WNV infection triggers persistent changes in the enteric neuronal network. Our previous work described a dysmotility syndrome after WNV infection with preferential effects in the small intestine (2). WNV antigen localizes to enteric neurons of the small intestine during the acute phase of infection (at 6 days post infection [dpi]) after subcutaneous inoculation (2); however, the viral tropism for specific neuronal cell populations is unclear. To address this question, we inoculated ChAT-eGFP reporter mice, which identify cholinergic neurons, with WNV (Figure 1A) and costained whole-mount tissue preparations of small intestines with antisera against WNV (15), as well as for other neuronal cell subsets. WNV antigen was detected at similar percentages in calretinin+, ChAT+, and nNOS+ neurons (Figure 1, B and C, and Supplemental Figure 1, A–C). WNV antigen in the myenteric and submucosal plexuses varied along the length of the small intestine, with the middle (jejunal) and distal (ileal) segments showing the highest penetrance (~70%–90% of mice) and the proximal (duodenal) region having less penetrance (20%–33% of mice) (Figure 1D and Supplemental Figure 1D). Thus, we focused subsequent analyses on the middle and distal regions of the small intestine.

WNV infection induces changes in ENS neuronal networks. (A) Nine- to 10-week-old C57BL6/J male mice were inoculated in the footpad with 102 FFU WNV (New York 1999 strain), and a carmine dye transit assay and tissue collections were done at the indicated time points. The figure in A was created using BioRender software. (B and C) Whole-mount preparations of the muscularis externa from ChAT-eGFP reporter mice were isolated at 5 or 6 dpi and costained for WNV antigen, calretinin+, and nNOS+ neurons. (B) Blue, green, and white arrowheads indicate WNV antigen+ calretinin+ neurons, ChAT+ neurons, and nNOS+ neurons, respectively. Images are representative of 2 experiments. Scale bars: 50 μm. (C) Proportion of neuronal subgroups infected with WNV. (D) Percentage of mice that had WNV antigen (Ag) in the proximal (Prox), middle (Mid), and distal (Dis) regions of the small intestine (SI) at 6 dpi. (E–K) The muscularis externa with the attached layer containing the submucosal plexus (SMP) (G), the myenteric plexus (MP) (F and I–K), or both the myenteric plexus and submucosal plexus, as indicated (E and H), was isolated from the middle and distal small intestine of sham- and WNV-infected mice at 7 dpi (E–H) or at 15, 28, and 65 dpi (I–K) and stained for neuronal markers. (E and I) The total number of HuC/D+ neurons in (E) the submucosal plexus and myenteric plexus and (I) the myenteric plexus only was counted and is shown as the number of neurons per mm2. (F and G) The fraction of area that stained positively for nNOS, calretinin, and 5-HT in the myenteric plexus (F) or for calretinin in the submucosal plexus (G); values were normalized to those for sham-infected mice. Circles, squares, and triangles indicate nNOS+, calretinin+, and 5-HT+ neurons, respectively. (H) Images show staining for the indicated markers in the middle small intestine from sham- and WNV-infected mice at 7 dpi in either the myenteric plexus or the submucosal plexus. Scale bars: 100 μm. (J) nNOS+ and calretinin+ and (K) 5-HT+ cell areas; values were normalized to those for sham-infected mice. (I–K) Images show staining in the middle small intestine from sham- and WNV-infected mice at 65 dpi. Scale bars: 100 μm. (L–O) Analysis of neuron-specific RNA-Seq using TRAP in WNV- or mock-infected Snap25l10a mice at 6 dpi. (L) Principal component analysis (PCA). (M) Volcano plot of differential expression analysis (DEseq2) of translating ribosome affinity purification sequencing (TRAP-Seq) comparing WNV- and mock-infected samples. Red dots indicate a log2 fold change of greater than 1 and a FDR (P-adjusted) of less than 0.05, and blue dots indicate a log2 fold change of 1 or less and a P-adjusted FDR of less than 0.05. (N and O) Heatmap of differentially expressed genes in sham- and WNV-infected mice showing genes related to (N) the response to virus and (O) cytokines and chemokines. Expression levels were normalized across each gene and represent the average of 4 mice per condition. Data were pooled from the following number of experiments: (C and D) 2; (E–K) 3 (myenteric plexus) and 2 (submucosal plexus); (F) 3; (G) 2, (I) 3; (J) 2 (15 dpi), 3 (28 dpi), and 4 (65 dpi); and (K) 2 (15 dpi), 3 (28 dpi), and 2 (65 dpi). The numbers of mice per group were as follows: (C) n = 6; (D) n = 9; (E) n = 7–11; (F) n = 6–10; (G) n = 6–7; (I) n = 9–16; (J) n = 8–13; (K) n = 5–10. Column heights in C, E–G, and I–K indicate mean values. *P < 0.05, **P < 0.01, and ***P < 0.001, by 2-tailed Mann-Whitney U test.

