Transcriptional Analysis of Lymphoid Tissues from Infected Nonhuman Primates Reveals the Basis for Attenuation and Immunogenicity of an Ebola Virus Encoding a Mutant VP35 Protein

Zaire Ebola virus (EBOV) infection causes a severe and often fatal disease characterized by inflammation, coagulation defects, and organ failure driven by a defective host immune response. Lymphoid tissues are key sites of EBOV pathogenesis and the generation of an effective immune response to infection.

ABSTRACT

Infection with Zaire ebolavirus (EBOV), a member of the Filoviridae family, causes a disease characterized by high levels of viremia, aberrant inflammation, coagulopathy, and lymphopenia. EBOV initially replicates in lymphoid tissues and disseminates via dendritic cells (DCs) and monocytes to liver, spleen, adrenal gland, and other secondary organs. EBOV protein VP35 is a critical immune evasion factor that inhibits type I interferon signaling and DC maturation. Nonhuman primates (NHPs) immunized with a high dose (5 × 10⁵ PFU) of recombinant EBOV containing a mutated VP35 (VP35m) are protected from challenge with wild-type EBOV (wtEBOV). This protection is accompanied by a transcriptional response in the peripheral blood reflecting a regulated innate immune response and a robust induction of adaptive immune genes. However, the host transcriptional response to VP35m in lymphoid tissues has not been evaluated. Therefore, we conducted a transcriptional analysis of axillary and inguinal lymph nodes and spleen tissues of NHPs infected with a low dose (2 × 10⁴ PFU) of VP35m and then back-challenged with a lethal dose of wtEBOV. VP35m induced early transcriptional responses in lymphoid tissues that are distinct from those observed in wtEBOV challenge. Specifically, we detected robust antiviral innate and adaptive responses and fewer transcriptional changes in genes with roles in angiogenesis, apoptosis, and inflammation. Two of three macaques survived wtEBOV back-challenge, with only the nonsurvivor displaying a transcriptional response reflecting Ebola virus disease. These data suggest that VP35 is a key modulator of early host responses in lymphoid tissues, thereby regulating disease progression and severity following EBOV challenge.

IMPORTANCE Zaire Ebola virus (EBOV) infection causes a severe and often fatal disease characterized by inflammation, coagulation defects, and organ failure driven by a defective host immune response. Lymphoid tissues are key sites of EBOV pathogenesis and the generation of an effective immune response to infection. A recent study demonstrated that infection with an EBOV encoding a mutant VP35, a viral protein that antagonizes host immunity, can protect nonhuman primates (NHPs) against lethal EBOV challenge. However, no studies have examined the response to this mutant EBOV in lymphoid tissues. Here, we characterize gene expression in lymphoid tissues from NHPs challenged with the mutant EBOV and subsequently with wild-type EBOV to identify signatures of a protective host response. Our findings are critical for elucidating viral pathogenesis, mechanisms of host antagonism, and the role of lymphoid organs in protective responses to EBOV to improve the development of antivirals and vaccines against EBOV.

INTRODUCTION

Zaire ebolavirus (EBOV), a member of the single-stranded RNA Filoviridae family, is the causative agent of Ebola virus disease (EVD) (1–4). EVD case fatality rates range from 40% to 90%; the disease is characterized by high viral replication, dysregulated coagulation, unbridled inflammation, lymphopenia, and multiorgan failure leading to death (5–11). Antigen-presenting cells, notably dendritic cells (DCs) as well as monocytes and macrophages, are early and major targets of EBOV infection that act as vehicles for systemic dissemination (8, 12–15). While infection of monocytes/macrophages is associated with aberrant inflammatory responses, EBOV infection of DCs inhibits their maturation and ability to mobilize the host adaptive immune response (4, 8, 12–17). The 2013–2016 West Africa epidemic and the recent outbreaks in the Equateur, Kivu, and Ituri provinces of the Democratic Republic of the Congo (DRC) emphasize the need to understand the mechanisms governing EBOV virulence to develop effective prophylactics and postexposure therapeutics (1, 2, 18, 19).

Inhibition of DC maturation is mediated by the VP35 protein, which plays critical functions in the viral life cycle and in antagonizing the host innate antiviral immune response (20, 21). Viral double-stranded RNA (dsRNA) is detected by the RIG-1-like receptors (RLRs) RIG-1 and MDA5, which leads to the phosphorylation, activation, and translocation of interferon (IFN) regulatory factor 3 (IRF3) to the nucleus to induce the transcription of type I IFN and interferon-stimulated genes (ISGs) (22, 23). VP35 dysregulates this antiviral response by binding dsRNA and interacting with the host protein PACT, thereby preventing IRF3 phosphorylation, which ultimately inhibits RLR-mediated expression of IFN and ISGs (15, 24–29). Clear correlations have been shown between the levels of IRF3 phosphorylation, DC maturation, IFN production, and induction of ISGs in vitro (24, 26, 27, 30–39).

We recently reported that nonhuman primates (NHPs) infected with a high dose (5 × 10⁵ PFU) of a VP35 mutant (VP35m) (EBOV Mayinga backbone) harboring three mutations (F239A, R319A, and R332A) generated a regulated innate immune response and a protective robust adaptive immune response that protected them against back-challenge with a lethal dose of the Kikwit variant of wild-type EBOV (wtEBOV) (40). This protection was mediated by tightly controlled induction of innate antiviral immune pathways and expression of adaptive-immunity-associated transcripts. Evidence reflecting these changes at the protein level included transient activation of DCs and monocytes, increased frequencies of memory T cells and B cells, and the presence of virus-specific IgG titers in whole blood (WB). However, analysis of the host response in lymphoid organs, critical sites of early immune responses, was not performed.

As lymphoid organs largely function in immune defense and are early sites of viral replication and dissemination, these tissues significantly influence the course of EVD (13, 41, 42). In this study, we investigated the host response to a 2 × 10⁴ PFU dose of VP35m in lymphoid organs, including draining inguinal lymph nodes (IngLN), nondraining axillary lymph nodes (AxLN), and spleen tissues. By infecting macaques with a log-lower dose of virus than previously described and examining early time points postinfection, we aimed to capture early host responses to viral replication in these key tissues before systemic dissemination (40). Our results demonstrate a distinct transcriptional profile in lymphoid organs following VP35m infection compared to challenge with the wtEBOV Makona variant. Splenic transcriptional responses of animals infected with VP35m that survived back-challenge with wtEBOV Kikwit differed from those of the animal that succumbed to back-challenge. Overall, these data elucidate molecular mechanisms of EBOV immune evasion and highlight the pivotal role of early immune responses in lymphoid tissues modulating EVD progression.

RESULTS

VP35m induces an antiviral defense response in whole blood.

