity of DNMT, and loss of its activity can induce IFN-I responses by increasing production of dsRNA [26]. We then reanalyzed our RNA-seq data and found that DNMT1 has the highest expression in cancer cells (Fig. 6B, S6C). Western blot analysis also showed that the protein levels of CDK2 and DNMT1 were significantly reduced in NAT10-deficient cells compared to WT cells (Fig. 6C). Furthermore, DNMT1 expression was restored by overexpressing the CDK2 in NAT10 deficient cells (Fig. S6G). More importantly, overexpression of CDK2 in sgNAT10 cells could promote the development of tumors (Fig. S6G). Correlation analysis between NAT10 and several downstream genes performed with the GEPIA website [34] revealed statistically positive correlations between NAT10 and MYC, CDK2, DNMT1 (Fig. S6H). These data suggest the critical role of NAT10 in maintaining the expression of CDK2 and DNMT1 through MYC.
+
+Increased IFN-I response in cancer cells has been shown to occur in response to DNA demethylation caused by 5-azacytidine, which inhibits the activity of DNMT1 [27]. DNMT1 inhibition could trigger IFN-I response by inducing dsRNA. In our study, quantification of dsRNA performed by immunofluorescence using the dsRNA-specific J2 antibody showed a significantly higher abundance of dsRNA within sgNAT10 and sgCDK2 cells than those within WT cells. Restored MYC and CDK2 significantly abolished NAT10 deficiency-induced dsRNA production (Fig. 6D, 6E). It has been reported that dsRNA could be sensed by RIG-I and MDA-5, which triggers IFN-I response [28]. Therefore, our next objective was to investigate whether NAT10 deletion-induced dsRNA production predominantly activates IFN via the RIG-I or MDA-5 signaling pathway. GSEA enrichment analyzed by RNA-seq data showed that “RIG-I like receptor signaling pathway” were upregulated in sgNAT10 cancer cells compared to WT cells (Fig. S6I, S6J). Therefore, we silenced RIG-I in sgNAT10 TC1 cells and assessed the functionality of the IFN-I signaling pathway. Our results showed that deletion of RIG-I partially negated the elevated expressions of IFN stimulated genes (ISGs) induced by NAT10 deletion (Fig. 6F) [29]. Together, our results demonstrate that NAT10 modulates the IFN-I signaling pathway via RIG-I-mediated dsRNA sensing.
+
+Inhibition of NAT10 with Remodelin enhances response to ICIs therapy
+
+The above data indicate that NAT10 deletion enhances intratumoral IFN-I production and T cell infiltration, two biomarkers associated with sensitivity to ICIs therapy. Previous studies also reported that activating the IFN-I pathway and enhancing T cell infiltration could promote the therapeutic effect of ICIs therapy [30]. Therefore, we next investigated cooperation between NAT10 inhibitor and PD-1 treatment using syngeneic tumor models. Mice were gavaged with Remodelin once a day for 7 consecutive days. On day 7, we treated mice with isotype control (vehicle), or anti-PD-1 mAb (10 mg/kg, intraperitoneally (ip), twice a week for 2 weeks), and a humane endpoint was reached in a vehicle group mouse on day 29 (Fig. 7A). The results showed that either PD-1mAb or Remodelin effectively inhibited tumor growth compared to control group. Importantly, the tumor size of the combined treatment was much smaller than either of the other two groups (Fig. 7B, 7C), suggesting combining inhibition of NAT10 and PD-1 synergistically suppresses cancer growth in vivo.
+
+Furthermore, we investigated the immunological changes and our results present with a significantly increased number of tumor-infiltrating CD8+ T cells after single treatment of Remodelin or anti-PD-1 mAb compared to control, while the combination group has the much higher CD8+ T cells infiltration (Fig. 7D, 7E). Moreover, a remarkable increase in the number of IFN-γ-positive active CD8+ T cells was seen in the Remodelin single treatment group; the effects were significantly enhanced by the combination treatment (Fig. 7F, 7G). Importantly, the combination treatment secreted more IFN-γ in the tumor microenvironment (Fig. 7H, 7I). Considering the clinical setting, we further detected and analyzed the relationship of NAT10 and PD-L1 in lung cancer samples. Our results demonstrate a positive correlation between the expression levels of PD-L1 and NAT10 (Fig. S7A). Overall, these data suggest that inhibition of NAT10 enhances the efficacy of PD-1 blockade therapy in suppressing tumor growth.
+
+Intratumoral delivery of siNAT10-lipid nanoparticles (LNPs) for cancer immunotherapy
+
+Considering the limited absorption of Remodelin that could mitigate its therapeutic effect on tumors, we developed two commonly-used delivery systems, SM102 and PEI/PC7A nanoparticles, to enhance inhibitory efficiency of NAT10 expression both in vivo and in vitro. SM102, a cationic amino lipid approved for mRNA delivery in the Moderna COVID-19 vaccine, also functions as an ionizable component in LNPs for RNAi-based therapeutics [31]. Additionally, PEI/PC7A nanoparticle, composed of polyethyleneimine (PEI) and a pH-responsive PC7A polymer, is developed for efficient siRNA transfection [32]. The particle size of siNAT10 was determined using dynamic light scattering (DLS, Malvern) and confirmed to be approximately 160 nm (Fig. 8A, S8A). Confocal laser scanning microscopy (CLSM) analysis demonstrated the overlap of fluorescence signals representing lysosomes (red fluorescence) with siRNAs (green fluorescence) within 4 hours. Moreover, a significant amount of green fluorescence was observed outside the lysosomes, indicating the escape of siRNA from the lysosomes (Fig. 8B, S8B). Successful release of siRNA from endosomes and lysosomes indicated the formation of the RNA-induced silencing complex in the cytosol.