To assess the effect of WNV infection on the small intestine, we first quantified the number of neuronal bodies in whole-mount tissue preparations by staining for HuC/D, a pan-neuronal marker. In the myenteric plexus, we found that neuronal cell body numbers were decreased in the middle and distal regions of the small intestine at 7 dpi compared with mock-infected controls (Figure 1, E and H). Similarly, we observed lower numbers of HuC/D+ neurons in the submucosal plexus of the middle region of the small intestine (Figure 1E). The loss of neuronal bodies within intestinal ganglia after WNV infection was associated with a decreased density of their axonal fiber networks (referred to herein as neuronal network density) in the myenteric plexus and the inner circular muscle layer of the muscularis propria in the middle and distal regions of the small intestine at 7 dpi; this included a reduction of the pan-neuronal peripherin+ network and of calretinin+ and nNOS+ networks that are major neuronal cell subsets in the myenteric plexus and inner circular muscle layer (Figure 1, F and H, and Supplemental Figure 1, E–H). In contrast, the density of serotonergic myenteric neuronal networks (i.e., those secreting 5-hydroxytryptamine [5-HT]) in the myenteric plexus at 7 dpi was not different from that of sham-infected controls (Figure 1, F and H). In the submucosal plexus, the calretinin+ neuronal network density was also diminished with WNV infection, which corresponded to the decreased numbers of neurons (Figure 1, G and H). However, as submucosal plexus neurons do not contribute substantially to GI motility (16), we focused on the neuronal cells and networks in the myenteric plexus.

As dysmotility after WNV infection can persist through 65 days (Supplemental Figure 1I and (2)), we quantified the number of HuC/D+ neurons and assessed the neuronal network density in the myenteric plexus at later post-infection time points: 15 dpi (subacute phase), 28 dpi (chronic phase), and 65 dpi (late chronic, convalescent phase) (Figure 1A). By 65 dpi, the numbers of HuC/D+ neurons in the distal small intestine showed near-complete recovery, a process that began as early as 15 dpi (Figure 1I). However, this recovery did not occur to the same extent in the middle region of the small intestine, as fewer neurons were detected at 65 dpi than in sham-infected controls. Our analysis of neuronal networks showed a durable loss of the density of nNOS+ and calretinin+ neuronal networks (Figure 1J) at 15, 28, and 65 dpi in both the middle and distal regions of the small intestine. In contrast, we detected no differences in neuronal network density in the proximal region of the small intestine at 28 dpi, a region that had less viral antigen detected at 6 dpi (Supplemental Figure 1, C and L). Although skewing of neuronal cell subgroup proportions can affect GI motility (7, 17), we did not detect differences in the proportions of nNOS+ or calretinin+ neurons at any point after WNV infection (Supplemental Figure 1, J and K). However, WNV infection caused a marked loss in the density of 5-HT+ neuronal processes at 15, 28, and 65 dpi in the middle and distal small intestine (Figure 1K). This neuronal cell subset is important for the formation of new neurons after intestinal injury (18, 19).

To identify factors that might regulate the ENS response to WNV infection in the acute phase (6 dpi), we performed a translating ribosomal affinity purification (TRAP) of the muscularis externa of WNV- and mock-infected Snap25l10a GFP mice. Snap25l10a GFP mice express a GFP-tagged ribosomal subunit 10la in all neurons, which enables isolation of RNA predominantly from neurons (Supplemental Figure 1M). RNA-Seq showed increased expression of antiviral genes and pathways (e.g., IFN-stimulated genes [ISGs], Ifit family members, Stat1/2 pathways, and pattern recognition receptor signaling pathways) (Figure 1, L–N, and Supplemental Figure 1N). We also observed higher levels of mRNAs encoding Ccl6, Ccl3, Cxcl10, and Ccl2 cytokines, which stimulate chemotaxis of T cells, monocytes, and macrophages (Figure 1O). Furthermore, WNV infection led to an increase in transcripts associated with antigen presentation, including components of MHC class I (Tap1, B2m) and MHC class II (H2-DMb1) antigen processing (Supplemental Figure 1, O and P). These RNA-Seq data identified gene signatures in a neuron-enriched population from the muscularis externa with possible antiviral, immune cell trafficking, and immunomodulatory effects in response to viral infection.