We conducted a transcriptional analysis of whole blood (WB), axillary lymph nodes (AxLN), inguinal lymph nodes (IngLN), and spleen tissues collected at 2, 3, and 4 days postchallenge (dpc) from cynomolgus macaques challenged intramuscularly (i.m.) with 2 × 10⁴ PFU of VP35m (Fig. 1A). None of the animals exhibited detectable viremia or signs of severe systemic infection (Table 1) (40). Low levels of viral RNA were detected in tissues of one animal, while no viral RNA was detected in the blood of all animals (Table 2). Principal-component analysis (PCA) of WB samples from VP35m-infected macaques revealed that 0- and 3-dpc samples clustered similarly, while dimensional separation was noted for 2- and 4-dpc samples (Fig. 1B). Differentially expressed gene (DEG) counts demonstrated that transcriptional changes were mostly upregulated and predominantly induced at 2 dpc (Fig. 1C). Since few DEGs were detected at 3 and 4 dpc in WB samples, all DEGs at 2 to 4 dpc were combined for our subsequent analyses. Functional enrichment of WB DEGs using Metascape software indicated overrepresentation of Gene Ontology (GO) terms associated with antiviral immunity and innate immunity, e.g., “response to virus,” “cellular response to interferon gamma,” and “regulation of innate immune response” (Fig. 1D) (43). Notable DEGs that enriched to these GO terms included those involved in antiviral defense (e.g., MX1 and HERC5), sensing of viral nucleic acid (e.g., DHX58 and OAS3), type I and II interferon signaling (e.g., STAT1, STAT2, and IFNGR), and T cell function (e.g., IDO1 and LGLAS9).

FIG 1.

RNA-Seq and flow cytometry analysis of VP35m- and EBOV-infected whole blood. (A) Study experimental design. (B) Principal-component analysis (PCA) of all whole-blood (WB) samples from macaques challenged with VP35m from 0, 2, 3, and 4 dpc. (C) Number of differentially expressed genes (DEGs) expressed in WB at 2, 3, and 4 dpc (n = 1 each) relative to 0 dpc with a Venn diagram of DEGs at the corresponding days. (D) GO term network depicting functional enrichment of DEGs expressed in WB at 2 to 4 dpc using Metascape. Clustered nodes correspond to one GO term, with the interior pie chart representing the proportion of DEGs unique to either condition or common to both. The node size represents the number of DEGs associated with the GO term. Gray lines represent shared interactions between GO terms, with density and number indicating the strength of connections between closely related GO terms. (E) Venn diagram of DEGs detected in the whole blood of VP35m (2 to 4 dpc), EBOV Makona (4 dpc), and EBOV Kikwit (4 dpc) infections. No DEGs were detected before 4 dpc in EBOV Makona and EBOV Kikwit. Violin plots represent the log₁₀ fold change (relative to 0 days postinfection) of the DEGs shared by all three different infections. (F) Number of viral transcripts (in RPKM) detected in WB for VP35m, Makona, and Kikwit infections at 2 to 4 dpc, 2 to 4 dpc, and 4 dpc, respectively. (G) Flow cytometry analysis of PBMCs from macaques following VP35m infection. Innate immune and adaptive immune cell frequencies are represented as frequencies of the total live-cell population. Red represents an elevated frequency, and blue represents a reduced frequency. Each column represents the mean frequency.

TABLE 1.

Clinical findings in VP35m-challenged cynomolgus macaques

Animal ID (wt)	Target dose(s)	VP35m challenge			wtEBOV back-challenge
		Viremia (log10 PFU/ml)	Clinical signb	Outcome	Viremia (log10 PFU/ml) (dpc)	Clinical sign(s) (dpc)b	Outcome
Serial sacrifice study
Cyno1 (5.44 kg)	20,000 PFU VP35m	None	None	Euthanized at 2 dpc for tissue collection
Cyno2 (5.26 kg)	20,000 PFU VP35m	None	None	Euthanized at 3 dpc for tissue collection
Cyno3 (4.90 kg)	20,000 PFU VP35m	None	None	Euthanized at 4 dpc for tissue collection
Low-dose study
Cyno4a (4.38 kg)	20,000 PFU VP35m, 1,000 PFU wtEBOV Kikwit	None	None	Survived to 28 dpc	4.12 (6), 7.18 (9)	Fever (6), anorexia (8–9), mild petechial rash (7), moderate petechial rash (8, 9), depression (7–9), leukopenia (6), thrombocytopenia (6), lymphopenia (6), BUN +++ (9), CRE +++ (9), ALT ++ (6) and +++ (9), AST ++ (6) and > (9), ALP +++ (9), GGT +++ (9), CRP increase (6, 9)	Euthanized at 9 dpc
Cyno5a (7.40 kg)	20,000 PFU VP35m, 1,000 PFU wtEBOV Kikwit	None	None	Survived to 28 dpc	None	None	Survived to 28 dpc
Cyno6a (5.48 kg)	20,000 PFU VP35m, 1,000 PFU wtEBOV Kikwit	None	None	Survived to 28 dpc	None	None	Survived to 28 dpc

^aHistorical sample (39).

^bThe day after challenge is in parentheses. Fever is defined as a temperature greater than 2.5°F above the baseline, at least 1.5°F above the baseline and ≥103.5°F, or 1.1°F above the baseline and ≥104°F. Lymphopenia, leukopenia, and thrombocytopenia are defined by a >40% drop in numbers of lymphocytes, leukocytes, and platelets, respectively. Abbreviations: BUN, blood urea nitrogen; CRE, creatinine; ALT, alanine aminotransferase; AST, aspartate aminotransferase; ALP, alkaline phosphatase; GGT, gamma-glutamyltransferase; CRP, C-reactive protein. Plus signs indicate increases in liver enzymes (ALT, AST, ALP, and GGT) or renal function test values (BUN and CRE): +, 2- to 3-fold increase; ++ >3- up to 5-fold increase; +++, >5-fold increase. CRP increase refers to values of >20. Shading indicates that data were not available for the given animals.

TABLE 2.

Viral RNA loads in VP35m-challenged macaques^a

^aViral RNA load measures as the log(average copies/g or copies/ml). * denotes historical sample (39). AxLN, axillary lymph node; IngLN, inguinal lymph node; LN, lymph node; DPC, days postchallenge; LOD, limit of detection. Gray boxes indicate data that were not obtained for the given tissue.

We also compared DEGs detected in WB at 2 to 4 dpc following VP35m infection to those detected at 2 to 4 dpc with wild-type EBOV (wtEBOV) Kikwit (Kikwit) and wtEBOV Makona (Makona) (Fig. 1E) (8, 44). A much larger number of DEGs was detected following Kikwit challenge than with both VP35m and Makona infections (Fig. 1E). The 29 DEGs common to all three infections mainly enriched to GO terms related to inflammatory processes (e.g., SERPINB1), cytosolic nucleic acid detection (e.g., DDX60L and IFIT5), and other antiviral defense mechanisms (e.g., GBP3 and HERC6). Greater upregulation of these genes was observed following Makona and Kikwit infections than following VP35m infection, which corresponded to increased viral reads early after challenge (Fig. 1E and F).

VP35m activates immune cells in the blood.

