Abstract Lipid droplets (LDs), as innate immune hubs, function in the front line of antimicrobial defense involved in the host-pathogen arms race. Particularly for intracellular bacterial pathogens, the endowed capacity to resist host LDs can effectively facilitate pathogen in vivo colonization and evasion from the host’s innate immune response. Here, to investigate the genetic mechanisms of intracellular bacteria response to host LDs, we utilized transposon insertion sequencing to dissect critical fitness determinants of Edwardsiella piscicida under the treatment of LDs isolated from its native host, turbot. Targeted metabolomics indicated that LD challenge resulted in the accumulation of intracellular arginine. The core arginine metabolism regulatory factor, ArgR, was found to play a pivotal role in combating LDs, emphasizing the importance of orchestrating intracellular arginine levels for bacterial LD adaptation. Specifically, ArgR enhanced the expressions of genes involved in arginine catabolism (speA/B and arcC) and diminished gene transcripts associated with arginine import (artP) and synthesis (argD/E/H). Furthermore, ArgR contributed to the pathogenesis of E. piscicida, promoting the proliferation in host cells and virulence in turbot. Collectively, our results shed light on the underlying mechanism of intracellular pathogens resisting LDs during bacterial infections and highlighting the crucial role of arginine in the host-pathogen interactions. Subject terms: Pathogens, Bacterial infection __________________________________________________________________ The arginine metabolism regulatory factor, ArgR, contributes Edwardsiella piscicida to combating Lipid droplet-mediated bactericidal effects, emphasizing the importance of orchestrating intracellular arginine levels in the host-pathogen interactions. Introduction Lipid droplets (LDs) are phospholipid-monolayer-bound organelles found in various eukaryotic cells, primarily responsible for lipid storage and maintenance of the cellular membrane system stability^[34]1. Despite being composed mainly of inert neutral lipids, the versatile proteins accommodated on the LDs surface endow LDs to directly participate in innate immune response^[35]2. Upon infection with bacteria, viruses, or parasites, LD formation and accumulation are induced, constituting the front line in the host-pathogen arms race. For instance, viperin located on LDs exerts antiviral activity against hepatitis C virus and dengue virus^[36]3; histones on LDs facilitate Drosophila embryos to resist bacterial infections^[37]4; LDs contribute to cross-presentation of exogenous antigens^[38]5,[39]6. Especially, LDs with anti-bacterial activity, referred to as defensive-LDs, have been recognized as critical elements of innate immune mechanisms^[40]7. LDs are vital for both pathogens and hosts involved in the dynamic process of bacterial infection^[41]8. For intracellular pathogens, LDs serve as a valuable nutritional and energy source to contribute to their resilience against stressful intracellular environments^[42]9,[43]10. Moreover, intracellular pathogens can manipulate LD generation within host cells to promote their proliferation. For instance, Salmonella utilizes its type III secretion system (T3SS) effector, SseJ, to esterify cholesterol and increase LD production, thus stabilizing Salmonella-containing vacuoles and aiding in vivo survival^[44]11. Furthermore, S. Typhimurium-induced LD formation was rapid and time-dependent with the participation of T3SS to promote bacterial replication^[45]12. Burkholderia pseudomallei induced LD accumulation to interfere with macroautophagic/autophagic flux, which blocked autophagy-dependent inhibition of infection^[46]13. Simultaneously, LDs also actively participate in the innate immune response to support host antimicrobial defense. Emerging evidence has revealed that LDs function as key signaling platforms where viperin recruits IRAK1 and TRAF6 to amplify the IFN-mediated immune response^[47]14. Remarkably, antimicrobial proteins would be recruited and augmented on the LD surface to directly kill intracellular pathogens. In mammalian cells, the synthesis of cationic antimicrobial peptides (CAMPs) was induced by LPS treatment, which were then loaded on LDs to inhibit the proliferation of bacterial pathogens^[48]15,[49]16. To succeed in escaping host immune clearance, intracellular pathogens have to evolve the ability to counteract LDs, however, the mechanisms by which pathogens directly resist antibacterial substances on LDs are not yet fully elucidated. Edwardsiella piscicida, a Gram-negative bacterium belonging to the Enterobacteriaceae family, can infect various commercially important aquatic species, including turbot, sea bass, and eel, posing a significant potential threat to the global aquaculture industry^[50]17–[51]19. As a typical intracellular pathogen, E. piscicida leverages the effectors of T3SS and type VI secretion systems (T6SS) to hijack host cells and interrupt immune response, establishing colonization within host epithelial cells and macrophages^[52]20,[53]21. Additionally, E. piscicida has evolved specific cell wall modification mechanisms for combating environmental stress and resisting antimicrobial substances^[54]22,[55]23. In this study, transposon insertion sequencing (Tn-seq) was employed to investigate the key fitness determinants governing the growth of E. piscicida upon encountering the turbot hepatic LDs. The presence of LDs resulted in the excessive accumulation of intracellular arginines, which was detrimental to bacterial resistance against the killing effects of LDs. The transcriptional regulator ArgR coordinated the expression of arginine metabolism-associated genes as well as the intracellular level of arginine, contributing to E. piscicida’s adaptation to host LDs. These findings highlight the pivotal role of arginine metabolism in LD-mediated host-pathogen interaction and further expand our understanding of the lifestyle of intracellular pathogens within host cells, which could be exploited as the antimicrobial strategy. Results Intracellular E. piscicida resists LD-mediated bactericidal effects When intracellular pathogenic bacteria enter host cells through phagocytic vesicles either actively or passively, LDs armed with antimicrobial proteins can attract and directly contact the pathogenic bacteria to kill them, facilitating innate immunity^[56]15. Proteomic analysis identified several antimicrobial proteins, such as histones, on turbot liver LDs following stimulation by inactivated E. piscicida, supporting that LDs function as immune response organelles (Supplementary Fig. S[57]1). To evaluate the resistance of E. piscicida against LDs, its growth curve was measured after the LD challenge. Hepatic LDs were extracted from turbot, which had been intraperitoneally injected with heat-inactivated bacteria as an immunity stimulus. The intracellular