Cell type specific gene expression profiling reveals a role for the complement component C3A in neutrophil migration to tissue damage

Following acute injury, leukocytes rapidly infiltrate into tissues. For efficient recruitment, leukocytes must sense and respond to signals from both from the damaged tissue and from one another. However, the cell type specific transcriptional changes that influence leukocyte recruitment and wound healing have not been well characterized. In this study, we performed a large-scale translating ribosome affinity purification (TRAP) and RNA sequencing screen in larval zebrafish to identify genes differentially expressed by neutrophils, macrophages, and epithelial cells in the context of wounding. We identified the complement pathway and c3a.1, homologous to the C3A component of human complement, as significantly increased in neutrophils in response to a wound. We report that c3a.1−/− zebrafish larvae have impaired neutrophil responses to both tail wounds and localized bacterial infections, as well as increased susceptibility to infection due to a neutrophil-intrinsic function of C3A. We further show that C3A enhances migration of human primary neutrophils to IL-8 and that c3a.1−/− larvae have impaired neutrophil migration in vivo, and a decrease in neutrophil directed migration speed early after wounding. Together, our findings suggest a role for C3A in mediating efficient neutrophil migration to damaged tissues and support the power of TRAP to identify cell-specific changes in gene expression associated with wound-associated inflammation.

Introduction sa31241 were isolated through the Sanger Zebrafish Mutation Project, Wellcome Sanger during quality control assessments because it clustered most closely with other samples 129 sequenced at the same time that had been generated via a different protocol. 130 Differentially expressed genes identified by RNA-seq were called using the DESeq2 R package 131 (42). The design formula for the generalized linear model used with DESeq2 was "~ replicate + 132 condition" where "condition" was the combination of cell type and treatment for each sample. 133 Statistical testing for differential expression within each cell type was performed using the Wald 134 test implemented in the DESeq2 package. Translating RNAs with at least a 2-fold change in 135 their relative abundance with a Benjamini-Hochberg corrected P value (FDR) ≤ 0.05 were 136 considered statistically significant. 137 Human homologs of zebrafish genes were extracted from Ensembl using the BioMart tool. Gene 138 Set Enrichment Analysis(43, 44) was performed by comparing gene expression data mapped to 139 these human homologs to Hallmark gene sets (v6.2) from the Molecular Signatures Database 140 (Broad Institute)(45). The gsea3 java release was run using all default settings. Heatmaps were 141 generated with Multiple Experiment Viewer (MeV) and Venn diagrams were generated and 142 overlaps determined by BioVenn(46). 143 together, as previously described (48). Primer sequences used in this study can be found in 152 Supplemental Table 2. 153

Tail transection 154
C3a.1 +/adults were in-crossed. 3 dpf larvae were wounded by tail transection using a no. 10 155 Feather surgical blade. To visualize neutrophils in the wound microenvironment, the larvae 156 were fixed at 2 hpw or 8 hpw in 4% paraformaldehyde in 1X PBS overnight at 4°C. Sudan 157 Black B staining was performed as described previously (50). Fixed larvae were imaged using 158 a zoomscope (EMS3/SyCoP3; Zeiss; Plan-NeoFluar Z objective) and then genotyped as above. 159 For macrophage quantification, c3a.1 +/adults carrying a mpeg1:GFP transgene (13) were in-160 crossed. At 3 dpf, larvae were pre-screened for fluorescence on a zoomscope. Tail wounding 161 was then performed as described above and the larvae were fixed in 1.5% formaldehyde 162 overnight at 4°C. Fixed larvae were imaged using a zoomscope and genotyped as above. All 163 image analysis was performed using Zen 2012 (blue edition, Carl Zeiss), blinded to genotype. 164