We performed flow cytometry of peripheral blood mononuclear cells (PBMCs) to gain an understanding of immune cell dynamics and integrate these findings with those of our transcriptional analysis (Fig. 1G). Overall, changes in immune cell frequencies paralleled innate and adaptive immune transcriptional signatures. Total levels of monocytes increased, with nonclassical (CD16⁺) monocyte populations increasing at 6 to 14 dpc, while activated (CD86⁺) monocyte subsets increased at 3 dpc (Fig. 1G). Activated dendritic cell (DC) subsets were also detected in all animals by as early as 3 dpc (myeloid DCs [mDCs] and other DCs) and as late as 6 dpc (plasmacytoid DCs [pDCs]). B cell populations, including proliferating B cells, generally increased at 6 to 14 dpc. In contrast, more dynamic and rapid changes were seen in CD4 and CD8 T cell subsets. Nearly all CD4 T cell subsets analyzed were elevated at 3 to 6 dpc. Conversely, increases in the levels of memory and naive CD8 T cells were not observed until 10 to 28 dpc.

VP35m induces overlapping transcriptional responses in different lymphoid tissues.

To uncover tissue-specific mechanisms of VP35m attenuation, we then evaluated transcriptomes of noninfected versus VP35m-challenged macaque AxLN, IngLN, and spleen tissues. Since animals were not perfused prior to lymphoid tissue collection, DEGs detected in the WB at the same time points were subtracted from the DEGs detected in tissues to focus the analysis on lymphoid tissue-specific transcriptional changes (Fig. 2A). PCA of expressed genes showed that AxLN and IngLN clustered separately from spleen samples obtained from infected animals (Fig. 2B). The 500 most variable genes enriched to GO terms associated with vascular development, cytokine signaling, innate immune processes, and stress (Fig. 2C). A substantial number of DEGs were expressed in the lymph nodes (LNs) (>2,000 each), while a smaller number (∼500), mostly downregulated, were expressed in the spleen (Fig. 2D). In line with the PCA (Fig. 2B), AxLN and IngLN samples exhibited the greatest overlap, sharing about 60% of total DEGs (Fig. 2E). A considerable portion of DEGs (292) detected in the spleen was shared with both LNs (Fig. 2E).

FIG 2.

RNA-Seq analysis of VP35m-infected tissues. (A) Venn diagram of DEGs detected in each VP35m-infected lymphoid tissue at days 2, 3, and 4 dpc and whole blood at days 1 to 4 postinfection. (B) PCA of axillary lymph node (AxLN), inguinal lymph node (IngLN), and spleen grouped by naive (0 dpc) and infected (2, 3, and 4 dpc) conditions. (C) Functional enrichment of the 500 most variable genes extracted from PCA. Horizontal bars represent the number of genes mapping to each GO term, with the color intensity representing the negative log of the FDR-adjusted P value [−log(q-value)]. (D) Number of differentially expressed genes (DEGs) for AxLN, IngLN, and spleen relative to naive tissues (n = 3 per tissue). (E) Venn diagram of total DEGs from each tissue from all time points (2 to 4 dpc). (F) Bubble plot representing functional enrichment of total DEGs from each tissue from all time points (2 to 4 dpc). The color intensity of each bubble represents the negative log of the FDR-adjusted P value (−logqvalue), and the relative size of each bubble represents the number of DEGs belonging to the specified Gene Ontology (GO) term. MAPK, mitogen-activated protein kinase.

Functional enrichment of total DEGs in each tissue revealed that DEGs detected in IngLN, AxLN, and spleen enriched to many similar pathways (Fig. 2F). DEGs in all tissues enriched to GO terms associated with gene expression (“viral gene expression”), innate immunity (“neutrophil degranulation”), cellular stress (“apoptotic signaling pathway”), and adaptive immunity (“lymphocyte activation” and “TCR [T cell receptor] signaling pathway”). Only spleen DEGs enriched to GO terms associated with “type I interferon signaling,” “response to interferon gamma,” and “B cell-mediated immunity.” Although DEGs detected in all three lymphoid tissues enriched to similar GO terms, they exhibited a mixture of decreased and increased expression in LNs, and DEGs were more frequently downregulated in the spleen (Fig. 3 and 4).

FIG 3.

Comparison of “lymphocyte activation” and “cytokine-mediated signaling pathway” DEG expression patterns in VP35m-challenged tissues. Heat maps represent the expression of DEGs from the GO terms “lymphocyte activation” for AxLN (A), IngLN (B), and spleen (C) and “cytokine-mediated signaling pathway” for AxLN (D), IngLN (E), and spleen (F). Expression is represented as the normalized RPKM. The range of colors for each heat map is based on scaled and centered RPKM values of the represented DEGs. Red represents upregulation, and blue represents downregulation. The first columns of each heat map represent the median RPKM of uninfected tissue samples. Each column of “infected” represents the RPKM value for one infected tissue sample at 2, 3, and 4 dpc.

FIG 4.

Comparison of DEG expression patterns for GO terms common or unique to multiple VP35m-challenged tissues. (A to C) Heat maps representing DEGs from the GO terms “viral gene expression” and “translation” for AxLN (A), IngLN (B), and spleen (C). (D and E) Heat maps representing DEGs from the GO term “myeloid cell differentiation” for AxLN (D) and spleen (E). (F) Heat map representing the DEGs from GO terms unique to the spleen in Fig. 1D: “response to virus,” “type I interferon signaling pathway,” and “response to interferon gamma.” Expression is represented as the normalized RPKM. The range of colors for each heat map is based on scaled and centered RPKM values of the represented DEGs. Red represents upregulation, and blue represents downregulation. The first columns of each heat map represent the median RPKM of uninfected samples. Each column of “infected” represents the RPKM value for one infected tissue sample at 2, 3, and 4 dpc.

DEGs shared or unique to VP35m-infected tissues reflect characteristics of EVD.

DEGs in all three lymphoid tissues enriched to “lymphocyte activation” (Fig. 2F and Fig. 3A to C). PRF1, which encodes perforin, was the only upregulated DEG shared by all tissues and is involved in NK and T cell cytotoxicity. Genes that play a role in T cell- and B cell-mediated immunity were detected in both LNs, with some upregulated (e.g., ZBTB7B and TCIRG1) and others downregulated (e.g., CD28 and SELENOK) (Fig. 3A and B). CCL19, which elicits T cell and B cell migration to secondary lymphoid organs via the chemokine receptor CCR7, was also upregulated in AxLN and IngLN. Upregulated DEGs unique to either IngLN (e.g., JAK3 and TCF7) or AxLN (e.g., SLA2 and ZAP70) were mostly associated with T cell activation. In contrast, DEGs enriching to “lymphocyte activation” in the spleen were largely downregulated and played a role in the response to type I IFN signaling (e.g., IRF4), antibody production (e.g., IGHV3-30), antigen presentation (e.g., HLA-DMB), and cell metabolism (e.g., IMPDH1) (Fig. 3C). Antiapoptotic DEGs such as BAK1 in the spleen and TNFRSF1B in the LNs were upregulated.