pathogen Salmonella Typhimurium SL1344, aquatic pathogen Aeromonas veronii GD2019, and enterohemorrhagic Escherichia coli O157:H7 EDL933 were used as controls. Sensitivity paper discs incubated with the purified LD suspensions were used to evaluate LD-mediated bactericidal effects. E. piscicida, A. veronii, and Salmonella exhibited apparent resistance against LD-mediated bactericidal effects and proliferated from 9 h post-treatment, while the growth of E. coli was almost completely inhibited (Fig. [58]1A, B). This suggested that certain pathogenic bacteria generally evolved a certain degree of resistance to LDs and antimicrobial substances on their surface to conquer host innate immune clearance. We observed that the growth inhibition caused by LDs primarily occurred during the early stages (0–12 h) and bacteria eventually overcame the LD-mediated bactericidal effects (Supplementary Fig. S[59]3A), probably due to the nutrients, such as lipids, within the LDs promoting the secondary growth of remnant when the antimicrobial substances were depleted^[60]14,[61]24. Noteworthily, no significant difference in bactericidal activity was detected between turbot liver LDs with or without bacterial stimulation (Supplementary Fig. S[62]2), in line with no substantial change in the abundance of antimicrobial proteins, such as histones (Supplementary Fig. S[63]1A). Fig. 1. E. piscicida counteracts LD-mediated bactericidal effects. [64]Fig. 1 [65]Open in a new tab A, B Growth profiles of different gram-negative bacteria in the presence of turbot liver LDs. E. piscicida and A. veronii were cultured in MinA minimal medium treated with LDs or PBS at 30 °C, while S. Typhimurium and E. coli were cultured at 37 °C. The value of OD[600] was measured every 3 h (A) and the relative CFU (B) was calculated by CFUs per microliter of the LD-treated group divided by CFUs per microliter of the corresponding PBS-treated group at the indicated time points (n = 3, 1 technical replicate per experiment). C CFUs of E. piscicida and E. coli cultured for 9 h in the presence of LDs or PBS (n = 6, 3 technical replicates per experiment). The results are shown as the mean ± S.D. (*P < 0.05; **P < 0.01; ***P < 0.001; ns not significant by Student’s t test). D Inhibition zone assays of E. piscicida and E. coli. The sensitivity paper discs were incubated with purified LDs suspension with a concentration of 5 mg protein per microliter and then placed on LB plates spread with bacteria. White scale bars, 1.5 cm. The image shown is representative of three independent experiments. As a typical intracellular pathogen as well as the natural infectious agent of turbot, E. piscicida possessed the most apparent resistance to LDs among tested pathogens. The tested E. piscicida strains of EIB202, PPD130/90, and EIB107 all exhibited relatively stable growth after 9 h of LD challenge (Fig. [66]1C). The inhibition zone assays with LD-incubated discs were then conducted. A distinct antibacterial zone was observed on the agar containing E. coli, whereas only a minimal inhibition zone (0.9 cm) occurred on the agar containing E. piscicida, indicating that E. piscicida effectively resists LD-mediated bactericidal effects (Fig. [67]1D). Tn-seq analysis of essential genes in E. piscicida resisting LD-mediated bactericidal effects Adaptability to host LD-mediated bactericidal effects in pathogens may involve a series of delicate mechanisms. To this end, transposon-insertion sequencing (Tn-seq) was employed to comprehensively identify the essential genes determining the pathogen’s combat LD-mediated bactericidal effects (Fig. [68]2A). Our previously constructed sequence-defined transposon mutant library of E. piscicida EIB202 was utilized for screening fitness determinants under the pressure of LDs^[69]25,[70]26. The reads per gene in the sample following growth in LD challenge condition were compared to those of the control sample grown in the normal DMEM to calculate the fold change (FC) (Fig. [71]2B). A total of 41 genes were identified with significant fitness differential after LD challenge (log[2]FC > 1 or log[2]FC < −1, p < 0.05) (Supplementary Table [72]S1). Among them, fewer insertions were identified in 23 genes, indicating that the disruptions of these genes might impede the growth of EIB202 in the presence of LDs. Fig. 2. Tn-seq analysis of essential genes in E. piscicida resisting LD-mediated bactericidal effects. [73]Fig. 2 [74]Open in a new tab A Schematic flow chart of the Tn-seq analysis. Hepatic LDs were purified from fractionated turbot liver by centrifugation in sucrose density. Each defined Tn library of E. piscicida EIB202 was grown in normal DMEM (control) or DMEM supplemented with LDs for 9 h at 30 °C. B Volcano plots illustrate the FC of the abundance of transposon insertion mutants grown in DMEM with LDs compared to those grown in normal DMEM. Genes with significantly highlighted with the cut-off of log[2]FC > 1 or < -1, and P < 0.05. C GO biological process and D KEGG pathway enrichment analysis of highlighted genes in B. The enrichment ratio and counts of indicated genes in each category are shown. P-values were calculated using hypergeometric distribution and q-values representing false discovery rate were calculated by Benjamini-Hochberg procedure. E PPI network of genes exported from the STRING database. Each node represented the highlighted genes in B. Clusters were computed and manually labeled by the k-means algorithm. We further explored the annotated functions of these genes (Fig. [75]2B). Nearly half of them are involved in classical bacterial stress response and tolerance pathways, including genes associated with extracytoplasmic function sigma factors (rpoS, rseB), two-component systems (cpxR, cpxP, qseC), bacterial stress response (hflC, hflK, sspB, rcsC), and DNA damage repair (recD). Additionally, several T3SS genes (esrB, escB, esaU) also exhibited noticeable differences in the presence of LDs, supporting the notion that T3SS effectors participate in the interaction between pathogen and host LDs^[76]11,[77]27. Most importantly, significant fitness changes occurred in some genes related to basal metabolism, particularly those associated with specific amino acid metabolism pathways (such as argR), suggesting that metabolic reprogramming was undertaken to cope with LD-mediated bactericidal effects. These certain amino acid metabolic pathways may be closely connected with the LD resistance of E. piscicida. Gene Ontology (GO) biological process and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were further performed to cluster the specific pathways in which the differentially expressed genes are involved (Fig. [78]2C, D). In the annotated functional categories, pathways related to chemotaxis, cell motility, and cell wall synthesis were enriched as well as CAMP resistance and oxidative stress response. Specifically, pathways associated with carbon metabolism, the