WNV infection triggers persistent changes in intestinal glial cell networks. Enteric glia provide structural and metabolic support for enteric neurons and contribute to neurogenesis (20–25). To determine whether the glial network is affected by WNV infection, we evaluated intestinal whole mounts from WNV-infected mice at 7 dpi by coimmunostaining with antisera against WNV and antibodies against the pan-glial marker S100β (20). Despite limited detection of WNV-antigen+ glial cells (Supplemental Figure 2A), the S100β+ glial network density in the myenteric plexus was markedly diminished in WNV-infected mice at 7 dpi (Figure 2, A and B), with sustained reductions also observed at 28 and 65 dpi (Figure 2C). However, not all neuronal cell–associated networks in the ENS showed diminished density following WNV infection. ICCs located in the smooth muscle layer act as transducers of signals from enteric neurons to smooth muscle cells (8). Although WNV antigen localized sporadically to ICCs in the circular muscle layer of the small intestine at 6 dpi (Supplemental Figure 2B), their density, as judged by cKit staining (26), was like that in sham-treated mice at 15, 28, or 65 dpi (Supplemental Figure 2C).

WNV infection affects enteric glial networks. (A–C) The muscularis externa was isolated from middle and distal regions of the small intestine of sham- or WNV-infected C57BL/6J mice at (A and B) 7 dpi or (C) 15, 28, and 65 dpi and stained for glia (S100β). The fraction of area that stained positive for S100β was determined, and the values were normalized to sham-infected mice. Representative images show S100β staining in the middle region of the small intestine in sham- and WNV-infected mice at (A) 7 dpi or (C) 65 dpi. (A and C) Scale bars: 100 μm. Original magnification, ×2.5 (enlarged insets). Data were pooled from (A and B) 2 experiments (n = 5–10) and (C) (left to right) 2, 3, and 4 experiments (n = 6–10, 12–13, and 13–16). Column heights indicate the mean values. *P < 0.05 and **P < 0.01, by 2-tailed Mann-Whitney U test.

Monocyte and macrophage infiltration is not required for ENS damage or intestinal dysmotility following WNV infection. Monocytes and monocyte-derived macrophages have been shown to injure the intestine in the context of herpesvirus infection (14, 27). Given our data showing higher levels of myeloid cell chemoattractant mRNAs (e.g., Ccl2, Ccl3, Ccl6, and Ccl9) in the muscularis externa of WNV-infected mice (Figure 1O), we quantified monocyte accumulation at 6 dpi in the small intestine of WNV-infected mice using heterozygous Ccr2-GFP reporter mice after staining whole-mount preparations for Iba1, a marker of monocyte-derived macrophages and endogenous muscularis macrophages, but not recent monocyte immigrants (28, 29). The number of both monocytes (Ccr2-GFP+Iba1– cells) and macrophages (Ccr2-GFP+Iba1+ cells) was increased in the proximity of WNV-infected neurons at 6 dpi; the elevation of monocytes persisted up to 15 dpi in the myenteric plexus and the circular inner smooth muscle layer of the muscularis externa (Figure 3, A–E). Elevated numbers of macrophages (Iba1+ cells) persisted through 65 dpi, with a peak at 15 dpi (Figure 3, C, F, and G). Fate-mapping studies using Ccr2 CreER YFP reporter mice demonstrated that the increased number of macrophages was due to infiltrating monocytes, since most of the Iba1+ cells also expressed YFP (Supplemental Figure 3A).

WNV infection promotes infiltration of monocytes into the intestine. (A–E) Whole-mount preparations of the muscularis externa were isolated from the middle and distal regions of the small intestine from WNV-infected heterozygous Ccr2-GFP mice at (A and B) 6 or (C) 15 dpi and stained for (A) neuron (HuC/D) and macrophage (Iba1) markers, (B) WNV antigen and macrophage markers, or (C–E) macrophage markers. Yellow arrowheads indicate monocytes (CCR2 GFP+Iba1– cells). Scale bars: 100 μm. (A–C) Images were obtained from the myenteric plexus of the middle region of the small intestine from at least 2 experiments. (D) Monocytes (Ccr2 GFP+Iba1–) in the myenteric plexus are shown as the numbers of cells per mm2. (E) The fraction of Ccr2 GFP+ area (representing monocytes and/or monocyte-derived macrophages) in the myenteric plexus of WNV- or sham-infected mice. (F and G) Muscularis externa of the middle and distal small intestines from sham- or WNV-infected mice harvested at 15, 28, or 65 dpi were stained for Iba1+ macrophages. Macrophages in (F) the myenteric plexus and (G) the circular muscle layer are shown as the number of Iba1+ cells per mm2. Images of Iba1 staining in sham- or WNV-infected mice at 65 dpi. Scale bars: 100 μm. (H–J) GI transit was measured after oral gavage of carmine red dye (H) in sham- or WNV-infected mice (at 7 dpi) after treatment with anti-CCR2 or isotype mAbs (I) in WNV-infected Ccr2+/– and Ccr2–/– mice, and (J) in sham- or WNV-infected mice after treatment with anti-CSF1R or an isotype control mAb. (K–M). Whole-mount preparations of the muscularis externa were isolated from the middle region of small intestine of WNV-infected mice treated with anti-CSF1R or isotype mAbs and stained for (K) nNOS+ and calretinin+ neurons, (L) 5-HT+ neurons, or (M) S100β+ glia. Scale bars: 100 μm. The fraction of the area that stained positive for calretinin, nNOS, 5-HT, or S100β; values were normalized to those for sham-infected mice treated with an isotype control mAb. Data were pooled from (D and E) 2; (F and G) 2 (15 dpi); 3 (28 dpi) and 4 (65 dpi); (H) 3; (I) 1; (J) 3; and (K–M) 3 experiments. The numbers of mice per group were as follows: (D and E) n = 4–7; (F) n = 9–13; (G) n = 10–12; (H) n = 5–20; (I) n = 7–10; (J) n = 7–16; (K) n = 6–13; (L) n = 5–10; and (M) n = 7–13. Column heights in D–G and J–L indicate mean values, and lines in H–J indicate median values. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by (D, F and G) 2-tailed Mann-Whitney U test and (L) Kruskal-Wallis ANOVA with Dunn’s post test.