Pseudomonas infections 165
infected larvae were placed into individual wells of a 96 well plate and survival was monitored 176 daily for 5 dpi. For neutrophil recruitment experiments, larvae were fixed at 1 hpi or 6 hpi in 4% 177 paraformaldehyde in 1X PBS overnight at 4°C. Sudan Black B staining was performed as 178 described previously (50), and injection success was further confirmed by visualization of GFP-179 positive bacteria in the otic vesicle on a spinning disk confocal microscope (CSU-X, Yokogawa) 180 as described below, without mounting in agarose. Imaging of the otic vesicle region for 181 neutrophil enumeration was performed using a zoomscope (EMS3/SyCoP3; Zeiss; Plan-182 NeoFluar Z objective). Image analysis was performed using Zen 2012 (blue edition, Carl Zeiss). 183 Photoconversion 184 Adult c3a.1 +/zebrafish carrying an mpx:Dendra2 transgene (54) were in-crossed and embryos 185 collected and incubated to 3 dpf. Larvae were prescreened for fluorescence using a 186 zoomscope (EMS3/SyCoP3; Zeiss; Plan-NeoFluar Z objective) and mounted in ZWEDGI 187 devices as previously described (55). An imaging sequence was performed for each larva 188 comprising an initial series of 2 overlapping Z-stacks of the region of the caudal hematopoietic 189 tissue (CHT) and photoconversion of the neutrophils within the CHT. This was followed by a 190 second series of 2 overlapping Z-stacks to confirm that photoconversion occurred. 191 Photoconversion was performed using a laser scanning confocal microscope (FluoView 192 FV1000; Olympus) with numerical aperature (NA) 0.75/20X objective. The following stimulation 193 settings were used: 40% 405 nm laser transmissivity, 10 µs/pixel dwell time, and 45 second 194 total stimulation time. Larvae were removed from the ZWEDGI devices following 195 photoconversion and subjected to wounding by tail transection as above. Larvae were 196 subsequently imaged live at 3 hpw using a spinning disk confocal microscope as described 197 below and then genotyped as above. Image analysis was performed using Zen 2012 (blue 198 edition, Carl Zeiss), blinded to genotype. 3 dpf c3a.1 +/+ or -/larvae carrying a mpx:mcherry transgene (56) were pre-screened for 201 fluorescence on a zoomscope (EMS3/SyCoP3; Zeiss; Plan-NeoFluar Z objective). For imaging 202 over 1-3 hours, larvae were mounted in ZWEDGI devices as previously described (55) and 203 retained in place using 2% low melting point agarose applied to the head. Images were 204 acquired every 3 minutes using a spinning disk confocal microscope (CSU-X, Yokogawa, NA 205 0.3/10X EC Plan-NeoFluar objective) with a confocal scanhead on a Zeiss Observer Z.1 206 inverted microscope equipped with a Photometrics Evolve EMCCD camera. Each image 207 comprised a 50 µm z-stack, with 11 slices taken at 5 µm intervals. Images were analyzed and 208 maximum intensity projections were made using Zen 2012 (blue edition) software (Carl Zeiss). 209 To track cell motility, time series were analyzed in Imaris (Bitplane) and neutrophil mean track 210 speed, track displacement, and track straightness, as well as instantaneous velocity for each 211 neutrophil at each point in the time series, were calculated using the "spots" tool as previously 212 described (57). To count total neutrophils and quantify neutrophil distribution in 213 photoconversion experiments, 12 overlapping images were acquired to capture the full length 214 and width of each larva and image analysis and neutrophil counts were performed using the 215 Chemotaxis was assessed using a microfluidic device as described previously (58). In brief, 225 For quantification of neutrophil instantaneous speed over time, a linear mixed effect regression 250 model was used. Genotype and time were treated as fixed effects, with experimental replicate, 251 fish, and neutrophil (within fish) treated as random effects. Statistical analyses were performed 252 in R version 3.5.1, using the associated lme4 package. Reported P values are 2-sided and 253 level of statistical significance preset to 0.05, with no adjustment for multiplicity. 254 Results 255 TRAP-RNAseq identifies genes differentially regulated in neutrophils, macrophages, and 256 epithelial cells in response to wounding. Communication between multiple cell types, 257 including leukocytes and epithelial cells, is essential to allow cells of the innate immune system 258 to effectively navigate complex interstitial tissues to reach the wound microenvironment (60). 259 However, the relative transcriptional contributions of each cell type are incompletely understood. 260 To identify cell-specific signals that are differentially expressed in different cell types in response 261 to wounding, we performed a large-scale translating protein affinity purification and RNA 262 sequencing (TRAP-RNAseq) screen (Fig. 1A). Briefly, 3 dpf transgenic zebrafish larvae 263 expressing an EGFP-tagged copy of the ribosomal subunit L10a specifically in neutrophils, 264 macrophages, or epithelium were subjected to multiple fin tissue wounds. 3 hours later, larval 265 tissue was homogenized and ribosomes were isolated with α-GFP immunoprecipitation. RNA 266 was then extracted. Illumina sequencing confirmed expression of a priori-selected, known cell-267 type-specific genes across all analyzed samples, validating our method (Fig. 1B). We then 268 focused our analysis solely on zebrafish genes that have identified human homologs. From 269 these genes, 299 were identified to be at least 2-fold changed (upregulated or downregulated) 270 in neutrophils, 301 in macrophages, and 717 in epithelial cells. In neutrophils, we were 271 surprised to find that only a single gene was statistically significantly upregulated in response to 272 wounding: c3a.1 (Fig. 1D). Although other changes in gene expression did not reach statistical 273 significance, which is likely due to high variability among samples and small numbers of replicates performed, we expect that 2-fold differential expression is potentially biologically 275 relevant. It is interesting to note that relatively few genes were more than 2-fold differentially 276 expressed in more than one cell type ( Fig. 1C and Supplemental Table 3). wounded fish, compared with unwounded controls ( Fig. 2A). Although only c3a.1 showed a 283 statistically significant increase in mRNA expression in neutrophils following wounding, other 284 complement pathway components, including c5 and c9, also showed trends toward increased 285 expression in neutrophils (Fig. 2B). Non-significant increases in c3a.1, c5 and c9 expression 286 were also evident in macrophages. Further, complement factor B (cfb) was one of only 3 287 genes that were differentially expressed in all 3 cell types (Fig. 1C). Taken together, these data (sa31241, Sanger) (61) (Fig. 3A). This premature stop codon occurs prior to the predicted 294 thioester bond and α2 macroglobulin-complement domains of the C3A protein that characterize 295 an anaphylatoxin (62). qPCR of cDNA from pooled 3 dpf c3a.1 -/larvae confirmed loss of c3a.1 296 mRNA, compared with c3a.1 +/+ controls, and showed no significant compensatory upregulation 297 of the other c3a orthologues expressed at this stage of larval development (Fig. 3B).