+
+RT-qPCR analysis was applied to assess the inhibitory efficiency of nanoparticles on NAT10 expression. The results demonstrated a significant reduction in NAT10 mRNA expression levels with both SM102 and PEI/PC7A/siNAT10 nanoparticles compared to only siNAT10 transfection (Fig. 8C, S8C). Furthermore, our in vivo experiments revealed that PEI/PC7A/siNAT10 has more effective inhibition on tumor growth than SM102 (Fig. S8D, S8E). Consequently, we employed PEI/PC7A/siNAT10 nanoparticles to evaluate its tumor inhibitory effect for the following study.
+
+Moreover, western blotting results further confirmed significant NAT10 protein expression suppression by PEI/PC7A/siNAT10 nanoparticles (Fig. 8D). And intratumoral delivery of PEI/PC7A/siNAT10 nanoparticles treatment significantly reduced TC1 tumor growth in C57/BL6N mice, showing much more superior efficacy compared to Remodelin (Fig. 8E). Subsequently, we combined PEI/PC7A/siNAT10 nanoparticles with ICIs therapy to enhance the effect of inhibiting tumors. Our results demonstrated the effective tumor growth inhibition with both PD-1mAb and PEI/PC7A/siNAT10 nanoparticles, with the combined treatment resulting in much smaller tumor sizes compared to other groups (Fig. 8F). Furthermore, we observed a significant increase in tumor-infiltrating CD8+ T cells following PEI/PC7A/siNAT10 nanoparticle treatment or combination therapy, with the combination group exhibiting substantially higher CD8+ T cell levels (Fig. 8G). Additionally, a notable rise in IFN-γ-positive active CD8+ T cells was observed in the PEI/PC7A/siNAT10 nanoparticle single treatment group, with significantly enhanced effects noted in the combination therapy group (Fig. 8H). Overall, these findings suggest that NAT10 suppression by nanoparticles enhances the therapeutic effects of ICIs in controlling tumor growth.
+
+# Discussion
+
+mRNA ac4C writer NAT10 may reshape the tumor immune microenvironment. Our results reveal that targeting NAT10 not only controls tumor growth but stimulates an antitumor immune response to achieve maximal therapeutic effects. This strategy, when combined with an ICI, could approach a cure. Our approach specifically targets NAT10 to ablate tumors by downregulating mRNA stability and translation efficiency. Additionally, targeting NAT10 induces ERV-mediated dsRNA, thereby cross-priming T cell activation through a type I IFN response. Moreover, we identify that the PEI/PC7A/siRNA nanoparticles as a potent inhibitor of NAT10 expression promote robust anti-tumor activity (Fig. <span class="InternalRef" refid="Fig9">9</span>).
+
+NAT10, the only known ac4C “writer” protein and a predominantly nuclear protein, is characterized by a unique RNA cytosine acetyltransferase domain [<span citationid="CR33" class="CitationRef">33</span>]. In this study, we used an acRIP-seq method to profile the changes in global mRNA acetylation modification patterns on NAT10 deletion and identified undefined typical GAGGAGA motifs. Specifically, acetylation modification at mRNA 3’UTR enables the Myc stability, promoting translation. Myc methylation at 3’UTR may allow upregulate CDK2-DNMT1 to prevent the ERV associated with the role of dsRNA structure. Indeed, NAT10 or CDK2 inhibition in cancer cells promoted dsRNA formation which underlies RIG-I activation in cancer cells.
+
+Our study demonstrates that tumor elimination induced by NAT10 deletion relies on IFN-I responses. Multichannel imaging and Image analysis revealed that the changes in T cell subpopulations seen after NAT10 deletion are associated with immunological memory, which effectively protected the host from challenge with the corresponding WT cancer cells. Current treatment options used high-dose cytotoxic chemotherapies that dampen immune responses. Interestingly, we found that neither NAT10 deletion nor inhibitor treatment perturbed T cell function.
+
+How does NAT10 inhibition in cancer cells elicit a distinct response in T cells? In this study, we showed that RIG-I-dependent dsRNA sensing by cancer cells is critical for the effects of T cell priming. Notably, cancer cells express higher levels of NAT10 relative to normal counterparts from healthy donors. On NAT10 deficiency, cancer cells accumulate cytosolic dsRNA, providing abundant substrate for RIG-I signaling. Such changes in dsRNA are partially due to DNA demethylation induced by the loss of DNMT1 seen after NAT10 inhibition. Interestingly, GSEA of RNA-seq from NAT10 deficiency versus control cancer cells showed significant downregulation of CDK2 and DNMT1, confirming an association between NAT10 inhibition and DNA demethylation.