DEGs detected in each of the three lymphoid tissues also enriched to the GO term “cytokine-mediated signaling pathway,” with the most overlap occurring between the lymph nodes (Fig. 3D and E). All DEGs detected in AxLN and IngLN exhibited similar directions of change (Fig. 3D and E). Upregulated DEGs detected solely in LNs mapped to GO terms that play a role in innate antiviral signaling (e.g., IFNAR1 and MAVS), proinflammatory (e.g., IL17RA and TRAF2,) and anti-inflammatory (e.g., SIGIRR and RNF31) processes, and chemotaxis (e.g., ITGAX and CCL21) (Fig. 3C and D). Downregulated DEGs shared by these two tissues were also associated with inflammation (e.g., NFKB1 and ANAX1) and chemotaxis (e.g., CCL11). Conversely, downregulated DEGs common to all tissues included proinflammatory (e.g., NFKBIA and FOS), angiogenic (e.g., HIF1A and S1PR1), proapoptotic (e.g., FOS), and chemotactic (e.g., CCL3L1 and CCL3) genes, suggesting a general downregulation of inflammation and apoptosis following VP35m infection. Additional DEGs that play a role in antiviral immunity (e.g., ISG15 and IFI27) and lymphocyte regulation (e.g., IL-4 and IRF4) were unique to the spleen (Fig. 3F).

DEGs detected in each of the three tissues also enriched to the GO terms “translation” and “viral gene expression” (Fig. 4A to C). These DEGs were predominantly downregulated in all tissues. Common downregulated DEGs encoded ribosomal proteins (e.g., RPL29) and translation-regulatory factors (e.g., EIF1) or played a role in T cell activation (CD28). Several downregulated DEGs shared by LNs also function in transcription and translation regulation (e.g., ZNF639 and RGS2) (Fig. 4A and B). In contrast, the few upregulated DEGs shared between the AxLN and IngLN were involved in antiviral immunity (TRIM14) and RNA metabolism (e.g., CNOT3). Upregulated DEGs unique to each tissue were related to general metabolic (e.g., DPH1 in AxLN) and translation (e.g., ABTB1 in IngLN) processes (Fig. 4A and B). DEGs detected solely in the spleen were involved in RNA interference (e.g., RBM3 and AGO2) and cellular proliferation (e.g., BTG2 and RNF139) (Fig. 4C).

Only AxLN and spleen DEGs enriched to the GO term “myeloid cell differentiation,” with AxLN samples possessing a greater number of DEGs than spleen samples (Fig. 4D and E). The few shared DEGs were downregulated in both tissues and played a role in cell growth and proliferation (e.g., KLF10 and JUNB) and the stress response (e.g., KLF2) (Fig. 4D and E). DEGs mapping to these GO terms and detected only in the AxLN were associated with T cell activation and development (e.g., JAK3 and TCIRG1), myeloid cell migration (e.g., CSF1R), and innate immune cell differentiation (e.g., PTPN2) (Fig. 4D). DEGs related to B cell signaling (LYN) and VDJ recombination (HMGB2) were downregulated in AxLN. The few DEGs unique to the spleen were downregulated and related to myeloid cell proliferation (PURB) and angiogenesis (MYC) (Fig. 4E).

Examination of DEGs unique to the spleen revealed increased expression of pathways involved in antiviral defense (“response to virus,” “type I interferon signaling pathway,” and “response to interferon gamma”) (Fig. 4F). Analogous to the WB analysis, DEGs related to viral nucleic acid detection (e.g., DDX58 and DDX60) and interferon-induced responses (e.g., ISG15 and IFI27) were upregulated; downregulated DEGs played a role in inflammation (e.g., CCL4), chemotaxis (e.g., CXCR4 and VIM), and antigen presentation (e.g., HLA-C and HLA-B).

To identify clusters of genes similarly and differently expressed across all three tissues, we applied a two-ways forward-regression model using MaSigPro (Fig. 5). We identified 4 significant gene clusters (Fig. 5A to D). Genes in cluster 1 were sharply downregulated at 2 dpc for all tissues (Fig. 5A). In contrast, genes in clusters 2 and 4 were upregulated at 2 dpc in AxLN and IngLN before plateauing and upregulated at 3 dpc in spleen before declining at 4 dpc (Fig. 5B and D). Genes in cluster 3 were progressively upregulated in AxLN and IngLN through infection but featured a decrease at 2 dpc for spleen before returning to baseline levels (Fig. 5C).

FIG 5.

Two-ways forward-regression analysis of VP35m-infected lymphoid tissues. (A to D) Gene clusters 1 to 4 identified by MaSigPro. (E) Bubble plot depicting functional enrichment of genes belonging to clusters 1 to 4 identified in panels A to D. The size of the bubble represents the number of genes enriching to the given GO term, while the color represents the FDR-adjusted P value [−log(q-value)]. (F to H) Heat maps representing genes enriching to GO terms “lymphocyte activation” for clusters 2 and 3 (F) and “defense response to virus” for cluster 2 (G) and cluster 4 (H). Expression is represented as the normalized RPKM. The range of colors is based on scaled and centered RPKM values of the represented DEGs. Red represents upregulation, and blue represents downregulation. The first columns of each heat map represent the median RPKM of uninfected samples. Subsequent columns represent the RPKM of one infected tissue on the indicated day (2, 3, or 4 dpc).

The 73 genes in cluster 1 play a role in translation (Fig. 5E), while genes in clusters 2 to 4 enriched to GO terms associated with innate and adaptive immune processes. Cluster 2 mainly consisted of genes involved in leukocyte activation (e.g., “lymphocyte activation”), antiviral defense (e.g., “defense response to virus”), and stress responses (“regulation of cellular response to stress”) (Fig. 5E). Similar findings were determined for genes in cluster 3, although these genes were also involved in innate immune processes such as “myeloid leukocyte activation.” Genes enriching to “lymphocyte activation” were involved in T cell-mediated (e.g., cluster 2 genes IL27RA and ZAP70 and cluster 3 genes TREML2 and TNFSF9) and humoral (e.g., cluster 2 genes FCRL3 and GON4L and cluster 3 genes LYL1 and IGHV3-21) immunity (Fig. 5E and F). Genes in clusters 2 and 4 also enriched to GO terms related to antiviral defense (e.g., cluster 2 genes MAVS and OAS3 and cluster 4 genes AIM2 and DDX60) in addition to other innate processes such as “response to wounding” and “positive regulation of cytokine production” (Fig. 5E and G).

VP35m challenge induces fewer changes in the expression of proangiogenic, -apoptotic, and -inflammatory genes than wild-type EBOV challenge.

