Temporally integrated single cell RNA sequencing analysis of controlled and natural primary human DENV-1 infections

Controlled dengue human infection studies present an opportunity to address many longstanding questions in the field of flavivirus biology. However, limited data are available on how the immunological and transcriptional response elicited by an attenuated challenge virus compares to that associated with a wild-type DENV infection. To bridge this knowledge gap, we utilized scRNAseq to analyze PBMC from individuals enrolled in a DENV-1 controlled human challenge study and from individuals experiencing a natural primary DENV-1 infection. While both controlled and natural DENV infection resulted in overlapping patterns of inflammatory gene upregulation, natural DENV infection was accompanied with a more pronounced suppression in gene products associated with protein translation and mitochondrial function, principally in monocytes. This suggests that the immune response elicited by controlled and natural primary DENV infection are similar, but that natural DENV infection has a more pronounced impact on basic cellular processes to induce a multi-layered anti-viral state

Dengue is one of the most widespread vector-borne viral diseases in the world. The causative agent-75 dengue virus (DENV) -is a positive-stranded RNA virus maintained in an anthroponotic cycle between 76 the Aedes aegypti mosquito and humans (1). Consisting of four co-circulating but immunologically and 77 genetically discrete serotypes , DENV is thought to infect up to 300 million 78 individuals yearly (2, 3). Although the majority of DENV infections are subclinical, as many as 100 79 million infections every year result in symptomatic dengue fever. In its most severe manifestation, 80 dengue fever can progress to dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS) (4-7). 81 While the pathogenesis of severe dengue is complex and may involve some degree of genetic 82 predisposition, severe symptoms are more likely to occur in individuals previously infected with a 83 heterologous viral serotype compared to individuals without any preexisting DENV immunity. Despite 84 decades of study, the precise mechanisms underpinning this unique epidemiological feature of DENV 85 infection remain unresolved and continues to impede the development of an effective DENV vaccine. conventional monocytes respond most robustly to infection across all subjects and study groups from an 135 unbiased transcriptional perspective. Using these data, conserved Differentially Expressed Genes 136 (cDEGs) induced or suppressed by natural or experimental primary DENV were identified, and the 137 overlap between the two arms of the study assessed. The infection-induced cDEGs associated with 138 experimental DENV infection were found to reflect a subset within the larger gene set associated with 139 natural primary DENV infection, primarily corresponding to gene products associated with a cellular 140 response to systemic inflammation and interferon (IFN) production. In contrast, the number infection-141 suppressed cDEGs was higher in cells obtained following natural primary DENV infection than in cells 142 obtained following experimental DENV infection. Infection-suppressed cDEGs primarily corresponded 143 to gene products associated with protein translation/elongation and mitochondrial function, two cellular 144 processes known to be suppressed by prolonged IFN signaling. These results are consistent with the 145 concept that the immune response elicited by DHIM represents a tempered version of that generated in 146 response to a natural DENV infection, and that the more pronounced inflammation associated with 147 natural primary DENV infection has a correspondingly more pronounced impact on basic cellular 148 processes to induce a multi-layered/systemic anti-viral state. These data provide insight into the 149 molecular level response to DENV infection, and how viral pathogenesis correlates with immune 150 activation and cellular pathophysiology. 151