+
+Additionally, effective inhibition of NAT10 <em>in vivo</em> can enhance the tumor immune response. Remodelin hydrobromide, an orally active and selective NAT10 inhibitor, has been identified as such [<span citationid="CR34" class="CitationRef">34</span>]. Upon intragastric administration, Remodelin inhibits tumor growth by activating host immunity. However, its oral bioavailability is low, resulting in limited efficacy against tumors. Lipid nanoparticles (LNPs) have gained clinical approval as carriers for siRNA and mRNA. Among LNPs' critical components, ionizable lipids are pivotal in determining RNA delivery efficiency. We developed two delivery systems: SM-102 and PEI/PC7A/siNAT10 nanoparticles. Our findings indicate that PEI/PC7A/siNAT10 effectively penetrates cell membranes, inhibits NAT10 expression, and suppresses tumor growth. Notably, PEI/PC7A/siNAT10 outperforms SM-102 and Remodelin. Combined with ICIs therapy, PEI/PC7A/siNAT10 stimulates potent anti-tumor immunity, effectively suppressing tumor growth.
+
+Collectively, we demonstrate a biological role for NAT10 in cancer. We develop a PEI/PC7A/siNAT10 nanoparticles blocking NAT10 activity <em>in vitro</em> and <em>in vivo</em>. Our study also prompts an appraisal of anticancer drugs with consideration of their impact on immune cells within the tumor microenvironment and provides a rationale for further evaluation of NAT10 inhibition combined with a PD-1/PD-L1 inhibitor against “cold” tumor.
+
+# Materials and methods
+
+## Analysis of tumor-infiltrating immune and prognostic model
+We analyzed the patterns of immune cell infiltration according to the immune cell biomarker previously reported [18]. The algorithm was operated with the R-package and the data were visualized with R package ggplot2. TCGA-LUAD (n = 1,082) were applied to illustrate the potential prognostic significance of NAT10. According to the expression of NAT10, patients were divided into high- or low-expression group. The CIBERSORT algorithms was used to calculate the proportion of immune cell infiltration of different groups. The diagram was drawn by using the ggplot2 package. Additionally, survival analysis was performed using R 'survival' package. The ggplot2 and survminer packages were used to create survival curves between different groups. In addition, the ROC curve was generated using the R package survival ROC to detect the prognostic value for NAT10 expression.
+
+## Cell culture
+The MCA205 murine fibrosarcoma, TC1 murine lung epithelial, and HEK293 human embryonic kidney cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin/streptomycin (Gibco). Cultures were maintained in a humidified atmosphere containing 5% CO₂.
+
+## Construction of stable cell lines with CRISPR/Cas9 system
+Deletion of NAT10, MYC, RIG-I or CDK2 was achieved using LentiCRISPR v2 (Addgene, Cambridge, MA, USA), which carries expression cassettes for Streptococcus pyogenes CRISPR-Cas9 and a chimeric guide RNA selected from the Guide Design Resources at http://crispr.mit.edu. HEK293 cells were co-transfected with three plasmids: pMD2.G (Addgene, cat#12259), psPAX2 (Addgene, cat#12260), and either LentiCRISPR v2 or a control vector, using Lipofectamine 3000 (Thermo Fisher Scientific) for 48 hours. Viral stocks generated were used to infect target cells. Post-infection, cells were cultured in puromycin (4 µg/ml, InvivoGen, cat#ant-pr-1) for at least seven days. Monoclonal cells obtained using FACSAria™ III cell sorter (Becton Dickinson, San José, CA, USA) were plated in a 96-well plate. Sequences synthesized in this study are provided in Supplementary Table S1. NAT10 heterozygous knockout, MYC, CDK2 and RIG-I knockout, and control cell lines were further validated through Western blot analysis of NAT10 expression.
+
+## Overexpression vectors and transfection
+To achieve overexpression of NAT10, MYC-HA, and CDK2-GFP, the dCAS9-VP64-GFP plasmid (Addgene, cat#61422) was digested with BamH I (NEB, cat#R0136S) and Nhe I (NEB, catalog no. R0131), and the VP64 sequence was replaced with the cDNA sequences corresponding to the genes of interest. Subsequently, 293T cells were transfected with the dCAS9-VP64-GFP plasmid, along with packaging plasmids psPAX2 (Addgene, plasmid cat#12260) and envelope pMD2.G (Addgene, cat#12259), using Lipofectamine 3000 (Invitrogen, catalog no. L3000-015) according to the manufacturer’s instructions. After 48 hours, lentivirus was harvested from the cell culture medium, followed by collection of lentiviral particles via centrifugation (5,000 rpm/10 minutes) and filtration through a 0.45 µm sterile filter (Merck Millipore Ltd. PR05543). The collected lentivirus was stored at -80°C. Transfected cells were subsequently isolated using fluorescence-activated cell sorting (FACS).
+
+## Tumor models
+C57BL/6J background mice and athymic nude BALB/c mice (nu/nu) aged 6–8 weeks, sourced from Vital River Laboratory Animal Technology Company (Beijing, China), were housed under specific pathogen-free (SPF) conditions at the Laboratory Animal Center of Shandong University. Following grouping, cells (2×10⁶ cells per mouse) were subcutaneously implanted. Tumor dimensions were measured daily using vernier calipers, and tumor size was calculated by multiplying the length by the width. After 9–13 days, tumors were harvested for RNA sequencing, flow cytometric analysis, tissue immunofluorescence staining, and ELISpot analysis. Survival analysis involved intravenous injection of cancer cells, was conducted with daily recording of mouse mortality. In the CD8⁺ cell blocking assay, anti-CD8 antibodies (200 µg/mouse, BE0004-1, BioXCell) were intravenously injected at the specified time point. Tumor growth curves were presented with error bars representing mean ± SD at each time point. Kaplan–Meier survival curves were generated. Animals were euthanized with CO₂ when tumor volume reached 300 mm².
