Innate immune response to homologous rotavirus infection in the small intestinal villous epithelium at single-cell resolution

Abstract

“Bulk” measurements of antiviral innate immune responses from pooled cells yield averaged signals and do not reveal underlying signaling heterogeneity in infected and bystander single cells. We examined such heterogeneity in the small intestine during rotavirus (RV) infection. Murine RV EW robustly activated type I IFNs and several antiviral genes (IFN-stimulated genes) in the intestine by bulk analysis, the source of induced IFNs primarily being hematopoietic cells. Flow cytometry and microfluidics-based single-cell multiplex RT-PCR allowed dissection of IFN responses in single RV-infected and bystander intestinal epithelial cells (IECs). EW replicates in IEC subsets differing in their basal type I IFN transcription and induces IRF3-dependent and IRF3-augmented transcription, but not NF-κB–dependent or type I IFN transcripts. Bystander cells did not display enhanced type I IFN transcription but had elevated levels of certain IFN-stimulated genes, presumably in response to exogenous IFNs secreted from immune cells. Comparison of IRF3 and NF-κB induction in STAT1^−/− mice revealed that murine but not simian RRV mediated accumulation of IkB-α protein and decreased transcription of NF-κB–dependent genes. RRV replication was significantly rescued in IFN types I and II, as well as STAT1 (IFN types I, II, and III) deficient mice in contrast to EW, which was only modestly sensitive to IFNs I and II. Resolution of “averaged” innate immune responses in single IECs thus revealed unexpected heterogeneity in both the induction and subversion of early host antiviral immunity, which modulated host range.

Following virus infection, eukaryotic cells respond by the activation of innate immunity in a cell type- and strain-specific manner. The early host innate response includes activation of the transcription factors IRF3 and NF-κB, which induce transcription of discrete and overlapping sets of genes, including those encoding type I IFNs and viral stress-induced genes (vSIGs) (1, 2). Following IFN secretion and binding to cognate receptors, there is feed-forward amplification of expression of IFN and several hundred IFN-stimulated genes (ISGs), resulting in the establishment of an antiviral state. Such early responses to the presence of viral particles are spatially distinct from those that occur in adjacent bystander cells lacking direct exposure to virus. The early host innate immune responses to virus generally have been measured using averaged signals from “bulk” populations of cells or whole tissues. Such analyses cannot reveal hierarchical responses in individual infected or bystander cells. Here, we used a single-cell analytic strategy to examine the diversity in the early host antiviral innate immune response to rotavirus (RV) in suckling mice.

RVs cause severe dehydrating diarrhea in the young of many mammalian species, resulting in more than 400,000 deaths of children annually (3). Homologous RV (RV derived from the infected host species) replicates efficiently within mature absorptive villous enterocytes of the small intestine but in the immune competent host, does not replicate efficiently in crypts or colon or at extraintestinal systemic locations (4). RV has evolved mechanisms to effectively block IRF3 and NF-κB–dependent IFN induction, as well as feedback-dependent amplification stages of the innate antiviral response (5). Specifically, the RV nonstructural protein NSP1 mediates degradation of IRF3 and/or NF-κB regulatory factor β-TrCP, depending on the viral strain and host cell involved (6, 7). In addition, RV also inhibits STAT1-dependent pathways in response to IFN (8). It is not clear which of these several inhibitory processes homologous RV uses, if any, to enhance replication during natural infection. However, several lines of evidence suggest that to replicate efficiently in the gut and spread to other homologous hosts, RV must circumvent host- and cell-type–specific IFN restriction (8, 9).