To assess the role of the infiltrating monocytes in delaying GI transit in WNV-infected mice, we inhibited their migration into the intestines using an anti-CCR2–blocking mAb (30) or Ccr2–/– mice (Supplemental Figure 3, B and C). We found that reduced accumulation of monocytes during acute WNV infection did not prevent damage to neuronal or glial networks at 7 dpi and did not affect the delayed GI transit time phenotype at 7 or 15 dpi (Figure 3, H and I, and Supplemental Figure 3, D–G). Neutrophil infiltration causes injury to enteric neurons and GI dysmotility after herpesvirus infection in mice (13) and can contribute to WNV-induced pathogenesis in the brain (31–33). However, complete and partial depletion of neutrophils and monocytes, respectively, with an anti-Ly6G/anti-Ly6C mAb (Gr-1) did not improve WNV-induced GI tract dysmotility at 7 dpi (Supplemental Figure 3, B and H–J). We confirmed this result by injecting Ccr2–/– mice with anti-Ly6G/anti-Ly6C mAb to ensure that a lack of monocyte infiltration during WNV infection was not compensated by an increased influx of neutrophils (34) (Supplemental Figure 3, B, K, and L). Thus, WNV-induced intestinal dysmotility appeared to be independent of both infiltrating monocytes and neutrophils.

Resident muscularis macrophages may reduce excessive damage to neuronal and glial networks in WNV-infected mice. Resident macrophages of the GI tract have been proposed to protect the ENS during bacterial infection by preventing neuronal cell death (12). To address their role after WNV infection, we injected mice with anti-CSF1R mAb, which depletes muscularis macrophages, as confirmed by quantification of Iba1+ cells in the myenteric plexus (Supplemental Figure 3, M and N). Animals depleted of muscular macrophages did not show improvement in delayed GI transit times at 7 dpi compared with those given an isotype control mAb (Figure 3J). This GI defect was not caused by a compensatory infiltration of monocytes or monocyte-derived macrophages, as WNV-infected Ccr2–/– mice treated with anti-CSF1R antibody had similar GI transit delays (Supplemental Figure 3O). However, mice deficient in resident macrophages showed a greater loss of neuronal and glial networks than did isotype control mAb–treated mice after WNV infection (Figure 3, K–M). Thus, resident muscularis macrophages appeared to prevent excessive damage to neuronal and glial networks during WNV infection.

Damage to neuronal and glial networks is caused by T cells. We previously noted that CD8+ T cells probably contribute to GI tract dysmotility after WNV infection (2). However, these studies were performed with Cd8α–/– mice, which lack CD8+ T cells, but retain other T cell subsets (35). Moreover, the effects of WNV infection on the neuronal network in Cd8α–/– mice were not evaluated. To measure GI tract motility and analyze the neuronal and glial networks in the absence of all T cells, TCRbd–/– mice, which lack both αβ and γδ T cells, were inoculated with WNV. Since mice lacking T cells develop uncontrolled CNS infection and succumb within 10–14 days (36–40), we only performed analyses at 7 dpi. Notably, WNV-infected TCRbd–/– mice did not show GI tract dysmotility or reduced density of nNOS+ and calretinin+ neuronal networks in the myenteric plexus as compared with WT littermate controls; infected TCRbd–/– mice appeared similar to sham-infected WT or TCRbd–/– mice (Figure 4, A and B), despite the high levels of viral antigen in the myenteric plexus (Figure 4C and Supplemental Figure 4A). Similarly, at 7 dpi, the glial network in the myenteric plexus was unaffected in WNV-infected TCRbd–/– mice (Figure 4D). In the absence of T cells, the numbers of HuC/D+ neurons in the submucosal plexus also were not affected by WNV infection, although the density of the calretinin+ network was decreased (Supplemental Figure 4B). These results are consistent with a role for T cells in mediating the injury of motor neurons and glial cells in the myenteric plexus and development of GI tract dysmotility during the acute phase of WNV infection. However, submucosal neurons may be injured by T cell–independent mechanisms. Because TCRbd–/– mice also have some defects in B cell development and differentiation (41), we performed additional experiments in μMT mice that lack mature B cells and antibody. As WNV-infected μMT and WT littermate mice showed similarly delayed GI transit times at 7 dpi (Supplemental Figure 4C), mature B cells or antibodies were not required for WNV-triggered GI dysmotility.