+
+## Multichannel imaging and Image analysis
+Multichannel imaging was conducted using a Vectra Polaris Imaging System (Akoya Biosciences). Slides were captured at 200× magnification. Image analysis was performed using QuPath version 0.4.3 (Queen’s University) [35]. Tissue sections were divided into tumor and stroma regions based on pan-CK staining. Cell segmentation employed an algorithm based on nuclear DAPI staining. Fluorescence intensity of cells was quantified for each marker. Cells were classified into distinct phenotypic classes using positivity thresholds for individual markers, determined by cytoplasmic or nuclear staining intensity, and evaluated across all samples. Cell count, density, and percentage in different regions were calculated for each phenotype.
+
+## RNA sequencing and data analysis
+Cells and tumor tissues were lysed directly after grinding, and total RNA extraction was carried out using the RNeasy Mini Kit (QIAGEN, cat#74104). Six hundred nanograms of total RNA were reversely transcribed into cDNA using ProtoScript II Reverse Transcriptase (New England BioLabs, cat#E7420L). The resulted double-stranded cDNA was purified with Agencourt AMPure XP Beads (Beckman, cat#A63881) and then ligated with paired-end adaptors using Multiplex Oligos for RNA sequencing. Sequencing was conducted on an Illumina HiSeq 10X platform, and data analysis was performed using the Linux system. Differentially expressed genes (DEGs) were identified using the R language, including the “edgeR” and “gplots” packages.
+
+## Gene set enrichment analysis (GSEA)
+GSEA analysis was conducted using GSEA 4.1.0 software following the guidelines provided on the official website. The complete normalized RNA expression count matrix, including all genes rather than just differentially expressed ones, was utilized as input. The matrix was partitioned into two groups: (1) KO-High group; (2) WT-Low group. Hallmarks were chosen from the gene sets database, and 1,000 permutations were performed based on default weighted enrichment statistics.
+
+## RNA extraction and RT-qPCR
+Following the manufacturer's protocol, cell pellets were collected and subjected to total RNA extraction using NucleoZol (MNG, Cat#740404.200). The extracted RNA was then reversely transcribed into cDNA using the One Step PrimeScript RT-PCR kit (TaKaRa, Cat#36110A) for subsequent qPCR analysis. Gene-specific primers listed in Table S2 and SYBR Green qPCR mix (Bimake. cn, Cat#B21202) were employed for PCR amplification and detection on the Light Cycler Real-Time PCR System (Roche). RT-qPCR data were normalized to GAPDH and presented as fold changes in gene expression relative to the control sample.
+
+## Protein extraction and Western blot analysis
+Cells, tumor tissues, or paired adjacent tissues were collected and lysed using the lysis buffer from Bestbio Company (China). Protein concentration was determined with the BCA kit (Beyotime, cat# P0011). Equal amounts of protein from each sample were loaded onto SDS-PAGE gels and subsequently transferred to PVDF membranes. The PVDF membranes were then blocked in 5% non-fat milk for 1 hour at room temperature. After being washed with PBST, the membranes were incubated with the primary antibodies as follow: anti-NAT10 (1:1000, Abcam, cat#ab194297 ), anti-Myc (1:1000, CST, cat#18583), anti-CDK2 (1:1000, CST, cat#2546), anti-DNMT1 (1:1000, CST, cat#5032), anti-RIG-I (1:1000, CST, cat#3743), anti-HA-tag (1:1000, CST, cat#3724), anti-Tubulin (1:1000, CST, cat#2146), anti-β-Actin (1:1000, CST, cat#4970). On the next day, the members were washed with TBST three times and incubated with anti-rabbit IgG, HRP-linked antibody (1:2000, CST, cat#7074) at room temperature for 50 minutes. Membranes were imaged using the ChemiDoc XRS + system (Bio-Rad, USA).
+
+## CCK8 and colony formation assay
+After chemical inhibition of NAT10 by Remodelin (10 µM or 20 µM, MCE, cat#HY-16706A ) or genomic depletion, the proliferation assays of cancer cells were detected by CCK8 and colony formation assays. In brief, 10000 cancer cells were seeded. And the CCK8 detection reagent was added in 96-well plates. After 4 hours, the absorbance was detected by a microplate reader (Biotek, HIMFD, USA) according to the manufacturers’ instructions (Bestbio Company, China). For the colony formation assays, 2000 cells were plated in the six-well plates. Severn days later, colonies were fixed with 4% PFA, stained with a 0.5% crystal violet staining solution (Beyotime Company, China) for 30 min and counted with microscopy.