We next compared the transcriptional responses in VP35m-challenged lymphoid tissues to those recently reported in lymphoid tissues of cynomolgus macaques infected with Makona (1 to 4 dpc) (45) (Fig. 6). DEGs detected in WB were removed from those detected in tissues to focus our analysis on tissue-specific transcriptional changes. To gain a general understanding of VP35m viral replication dynamics in our NHP model, we compared viral reads in Makona- versus VP35m-infected tissues and detected higher viral transcript levels in the spleen for both virus variants (Fig. 6A). Makona-infected versus VP35m-infected macaques had increased splenic viral reads. Comparison of tissue DEGs showed a significant overlap that was most notable in AxLN and IngLN (Fig. 6B to D). Functional enrichment of unique and shared DEGs was performed to understand the biological implications of these transcriptional profiles (Fig. 6E to G and Fig. 7).

FIG 6.

Transcriptional responses in VP35m- and wild-type EBOV (wtEBOV)-challenged lymphoid tissues are distinct. (A) Number of viral transcripts (in RPKM) detected in VP35m- and Makona-infected lymphoid tissues at 2 to 4 dpc. * denotes a statistical significance value of <0.05 (P value of 0.0128). (B to D) Venn diagrams of DEGs from AxLN (B), IngLN (C), and spleen (D) challenged with VP35m compared to the respective tissues challenged with wtEBOV Makona. (E to G) Heat maps representing functional enrichment of DEGs unique to Makona-infected tissue, unique to VP35m-infected tissue, or common to both infections for AxLN (E), IngLN (F), and spleen (G). The color intensity represents the statistical significance, shown as the negative log of the FDR-adjusted P value [−log(q-value)], with the range of colors based on the GO terms with the lowest and highest −log(q-value) values for the entire set of GO terms per tissue. Numbers enriching to each GO term per column are represented in each box; blank boxes indicate no statistical significance.

FIG 7.

Comparison of GO term differentially expressed genes (DEGs) in Makona- and VP35m-infected tissues. (A to C) Heat maps for AxLN infections representing DEGs from the GO terms “response to wounding,” “blood coagulation,” “angiogenesis,” and “acute inflammatory response” unique to AxLN-Makona infection (A) and DEGs from the GO term “T cell proliferation” that are shared between Makona and VP35m infections (B) and unique to VP35m infection (C). (D to G) Heat maps for IngLN infections representing IngLN tissue DEGs from the GO terms “blood vessel development” unique to IngLN-Makona (D), “cellular response to type I interferon” and “response to interferon gamma” unique to IngLN-Makona (E), “lymphocyte activation” common to both infections (F), and “apoptotic signaling” common to both infections (G). (H to K) Heat maps for spleen infection representing DEGs from the GO terms unique to spleen-Makona in Fig. 3E (H), “apoptotic signaling” common to both infections and unique to spleen-VP35m (I), “lymphocyte activation” common to both infections and unique to spleen-Makona (J), and “lymphocyte activation” common to both infections and unique to spleen-VP35m and “B cell-mediated immunity” unique to spleen-VP35m (K). Expression is represented as the normalized RPKM. The range of colors for each heat map is based on scaled and centered RPKM values of the represented DEGs. Red represents upregulation, and blue represents downregulation. The first columns of each heat map represent the median RPKM of uninfected samples. Each column of “infected” represents the RPKM value for one infected tissue sample at 2, 3, and 4 dpc. Each column represents the median RPKM of infected tissue samples taken at 2, 3, and 4 dpc for a singular infection (i.e., Makona or VP35m).

DEGs detected in AxLN following both VP35m (AxLN-VP35m) and Makona (AxLN-Makona) infections enriched to GO terms related to gene expression and innate immune responses, such as “translation,” “cytokine-mediated signaling pathway,” and “myeloid leukocyte activation” (Fig. 6E). DEGs detected only in AxLN-Makona infection enriched to inflammatory, stress, and cardiovascular processes as well as humoral immunity. Notable DEGs enriching to these GO terms played a role in blood vessel development (e.g., SERPINF1 and ENG), endothelial growth factor signaling (e.g., VEGFB and PDGFRB), cell adhesion (e.g., CD81 and ICAM1), migration (e.g., CCL2 and CXCL13), and complement (e.g., C1S and CR2) (Fig. 7A). DEGs detected only in AxLN-VP35m also enriched to GO terms associated with cell-cell interactions as well as those associated with T cell proliferation (Fig. 6D and Fig. 7B and C).

While DEGs detected solely in Makona-IngLN enriched to GO terms associated with inflammation, vascular development, and interferon signaling, common DEGs detected in VP35m-IngLN and Makona-IngLN indicated primarily innate and adaptive immunity gene signatures (Fig. 6E). As described above for AxLN-Makona, many upregulated genes played a role in angiogenesis (e.g., FGFR1 and PLXND1) and type I IFN signaling (e.g., IFIT2 and IFI27) (Fig. 7D and E). DEGs unique to IngLN-VP35m and common to both IngLN-Makona and IngLN-VP35m also enriched to “lymphocyte activation” (e.g., CD274, VSIR, TNFSF13B, and VCAM1) (Fig. 7F). Finally, shared DEGs enriched to apoptosis, inflammation, and hypoxia signaling pathways (Fig. 7G).

For spleen samples, DEGs following VP35m or Makona infection primarily enriched to GO terms associated with gene expression and signaling, such as “viral gene expression,” “cytokine-mediated signaling pathway,” and “apoptotic signaling pathway” (Fig. 6F). Unlike DEGs unique to spleen-VP35m, DEGs unique to spleen-Makona were linked to antiviral defense and inflammatory pathways such as “cellular response to dsRNA” and “IκB kinase/NF-κB signaling” (e.g., TNFSF4) (Fig. 6F and Fig. 7H). VP35m-spleen DEGs enriched to the GO terms “apoptotic signaling pathway” (e.g., CASP10) and “lymphocyte activation” (e.g., BCL6 in VP35m-infected animals and LCP1 in Makona-infected animals) (Fig. 7I to K).

VP35m-mediated protection against wtEBOV back-challenge is associated with a robust adaptive immune response in the spleen.

An additional three cynomolgus macaques were challenged with a low dose of VP35m and then back-challenged with a lethal dose (1,000 PFU) of Kikwit 28 days later (Fig. 1A). Two of three macaques survived back-challenge up to 28 dpc with no signs of illness, while one macaque succumbed (Fig. 1A; Table 1) (40). High levels of viral RNA were detected in the AxLN, IngLN, and spleen of the nonsurvivor (Table 2). Spleen samples taken at the time of euthanasia (9 dpc for the nonsurvivor and 28 dpc for survivors) for survivors were analyzed to delineate gene signatures associated with disease outcome. PCA shows that splenic transcriptional signatures were unique for the nonsurvivor (Fig. 8A). Considerably more DEGs were detected in the nonsurvivor (2,219) than in the survivors (502) (Fig. 8B).

FIG 8.