Damage to the neuronal and glial networks is caused by CD4+ and CD8+ T cells. (A and F) GI transit was measured after oral gavage of carmine red dye. (A) Transit time for sham- or WNV-infected WT or TCRbd–/– mice at 7 dpi. (B–D, G, and H) The muscularis externa was isolated from (B–D) middle and distal regions of the small intestine of sham- or WNV-infected WT or TCRbd–/– mice at 7 dpi, (G and H) middle regions of the small intestine of sham- or WNV-infected WT mice at 7 dpi that were treated with anti-CD4 and/or anti-CD8β or isotype control mAbs and stained for (B and G) calretinin+ and nNOS+ neurons, (C) WNV antigen, (D) S100β+ glia, or (H) and S100β+ glia and WNV antigen. The fraction of the area that stained positive for calretinin, nNOS, or S100β was determined, and the values were normalized to values for (B and D) WT sham-infected mice or (G and H) animals treated with an isotype control mAb. Representative images from the myenteric plexus of the middle region of the small intestine. Scale bars: 100 μm. Original magnification, ×2.5 (enlarged insets). (C) Data are presented as the percentage of WNV antigen+ area in the field of view. (E) Counts of live CD45+TCRβ+CD4+ or CD8+ T cells in the muscularis of sham- or WNV-infected C57BL6/J mice at 7 dpi. (F) Transit time for sham- or WNV-infected mice at 7 dpi; mice were treated with anti-CD4 (α-CD4) and/or anti-CD8β or isotype control mAbs. Data were pooled from (A) 3; (C–E, and G) 2; and (F) 4 experiments. The numbers of mice per group were as follows: (A) n = 7–13; (C and D) n = 5–8; (E) n = 6–7; (F) n = 8–18; (G) n = 7–8; (H) n = 6–8. Lines in A, E, and F and column heights in B–D, G, and H indicate mean values. *P < 0.05, **P < 0.01, and ***P < 0.001, by (A, B, D, G, and H) Kruskal-Wallis ANOVA with Dunn’s post test (all groups compared with each other); (F) Kruskal-Wallis ANOVA with Dunn’s post test (compared with the isotype control group); and (C) 2-tailed Mann-Whitney U test.

We next evaluated the specific roles of CD4+ and CD8+ T cells in WNV-triggered intestinal dysmotility. Flow cytometric analysis and whole-mount staining showed that during the acute phase of WNV infection (days 6 and 7), CD4+ and CD8+ T cells accumulated in the areas of the muscularis externa adjacent to damaged neurons (Figure 4E and Supplemental Figure 4, D and E). To assess the individual contributions of CD4+ and CD8+ T cells to WNV-induced GI dysmotility, we treated mice with depleting mAbs that target these T cell subsets (Supplemental Figure 4F). We chose this approach because Cd4–/– mice also have altered CD8+ T cell development due to lineage commitment effects during thymopoiesis (42). Mice treated with anti-CD4 or anti-CD8β antibodies all demonstrated targeted cell depletion in the muscularis externa of the small intestine, the spleen, and Peyer’s patches (Supplemental Figure 4, G–I). Although treatment with either anti-CD4 or anti-CD8β mAbs alone did not rescue intestinal motility at 7 dpi, administration of both anti-CD4 and anti-CD8β mAbs restored intestinal motility to homeostatic levels in most animals (Figure 4F), and this was associated with improved neuronal and glial networks in the myenteric plexus after WNV infection (Figure 4, G and H). These results were similar to those observed in WNV-infected TCRbd–/– mice, supporting a role for both CD4+ and CD8+ T cells in the damage of neuronal and glial networks and intestinal dysmotility after WNV infection.

Antigen-specific CD4+ and CD8+ T cells cause neuronal and glial injury after WNV infection. To further dissect the specific contribution of CD4+ and CD8+ T cells in WNV-infected mice to neuronal and glial damage, we isolated WNV-primed CD4+ and CD8+ T cells from WT mice at 7 dpi and adoptively transferred them into WNV-infected TCRbd–/– mice at 2 dpi (Supplemental Figure 5A). Subsequently, we analyzed intestinal tract motility and injury to neurons and glia at 7 dpi (Figure 5, A–E) after confirming the efficiency of the T cell transfers (Supplemental Figure 5, B and C). More than 50% of TCRbd–/– mice that received CD8+ T cells developed severe GI tract dysmotility (≥360 minutes) and showed damage to neuronal and glial networks, whereas sham or WNV-infected TCRbd–/– mice without T cell transfers did not (Figure 5, A–E). Although most WNV-infected mice injected with CD4+ T cells did not show delayed GI tract transit times, we still observed injury to neurons (Figure 5, A–E). Together, these results suggest that, while both CD4+ and CD8+ T cells could injure the neuronal network, CD8+ T cells triggered greater damage resulting in intestinal dysmotility.