+
+### Flow cytometry
+Mice were sacrificed at appropriate time, and the tumors were collected and separated into single cells. Briefly, tumors were excised and minced. Then Liberase TL Research Grade 10 (2 µg/mL, Roche, cat#05401020001) and DNase I (Roche, cat#70271500) was used for the digestion. After being filtered with strainer, the cell suspensions were stimulated with brefeldin A (BFA, PeproTech, 10 mg/mL), phorbol myristate acetate (PMA, PeproTech, 100 µg/mL) and ionomycin (PeproTech, 1mg/mL) at 37℃for 4 h. For the cell surface staining, cells were stained for cell markers including cell death dye (1:300, Invitrogen, eBioscience™ Fixable Viability Dye eFluor™ 780, cat#2633409), CD45.2 (1:100, BioLegend, cat#109814), CD8a (1:100, BioLegend, cat#B373965), CD11c (1:100, Invitrogen, cat#2400633), and IA/IE (1:100, BioLegend, cat#107608) at 4℃ for 30 min. As for intracellular staining, cells were stained with IFN-γ (1:100, Invitrogen, cat#2481435) and Granzyme B (1:100, BioLegend, cat#515406) after being treated with fixation/permeabilization kit at 4℃ for 30 min. Cells were analyzed using a Gallios flow cytometer (Beckman Coulter, USA) and the results were analyzed by Flowjo.
+
+## Immunohistochemical (IHC) analysis
+The pathology sections of patients was obtained from Qilu Hospital of Shandong University. After dewaxing, dehydration, and antigen retrieval, paraffin-embedded slides (4 µm) were blacked and labeled with anti-NAT10 (1:250, Abcam, cat#ab182744), anti-CD8a (1:250, Abcam, cat#ab182744), and anti-PD-L1 (1:250, Abcam, cat#ab213524) antibody. The next day, the slides were incubated with the secondary antibody labeled with HRP (Shanghai Gene Company, cat#GK500705) for 1 hour, stained with DAB and counterstained with hematoxylin. The images were detected by a microscopy.
+
+### Immunofluorescence staining and imaging
+Tumors were collected at the appropriate time and fixed in 4% paraformaldehyde for 24 hours. Subsequently, the tumors were embedded in OCT after dehydration in a 30% (wt/vol) sucrose solution. Following sectioning into 4.5 mm thick slices, the sections were blocked in 10% goat serum in PBS. Primary antibodies against CD8a (1:100, Abcam, cat#ab217344) were then used to incubate the tumor sections. The next day, secondary antibodies (1:500, Invitrogen, cat#A32732) were applied to the sections for 1 hour. Nuclei were stained with DAPI, and the results were visualized using a confocal microscope and analyzed with ImageJ software. dsRNA was detected by the J2 antibody [1:250 dilution, 4 mg/ml, English and Scientific Consulting Kft (SCICONS), cat#10010200].
+
+### T-cell proliferation assay
+Freshly purified splenocytes were isolated from C57BL/6N mice. Splenocytes were labeled with cell proliferation Dye eFluor™ 670 (eBioscience, cat#65-0840-85) at 5 µM for 10 min at 37°C, and then resuspended in the RPMI media containing 10% FBS, 1% penicillin/streptomycin, 0.5 µg/ml purified anti-mouse CD3 Antibody (Biolegend, cat#100238) and 0.5 µg/ml anti-mouse CD28 Antibody (Biolegend, cat#102112). 3^10⁵ purified splenocytes were then co-cultured with 1^10⁴ WT or NAT10 deficient TC1/MCA205 cancer cells in 96-well round-bottom plates. The unstimulated splenocytes were used as a negative control, and those stimulated with CD3 and CD28 antibodies were used as a positive control. After 72 h, cells were collected and stained, and the dilution of cell proliferation Dye eFluor™ 670 in CD4⁺ T (Biolegend, cat#100428, 1:100) or CD8⁺ T (Biolegend, cat#100708, 1:100) cells was determined by flow-cytometric analysis.
+
+### Enzyme-linked immune spot (ELISpot) assay
+IFN-γ secretion was assessed using BD ELISpot assay kits (BD Biosciences, cat#551881) according to the manufacturer's instructions. Tumors were aseptically harvested and processed into a single-cell suspension. Cells were then plated at a density of 2x10⁶ per well in ELISpot plates precoated with capture antibodies and incubated in a humidified 5% CO₂ incubator at 37°C for 20 hours. After incubation, cells were removed, and the plate was washed three times. IFN-γ production was detected by incubating with a detection antibody for 2 hours, followed by three washes and incubation with an HRP-linked secondary antibody for 2 hours. Color development was achieved by adding 100 µL of Final Substrate Solution (AEC). Red dot signals were visualized using the CTL ImmunoSpot® S6 Analyzers (LLC, OH, USA).
+
+## Dual-luciferase reporter assay
+The promoter activity of NAT10 in TC1 cells was assessed using a luciferase assay. In brief, pEZX-MT06-MYC-WT-Luc or pEZX-MT06-MYC–Mut-Luc were cloned into pEZX-MT06 Reporter Vector pGL4.0 (GeneCopoeia, cat#NM_001177354.1). Nat10 coding DNA sequence was cloned into dCAS9-VP64-GFP (Addgene, cat#61422). TC1 WT cells were seeded at a density of 2^10⁵ cells per well in 24-well plates and incubated overnight prior to transfection. Subsequently, cells were co-transfected with pEZX-MT06-MYC, Renilla luciferase plasmids, and either VP64-NAT10 plasmids or empty plasmids using Lipofectamine 3000 (Invitrogen, cat#L3000-015). After 24 hours, firefly luciferase and Renilla luciferase activities were assessed using the Dual-Luciferase reporter system (Promega, cat#E1960), and efficacy was determined by calculating the ratio of firefly luciferase to Renilla luciferase activity.