The nonsurvivor of back-challenge lacks an adaptive immune response in the spleen. (A) PCA of naive (n = 4), survivor (n = 2), and nonsurvivor (n = 1) spleen tissues following challenge with VP35m and back-challenge with EBOV Kikwit. Naive spleen tissues were derived from a recent study (45). DPC, days postchallenge. (B) Venn diagram of total differentially expressed genes (DEGs) from the nonsurvivor and survivors. (C) Heat map representing functional enrichment of DEGs to survivor-unique, nonsurvivor-unique, and shared DEGs following back-challenge. The color intensity represents the statistical significance {shown as the negative log of the FDR-adjusted P value [−log(q-value)]}, with the range of colors based on the GO terms with the lowest and highest −log(q-value) values for the entire set of GO terms per tissue. Numbers enriching to each GO term per column are represented in each box; blank boxes indicate no statistical significance. (D and E) Heat maps representing DEGs enriching to all common GO terms identified in panel C (D) and “lymphocyte activation” for nonsurvivors (E). Expression is represented as the normalized RPKM. The range of colors is based on scaled and centered RPKM values of the represented DEGs. Red represents upregulation, and blue represents downregulation. The first columns of each heat map represent the median RPKM of uninfected samples. Subsequent columns represent either the RPKM of the nonsurvivor’s infected spleen or the median RPKM of the survivors’ spleens.

Enrichment of DEGs was performed to further elucidate distinct gene signatures associated with disease outcome following back-challenge (Fig. 8C). DEGs unique to the nonsurvivor overwhelmingly enriched to GO terms related to innate and adaptive immunity, such as “myeloid leukocyte activation” and “adaptive immune response” (Fig. 8C). Small numbers of DEGs mapping to these networks were also shared with survivors (Fig. 8C). Shared downregulated DEGs played a role in cell recruitment (e.g., GPR183 and CXCL8), regulation of inflammation (e.g., NFKBIZ and SMAD7), and adaptive immune responses (e.g., IGLV3-19 and CD28); upregulated common DEGs were involved in apoptosis (e.g., LGALS3 and BRI3) and T cell signaling (e.g., IL7R and RIPK2) (Fig. 8D). Several innate antiviral defense genes like MAVS were upregulated in the nonsurvivor (Fig. 8D). The 60 most downregulated DEGs detected only in the nonsurvivor enriched to “lymphocyte activation” and included genes important for humoral (e.g., CD27 and MZB1) and cellular (e.g., CD40LG and CCR7) immunity as well as innate immune processes like complement activation (e.g., CR2 and CR1) and NK cell-mediated immunity (e.g., NCR1 and CD244) (Fig. 8E).

We then compared the transcriptomes of survivors to that of the nonsurvivor following wtEBOV Kikwit back-challenge (Fig. 9). DESeq2 analysis was performed given that we had access to tissues from only one nonsurvivor and two survivors. We identified ∼1,258 significant genes (SGs) (log₂ fold change cutoff of 6 in either direction) in the nonsurvivor, with 929 downregulated and 329 upregulated relative to the survivors. SGs downregulated in the nonsurvivor enriched to “adaptive immune response,” “leukocyte migration,” and “chemokine-mediated signaling” GO terms. Analysis of individual SGs indicated a lack of antibody production (e.g., JCHAIN and CD19) and T cell activation (e.g., LTA, GZM, and CD28) (Fig. 9A to C). In contrast, SGs upregulated in the nonsurvivor enriched to GO terms characteristic of fatal EVD: “blood vessel development,” “response to wounding,” and “regulation of hemostasis” (Fig. 9D). The increased expression levels of SGs related to blood coagulation (e.g., PLG, FGA, and FGG), cellular stress (e.g., CRP), and angiogenesis (e.g., TFPI2) further support this observation (Fig. 9E).

FIG 9.

The transcriptional response to back-challenge is distinct in nonsurvivors. (A and B) Functional enrichment of downregulated (A) and upregulated (B) genes in the nonsurvivor relative to survivors using DESeq2 analysis without replicates. Horizontal bars represent the number of genes mapping to each GO term, with the color intensity representing the negative log of the FDR-adjusted P value [−log(q-value)]. (C to E) Heat maps representing DEGs from the GO terms “adaptive response” (C), “leukocyte migration” and “chemokine-mediated signaling pathway” (D), and “blood vessel development,” “response to wounding,” and “regulation of hemostasis” (E). Red represents upregulation, and blue represents downregulation. The first columns of each heat map represent the median RPKM of uninfected samples. Subsequent columns represent the RPKM for the nonsurvivor and the median RPKM for survivors. Flow cytometry analysis following VP35m challenge separated DEGs into animals surviving (n = 2) and succumbing (n = 1) to later back-challenge (n = 1) (Fig. 4). (F and G) Innate and adaptive immune cell frequencies in survivors (F) and adaptive immune cell frequencies in the nonsurvivor (G), represented as frequencies of the total live-cell population. Red represents an elevated frequency, and blue represents a reduced frequency. Each column for survivors represents the mean frequency. Due to poor data quality, innate immune cell data were not collected for the nonsurvivor. CM, central memory; EM, effector memory.

We next used flow cytometry to analyze circulating immune cells (Fig. 9F and G). Due to a paucity of cells, we were unable to analyze innate immune cells collected from the nonsurvivor. In survivors, we detected decreases in total classical (CD16⁻) and nonclassical (CD16⁺) monocyte populations as well as activated (CD86⁺) subsets at 6 dpc (Fig. 9F). Frequencies of activated myeloid DCs also decreased following back-challenge, while activated classical monocyte and pDC subsets increased at 6 dpc. Changes in adaptive immune cells were more dynamic for survivors, with slight increases in proliferating T and B cell subsets 3 to 6 days following back-challenge (Fig. 9F). In contrast to survivors, the nonsurvivor experienced an overall decrease in proliferating T cell subsets, while the frequency of proliferating B cells increased (Fig. 9F and G).

DISCUSSION

Several in vitro and small-animal studies clearly demonstrate the role of EBOV VP35 in immune evasion (15, 24, 26, 27, 32–36, 39). We recently showed that NHPs infected with a high 5 × 10⁵ PFU dose of recombinant EBOV expressing mutant VP35 protein (VP35m) generated innate and adaptive immune responses that fully protected these animals against a subsequent wild-type EBOV (wtEBOV) challenge (40). To improve our understanding of VP35m-mediated immune protection early in infection, we analyzed the host transcriptional response at major sites of EVD pathogenesis at 2 to 4 dpc. Axillary lymph node (AxLN), inguinal lymph node (IngLN) (draining LN), and spleen tissues were collected from cynomolgus macaques challenged intramuscularly with a lower 2 × 10⁴ PFU dose of VP35m to capture initial virus-host interactions before widespread viral dissemination. EBOV replicates early in draining lymph nodes and is believed to induce changes in gene expression at these sites before the onset of clinical symptoms, viremia, and differential gene expression in peripheral blood (13, 14). Moreover, as seen in our recent study with wtEBOV Makona, earlier transcriptional responses were seen in lymph nodes and spleen 2 to 4 days following infection with VP35m than in whole blood (WB), confirming the role of these tissues in determining the trajectory of the host response (45).