respiratory electron transport chain, and amino acid synthesis exhibited high enrich scores, indicative of the vital role of metabolic reprogramming in response to LDs. Among them, the term “arginine biosynthesis” was enriched in both the KEGG pathway and GO biological process functional analysis. To visually explore the functional relevance among the enriched genes, the STRING database was queried to display the genes with established associations using a protein-protein interaction (PPI) network (Fig. [79]2E). Three interconnected clusters, including the stress response regulation and T3SS, were highlighted, of which arginine metabolism formed a distinct cluster within amino acid metabolism. These results supported the essential role of the arginine metabolism pathway in the resistance to LD-mediated bactericidal effects. Validation of genes revealed by Tn-seq analysis for resisting LD-mediated bactericidal effects The sequence-defined transposon library enables us to assess the impact of individual gene mutations conveniently. For the growth test in the presence of LDs, 18 insertional mutants that exhibited significant fitness differences were selected, meanwhile, four randomly selected insertional mutants (ETAE_1610::Tn, ETAE_2755::Tn, ETAE_2004::Tn, ETAE_2009::Tn) with no influences in Tn-seq analysis were used as controls (Fig. [80]3A). In total, 16 mutants exhibited significant differences in growth at 9 h post-LD-treatment, which was consistent with the trends observed in Tn-seq analysis and proved the reliability of our screening. Among the genes, the insertions of regulatory factors argR, cpxR, esrB, and rpoS resulted in an apparent growth reduction under the pressure of LDs, indicating their potential regulation roles in the resistance to LD-mediated bactericidal effects (Fig. [81]3A). The multidrug efflux component acrA was also found to determine the resistance of E. piscicida to LDs. Fig. 3. Validation of critical genes revealed by Tn-seq analysis for resisting LD-mediated killing effects. [82]Fig. 3 [83]Open in a new tab A CFUs of WT E. piscicida and indicated transposon insertion mutants grown in MinA medium with LDs or PBS at 30 °C for 9 h. The corresponding insertion mutants were picked from the defined Tn-library according to our Tn-seq results, including genes with more insertions (log2FC > 1, purple), genes with fewer insertions (log2FC < −1, yellow), and not significantly changed genes (0.5 < FC < 2, black) (n = 3, 3 technical replicates per experiment). B Normalized growth curve of E. piscicida variants in the presence of LDs. WT, ΔrpoS, ΔargR, ΔcpxR, ΔesrB, ΔacrA and five corresponding complementary strains were grown in MinA medium supplemented with LDs at 30 °C for 12 h. The relative CFU of each strain was normalized to corresponding CFUs per microliter in MinA medium with PBS (control). (n = 3, 3 technical replicates per experiment). C Inhibition zone assays of E. piscicida variants in the presence of paper discs containing LD suspension. Bar plot represents the measured diameters of the indicated deletion mutants and their complementary strains (cpl) as well as E. piscicida WT (n = 3, 1 technical replicate per experiment). White scale bars, 1.5 cm. The disc size, 0.6 cm. The image shown is representative of at least three independent experiments. D The relative transcripts of argR and cpxR in E. piscicida WT in the presence and absence of LDs for 9 h. gyrB was used as a negative control (n = 3, 3 technical replicates per experiment). The results are shown as the mean ± S.D. (*P < 0.05; **P < 0.01; ***P < 0.001; ns not significant by Student’s t test). To further investigate their impacts on the growth of E. piscicida responding to LD challenge, the in-frame deletion strains of indicated five genes (ΔargR, ΔrpoS, ΔesrB, ΔcpxR, ΔacrA) were constructed. The growth trends of all five deletion strains were observed to exhibit noticeable growth inhibition within the first 9 h post-LD-treatment than WT did. At 9 h post-LD challenge, the difference in colony-forming units (CFU) between the deletion strains and WT reached the maximum (Fig. [84]3B). In line with the performance of WT, ΔargR, ΔesrB, and ΔcpxR possessed the growth recovery after 9 h, while the growth of ΔacrA and ΔrpoS was constantly inhibited by the same concentration of LD (Fig. [85]3B). Furthermore, the deletion mutants formed larger inhibition zones (1.3-1.7 cm) than WT did, while the complementary strains rescued the impaired resistance to LD-mediated bactericidal effects (Figs. [86]1B and [87]3C). These observations were congruent with results in the susceptibility assays and further validated their critical roles in the process of bacterial resistance to LDs. We then investigated the transcription levels of identified candidates in E. piscicida upon the treatment of LDs. In the presence of LDs, the transcription level of argR was up-regulated by up to 4.5-fold and that of cpxR was enhanced to approximately 2-fold (Fig. [88]3D). The transcription levels of rpos, esrB, and acrA exhibited no significant changes in the treatment of LDs, suggesting their contribution to LD resistance potentially at post-transcriptional levels. Considering that ArgR serves as a central regulator for arginine metabolism and is conserved in enteric pathogens (Supplementary Fig. S[89]5), we thus supposed that arginine metabolism modulated by ArgR contributes to the resistance mechanism against LD-mediated bactericidal effects. Arginine influences the resistance of E. piscicida to LD-mediated bactericidal effects in an intracellular concentration-dependent manner Our Tn-seq data gave us the clue that genes involved in arginine metabolism may have important impacts on the growth of E. piscicida under LD challenge (Figs. [90]2B and [91]3A). To further analyze the correlation between arginine metabolism-related genes and their fitness changes, a gene-set enrichment analysis (GSEA) was performed from our Tn-seq data. Each gene was ranked using the normalized log[2] FC and then running enrichment scores were calculated for the defined gene sets “arginine import or synthesis” and “arginine export or catabolism” (Fig. [92]4A; Supplementary Table [93]S2). The running enrichment score curves for the two gene sets exhibited opposite trends: genes associated with arginine import and synthesis possessed an upward trend with positive enrichment scores, indicating that the disruptions of these genes increased E. piscicida fitness under LD challenge. Conversely, a downward trend with negative enrichment scores was identified among genes related to arginine degradation and export, suggesting that the interference of these genes impaired LD resistance of E. piscicida. Fig. 4. Genes involved in arginine metabolism contribute to resisting LD-mediated killing effects. [94]Fig. 4 [95]Open in a new tab A GSEA of genes involved in arginine import or synthesis (red) and export or catabolism (light