Title : 2 Identification of organ-specific transcriptomic shifts in the vasculature during 3 systemic inflammation using TrendCatcher 4

44 A recent analysis of endothelial cell (EC) gene expression suggests that ECs of distinct organs 45 vary in their gene expression profiles and respond distinctly to the systemic inflammatory stimulus 46 of endotoxemia (Jambusaria et al., 2020). There is a need to systematically define tissue-specific 47 gene expression dynamics in response to inflammation but such an analysis is in part limited by 48 the availability of appropriate algorithms to analyze differential expression across a time course. 49 Here, we present TrendCatcher (https://jaleesr.github.io/TrendCatcher_1.0.0/), an R package 50 designed for time course RNA-seq data analysis which identifies distinct dynamic transcriptional 51 programs. When applied to ECs, we observed that approximately 85-95% of EC genes in all three 52 organs followed a biphasic response following endotoxemia. The rapid upregulation of innate 53 immune response and bacterial response genes occurred within 6 hours in all three vascular beds 54 but the subsequent upregulation of reparative EC mitosis and cell cycle genes occurred most 55 rapidly in lung ECs (24-48h) and was most delayed in cardiac ECs (72h-168h). The distinct 56 kinetics of EC inflammatory injury and regeneration identify vascular-bed specific temporal 57 windows for targeted therapeutics. 58


59
Endothelial cells (ECs) form the inner cellular lining of all blood vessels and establish a barrier 60 between the circulating blood and the underlying parenchyma (Aird, 2012;Bautch, 2011). ECs 61 perform a wide range of critical functions in many physiological processes, such as maintaining 62 intravascular homeostasis, trafficking nutrients and waste between blood and the parenchymal 63 tissues, as well as initiating and regulating immune responses (Liao, 2013). Despite sharing

74
To investigate the transcriptomic basis of EC heterogeneity, unbiased transcriptomic 75 profiling is essential for providing precise insights into the underpinning molecular mechanisms.

96
Recent studies on organ-specific EC transcriptomic profiles have focused on differences 97 between ECs at baseline but there is a growing interest in how organ-specific EC heterogeneity

147
(more complicated dynamic response, e.g., a combination of 2 or more basic types of trajectories).

148
As shown in Figure 2A-C, compared to DESeq2, DESeq2Spline and ImpulseDE2, TrendCatcher 149 achieved higher accuracy in a mixed simulated dataset. We also tested each model's 150 performance on a varying number of time points. As shown in Figure 2D, TrendCatcher had the 151 highest prediction accuracy across all time points tested, with accuracies of more than 75%.

152
We also generated 1,000 gene trajectories for each type of dynamic profile to compare

182
TrendCatcher also assigned trajectory pattern types to all dynamic genes identified from 183 post LPS-injury, and provided hierarchical pie charts to visualize the composition of trajectory 184 patterns across the three organs. Each dynamic gene's trajectory pattern was characterized using 185 a hierarchical scheme, a master-pattern type and a sub-pattern type. A master-pattern type 186 describes the basic type of gene expression trend, for example, "up-down" represents a biphasic 187 8 response, upregulation followed by downregulation. The sub-pattern type shows the time-188 dependent dynamic behavior of a gene transcriptional response. For example, 189 "0h_up_24h_down_168h" belongs to an "up-down" master-pattern type and represents peak 190 gene upregulation at 24 hours, followed by downregulation until 168 hours.

191
We observed the same 8 major master-patterns across the three organs ( Figure 3A) but 192 with sub-pattern composition shifts (Figure 3-figure supplement 1-3), especially for "up-down" 193 biphasic response with distinct upregulation time. As the inner hierarchical pie charts of Figure   194 3B-D show, we found over 85-95% of all the dynamic genes from brain, heart and lung ECs had   response to interferon-beta (14 genes), response to interferon-gamma (17 genes) (Figure 4B).

215
These shared early activation pathways suggest an organ-universal mechanism to LPS-induced 216 inflammatory injury, in which ECs across multiple major organs activate self-defense pathways 217 shortly after injury, with a strong enrichment of interferon response genes.

218
We also assessed organ-specific pathways that were upregulated early on at 6 hours after 219 LPS injury. Brain-specific inflammatory responses were enriched for increased neutrophil 220 migration and regulation of angiogenesis ( Figure 4C). However, pathways related to Wnt 221 signaling were downregulated (Figure 4-figure supplement 1B), which is important because 222 Wnt signaling is a key regulator of the integrity of the blood brain barrier (BBB) (Laksitorini,

230
At 24h, brain ECs upregulated ribosome biogenesis processes, mirroring the upregulation 231 of these processes that we had observed in heart ECs at the earlier 6h time point (Figure 4-figure   232 supplement 2B). At this time point, heart ECs concomitantly upregulated ER stress response 233 pathways as well as proteasomal degradation pathways (Figure 4-figure supplement 2C). In 234 lung ECs, the bulk of upregulated genes occurred at the 6h time point but we also identified 235 selected genes peaking later at 24h, however they were also related to pro-inflammatory 236 processes such as leukocyte migration (Figure 4-figure supplement 2D), suggesting that 237 compared to the heart and the brain, the lung endothelium showed the most prominent sustained 238 upregulation of pro-inflammatory genes during 6h-24h.

256
The upregulation of EC migration and angiogenesis during the late phase (72h-168h) suggests 257 that after inflammatory injury and proliferation, repair of the lung endothelial barrier likely involves 258 EC migration.

259
We also applied the Time-Heatmap analysis to brain ECs and heart ECs. We observed a 260 similar juxtaposition of early host-defense responses followed by EC proliferation and 261 regeneration (Figure 5-figure supplement 1 and 2). However, the regenerative responses in 262 heart ECs showed a significantly delayed activation which was primarily seen at 168h, i.e., 1 week 263 after the injury.

264
For mechanistic analyses, it is also critical to identify individual genes which drive the 265 biological processes which is why we developed a function allowing for a visualization of fold-266 11 change (logFC) expression of enriched gene sets from selected target GO terms. As shown in 267 Figure 6A, during the 0-6h time window the GO term defense response was enriched in brain, 268 heart and lung ECs, but in each vascular bed different sets of dynamic genes were involved. For 269 brain ECs (Figure 6A), the highly dynamic DDEGs were tumor necrosis factor (Tnf), aconitate    Figure 7A, Supplementary file 5). We found the following selected genes which constitute the 288 organ-specific EC signatures during homeostasis are modulated during inflammatory injury. For 289 brain ECs, 7 out of 10 highly expressed brain EC signature genes were disrupted by the 290 inflammatory process (Figure 7A, Supplementary file 5), most of which showed rapid 12 downregulation. Only apolipoprotein D followed an "up-down" pattern, which was upregulated in 292 the first 72 hours after inflammation and returned to baseline after 72 hours. For heart ECs ( Figure   293 7B, Supplementary file 5), butyrophilin like 9 (Btnl9) and aquaporin-7(Aqp7) followed a 294 monotonic increase pattern and increased almost 2-fold at 168 hours after inflammatory 295 activation. For lung ECs (Figure 7C, Supplementary file 5), only resistin like alpha (Retnla) was 296 upregulated in the first 6 hours and then downregulated to baseline. In contrast to brain ECs, the 297 top 10 signature genes for heart and lung ECs remained stable during the injury phase suggesting 298 that brain ECs exhibit the greatest vulnerability to losing their homeostatic brain EC signature in

404
There are two types of break points, type I means gene up-regulated followed by a down 435 regulation and type II means gene expression level down regulated and then followed by 436 an upregulation. By screening along the break point, the master-pattern and sub-pattern 437 were assigned to each gene.