Subject selection and characterization 155
The primary objective of this study was to determine the kinetic transcriptional signature associated with 156 experimental DENV-1 infection and to determine how closely this profile correlates with the 157 transcriptional signature accompanying natural primary DENV-1 infection. To this end, three subjects 158 from the SUNY/WRAIR DENV-1 DHIM study were selected for analysis ( Figure 1A) (10). All 159 subjects included in this study received 3.25 x 10 3 PFU of the 45AZ5 DENV-1 challenge virus strain 160 following extensive pre-screening to ensure an absence of preexisting DENV immunity. All three 161 subjects exhibited significant DENV-1 RNAema between 5 and 15 days post-challenge and 162 demonstrated classic staggered IgM/IgG seroconversion between study days 13 and 16 ( Figure 1A). 163 PBMC from a total of 8 time points per subject were selected for this study, corresponding to study days 164 0, 2, 4, 6, 8, 10, 14/15, and 28 ( Figure 1B). In addition, PBMC from two children experiencing 165 serologically-confirmed natural primary DENV-1 infections were selected for analysis ( Figure 1B In order to assess the transcriptional signature associated with experimental and natural primary DENV 174 infection, high-throughput scRNAseq analysis was performed using the 10xGenomics 5' capture gene 175 expression platform with both TCR and BCR recovery. Samples were sequenced to achieve an average 176 depth of 108,000 reads per cell, with an average of 5,707 cells captured per library (Supplemental 177 Table 2). This final dataset contains a total of 171,208 high quality cells and 22 statistically distinct 178 populations corresponding to all major anticipated leukocyte subsets (Figure 2A, Figure 2B, 179 Supplemental Table 3, Supplemental Table 4 plasmablasts, 15 monoclonal antibodies (mAbs) were synthesized from DHIM subject 0002 (5 IgM  209 isotype, 5 IgG isotype, 5 IgA isotype) and tested for their ability to bind DENV-1, -2, -3, and -4 210 (Supplemental Table 5). Consistent with previously published reports describing the antigen 211 specificity of natural or experimental primary DENV infection (14, 15), 2 of 15 (13%) of the tested 212 mAbs exhibited DENV-binding activity, including serotype specific and cross-reactive profiles 213 (Supplemental Table 6). While the overall magnitude of plasmablast expansion observed following 214 DHIM-1 challenge is lower than what has been described in natural primary DENV-1 infection (13, 14, 215 16, 17), these data suggest that the B cell activation profile and antigen specificity associated with 216 DHIM mirrors that observed in natural primary DENV-1 infection. 217   Table 7). The genes that were 257 most consistently and significantly upregulated following experimental DENV-1 infection primarily 258 corresponded to interferon-induced gene products (e.g. IFITM1, IFI6, ISG15, and MX1) and other genes 259 associated with acute inflammation (e.g. TRIM22 and LY6E) (Supplemental Table 7). The expression 260 of these gene products consistently peaked on study days 10 -14, and trended towards baseline by study 261 day 28 ( Figure 4C, Supplemental Table 7). The few cDEGs that were repressed in response to 262 experimental DENV-1 infection primarily corresponded to ribosomal protein subunits (e.g. RPL4, RPL5 263 and RPL6) translation elongation products (e.g. EEF2, EIF3L and EIF4B) mitochondrial associated gene 264 products (e.g. MT-CYB and MT-ND4) ( Figure 4D, Supplemental Table 7). This transcriptional profile 265 is consistent with canonical IFN-associated suppression of protein translation and mitochondrial 266 biogenesis and highlights a key mechanistic response to acute viral infection (18,19).  Table 8). In contrast to the modest number of cDEGs 293 suppressed in response to experimental DENV1 infection, a significant fraction of the cDEGs identified 294 in natural primary DENV-1 infection were suppressed relative to baseline ( Figure 5B). However, 295 similarly to those suppressed conserved DEGs identified following experimental DENV-1 infection, the 296 suppressed cDEGs identified following natural primary DENV-1 infection primarily corresponded to 297 ribosomal protein subunits (e.g. RPL4, RPL5, and RPL6), translation elongation products (e.g. EEF2, 298 EIF3L, and EIF4B) and mitochondrial associated gene products (e.g. MT-CYB, and MT-ND4) ( Figure  299 5D, Supplemental Table 8). These data suggest that natural primary DENV infection not only induces 300 the expression of interferon-induced gene products in a wide range of cell type, but that the 301 physiological consequence this response is profound and widely distributed. 302 303

Transcriptional profile overlap between controlled and natural primary DENV infection 304
Having identified cell-specific conserved sets of differently expressed genes present in PBMC 305 subpopulations following experimental and natural primary DENV infection, we assessed the overlap 306 between these two datasets to compare the conserved transcriptional response to experimental and 307 natural primary DENV infection. In light of the fact that cells obtained 10 days post infection in the 308 DHIM study contained the most cDEGs relative to baseline, this time point was selected for comparison 309 against the natural primary infection dataset. 310