blue). Background genes were chosen from Tn-seq analysis and genes without any reads were excluded. The ranked list metric represents normalized log[2]FC in Tn-seq and the running score was calculated by R package clusterprofiler. B Volcano plot illustrating the abundance changes of amino acid-related metabolites in E. piscicida grown in LD-treated MinA minimal medium compared to those grown in PBS-treated medium. A total of 22 amino acid-related metabolites were quantified by targeted metabolomics using UPLC-MS/MS. Metabolites with significantly highlighted by the cut-off of log[2]FC > 1 or <1, and p < 0.05. C Schematic of bacterial arginine metabolism and the regulation of ArgR. D CFUs of WT E. piscicida and arginine metabolism-related transposon insertion mutants (tn) and corresponding complementary strains (cpl) in the presence and absence of LDs at 30 °C for 9 h. The complementary strains were constructed via pUTt vectors carrying the complemented genes. The results are shown as the mean ± S.D. (n = 3, 3 technical replicates per experiment, *P < 0.05; **P < 0.01; ***P < 0.001; ns not significant by Student’s t test). E Circos plot of in silico footprint assay result of ArgR. Experimental validated ArgR binding motif 1 (red track) and ArgR binding motif 2 (green track) in E. coli were used for predicting the ArgR binding sites across E. piscicida EIB202 genome. Motif scores (blue track) were calculated by the average of two binding motif scores. The motif scores were computed with FIMO software and genes related to arginine metabolism were highlighted in the inner track. A motif score cutoff of 0.5 was applied to highlight significant results. We thus hypothesized that excess intracellular arginine might interfere with the process through which E. piscicida combats LD-mediated bactericidal effects. To verify if intracellular arginine level is correlated with the response of E. piscicida against LDs, WT pellets were harvested following the incubation with LDs and the levels of amino acids metabolites were measured by targeted metabolomics (Supplementary Table [96]S3). By a threshold of FC less than 0.5 and greater than 2, the intracellular arginine was significantly augmented compared to the control group (Fig. [97]4B). Given that the majority of amino acid metabolites exhibited no abundance changes, E. piscicida was speculated to undergo the metabolism reprogramming of arginine under the pressure of LDs. As a repressor of arginine biosynthesis and an activator of arginine catabolism, argR is responsible for the balance of intracellular arginine^[98]28. Considering the fine-tuning of arginine level is critical for bacteria in a stressful environment^[99]29,[100]30, this result supported the observed fitness defect of ΔargR as well as the up-regulation of argR. To decipher the role of arginine metabolism in resisting LD-mediated bactericidal effects, we then focused on the performances of the genes involved in arginine metabolism (Fig. [101]4C). Initially, eight insertional mutants related to arginine biosynthesis from glutamate (argD::Tn, argH::Tn, argE::Tn), arginine deiminase pathway (arcC::Tn), arginine decarboxylation (speA::Tn, speB::Tn) along with arginine exportation (argO::Tn), and importation (artP::Tn) were selected and their growth was tested in the presence of LDs. The insertion in arginine exporter argO impaired the resistance of E. piscicida to LDs, whereas the blockage of the arginine generation (argD, argE, and argH) enhanced the viability under the pressure of LDs, indicating the importance of the appropriate level of intracellular arginine (Fig. [102]4D). Similarly, the insertional mutations of speA and speB, involved in converting arginine to putrescine/spermidine, resulted in weakened growth after LD challenge (Fig. [103]4D). However, the mutation of arcC, which converts arginine to ornithine, had no significant influence on the resistance to LDs. Additionally, we successfully obtained the corresponding deletion mutants except for speA/argE and they exhibited similar tendency against LD-mediated bactericidal effects (Supplementary Fig. S[104]3B). These findings suggested that the elimination of the excessive intracellular arginine is vital for the LD resistance of E. piscicida. To further investigate the role of ArgR involved in arginine metabolism under LD challenge, we initially utilized the conserved ArgR binding motif (previously characterized in E. coli) to scan the E. piscicida genome (Fig. [105]4E). The in silico DNA footprint analysis indicated that ArgR potentially binds to the promoter regions of genes involved in arginine metabolism, including argE/H/D, speA/B, arcC, and artP. (Supplementary Table [106]S4). Electrophoretic mobility shift assay (EMSA) was further conducted and confirmed the binding ability of ArgR to the promoters of speA/B, argE/H, argD, and artP (Supplementary Fig. S[107]4). We then compared the transcription levels of arginine metabolism-related genes in WT and ΔargR under LD challenge (Fig. [108]5A). In congruent with increased intracellular arginine levels observed in metabolomics, LD challenge significantly enhanced the transcription levels of tested genes in WT, reflecting the activation of arginine metabolism. The absence of argR resulted in the down-regulations of speA/B and arcC and the up-regulations of argE/H/D and artP, indicating the role of ArgR in repressing arginine import or synthesis and activating arginine export or catabolism (Fig. [109]4C). The presence of argR had no influences on argO transcripts, aligning with no predicted AgrR-binding sites in argO’s promoter region. More importantly, LD challenge amplified the activation of argE/D/H and artP and the repression of speA/B caused by the absence of argR. Collectively, it can be concluded that LD challenge stimulates intracellular arginine accumulation and argR is responsible for the maintenance of arginine homeostasis, conferring LD resistance on E. piscicida. Fig. 5. Arginine attenuates E. piscicida’s resistance to LD-mediated killing effects in an intracellular concentration-dependent manner. [110]Fig. 5 [111]Open in a new tab A The relative transcripts of arginine metabolism-related genes in WT, ΔargR, and argR^+ grown in MinA medium with LDs or PBS for 9 h. gyrB was used as a negative control (n = 3, 3 technical replicates per experiment). B CFUs of WT E. piscicida grown in MinA medium with LDs or PBS at 30 °C for 9 h and supplemented with different concentrations of arginine (n = 6, 3 technical replicates per experiment). C CFUs of E. piscicida transposon insertion mutant artP::Tn and the complementary strains artP^+ grown in MinA medium with LDs or PBS in the presence and absence of 200 μM arginine (n = 6, 3 technical replicates per experiment). D, E Intracellular arginine levels and relative transcription levels of arginine metabolism-related genes throughout infection. RAW264.7 cells were infected with