+
+## Acetylated RNA Immunoprecipitation Sequencing (acRIP-seq) and acRIP-qPCR
+acRIP-seq and data analysis was conducted by Guangzhou Epibiotek Co., Ltd. The WT and sgNAT10 TC1 cells were subjected to acRIP-seq. Total RNA was extracted and purified from WT and sgNAT10 TC1 cells using TRIzol reagent (Invitrogen). One hundred micrograms of total RNA was fragmented into 100–200 nt RNA fragments using 10X RNA Fragmentation Buffer (100 mm Tris-HCl, 100 mm ZnCl₂ in nuclease-free H₂O), followed by termination of the reaction with 10XEDTA. Immunoprecipitated RNA fragments were obtained by incubating fragmented RNA with anti-ac4C monoclonal antibody for 3 h at 4°C, followed by incubation with protein A/G magnetic beads (Invitrogen, Cat#8880210002D/10004D) for 2 h at 4°C, as per the EpiTM ac4C immunoprecipitation kit protocol (Epibiotek, R1815). The library was prepared using the smart-seq method. Both the input samples without IP and the ac4C IP samples were subjected to 150-bp, paired-end sequencing on an Illumina NovaSeq 6000 sequencer.
+
+The RIP-qPCR assay was conducted to confirm the interaction between NAT10 and Myc mRNA using the RIP Kit (BersinBio, Cat# Bes5101). TC1 cells were lysed using a polysome lysis buffer containing protease and RNase inhibitors. DNase was added to degrade the DNA at 37°C for 10 min. NAT10 or IgG antibodies were added to the samples and incubated at 4°C for 16 h in a vertical mixer. Subsequently, the samples were incubated with protein A/G beads for 1 h. Following the manufacturer’s instructions, the beads containing the immunoprecipitated RNA-protein complex were treated with proteinase K to remove proteins. The target RNAs were then extracted using the phenol-chloroform method, amplified by PCR (Table S3), and detected using DNA gel electrophoresis with normalization to their input group.
+
+### mRNA Stability Assay
+TC1 cells were cultured in complete DMEM medium supplemented with 5 µg/ml actinomycin D (Sigma, Cat#A9415) for 0, 4, 8, 12, 16, and 24 hours. At the specified time points, cells were harvested, and total RNA was extracted following the protocol outlined in the "RNA extraction" section for subsequent real-time PCR analysis (Table S2).
+
+## Synthesis of SM-102
+NAT10 siRNA and FAM-labeled siRNA-NC were synthesized by Atantares. Ionizable lipids SM-102 (Cat#O02010), 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC, Cat#S01005) and 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2000, Cat#O02005) were purchased from AVT (Shanghai) Pharmaceutical Tech Co., Ltd. Cholesterol (Cat#A90286) was purchased from Innochem. Polyetherimide (Cat#61128-46-9) was purchased from MACKLIN. 2-(azepan-1-yl) ethanol (Cat#35984E), Methacryloyl chloride (Cat#90100B) and 2-Bromoisobutyryl Bromide were purchased from Adamas.
+
+## Synthesis of PC7A
+For synthesis of 2-(Hexamethyleneimino) ethyl Methacrylate Monomer (C7A-MA), 2-(azepan-1-yl) ethanol (5.0 g) and triethylamine (7.0 g) were dissolved in 100 mL of dried tetrahydrofuran and cooled to 0°C in an ice bath. Methacryloyl chloride (4.0 g) was dissolved in 15 mL THF and subsequently dropped into previous solution. This reaction conducted at room temperature under stirring for 8 h. For synthesis of PC7A polymer, 0.5 g C7A-MA, 8.5 g CuBr and 11.6 mg initiator were dissolved in 0.5 mL of dried THF. After undergoing three rounds of freeze−pump−thaw, 10.3 mg N,N,N′,N″,N″-pentamethyldiethylenetriamine was introduced. Subsequently, the polymerization process was conducted at a temperature of 70°C for a duration of 10 hours. The resulting reaction mixture was then dissolved in acidic water with a pH of 4 and dialyzed in distilled water, utilizing a cut-off molecular weight of 3500 Da, to eliminate any unreacted monomers and copper. Finally, the product was obtained through the process of lyophilization.
+
+### Preparation of Lipid Nanoparticles
+siRNA was encapsulated lipid nanoparticles (LNPs) as described [36]. Briefly, siRNAs were dissolved in sodium acetate (pH = 4) and combined with a lipid solution at an amine-to-phosphate (N/P) ratio of 8. The lipid stock solutions were prepared with a total lipid concentration of 12.5 mM by dissolving SM102, DSPC, cholesterol, and DMG-PEG-2000 in ethanol at a molar ratio of 50:10:38.5:1.5. Ultrafiltration centrifugation (3500G, 40 min) was used to remove unentrapped siRNA from the LNPs.