311
Although the overall number of infection-induced cDEGs identified in cells from natural primary 312 DENV-1 infection was greater than that observed following experimental DENV-1 infection, the cell 313 population specific frequencies of infection-induced cDEGs in each population were highly correlated 314 between groups (R 2 = 0.6825, p < 0.0001) ( Figure 6A). Furthermore, the infection-induced cDEGs 315 identified following experimental DENV were determined to consistently represent a subset of the 316 cDEGs induced in natural primary DENV-1 infection ( Figure 6B, Table 1). Gene ontogeny analysis 317 revealed that the infection-induced cDEGs that were found in common to experimental and natural 318 primary DENV infection datasets fell into classic interferon response pathways and negative regulation 319 of viral replication ( Table 2). Those cDEGs which were unique to natural primary DENV-1 infection 320 corresponded to gene pathways involved in cytokine secretion, antigen processing/presentation, and 321 detection of viruses ( Table 2). 322

323
In contrast to the high degree of correlation observed in the cell-population specific frequencies of 324 infection-induced cDEGs, infection-suppressed cDEGs were overwhelming restricted to those cells 325 obtained following natural primary DENV infection ( Figure 6C). However, the few infection-326 suppressed cDEGs observed in cells following experimental DENV infection generally represented a 327 subset of those suppressed following natural primary DENV infection ( Figure 6D, Table 3). Infection-328 suppressed cDEGs overwhelming represented gene products associated with protein 329 translation/elongation, as well as mitochondrial function and biogenesis ( Table 3, Supplemental Table  330 8). These data suggest that both natural and experimental primary DENV-1 infection upregulate a 331 similar acute transcriptional program in similar cells, but that the greater magnitude of the that response 332 elicited by natural primary DENV-1 infections has a significantly more profound physiological response 333 on basic cellular processes, broadly suppressing protein translation and mitochondrial function. 334 Furthermore, these data indicate that monocytes are uniquely responsive to both experimental and 335 natural primary DENV infection. 336