E. piscicida (WT) at an MOI of 10, and then cells were lysed before centrifuged to collect intracellular bacteria. gyrB was used as a negative control (n = 3, 2 technical replicates per experiment) (D). Intracellular bacteria were collected for qRT-PCR analysis. gyrB was used as a negative control (n = 3, 3 technical replicates per experiment) (E). The results are shown as the mean ± S.D. *P < 0.05; **P < 0.01; ***P < 0.001; ns not significant by Student’s t test. Given that the intracellular arginine directly participates in bacterial pH regulation, biofilm formation, and resistance to antimicrobial substances^[112]30–[113]32, we further aimed to investigate the direct impact of arginine on the LD resistance of E. piscicida. To this end, a gradient concentration of 0, 100, 200, and 300 μM arginine was added into the medium and the growth of E. piscicida was assessed in the presence and absence of LDs, respectively. All the gradient arginine concentrations did not influence the growth of WT without LD challenge (Fig. [114]5B). In the presence of LDs, the growth of WT gradually slowed down with the increased concentration of arginine, suggesting excessive arginine attenuated the LD resistance of E. piscicida (Fig. [115]5B). Furthermore, when the extracellular arginine transport gene artP was mutated, the addition of arginine no longer had a significant impact on the LD resistance of E. piscicida (Fig. [116]5C). The complement of artP restored the weakened LD resistance as WT did (Fig. [117]5C). These results proved that excessive intracellular arginine impairs the LD resistance of E. piscicida. To investigate the intracellular arginine levels within host cells during E. piscicida infection, enzyme-linked immunosorbent assay (ELISA) was performed and indicated that intracellular arginine levels in RAW264.7 gradually accumulated throughout infection, suggesting that arginine may participate in the LD-mediated defense mechanisms (Fig. [118]5D). To cope with the augmented arginine within host cells, qRT-PCR analysis further revealed the altered expression of argR and other arginine metabolism-related genes in E. piscicida throughout the infection (Fig. [119]5E). During the early phase of the infection (0–2 h), all arginine metabolism-related genes were upregulated in E. piscicida, consistent with the stimulated arginine metabolism in response to LD challenge. At 6 h post-infection, the expression of argD/E/H and artP was significantly reduced, while the transcripts of argO and speB remained at relatively high levels. It might be explained by the fact that E. piscicida attenuates de novo arginine synthesis/import and promotes arginine degradation/export to maintain arginine homeostasis and achieve immune evasion. argR determines the ex vivo and in vivo pathogenesis of E. piscicida To investigate the impact of LD resistance essential genes on the pathogenesis of E. piscicida, arginine metabolism-related mutants (ΔargR, ΔargO, ΔspeB), as well as global stress tolerance-related mutants (ΔacrA and ΔcpxR), were employed to infect RAW264.7 cells at a multiplicity of infection (MOI) of 10. Given that esrB and rpoS have been previously validated as critical virulence determinants of E. piscicida^[120]33,[121]34, they were not included. Similar to the pathogenicity attenuated strain ΔT3SS losing T3SS function, CFU counts of ΔargR, ΔargO, ΔspeB, ΔacrA, and ΔcpxR significantly decreased inside macrophages compared to that of WT while the corresponding complement strains restored the intracellular CFUs, reflecting their reduced colonization and proliferation in host cells (Fig. [122]6A). Fig. 6. ArgR determines the ex vivo and in vivo pathogenesis of E. piscicida. [123]Fig. 6 [124]Open in a new tab A Intracellular replication ability of E. piscicida variants. RAW264.7 cells were infected at an MOI of 10. The CFU of intracellular E. piscicida variants at 6 h post-infection was normalized to cell numbers to determine the bacterial proliferation capacity (n = 6, 3 technical replicates per experiment). B Survival curve of turbot infected with E. piscicida variants. A total of 15 turbots per group were used. The significance was determined by the Log-rank test by running the survival R package. C CFUs recovered from liver samples of turbot with infected E. piscicida variants. Livers were harvested at 14 days post-injection and CFUs were counted according to the colonies on DHL plates. The central line within each box represents the median, while the box represents the interquartile range (IQR), and the whiskers extend to 1.5 times the IQR (n = 6, 3 technical replicates per experiment). (*P < 0.05; **P < 0.01; ***P < 0.001; ns not significant by Student’s t test). To determine the role of argR, argO, speB, acrA, and cpxR during the in vivo invasion process, the mutants were intraperitoneally injected into turbot at a dose of 5 × 10^6 CFU/fish, and then the survival curve was plotted (Fig. [125]6B). From the curve, turbot infected with WT began to show signs of mortality at 10 days post-injection (dpi) and were with complete mortality by 14 dpi. In contrast, turbot from ΔargR, ΔargO, ΔspeB, ΔacrA, and ΔcpxR infected groups exhibited a delayed onset of mortality as well as higher survival rates (Fig. [126]6B). Subsequently, liver tissue was collected at 10 dpi and the colonization was counted by spreading the hepatic homogenate on deoxycholate hydrogen sulfide lactose (DHL) agar. Bacterial load of ΔargR, ΔargO, ΔspeB, ΔacrA, and ΔcpxR were observed with a significant reduction in the liver (Fig. [127]6C). These findings demonstrated that argR, argO, speB, acrA, and cpxR function as virulence determinants in the pathogenesis of E. piscicida. Discussion Substantial evidence from previous studies has indicated the vital role of LDs in the interactions between eukaryotic host cells and pathogens. LDs serve as the potential nutrition and energy source for intracellular pathogens’ colonization and proliferation, in one respect, however, various antimicrobial proteins accommodated on LDs highlight the immune defense mediated by LDs^[128]4,[129]15. In this study, we utilized Tn-seq to identify critical fitness determinants of E. piscicida counteracting LD-mediated bactericidal effects. The presence of LDs resulted in the accumulation of intracellular arginine. The maintenance of intracellular arginine homeostasis by ArgR-regulated arginine metabolism-associated genes was vital for E. piscicida’s adaptation to host LDs. Furthermore, ArgR promoted the pathogenesis of E. piscicida, including the proliferation in cells and virulence in turbot. Similar to LDs in mammalian cells, hepatic LDs isolated from turbot exhibited antimicrobial effects on numerous gram-negative bacteria. The proteome analysis indicated that turbot liver LDs are abundant with histones that exhibit antimicrobial activities^[130]16. Recent findings also proved that under the stimulation of bacterial pathogens, CAMPs