The host transcriptional response in WB at early time points following VP35m infection was markedly different from that observed following either EBOV Makona or EBOV Kikwit infection, particularly in terms of the reduced magnitude of induced DEGs, which enriched almost exclusively to antiviral defense GO terms. The proinflammatory and antiviral DEGs detected early in the WB of VP35m-infected animals and shared with wild-type infections were also less deregulated. Of notable interest, EBOV viral reads were lower in WB and tissues obtained from VP35m-infected animals than in wild-type-challenged animals.

Among VP35m-infected lymphoid tissues, early transcriptional responses were largely overlapping in all three lymphoid tissues but distinct from other variants, with a particular upregulation of genes that play a role in innate antiviral defense and adaptive immunity. Genes involved in EVD-related pathology, including apoptosis, inflammation, and angiogenesis, were both up- and downregulated in all three tissues and among the most variable genes detected. This is dissimilar from transcriptional responses detected in Makona-infected tissues, which featured an overwhelming upregulation of genes related to severe EVD-associated apoptosis, coagulation, blood vessel development, and inflammation early in infection and either delayed or downregulated innate antiviral processes. Transcriptional findings in tissues indicated early activation of antiviral and innate immunity processes following VP35m infection that likely promoted adaptive immunity (46, 47). These findings complement our flow cytometry analysis of PBMCs from challenge, which demonstrated the early mobilization of activated monocytes, DCs, and both T and B lymphocytes. However, both responses are different from that of animals challenged with wild-type strains of EBOV, where innate and adaptive immune cell populations generally decline over the course of infection (8). These data suggest that the remarkably early innate antiviral defense in lymphoid tissues and WB may be the basis of VP35m attenuation and blunted replication. This early response may drive the induction of an adaptive immune response and the absence of EVD pathology that we observed here and a recent report (40).

Transcriptional indicators of an antiviral response in lymphoid tissues were induced early following VP35m infection. Our bivariate analysis comparing infected tissues to uninfected tissues involved the subtraction of WB DEGs to focus on tissue-specific transcriptional changes. Longitudinal analysis without WB gene exclusion revealed a robust induction of antiviral genes in all tissues, although upregulation was seen at 2 dpc for LNs, similar to WB transcriptomics, and at 3 dpc for the spleen. The differential pattern of antiviral gene expression in the spleen suggests a distinguishing role for the spleen in EVD pathology.

Genes related to T cell activation were both up- and downregulated DEGs in VP35m-infected lymphoid tissues, particularly the LNs. Our longitudinal analysis using a two-ways forward-regression strategy showed that T cell-related genes were upregulated earlier in LNs than in the spleen. Interestingly, we detected downregulation of CTLA-4 in all tissues and upregulation of CD274 (PDL-1) in LNs, which was also seen in Makona infection, suggesting alternative mechanisms of T cell regulation in VP35m infection. Previous transcriptional studies have correlated enhanced early T cell activation and cytotoxic responses with fatal cases of EVD (48). Studies in human cases have also correlated dysregulated T cell responses with poor EVD prognosis (49, 50). The reduced expression of T cell-related DEGs in VP35m-infected tissues at these early time points may suggest a more tightly regulated T cell response than with Makona infection, where the expression of DEGs that play a role in cytotoxic processes and T cell activation was mostly increased. An in-depth analysis of the specificity of the T cell response in VP35m-infected tissues should be performed to investigate and clarify this hypothesis.

Additionally, the expression of genes related to humoral immunity in VP35m-infected AxLN and IngLN showed a mixed pattern of gene induction and repression, whereas the expression of similar genes in the spleen was almost uniformly downregulated or present at lower levels. This is significant because the generation of an antibody response and the magnitude of the memory B cell proliferative burst are associated with survival in EBOV-infected patients and NHPs (51–55). These gene expression changes differ from those in our recent study where we reported an induction of genes important for both humoral and cellular immunity in PBMCs from VP35m high-dose-challenged animals that was paralleled by increases in IgG titers (40). However, the dose used in this study was 50-fold lower than that of the previous study.

Animals infected with a lower dose of VP35m were not uniformly protected following lethal challenge with wtEBOV Kikwit, providing us with a unique opportunity to explore transcriptional signatures associated with disease outcomes in the spleen. The one nonsurvivor harbored higher levels of viral RNA in infected tissues than survivors. Animals that survived back-challenge exhibited transcriptional changes indicative of an adaptive immune response that was absent in the nonsurvivor. Likewise, genes related to T and B cell activation were mostly downregulated, while those related to EVD (e.g., blood coagulation, stress, and innate immune cell mobilization) were overwhelmingly upregulated in the nonsurvivor. Our detection of decreased expression of FOXP3 may signify a lack of appropriate immune regulation in the nonsurvivor as well. Changes in peripheral immune cells indicated primarily poor and/or delayed induction of memory T and B cells in the nonsurvivor after back-challenge, supporting these transcriptional data. Moreover, tissue-specific changes in the nonsurvivor were more reminiscent of those seen in wtEBOV Makona infection. Collectively, these data suggest that a regulated adaptive immune response in lymphoid organs correlates with survival and that, at the time of collection of survivor samples (28 dpc), resolution of the host immune response was occurring.

Our study has several caveats. One caveat is the limitations imposed by biosafety level 4 (BSL4) safety regulations and practices. Maximum-containment settings with a lethal pathogen prevent the use of a large number of animals per cohort and the collection of many samples at various time points. There are also significant ethical considerations and limitations for NHP studies, hence the use of only three animals per tissue analyzed here in an attempt to reduce the overall number of animals used, which is strongly encouraged by our IACUC. Second, a limitation of our transcriptional comparison of VP35m- and Makona-infected lymphoid tissues is that the backbone of VP35m is derived from the EBOV Mayinga variant and not Makona. Additionally, NHPs were back-challenged with Kikwit following infection with VP35m. EBOV variants Mayinga, Kikwit, and Makona share 97% nucleotide sequence identity, yet small amino acid differences can elicit distinct transcriptional responses, as we reported recently (8, 56). The different doses may also have an effect. Thus, future studies would ideally compare host responses in lymphoid tissues challenged with VP35m and wtEBOV with identical doses of the same EBOV variants for the most representative comparisons and with sufficiently powered sample sizes to clearly define outcome-associated disease signatures. These studies will also perfuse animals prior to tissue collection to focus on tissue-specific transcriptional changes. Our strategy of subtracting WB DEGs from tissue DEGs inherently eliminates the detection of DEGs found in both the WB and tissues. Third, our understanding of the immune cell dynamics in the nonsurvivor of the back-challenge was limited by the lack of cells. This prohibits a comprehensive understanding of the nonsurvivor disease signature. Fourth, our results lack cell-specific resolution of transcriptional changes. While single-cell sequencing delivers data-rich outcomes, the throughput is relatively low, with typically only a few hundred cells being analyzed per experiment (57). Here, we analyzed bulk genetic signatures to understand transcriptional dynamics at the tissue level. Our future studies will seek to resolve the effects of VP35m at the cell-specific level in our nonhuman primate model.