+
+### Preparation of PEI/PC7A
+PEI and PC7A were dissolved in sterile water. Subsequently, mix the PEI, PC7A and siRNA in a weight ratio of 1.3:1:1 to form nanoparticles through electrostatic interaction with the negatively charged siRNA. Leave the nanoparticle for 5 minutes before using.
+
+### Size distribution of nanoparticles
+Size distribution and PDI were measured by Malvern Nano Sizer (Malvern Instruments Ltd) in double-distilled water.
+
+### LNPs and PEI/PC7A transfection
+Cells were counted using trypan blue dye. For Real-time PCR and Western blotting, 2 x 10⁵ cells were placed in 12-well plates overnight. Replace the medium with Opti-MEM and add LNPs or PEI/PC7A with a final concentration of siRNA of 20 nM. For immunofluorescence, cells were cultured in chamber slides overnight, and then added with 20 nM FAM-labeled siRNA for 4 h. Cells were stained with 50 nM Lyso-Tracker Red (Beyotime, Cat#C1046) and 10 µg/mL Hoechst (Beyotime, Cat#C1022) for 30 min. Immunofluorescence images were acquired on a Nikon A1 fluorescence microscope.
+
+# References
+
+1. Li, B., H.L. Chan, and P. Chen, *Immune Checkpoint Inhibitors: Basics and Challenges*. Curr Med Chem, 2019. 26(17): p. 3009–3025.
+2. de Miguel, M. and E. Calvo, *Clinical Challenges of Immune Checkpoint Inhibitors*. Cancer Cell, 2020. 38(3): p. 326–333.
+3. Benci, J.L., et al., *Tumor Interferon Signaling Regulates a Multigenic Resistance Program to Immune Checkpoint Blockade*. Cell, 2016. 167(6): p. 1540–1554.e12.
+4. Chen, Y.G. and S. Hur, *Cellular origins of dsRNA, their recognition and consequences*. Nat Rev Mol Cell Biol, 2022. 23(4): p. 286–301.
+5. Huang, R.X. and P.K. Zhou, *DNA damage response signaling pathways and targets for radiotherapy sensitization in cancer*. Signal Transduct Target Ther, 2020. 5(1): p. 60.
+6. Cheng, B., et al., *Recent advances in DDR (DNA damage response) inhibitors for cancer therapy*. Eur J Med Chem, 2022. 230: p. 114109.
+7. Cañadas, I., et al., *Tumor innate immunity primed by specific interferon-stimulated endogenous retroviruses*. Nat Med, 2018. 24(8): p. 1143–1150.
+8. Dorrity, T.J., et al., *Long 3'UTRs predispose neurons to inflammation by promoting immunostimulatory double-stranded RNA formation*. Sci Immunol, 2023. 8(88): p. eadg2979.
+9. Chiappinelli, K.B., et al., *Inhibiting DNA Methylation Causes an Interferon Response in Cancer via dsRNA Including Endogenous Retroviruses*. Cell, 2015. 162(5): p. 974–86.
+10. Guo, E., et al., *WEE1 inhibition induces anti-tumor immunity by activating ERV and the dsRNA pathway*. J Exp Med, 2022. 219(1).
+11. Alicea-Torres, K., et al., *Immune suppressive activity of myeloid-derived suppressor cells in cancer requires inactivation of the type I interferon pathway*. Nat Commun, 2021. 12(1): p. 1717.
+12. Im, J.H., et al., *Mechanisms of length-dependent recognition of viral double-stranded RNA by RIG-I*. Sci Rep, 2023. 13(1): p. 6318.
+13. Wu, B., et al., *Structural basis for dsRNA recognition, filament formation, and antiviral signal activation by MDA5*. Cell, 2013. 152(1–2): p. 276–89.
+14. Dai, E., et al., *Epigenetic modulation of antitumor immunity for improved cancer immunotherapy*. Mol Cancer, 2021. 20(1): p. 171.
+15. Luo, J., et al., *Emerging role of RNA acetylation modification ac4C in diseases: Current advances and future challenges*. Biochem Pharmacol, 2023. 213: p. 115628.
+16. Tang, Z., et al., *GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis*. Nucleic Acids Res, 2019. 47(W1): p. W556-w560.
+17. Larrieu, D., et al., *Chemical inhibition of NAT10 corrects defects of laminopathic cells*. Science, 2014. 344(6183): p. 527–32.
+18. Chen, B., et al., *Profiling Tumor Infiltrating Immune Cells with CIBERSORT*. Methods Mol Biol, 2018. 1711: p. 243–259.
+19. Wu, F., et al., *Single-cell profiling of tumor heterogeneity and the microenvironment in advanced non-small cell lung cancer*. Nat Commun, 2021. 12(1): p. 2540.
+20. Singh, N., et al., *Inflammation and cancer*. Ann Afr Med, 2019. 18(3): p. 121–126.
+21. Tokunaga, R., et al., *CXCL9, CXCL10, CXCL11/CXCR3 axis for immune activation - A target for novel cancer therapy*. Cancer Treat Rev, 2018. 63: p. 40–47.
+22. Li, X., J. Peng, and C. Yi, *Acetylation Enhances mRNA Stability and Translation*. Biochemistry, 2019. 58(12): p. 1553–1554.