Damage to the neuronal and glial networks is caused by WNV-specific CD8+ and CD4+ T cells. (A–E) CD4+ or CD8+ T cells from WNV-infected WT mice were isolated at 7 dpi and adoptively transferred into TCRbd–/– mice at 2 dpi. (A) GI transit time in recipient TCRbd–/– mice at 7 dpi, (B) proportions of mice with severe GI dysmotility (≥360 min), (C and D) analysis of neuronal (calretinin, nNOS) and glial (S100β) networks from the middle small intestine at 7 dpi, and (E) images obtained from the myenteric plexus of the middle region of small intestine. Scale bars: 100 μm. (F) Flow cytometric analysis of the muscularis externa or mucosa and lamina propria at 7 dpi. Cells were stained with mAbs against CD45, TCRβ, TCRγδ, CD8a, CD44, and WNV NS4B Db–restricted tetramers and gated on live CD45+TCRβ+CD8+ cells (see Supplemental Figure 4C). Graph shows the percentage of CD8+ T cells positive for NS4B. (G–I) Adoptive transfer of CD8+ T cells from P14 transgenic mice (targeting LCMV gp33 peptide) or WNV NS4B transgenic mice into TCRbd–/– mice. T cells were administered to TCRbd–/– mice 1 day prior to subcutaneous inoculation with WNV. (G) GI transit 7 dpi, (H and I) analysis of the neuronal (calretinin, nNOS) and glial network (S100β) in middle and distal small intestines at 7 dpi. (A and G) GI transit was measured after oral gavage of carmine red dye. (C, D, H, and I) Fraction of the area that stained positively for calretinin, nNOS, or S100β; values were normalized to (C and D) WT sham mice or (H and I) TCRbd–/– mice without adoptive transfer. Data were pooled from (A–D) 6; (F) 2; and (G–I) 3 experiments. The numbers of mice per group were as follows: (F) n = 6; (A) n = 8–13; (B) n = 8–13; (C and D) n = 8–13; (F) n = 6; (G) n = 5–9; (H and I) n = 6–10. Lines in G and F and column heights in C, D, H, and I indicate mean values. *P < 0.05 and **P < 0.01, by (C and D) ANOVA with Dunnett’s post test (comparison with the no-transfer group); (B) χ2 test with Bonferroni correction (proportions compared with the no-transfer group); and (G–I) Kruskal-Wallis ANOVA with Dunn’s post test (comparison with the no-transfer group).

Virus-specific and bystander effector CD8+ T cells can target infected cells for lysis and promote local inflammation (43). To determine the role of antigen-specific recognition of target cells by CD8+ T cells in ENS injury, we stained cells with Db-restricted tetramers that recognize an immunodominant peptide epitope (SSVWNATTAI) in the WNV NS4B protein (44). In WNV-infected small intestines, approximately 25% of CD8+ T cells in the muscularis externa (and 10% in the remainder of the small intestine) were specific for the NS4B immunodominant peptide (Figure 5F). To determine the contributions of antigen-specific and bystander CD8+ T cells to the damaging of neurons and glia after WNV infection, we utilized T cell receptor–transgenic (TCR-transgenic) mice in which the vast majority of CD8+ T cells were specific for the WNV NS4B peptide epitope or, as a control, for the lymphocytic choriomeningitis virus (LCMV) gp33 peptide epitope (KAVYNFATC) (45–47). Transgenic WNV NS4B or LCMV gp33 CD8+ T cells were adoptively transferred into TCRbd–/– mice, and, 1 day later, recipient animals were inoculated subcutaneously with WNV (Supplemental Figure 5, D and E). At 7 dpi, we measured intestinal tract motility, collected mesenteric lymph nodes to confirm T cell colonization (Supplemental Figure 5F), and analyzed the neuronal and glial networks from the middle and distal regions of the small intestine. Whereas TCRbd–/– mice given NS4B-specific CD8+ T cells showed delayed intestinal transit times and damage to neuronal and glial networks in the middle and distal regions of the small intestine, animals given LCMV gp33–specific (P14) CD8+ T cells did not develop dysmotility (Figure 5, G–I). Nonetheless, WNV-infected TCRbd–/– mice that received LCMV-specific CD8+ T cells showed some damage to neuronal networks (Figure 5H), suggesting either limited bystander injury mediated by antigen-nonspecific CD8+ T cells or expansion of WNV-specific CD8+ T cells from the small repertoire of endogenous TCRs in the LCMV gp33 (P14) TCR-transgenic mice on a WT C57BL/6 background.