C3a.1 mediates resistance to Pseudomonas aeruginosa infection in a neutrophil-
with ~50% mortality at 1 dpi. In comparison, we noted no significant change in susceptibility in is predominantly due to a neutrophil-intrinsic function of c3a, possibly due to the reduction in 349 numbers of neutrophils found at the infection site. 350 Depletion of c3a.1 does not alter neutrophil egress from hematopoietic tissues following 351 wounding. We next asked how C3A controls neutrophil mobilization from hematopoietic tissue. 352 In response to inflammatory signals, zebrafish neutrophils may be released directly from 353 hematopoietic tissues or recruited from a population of neutrophils already patrolling in 354 peripheral tissues (67). C3a has been shown in mice to help to retain immature neutrophils in 355 hematopoietic tissues (68, 69). C3 and C3A receptor-deficient mice subsequently have faster 356 and more pronounced neutrophil egress from bone marrow in response to inflammatory stimuli 357 (70). Although this finding is opposite to the decreased neutrophil numbers that we observe at a 358 wound in larval zebrafish (Fig. 1C-D), we still wanted to determine whether decreased 359 neutrophil numbers at inflammatory sites in c3a.1 -/larvae were due to a difference in neutrophil 360 recruitment from hematopoietic tissue. At 3 dpf, the primary organ of hematopoiesis in the 361 larval zebrafish is the caudal hematopoietic tissue (CHT), in which hematopoiesis resembles 362 that within the mammalian fetal liver (71). We crossed the c3a.1-deficient line to the 363 Tg(mpx:dendra2) line, in which neutrophils are labeled with the photoconvertible fluorophore 364 Dendra2, enabling fate tracking of neutrophils originating from the CHT over time (54). We 365 photoconverted neutrophils in the CHT and then subjected the larvae to tail transection. We 366 subsequently counted both the photoconverted neutrophils remaining in the CHT and those 367 mobilized to the periphery at 3 hpw (Fig. S4A). Neither the number of neutrophils retained in 368 the CHT nor the number mobilized neutrophils differed between c3a.1 +/+ and c3a.1 -/larvae ( skin graft sites (78, 79). Using microfluidic devices, we found that the total number of 388 neutrophils arriving at the source of an established gradient of IL-8 did not differ between C3A-389 pre-treated neutrophils and untreated controls, although there was considerable variation 390 between both technical and biological replicates (Fig. 5C). We therefore focused specifically on 391 the neutrophils in each device that eventually reached the chemoattractant source and 392 measured the time at which each neutrophil arrived at the source. Analysis of either the first 393 20% or first 50% of neutrophils to arrive at the source revealed that neutrophils pre-treated with 394 C3A arrived faster than control neutrophils (Fig 5D). These findings suggest that C3a may 395 prime neutrophils to respond more quickly to other exogenous cues.