+23. Li, J., et al., *The effects of MYC on tumor immunity and immunotherapy*. Cell Death Discov, 2023. 9(1): p. 103.
+24. Tadesse, S., et al., *Cyclin-Dependent Kinase 2 Inhibitors in Cancer Therapy: An Update*. J Med Chem, 2019. 62(9): p. 4233–4251.
+25. Jian, Y., et al., *Actin-like protein 6A/MYC/CDK2 axis confers high proliferative activity in triple-negative breast cancer*. J Exp Clin Cancer Res, 2021. 40(1): p. 56.
+26. Liu, J., et al., *Effect of CDK4/6 Inhibitors on Tumor Immune Microenvironment*. Immunol Invest, 2024. 53(3): p. 437–449.
+27. Huang, K.C., et al., *DNMT1 constrains IFNβ-mediated anti-tumor immunity and PD-L1 expression to reduce the efficacy of radiotherapy and immunotherapy*. Oncoimmunology, 2021. 10(1): p. 1989790.
+28. Barral, P.M., et al., *Functions of the cytoplasmic RNA sensors RIG-I and MDA-5: key regulators of innate immunity*. Pharmacol Ther, 2009. 124(2): p. 219–34.
+29. Xu, L., et al., *RIG-I is a key antiviral interferon-stimulated gene against hepatitis E virus regardless of interferon production*. Hepatology, 2017. 65(6): p. 1823–1839.
+30. Lu, C., et al., *Type I interferon suppresses tumor growth through activating the STAT3-granzyme B pathway in tumor-infiltrating cytotoxic T lymphocytes*. J Immunother Cancer, 2019. 7(1): p. 157.
+31. Huang, C., et al., *Liver-Specific Ionizable Lipid Nanoparticles Mediated Efficient RNA Interference to Clear "Bad Cholesterol"*. Int J Nanomedicine, 2023. 18: p. 7785–7801.
+32. Ye, M., et al., *Double-Network Nanogel as a Nonviral Vector for DNA Delivery*. ACS Appl Mater Interfaces, 2019. 11(46): p. 42865–42872.
+33. Xie, L., et al., *Mechanisms of NAT10 as ac4C writer in diseases*. Mol Ther Nucleic Acids, 2023. 32: p. 359–368.
+34. Dalhat, M.H., et al., *Remodelin, a N-acetyltransferase 10 (NAT10) inhibitor, alters mitochondrial lipid metabolism in cancer cells*. J Cell Biochem, 2021. 122(12): p. 1936–1945.
+35. Bankhead, P., et al., *QuPath: Open source software for digital pathology image analysis*. Sci Rep, 2017. 7(1): p. 16878.
+36. Ball, R.L., et al., *Lipid Nanoparticle Formulations for Enhanced Co-delivery of siRNA and mRNA*. Nano Lett, 2018. 18(6): p. 3814–3822.
+
+# Supplementary Files
+
+- [supplementarymaterials.doc](https://assets-eu.researchsquare.com/files/rs-4352052/v1/7507158f4459fdf45c954421.doc)
+
+- [SupplementaryTable.doc](https://assets-eu.researchsquare.com/files/rs-4352052/v1/721404cabd336585d971d4ce.doc)
+  - Additional file 1: Table S1. Sequence of Guide RNA (sgRNA) oligonucleotides.
+  - Additional file 2: Table S2. Primers designed for real-time PCR.
+  - Additional file 3: Table S3. Primers designed for RIP-qPCR.
+  - Additional file 4: Table S4. NAT10 siRNA and the primers for RT-qPCR.
+
+- [FigS1.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/1027ab8f845cab3c941ddc1e.tif)
+  - Additional file 5: Figure S1. Elevated NAT10 expression is associated with decreased survival and reduced infiltration of immune cells in cancer.
+
+- [FigS2.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/f9bc53876c67867970984856.tif)
+  - Additional file 6: Figure S2. NAT10 deficiency suppresses tumor growth via immune-dependent mechanisms.
+
+- [FigS3.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/5340e6b2ae9079a368215641.tif)
+  - Additional file 7: Figure S3. NAT10 deficiency triggers immune-response signaling and induces cellular immune responses in vivo.
+
+- [FigS4.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/b246cfb3e4a16b764bd4f898.tif)
+  - Additional file 8: Figure S4. NAT10 deficiency triggers IFN-I responses in cancer cells.
+
+- [FigS5.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/4ac332ed26b209b9ac96602f.tif)
+  - Additional file 9: Figure S5. NAT10 regulates the stability and translation efficiency of MYC mRNA.
+
+- [FigS6.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/cd28fcbd8cd425e61f886868.tif)
+  - Additional file 10: Figure S6. Depletion of NAT10 induces dsRNA-mediated RIG-I-dependent signaling through the Myc/CDK2/DNMT1 pathway.
+
+- [FigS7.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/0e1c5b78fcbbd652f7223ea1.tif)
+  - Additional file 11: Figure S7. Expression of NAT10 is positively correlated with PD-L1 in human lung cancer.
+
+- [FigS8.tif](https://assets-eu.researchsquare.com/files/rs-4352052/v1/b8df032187557f5d211d5a37.tif)
+  - Additional file 12: Figure S8. Intratumoral delivery of siNAT10-lipid nanoparticles (LNPs) for cancer immunotherapy.