To further assess the potential role of bystander CD8+ T cells, we crossed the P14 LCMV-transgenic mice with Rag1–/– mice to generate animals in which virtually every CD8+ T cell was specific for the LCMV gp33 peptide (Supplemental Figure 5G); these animals also lacked CD4+ T cells. We measured the intestinal transit in WT and P14 Rag1–/– mice 7 days before (baseline) and after WNV infection (Supplemental Figure 5, G–I). Notably, WNV-infected P14 Rag1–/– mice showed normal GI tract transit at day 7, comparable to baseline and sham-infected WT mice (Supplemental Figure 5H). Similarly, we did not observe damage to neuronal or glial networks in WNV-infected P14 Rag1–/– mice (Supplemental Figure 5I). Collectively, these results did not indicate a substantive role for bystander CD8+ T cells in neuronal or glial injury in the context of WNV infection.

We similarly assessed the role of bystander CD4+ T cells on ENS integrity and function after WNV infection, by crossing OT-II TCR transgenic mice with Rag1–/– mice to generate mice that lack CD8+ T cells and have CD4+ T cells that are specific for the OVA peptide 323-339 (ISQAVHAAHAEINEAGR) (Supplemental Figure 5J). At 7 dpi, none of the WNV-infected OTII Rag1–/– mice exhibited intestinal dysmotility, and neuronal networks appeared normal (Supplemental Figure 5, K and L). Thus, bystander CD4+ T cells also did not induce neuronal injury and GI dysmotility after WNV infection.

CD4+ and CD8+ T cells cause neuron and glia injury using multiple effector mechanisms. T cells can use a variety of mechanisms to clear WNV-infected neurons from the brain including cytotoxic granules (perforin/granzymes), proinflammatory cytokines (e.g., TNF and IFN-γ), and death receptor signaling pathways (Fas ligand [FasL] or TRAIL) (37, 48–52). To determine the contribution of these T cell effector mechanisms to ENS damage after WNV infection, we used a combination of genetic and pharmacological loss-of-function approaches to test the role of perforin (Prf1–/– mice), FasL (gld-mutant; Faslgld/gld mice), IFN-γ (Ifngr–/– mice), and TNF (blocking mAb, MP6-XT22 that inhibits both membrane-associated and soluble forms; ref. 53). At 7 dpi, WNV-infected Prf1–/–, Faslgld/gld, and Ifngr–/– mice all showed delayed GI transit like the WT littermate controls (Figure 6, A–C). WT mice treated with an anti-TNF mAb (MP6-XT22) also showed delayed GI transit times after WNV infection similar to controls (Figure 6D and Supplemental Figure 6A). In addition, we observed similar or even greater intestinal segment dilation in WNV-infected Prf1–/–, Faslgld/gld, Ifngr–/–, and anti-TNF–treated mice compared with WNV-infected control mice (Supplemental Figure 6B). Although we observed damage to the neuronal networks and high numbers of CD3+ T cells in the myenteric plexus in WNV-infected Prf1–/– and Faslgld/gld mice, the glial network appeared more intact in these 2 strains (Figure 6, E and F, and Supplemental Figure 6, C and D). These data suggest that loss of individual cytolytic pathways or effector cytokines was not sufficient to ameliorate the WNV-induced damage of neurons and intestinal dysmotility, although glia could be affected.

CD4+ and CD8+ T cells injure neurons and glia using multiple effector functions. (A–D, H, and K) GI tract transit was measured after oral gavage of carmine red dye at 7 dpi. Transit time for sham, WNV-infected WT, or WNV-infected (A) Prf1–/–, (B) Faslgld/gld, (C) Ifngr–/–, and (D) WT mice treated with anti-TNF or isotype control mAb or WNV-infected (H) Prf1–/– or (K) Faslgld/gld mice treated with anti-CD4, anti-CD8β, or isotype control mAb. (E, F, I, J, and L) The muscularis externa was isolated from the middle regions of small intestines from sham-infected, WNV-infected WT, or Prf1–/– mice (E), Faslgld/gld mice (F), WNV-infected Prf1–/– mice (I and J), or Faslgld/gld mice (L) treated with anti-CD4 or anti-CD8β mAb at 7 dpi and then stained. The fraction of the area that stained positive for calretinin, nNOS, or S100β was determined, and values were normalized to those for sham-infected WT mice. (J) Representative images were obtained from the myenteric plexus of the middle region of the small intestine. Scale bars: 100 μm. Original magnification, ×2.5 (enlarged insets). Data were pooled from (A–C, F, I, and J) 3; (D, G, and L) 2; (E and K) 5; and (H) 4 experiments. The numbers of mice per group were as follows: (A) n = 4–11; (B) n = 7–10; (C) n = 4–11; (D) n = 10; (E) n = 4–13; (F) n = 6–11; (G) n = 6; (H) n = 9–12; (I) n = 9–12; (K) n = 7–15; (L) n = 6–7. Lines indicate (A–D and G) median or (H and K) mean values, and column heights indicate the mean values. *P < 0.05 and **P < 0.01, by (A–F, I, and L) Mann-Whitney U test and (H and K) ANOVA with Dunnett’s post test (comparison with the isotype control group).