Loss of c3a.1 impairs neutrophil recruitment in vivo by decreasing neutrophil migration 397
speed early after wounding. Our data thus far suggest a role for C3A in priming neutrophils to 398 migrate effectively toward other chemotactic signals. Thus, we next tested whether the 399 impaired neutrophil recruitment phenotype we observed in c3a.1 -/zebrafish larvae was due to 400 alterations in the dynamics of interstitial migration to the wound. We took advantage of the Following tail transection, we imaged labeled neutrophils in the wound microenvironment for 1 406 hour and tracked the neutrophils using Imaris software (Bitplane) (Fig. 6A, still images, and  407 movie 2). We found that average neutrophil speed during the first 30 minutes after wounding is 408 significantly impaired in c3a.1 -/larvae, compared to c3a.1 +/+ controls. This change in neutrophil 409 migratory behavior is confined to the early post-wounding period, as when speed is averaged 410 over the first 60 minutes post-wound, it is not different between groups (Fig. 6B). The mean 411 displacement and track straightness traveled by the neutrophils also did not differ between 412 groups ( Fig. S5A-B). Decreased neutrophil speed in c3a.1 -/larvae is specific to neutrophil 413 directed migration, as neutrophil random migration in the absence of an inflammatory stimulus is 414 not impaired in c3a.1 -/larvae, and neutrophil random migration speed is in fact slightly faster in 415 c3a.1-deficient zebrafish than in c3a.1 +/+ controls (Fig. S5C). Finally, quantification of each 416 neutrophil's instantaneous speed at 3 minute intervals over the first hour post-wounding shows 417 that neutrophils in c3a.1-intact larvae rapidly achieve and maintain a steady speed toward the 418 wound. In contrast, neutrophils in c3a.1 -/initially migrate significantly more slowly, before 419 accelerating to reach c3a.1 +/+ speeds by about 30 minutes post-wound (Fig. 6C-D). Altogether, 420 these data support the idea that C3A primes neutrophil responses to damaged tissues.

Discussion 422
Here we report, for the first time, the results of a cell-specific translation profiling screen 423 designed to identify genes differentially expressed in the inflammatory context of wounding in 424 the larval zebrafish model. We have previously shown that the signals that guide neutrophils to 425 sites of sterile injury differ from those that regulate migration to bacterial infection; specifically, 426 that, while PI3K signaling is required in both contexts, tissue-generated H2O2 signaling is 427 required for neutrophil responses to wounds, but is dispensable for neutrophil responses to 428 infection (80). Our work here supports the increasing recognition that molecular drivers of 429  We were surprised to find relatively little overlap in the transcriptomes of neutrophils, 435 macrophages, and epithelial cells in that few genes identified by our screen were differentially 436 expressed in more than one cell type. This suggests the presence of a complex network of 437 inter-and intracellular signals, in which cross-communication among cell types is essential for 438 optimal leukocyte recruitment and subsequent wound healing. 439 We identified c3a.1 as the only gene significantly upregulated in neutrophils in response to 440 wounding. This finding is interesting because, while the complement system has been 441 implicated in multiple inflammatory contexts, including wounding, infection, and hematopoiesis, 442 it is best understood in infection, where it functions to opsonize bacteria for phagocytosis or kill 443 them directly via assembly of the membrane attack complex (19, 63). Similar to our finding that 444 c3a.1 -/zebrafish have impaired survival to bacterial infection, mice deficient in either C3 or the activation induced by deletion of the scavenger carboxypeptidase B2 displayed a survival 447 benefit in the context of polymicrobial sepsis (82), confirming a specific, protective role for C3A 448 in infectious inflammatory contexts. However, we find that, in larval zebrafish, C3A acts through 449 neutrophils, as C3A mutation had no further effect when neutrophils were defective. 450 The role of C3A in the context of sterile injury is less well understood. C3A signaling through 451 the C3A receptor (C3AR) is required for hepatocyte proliferation and liver regeneration following In summary, our data identify the complement pathway as a whole, and c3a.1 in particular, as 494 significantly upregulated in neutrophils in response to wounding. Our data further support a neutrophils for efficient migration to other chemoattractants, both in vivo in zebrafish and in vitro 497 in human primary neutrophils; however, the role of autocrine neutrophil C3A-C3AR signaling 498 warrants further investigation. On the basis of these observations, we conclude that C3A plays 499 an underappreciated role in mediating neutrophil migration. By exploiting the genetic resources 500 of the larval zebrafish model, we and others are well positioned to further investigate the role of 501 neutrophil-derived C3A in optimizing neutrophil recruitment to wounding. Finally, these results 502 support the power of TRAP in the identification of cell type specific changes in gene expression 503 that may influence inflammation and wound healing.