In conclusion, our data provide crucial insights into VP35-mediated immune antagonism early in infection in lymphoid organs, key sites of pathogenesis. Transcriptional responses indicate that an adaptive immune response is activated in lymphoid organs within days of exposure near the site of injection. Ultimately, this leads to the formation of a robust antiviral defense response, which may lead to protective, regulated humoral and cellular memory responses in macaques surviving back-challenge. Furthermore, there is a distinct lack of proangiogenic, -apoptotic, and -inflammatory DEGs in VP35m infection in comparison to wtEBOV infection. Lymphoid organs have key roles in VP35-mediated immune evasion early in infection. The data here are instrumental in increasing our understanding of EBOV immune evasion and will be critical for refining strategies for EBOV vaccine development and EVD therapeutics.

MATERIALS AND METHODS

Study design.

All work with infectious EBOV and cynomolgus macaques was conducted at the biosafety level 4 (BSL4) laboratory in the Galveston National Laboratory at the University of Texas Medical Branch. Six cynomolgus macaques (Macaca fascicularis) were challenged intramuscularly (i.m.) in the caudal thigh muscle with 2 × 10⁴ PFU of a recombinant EBOV (Mayinga variant backbone) containing three point mutations (F239A, R332A, and K319A) in VP35 (VP35m) (GenBank accession number NC_002549.1) (40). Peripheral whole blood (WB) and lymphoid tissues (axillary lymph node [AxLN], inguinal lymph node [IngLN], and spleen) were collected from three animals 2 days (n = 1), 3 days (n = 1), and 4 days (n = 1) after VP35m challenge (Fig. 1A) (40). The remaining three macaques were back-challenged 28 days later with 1,000 PFU of EBOV Kikwit (Kikwit) (GenBank accession number AY354458) (40). Two of the animals survived, while one succumbed at 9 days postchallenge (dpc) (see reference 40 for details). Blood and spleen samples were collected at the time of euthanasia (Fig. 1A). Clinical pathology data were acquired and viral RNA quantification by quantitative PCR (qPCR) was performed as described previously (39).

Library generation and sequencing.

The quality of RNA extracted from WB and tissue samples was determined using an Agilent 2100 bioanalyzer. cDNA libraries were constructed using the TruSeq stranded total RNA LT-LS kit after rRNA depletion. Adapters were ligated to cDNA products, and the subsequent ∼300-bp products were PCR amplified and selected by size exclusion. An Agilent 2100 bioanalyzer was used to quantify and assess the quality of libraries prior to sequencing on the single-end 75-bp Illumina NextSeq500 sequencing platform. Naive lymphoid tissues (0 dpc), infected lymphoid tissues (1 to 4 dpc), and infected whole blood (2 and 4 dpc) from EBOV Makona-infected cynomolgus macaques and whole blood collected from Kikwit-infected cynomolgus macaques were collected from previous studies, which were processed as described above (8, 45).

Bioinformatics analysis.

Data analysis was performed with the RNA-Seq workflow module of systemPipeR developed by Backman and Girke (58). RNA-Seq reads were demultiplexed, quality filtered, and trimmed using Trim Galore (average phred score cutoff of 30 and minimum length of 50 bp). The FastQC function was used to generate quality reports. Hisat2 was used to align reads to the reference genome of Macaca fascicularis (Macaca_fascicularis.Macaca_fascicularis_5.0.dna.-toplevel.fa), and Macaca_fascicularis.Macaca_fascicularis_5.0.94.gtf was used for gene annotation. To identify viral reads, the EBOV Mayinga genome was concatenated to the Macaca fascicularis reference and the EBOV Mayinga genome annotation file (GTF). The numbers and percentages of uniquely aligned reads for VP35m WB, VP35m tissue, Kikwit back-challenge tissue, Makona WB, and Makona tissue samples were 3.00 × 10⁷ and 93.5%, 3.16 × 10⁷ and 92.5%, 1.84 × 10⁷ and 83.0%, 2.01 × 10⁷ and 92.0%, and 3.68 × 10⁷ and 92.1%, respectively (40, 45). Statistical analysis of differentially expressed genes (DEGs) was carried out using the edgeR package. DEGs were defined as genes with a median reads per kilobase per million (RPKM) value of ≥5, a false discovery rate (FDR)-corrected P value of ≤0.05, and a log₂ fold change of ≥1 compared to naive tissues. DEGs detected in WB were excluded from those found in tissues to identify “tissue-specific” DEGs. The number of total viral reads was determined as the total number of normalized read counts mapping to all viral genes. The P value for the overlap of DEGs detected in different infections was calculated using the GeneOverlap package in R.

DESeq2 analysis without biological replicates was used to compare the spleen tissues obtained following back-challenge from the nonsurvivor (n = 1) to those from survivors (n = 2). Briefly, the fold change (average of survivors relative to the nonsurvivor) was calculated using RPKM values and the exactTest function. Cutoff values of a log₂ fold change greater than or equal to 6 or less than or equal to −6 were applied to analyze significant protein-coding genes (SGs) with notably higher or lower expression levels in survivors, respectively, and enriched as described below.

Functional enrichment of DEGs and SGs was carried out using Metascape to identify significant Gene Ontology (GO) biological processes (43). The Cytoscape network data integration and visualization tool was used to generate GO term networks (59). Heat maps, bubble plots, and Venn diagrams were generated using the R packages ggplot and Venn Diagrams. Graphs were generated with GraphPad Prism software (version 8).

Flow cytometry staining and analysis.

Flow cytometry data for peripheral blood mononuclear cells (PBMCs) were obtained in our recent study (40). Briefly, live cells were differentiated from dead cells with a fixable viability dye. Innate cell subsets were delineated using the following cell surface markers: CD3, CD20, HLA-DR, CD14, CD16, CD11c, CD123, and CD86. Live CD3⁻ CD20⁻ HLA-DR⁺ cells positive for CD11c, CD123, and CD14 were categorized as myeloid dendritic cells (DCs), plasmacytoid DCs, and monocytes, while all cells negative for these three markers were denoted “other DCs.” Monocytes were classified into CD16⁻ (classical) and CD16⁺ (nonclassical subsets). CD86 was used to identify activated monocyte and DC subsets. Adaptive immune subsets were identified by intracellular staining with a Ki67 proliferation marker and the following cell surface markers: CD4, CD8b, CD20, CD27, D28, and CD95. T cell populations (CD4⁺ or CD8⁺) were classified as naive (CD28⁺ CD95⁻), central memory (CD28⁺ CD95⁺), or effector memory (CD28⁻ CD95⁻). B cells (CD20⁺) were identified as memory (CD27⁺) and naive (CD27⁻) subsets based on CD27 expression.

Statistical analysis.

The statistical significance of viral read numbers was assessed with GraphPad Prism software using unpaired t tests with Welch’s correction.

Data availability.

The accession numbers for the RNA sequencing data reported in this paper are reported under NCBI BioProject accession number PRJNA680570.