Activated CD4+ and CD8+ T cells can both have cytolytic activity (54) and produce inflammatory cytokines after WNV antigen stimulation (37, 55). Flow cytometric analysis revealed that perforin was present in almost all CD8+ T cells and, on average, 25% of CD4+ T cells in the muscularis layer of WNV-infected mice at 7 dpi (Figure 6G). As multiple effector functions in different T cell populations could be used concurrently to target WNV-infected cells, we hypothesized possible redundancy in effector mechanisms that cause injury to enteric neurons. To test this hypothesis, prior to WNV infection, we depleted CD4+ or CD8+ T cells in either Prf1–/–, Faslgld/gld, or Ifngr–/– mice or in WT mice treated with blocking mAbs against IFN-γ (Supplemental Figure 6A). Depletion of CD4+ or CD8+ T cells did not mitigate GI transit defects in either WNV-infected Ifngr–/– or WT mice treated with an IFN-γ–blocking mAb (Supplemental Figure 6, E and F). However, depletion of CD8+, but not CD4+, T cells in Prf1–/– and Faslgld/gld mice normalized the GI transit time defect and prevented damage to neuronal networks (Figure 6, H–L, and Supplemental Figure 6G). These data suggest that (a) CD8+ T cells can utilize an alternative cytolytic mechanism in the absence of perforin and (b) CD4+ T cells require both perforin and FasL pathways to mediate WNV-induced damage and dysmotility in the context of a CD8+ T cell deficiency. Thus, T cells can use multiple effector mechanisms to target WNV-infected neuronal cells in the small intestine.

Mice lacking both perforin and FasL do not develop GI dysmotility and have intact neuronal and glial networks. FasL and perforin can function together to augment cytotoxic T cell responses (56). To more definitively determine whether perforin and FasL together are the dominant mechanisms causing neuronal and glial injury and ensuing GI dysmotility after WNV infection, we generated double-KO (DKO) mice by crossing Prf1–/– and Faslgld/gld mice (Figure 7A). We inoculated DKO (Prf1–/– Faslgld/gld) subcutaneously with WNV and at 7 dpi measured the GI transit time and analyzed neuronal and glial networks in the small intestine. Most WNV-infected DKO mice showed normal GI motility and an absence of bowel dilation (Figure 7, B and C). Consistent with these results, the neuronal and glial networks were intact in WNV-infected DKO mice and like those of uninfected WT mice (Figure 7, D and E), despite the presence of WNV antigen and CD3+ T cells in the myenteric plexus of DKO mice (Figure 7, F and G). These results indicate that in mice containing both CD4+ and CD8+ T cells, either FasL or perforin was sufficient to cause the WNV-triggered pathology in the gut. When both effector mechanisms or T cells were absent, WNV-induced defects in GI motility and neuronal and glial network injury were prevented.

Mice lacking perforin and Fas/FasL signaling do not develop WNV-triggered GI dysmotility or neuronal and glial network injury. (A) Scheme for the generation of Prf1–/– Faslgld/gld (DKO) mice. The figure in A was created using BioRender software. (B) GI transit time in WNV-infected WT or DKO mice at 7 dpi. (C) Proportions of WNV-infected WT and DKO mice showing abnormal bowel dilation in the small intestine at 7 dpi. (D–G) The muscularis externa was isolated from the middle regions of small intestines from sham- (D and E) or WNV-infected WT or DKO (F and G) mice at 7 dpi and stained. (D–F) The fraction of the area that stained positive for calretinin, nNOS, S100β, and WNV antigen was determined, and values were normalized to those for (D and E) sham-infected WT mice or (F) WNV-infected WT mice. (G) The numbers of CD3+ cells in the myenteric plexus were calculated by dividing the area positive for CD3 staining with the average size of CD3+ cells. Cell counts are expressed as the number of CD3+ cells per mm2. (D and G) Representative images of the myenteric plexus of the middle region of the small intestine. Scale bars: 100 μm. Original magnification, ×2.5 (enlarged insets). Data were pooled from (B, C, and F) 5 and (D, E, and G) 4 experiments. The numbers of mice per group were as follows: (B and C) n = 13–16; (D and E) n = 8–10; (F) n = 10–15; (G) n = 10. Lines in B and column heights in D–G indicate mean values. *P < 0.05, **P < 0.01, and ***P < 0.001, by (B, F, and G) 2-tailed Mann-Whitney U test and (D and E) Kruskal-Wallis ANOVA with Dunn’s post test.