DISCUSSION 338
In this study, we utilized high-throughput single-cell RNA sequencing (scRNAseq) technology to assess 339 the longitudinal transcriptional profile in both controlled and natural primary DENV-1 infection with 340 single-cell resolution. Core sets of conserved Differentially Express Genes that were induced or 341 suppressed by either natural or experimental primary DENV were identified, and the overlap between 342 the study groups assessed. The infection-induced cDEGs associated with experimental DENV infection 343 were found to reflect a subset within the larger gene set associated with natural primary DENV 344 infection, primarily corresponding to gene products associated with a cellular response to systemic 345 inflammation and interferon production. In contrast, infection-suppressed cDEGs were much more 346 common in cells obtained from natural primary DENV infection than in cells obtained from 347 experimental DENV infection. Infection-suppressed cDEGs primarily corresponded to gene products 348 associated with protein translation/elongation and mitochondrial function, two cellular processes known 349 to be suppressed by IFN signaling. These results are consistent with the concept that the immune 350 response elicited by DHIM represents a tempered version of that generated in response to a natural 351 primary DENV infection, but that the more pronounced inflammation associated with natural primary 352 DENV infection has a correspondingly more pronounced impact on basic cellular processes to induce a Bangkok, Thailand, the design of which has been previously described (33,34). In brief, the study 417 enrolled children who presented to the hospital with acute febrile illness. Blood samples were obtained 418 daily during illness and at early and late convalescent time points; the term 'fever day' is used to report 419 acute illness time points relative to Day 0, defined as the day of defervescence. The infecting virus type 420 (DENV-1-4) was determined by RT-PCR and/or virus isolation as previously described (35) Multi-sample integration, data normalization, dimensional reduction, visualization, and differential gene 466 expression were performed using the R package Seurat (v3.1.4) (38, 39). All datasets were filtered to 467 only contain cells with between 200-6,000 unique features and <10% mitochondrial RNA content. To 468 eliminate erythrocyte contamination, datasets were additionally filtered to contain cells with less than a 469 5% erythrocytic gene signature (defined as HBA1, HBA2, HBB). Data were scaled, transformed, and 470 variable genes were identified using the SCTransform() function. SelectIntegrationFeatures() and 471 PrepSCTIntegration() functions were used to identify conserved features for dataset integration, and 472 final dataset anchoring/integration were performed using FindIntegrationAnchors() and IntegrateData() 473 functions, with the day 0 DHIM samples and day 180 natural primary infection samples used as 474 reference datasets. PCA was performed using variable genes defined by SCTransform() additionally 475 filtered to remove TCR V/D/J or BCR k/l gene segments. The first 33 resultant PCs were used to 476 perform a UMAP dimensional reduction of the dataset (RunUMAP()) and to construct a shared nearest 477 neighbor graph (SNN; FindNeighbors()). This SNN was used to cluster the dataset (FindClusters()) with 478 default parameters and resolution set to 0.4. 479 480 Following dataset integration and dimensional reduction/clustering, gene expression data was 481 loge(UMI+1) transformed and scaled by a factor of 10,000 using the NormalizeData() function. This 482 normalized gene expression data was used to determine cellular cluster identity by utilizing the Seurat 483 application of a Wilcoxon rank-sum test (FindAllMarkers()), and comparing the resulting differential 484 expression data to known cell-linage specific gene sets. Differential gene expression analysis between Transfection grade plasmids were purified by maxiprep and transfected into a 293-6E expression 503 system. Cells were grown in serum-free FreeStyle 293 Expression Media (Thermo Fisher), and the cell 504 supernatants collected on day 6 for antibody purification. Following centrifugation and filtration, the cell 505 culture supernatant was loaded onto an affinity purification column, washed, eluted, and buffer 506 exchanged to the final formulation buffer (PBS). Antibody lot purity was assessed by SDS-PAGE, and 507 the final concentration determined by 280nm absorption. The clonotype information for all monoclonal 508 antibodies generated as part of this study is listed in Supplemental Table 5. 509 510 Viruses: DENV1-4 (strains Western Pacific 1974, S16803, CH53489, and TVP-360, respectively) 511 propagated in C6/36 mosquito cells were utilized for ELISA. Virus for ELISA was purified by 512 ultracentrifugation through a 30% sucrose solution and the virus pellet was resuspended in PBS. 513 514 Monoclonal antibody DENV-capture ELISA: Monoclonal antibody DENV reactivity was assessed 515 using a 4G2 DENV capture ELISA protocol. In short, 96 well NUNC MaxSorb flat-bottom plates were 516 coated with 2 µg/ml flavivirus group-reactive mouse monoclonal antibody 4G2 (Envigo Bioproducts, 517 Inc.) diluted in borate saline buffer. Plates were washed and blocked with 0.25% BSA + 1% Normal 518 Goat Serum in PBS after overnight incubation. DENV-1, -2, -3 or -4 (strains Western Pacific 1974, 519 S16803, CH53489, and TVP-360, respectively) were captured for 2 hours in the appropriate wells, 520 followed by extensive washing. Serially diluted monoclonal antibody samples were incubated for 1hr at 521 RT on the captured virus, and DENV-specific antibody binding quantified using anti-human IgG HRP 522 (Sigma-Aldrich, SAB3701362). Secondary antibody binding was quantified using the TMB Microwell 523      Supplemental Table 1. Natural primary DENV infection subject information 807 Supplemental Table 2. Sample sequencing metrics 808 Supplemental Table 3. Sample/population frequency: T cell populations 809 Supplemental Table 4. Sample/population frequency: B cells and myeloid linage cells 810 Supplemental Table 5. mAb sequence information 811 Supplemental Table 6. mAb EC50 from DHIM subject 002, day 15 post infection 812 Supplemental Table 7. Conserved differentially expressed genes: day 10 DHIM 813 Supplemental Table 8. Core differentially expressed genes: natural DENV-1 infection 814 Supplemental Table 9. Antibodies used for flow cytometry 815 816