were specifically recruited to LDs to enhance antimicrobial capacity^[131]2. These two kinds of protein can inhibit bacterial growth by polarization of bacterial membrane and disruption of membrane integrity as well as proton gradient. E. piscicida encodes multiple sets of cell wall modification genes, which might partly explain the enhanced resistance to LDs^[132]22,[133]23. However, although a conserved CAMP-resistance mechanism has also been identified in Salmonella^[134]23, S. Typhimurium SL1344 displayed weaker resistance to LDs compared to E. piscicida EIB202. What is more, Pseudomonas plecoglossicida, lacking such genes, also counteracted LD-mediated bactericidal effects. These results suggested that different bacteria employ diverse and intricate mechanisms upon encountering LDs, which are not solely limited to resist histones or CAMPs. As a result of the co-evolution with the host immune system, intracellular pathogens are urged to evolve strategies to resist LD-mediated bactericidal effects. From our Tn-seq results, E. piscicida employed several global regulators, including EsrB, CpxR, RpoS, and ArgR along with a multi-drug efflux complex component AcrA to cope with LD-mediated bactericidal effects. Of them, CpxR is responsible for envelope stress response and AcrA participates in the efflux of antibiotic substrates which are supposed to assist E. piscicida in resisting toxic substrates like CAMP or histones located on LDs^[135]35,[136]36. Our previous RNA-seq data of E. piscicida indicated that regulatory factors EsrB and RpoS, govern a bunch of genes involved in arginine metabolism and transportation which is crucial for bacterial stress response and virulence^[137]33,[138]34. Moreover, our Tn-seq analysis also discovered that T3SS proteins influenced the LD resistance of E. piscicida, which are mainly regulated by EsrB and RpoS^[139]33. These findings highlighted the potential intersecting transcription networks that govern LD resistance of E. piscicida. Arginine, as a crucial metabolite as well as a signaling molecule, has been validated to play a significant regulatory role in the interaction between pathogens and host immune response. During infection, arginine has been shown to mediate the metabolic reprogramming of cells and enhance host defense against bacterial pathogens as the sole substrate for the synthesis of nitric oxide (NO) by inducible nitric oxide synthase (iNOS)^[140]37. Alternatively, the conversion of arginine into ornithine catalyzed by Arg1 has been considered as the hallmark of M2-polarized macrophages^[141]38. On the other hand, arginine is also a key metabolite affecting bacterial stress tolerance and virulence^[142]31,[143]39. Arginine can serve as a buffering substance to help S. Typhimurium survive in extremely acidic environments and affect biofilm formation to alter antibiotic resistance in Staphylococcus aureus^[144]30,[145]31. Arginine derivatives, such as spermine and putrescine, prevent intracellular pathogens from clearance within macrophages^[146]40. During infection, ArgR of enterohemorrhagic E. coli sensed arginine fluctuations to directly regulate the expression of T3SS genes and E. piscicida reprogrammed its arginine metabolism to evade NLRP3 inflammasome^[147]39,[148]41. In conclusion, these findings contribute to revealing the pivotal role of arginine metabolism as a key determinant during the interaction between bacterial pathogens and host. Once entered into host cells, arginine in E. piscicida would be significantly augmented in response to LDs, and thus, E. piscicida has to undergo metabolic reprogramming to survive and achieve immune evasion. To cope with LD-mediated bactericidal effects, ArgR govens bacterial arginine homeostasis, by enhancing the expressions of genes involved in arginine catabolism (speA/B and arcC) and diminishing gene transcripts associated with arginine import (artP) and synthesis (argD/E/H). Nevertheless, our in vitro LD challenge assay primarily reflects the direct antimicrobial capacity of LDs. Besides, the isolated LDs contain few residual cytosolic fraction following centrifugation-extraction, such as proteins, amino acids, and sugars, which might participate in LD-mediated bactericidal effects. Therefore, future studies should focus more on the comprehensive LD-mediated innate immunity within host. Methods Bacterial strains and culture conditions The wild-type (WT) E. piscicida strain EIB202 (CCTCC M208068) was isolated from a diseased turbot in Yantai, China. Details of all bacterial strains and plasmids utilized in this study are presented in Supplementary Table [149]S5. The primers applied in this study are listed in Supplementary Table [150]S6. All E. piscicida strains were inoculated into lysogeny broth (LB) medium at 30 °C overnight and then were sub-cultured into a secondary medium at a dilution rate of 1:100. In LD-resistance assays, bacteria were sub-cultured in MinA minimal medium^[151]42 supplemented with10 μM MgSO[4] and 0.25% casamino acids. For bacterial growth, E. coli O157:H7 EDL933 and S. Typhimurium SL1344 were cultured at 37 °C. A. veronii GD2019 and P. plecoglossicida were grown at 30 °C. Antibiotics were added with the following concentrations when necessary: ampicillin (Amp, Sangon Biotech, A429319, 100 μg/ml), polymyxin (Col, Sangon Biotech, A610318, 30 μg/ml), kanamycin (Km, Sangon Biotech, 60206ES, 100 μg/ml), and gentamicin (Gm, Sangon Biotech, A428430, 20 μg/ml). Turbot hepatic LD purification The hepatic LDs were extracted and purified using a tissue homogenization sucrose gradient centrifugation method^[152]15. The liver harvested from turbot was perfused with pre-cold phosphate-buffered saline (PBS) buffer and washed twice. After being chopped into pieces smaller than 1 cm^3 in volume, the liver tissues were transferred into the Dounce tissue grinder with LD purification buffer A (25% (w/v) sucrose, pH 7.5, 10 mM HEPES) at a ratio of 1 g of tissue to 1 ml of buffer and homogenized on ice. The homogenate was then aliquoted into 2 ml Eppendorf (EP) tubes and centrifuged at 15,000 × g for 90 min at 4 °C to separate different cellular components. The top fraction containing floated LDs was collected and transferred to 1.5 ml EP tubes for purification. The LD purification buffer B (100 mM KCl, pH 7.5, 25 mM Tris-HCl) was added on top of the LD fractions at a volume ratio of 1:1 and the fractions were then centrifuged at 4000 × g for 5 min at 4 °C. The bottom fraction was carefully removed using a 23 G syringe followed by adding the same volume of buffer B. This step was repeated 2 or 3 times until the lower liquid fraction was completely transparent. Purified LD samples were then resuspended in sterile PBS and stored at −20 °C for subsequent experiments. To quantify LDs, a BCA protein assay kit (Sangon, C503021) was used to determine the concentration of lipid-protein. For quality control (QC) of each LD extraction batch, western blot was performed to detect the levels of the LD protein marker (PLIN2) and β-actin in both the LD and post-centrifugation cytosolic (CYTO) fractions. Customized antibody of PLIN2 was generated by Genscript. Rabbit polyclonal anti-β-actin (ABclonal, WH400517) and corresponding secondary antibody (YEASEN, 34850ES60) were used. All the antibodies were diluted at a ratio of 1:1000. LD protein samples were prepared using acetone precipitation method^[153]15. The LD fraction suspension was mixed with four volumes of ice-cold acetone and incubated at −20 °C for 24 h. After precipitation, the sample was centrifuged at 15,000 × g for 10 min at 4 °C, and the supernatant was discarded. The pellet was washed 2-3 times with ice-cold acetone and then resuspended in 10 mM Tris-HCl with 5% SDS. After sonication, the sample was warmed at 50 °C for at least 10 min. Protein samples were quantified using the BCA assay and subjected to 4D label-free quantitative proteomics. LD challenge assays Due to the physical nature of lipids, LDs can hardly be dissolved in the aqueous phase evenly and tend to float above. To overcome this problem, the drug sensitivity testing disc was used as the carrier of LDs in the medium. The sensitive disc-assay method for LDs was performed^[154]4. Generally, the protein concentration of purified LD suspension was quantified and then diluted by PBS to the protein concentration of 5 mg/ml in a 1.5 ml EP tube. Blank sensitivity discs were immersed in LD suspension and incubated for 1 h at 4 °C on a rotating mixer. Subsequently, discs that fully absorbed LDs were placed on a clean sterile plate to evaporate excess moisture for 30 min at 4 °C. For LD challenge assays in liquid cultures, discs loaded with LDs were immersed in the medium at a dose of 1 disc per 500 ml medium and incubated for 30 min. For control groups, discs incubated with sterile PBS were used. After the ingredients of LDs were fully released into the medium, the discs were removed and bacterial cultures grown to an optical density of OD[600] = 1 were diluted 1:100 into the MinA medium pretreated with LD- or PBS-loaded discs. Cultures were grown at 30 °C in a shaking incubator. Serial dilutions made in PBS were plated in triplicate on LB agar plates and CFUs per microliter were quantified after incubating at 30 °C overnight. The relative CFU was calculated by dividing by the CFUs per microliter of the corresponding PBS discs treated group. For determining the inhibition zone of LDs, sensitivity discs (size of 0.6 cm in diameter) were prepared as the same procedure mentioned above. Overnight cultures grown to OD[600] = 1 were diluted 1:100 in sterile PBS. After spreading the dilutions on 7 cm LB agar plates, the discs were placed and the static incubation was performed at 30 °C overnight before measuring the diameter of the inhibition zone around the discs. Growth profiles Growth curve experiments were performed using sterile, clear bottom, black 96-well microplates. Overnight cultures grown to OD[600] = 1 in LB medium were diluted 1:100 into MinA medium supplemented with LDs. The bacterial solution was aliquoted to 200 μl for each well. For OD[600] measurement experiments, The Infinite® F50 multi-mode plate reader (Bioscreen) was used to monitor bacterial growth. The temperature setpoint was 30 °C and preheated before measurements. Growth curves were measured every 1 h for up to 21 h. For CFU measurement, bacterial solution in each well of the microplate was sampled every 3 h after the incubation began. Serial dilutions were made in PBS and plated on LB agar for each sample. The normalized growth curve was constructed by calculating the relative CFU at every time point. LD resistance screening by Tn-seq analysis The saturated transposon insertion mutant library was initially inoculated into LB medium at 30 °C and agitated for 2 h until the OD[600] reached 1.0^[155]25,[156]26. Subsequently, bacterial cultures were centrifuged at 8000 × g for 2 min, washed twice with PBS, and diluted 1:100 into DMEM (Sangon Biotech, E600003) treated with discs pre-incubated in LD suspension with 5 mg/ml protein concentration (LD group) and PBS (Control group), respectively. After shaking for 9 h, bacteria were pelleted, plated on LB agar plates, and then cultured overnight at 30 °C to eliminate DNA contamination from dead cells before genomic DNA extraction. Tn-seq experiments procedures and data analysis were conducted^[157]43. Briefly, the fragmented genomic DNAs were subjected to end repairing, A-tailing, adapters, and P5/P7 primers, generating high-throughput sequencing libraries for sequencing on the Illumina MiSeq platform. The raw sequencing data was processed by Bowtie algorithm to generate sequences mapped to E. piscicida EIB202 genome and Transit software was employed to process locus tallying for each locus of EIB202^[158]44,[159]45. The sum of read counts of all locus within a gene represents the read counts of this gene and the FC of each gene was computed by the LD group read counts divided by the corresponding control group read counts after normalization. Construction of deletion and complementary strains The in-frame deletion mutants of the indicated genes were constructed using the sacB-based allelic exchange method^[160]46. Briefly, the suicide vectors pDM4, which contained the upstream and downstream sequences required for in-frame deletions of each gene (Supplementary Table [161]S5), were transformed into E. coli SM10 strain and then conjugated with E. piscicida EIB202. After the double-crossover recombination process, the deletion mutants were then selected on LB plates containing 12% sucrose and confirmed by PCR. For the construction of complementary strain, the pUTat vectors cloned with the intact gene sequence were generated followed by the transformation via electroporation into the corresponding deletion mutants (Supplementary Table [162]S5). Total RNA extraction and qRT-PCR Total RNA samples were extracted from E. piscicida that were grown in MinA medium for 9 h supplemented with LDs and PBS, using an RNA isolation kit (BioFlux, BSC52S1). The mRNAs were immediately reverse-transcribed into cDNAs using a FastKing RT kit (Tiangen, KR11601). The qRT-PCR experiments were performed on an Applied Biosystems 7500 real-time system (Applied Biosystems, CA). Comparative CT method was further used to determine the relative quantities of each transcript, normalized to the gyrB gene which serves as a control. All primers used for qRT-PCR are listed in Supplementary Table [163]S6. Targeted metabolomics of intracellular amino acid metabolites Bacteria grown in MinA medium supplemented with LDs and PBS for 9 h were harvested by centrifugation at 8000 × g at 4 °C. After washed twice with pre-cold PBS, the pellet samples of each group were stored in EP tubes at −80 °C. CFU of each sample was quantified by serial dilution and plating on LB agar. The quantified samples were thawed on ice and added with 120 μl of methanol solution containing internal standards. The samples were then subjected to ultrasonic disruption at 4 °C and centrifugated at 18,000 × g for 15 min. Subsequently, 10 μl of the supernatant was mixed with 70 μl of borate buffer (1/4, pH = 8.8) and 20 μl of 6-aminoquinolyl-N-hydroxysccinimidyl carbamate (AQC) derivatization reagent (1.5 mg/ml) followed by incubation for 10 min at 55 °C. After that, a 100 μl aliquot was taken, mixed with 900 μl of ultrapure water, and then subjected to UPLC-MS/MS analysis. DNA-binding analysis of ArgR To determine the potential binding sites of transcription factor ArgR on the genome of E. piscicida EIB202, previously validated ArgR DNA-binding motif in E. coli, ARG Box1 ([164]MX000116 in PRODORIC database) and ARG Box2^[165]47,[166]48, were used to predict the potential ArgR DNA-binding sites in E. piscicida. Binding motifs were downloaded as position weight matrices and screening was processed on the genome of E. piscicida EIB202 by FIMO on MEME Suite ver. 5.5.1. The motif binding score of each position was calculated according to the FIMO output scores and normalized to a range of 0-1. To validate the in silico predictions, the 6 × His-tagged ArgR was expressed in E. coli BL21 with pET28a vector and purified using nickel beads (Sangon Biotech, 20504ES08). The putative promoter region of the test gene for EMSA (−150 to +150 predicted to contain the AgrR binding site) was amplified by PCR. A universal primer sequence (5′-AGCCAGTGGCGATAAG-3′) was included in the gene-specific primers and then labeled 5′-biotin with a universal primer. EMSA was performed according to the protocol provided with the chemiluminescent EMSA kit (Beyotime, P0018FS). The purified ArgR was incubated with 0.25 ng biotin-labeled DNA probes in EMSA buffer (10 mM Tris-HCl, 50 mM NaCl, 50 μg/mL poly(d[IC]), 10% glycerol, pH 8.0) at 25 °C for 30 min. The samples were then dissolved in ice-cold 0.5×Tris-Borate-EDTA at 100 V on a 6.5% non-denaturing PAGE gel. Bands were detected using Omni-ECL™ Pico Light chemiluminescent kit (Epizyme, SQ202L). The plasmid construction and primers used for probe amplification are listed in Supplementary Table [167]S6. Cell infection RAW264.7 cells were cultivated in 6-well plates at a density of 1 × 10^6 cells per well and allowed to adhere overnight. Overnight bacterial cultures were washed three times with PBS and used to infect the cells at a specific MOI of 10 for RAW264.7 cells. After a 2-hour incubation period, the cells were washed twice with PBS and then exposed to a medium containing 1000 µg/ml of gentamicin for 15 min at 35 °C to eliminate extracellular bacteria. To detect intracellular bacteria, the medium was aspirated and cells were washed twice with sterile PBS. Then, 0.5% Triton X-100 (YEASEN, 20107ES76) cell lysis buffer was added to the cells and incubated for 25 min at 35 °C. Cell lysate samples were collected and serial dilutions were made with sterile PBS followed by the plating on agar plates for CFU enumeration. To assess bacterial proliferation, after eliminating extracellular bacteria, the medium was replaced with a fresh medium containing 10 µg/ml of gentamicin. The cells were further incubated for an additional 3.5 h at 35 °C. Following a total infection time of 5.5 h, the medium was aspirated and cells were washed twice with sterile PBS. Cell lysis buffer was added to cells followed by incubation for 25 min at 35 °C. Cell lysate samples were collected, serially diluted with sterile PBS, and plated on agar plates for bacterial enumeration. The bacterial numbers obtained from the agar plates were used to calculate CFUs. The CFUs were then converted to cell numbers to determine the bacterial proliferation capacity inside the cells. To assess gene expression during infection, macrophage-released E. piscicida was prepared^[168]41. Briefly, the infected RAW264.7 cells were lysed with 0.5% Triton X-100 at 2 h, 4 h, and 6 h post-infection. The supernatant was separated from cell debris by centrifugation at 600 × g for 5 min. The isolated supernatant was further centrifuged at 12,000 × g for 5 min to collect macrophage-released E. piscicida. The collected bacterial pellets were then used for mRNA extraction and gene expression analysis via qRT-PCR. To measure the changes in intracellular arginine levels during infection, RAW264.7 cells were infected with E. piscicida and then lysed with 100 μL of 0.5% Triton X-100. After centrifugation at 600 × g for 5 min, the supernatant was used to determine L-arginine concentration using an ELISA kit (Chenjun, BY-M02120). In vivo challenge assays in turbot Healthy turbots weighing 25.0 ± 3.0 g were chosen from a commercial farm in Yantai, China, and maintained in aerated tanks supplied with a continuous flow of seawater at 16 °C. After the acclimation of turbot in the laboratory for 7 days, all E. piscicida strains grown in LB medium at 30 °C overnight were harvested by centrifugation at 8000 × g for 2 min and resuspended by sterile PBS three times. The bacteria suspension was quantified to a concentration of 10^8 CFU/ml and was intraperitoneally injected into each turbot at a dose of 10^7 CFU/fish. A total of 15 fish were used for each group and the survival was observed every day for up to 16 dpi to construct a survival curve. On the 10th day post-injection, 3 remaining fish from each group were selected and sacrificed. The harvested liver tissue was homogenized and serially diluted followed by plating on 15 cm DHL agar with four replicates for each tissue sample before incubation at 30°C overnight. Statistics and reproducibility Data were presented as the mean ± S.D. of triplicate samples per experimental condition unless noted otherwise. Representative results are shown in the figures. Statistical analyses for all bar-plots and box-plots were performed by running unpaired one-tailed Student’s t test using Microsoft Excel (version 23.11) and the statistical analysis for the results of Fig. [169]6B was performed using the Log-Rank method from the R package survival (version 3.3.5). differences were considered significant at *P < 0.05, **P < 0.01, and ***P < 0.001. Ethics statement All animal procedures performed were authorized by the animal care committee of the East China University of Science and Technology (2006272). The Experimental Animal Care and Use Guidelines from the Ministry of Science and Technology of China (MOST-2011-02) were rigorously adhered to. Reporting summary Further information on research design is available in the [170]Nature Portfolio Reporting Summary linked to this article. Supplementary information [171]Supplementary Information^ (34.7MB, pdf) [172]42003_2025_7777_MOESM2_ESM.pdf^ (38.1KB, pdf) Description of Additional Supplementary Files [173]Supplementary Data 1^ (407.5KB, xlsx) [174]Reporting summary^ (1.6MB, pdf) [175]Transparent Peer Review file^ (1.9MB, pdf) Acknowledgements