In mice lacking type I and II IFN responses, selected heterologous RV strains replicate much more efficiently at systemic sites and cause a lethal disease (4). Heterologous simian RV (RRV) replicates ∼1,000 times more efficiently in the intestine of STAT1^−/− suckling mice than in wild-type (wt) mice (4). In contrast, homologous murine RVs replicate very efficiently in the intestine of wt suckling mice and replication is not substantially (∼10 times) enhanced in STAT1^−/− mice (10). A single infectious particle of murine RV is sufficient to initiate full-blown disease in the homologous host with an intact IFN response (11), whereas 10³ to 10⁶ infectious particles of a nonmurine strain are required to initiate infection, and this restriction is highly IFN dependent. How homologous RV successfully negotiates the early host antiviral response within intestinal epithelial cells (IECs) has not been well characterized, and such knowledge likely will improve our understanding of mucosal innate immunity as well as the strategies mucosal pathogens use to subvert it.

Results

Analysis of the Innate Response to RV in Single IECs.

Previously we demonstrated that unlike heterologous RRV, homologous murine RV replication in the gut is not highly sensitive to type I or II IFN receptors or downstream STAT1 signaling, indicating that RV encodes strategies to replicate with high efficiency in the presence of an intact host IFN response (4). To determine whether murine RV infection results in the suppression of IFN and ISG induction in the intestine, we infected suckling mice with the EW strain of murine RV and at 16 h post infection (hpi) analyzed bulk intestinal tissue lysates for the levels of RV VP6, and antiviral genes by quantitative RT-PCR (qRT-PCR) (Fig. 1A). Surprisingly, EW infection resulted in significant (>10×) induction of Ifn-β, Isg20, Mx2, and Ifi203. Increases in IRF7 and p56 were observed at 16 hpi, but not 6 hpi, by immunoblot analysis (Fig. 1B). Thus, despite the ability of murine RV to replicate with high efficiency in the murine small intestine, it efficiently induces type I IFN and antiviral genes.

Fig. 1.

Induction of intestinal innate immunity by homologous RV. Suckling mice were infected with EW. (A) Total intestinal RNA was analyzed by qRT-PCR for indicated transcripts at 16 hpi; results are presented as mean fold change over uninfected mouse intestines and SD from duplicate measurements. (B) Protein lysates were analyzed by immunoblot for indicated innate immune response markers.

To determine whether efficient replication of murine RV coincides with suppression of type I IFN and ISG induction in the IEC, we performed transcriptional analysis of single murine IECs using a microfluidics-based multiplex RT-PCR technique (12–14). Transcription at the single-cell level is stochastic and results in non-normal signal accumulation across cells (12). To improve the likelihood of identifying statistically significant changes and obtaining sufficient numbers of RV-infected cells, we purified and fractionated IECs from the proximal small intestine of infected and noninfected mice by flow cytometry and gating for selected surface markers. The isolated epithelial cells (Esa⁺CD45⁻) were sorted for CD26, an enterocyte brush-border peptidase, which enriched for RV-infected cells by ∼200-fold (Fig. 2A and Fig. S1), and CD44, highly expressed in the crypt base that is resistant to RV infection. The resultant doubly sorted single cells (>95% purity and ∼15% of the starting live singlet cells) were Esa⁺CD45⁻CD26⁺CD44⁻.

Fig. 2.

Single-cell analysis of RV-induced innate immune responses in intestinal villous epithelial cells. (A) Purification scheme used for isolation of RV-infected and uninfected villous cells by flow cytometry and relative RV expression in indicated IEC subsets. (B) Classification of the 81 transcriptional parameters chosen for quantitative RT-PCR into eight functional categories. (C) Distribution of positive Ct measurements from 976 single cells for 81 simultaneous parameters grouped into categories defined in B. Data are presented as median Ct and interquartile spread. Assays are presented left to right in the order shown in Table S2. (D) Multimodal signal distribution from single cells showing discrete signaling clusters of cell populations.

The intestinal epithelium contains cells along a maturational gradient beginning in the crypts and culminating with mature absorptive enterocytes at the villous tips. Although surface marker expression can define relatively homogenous cell phenotypes, diverse hierarchies are identified within such “pure” populations by single-cell analysis (15). We developed a panel of 81 transcripts representing eight prespecified functional transcriptional categories (Fig. 2B and Table S1). IEC and housekeeping marker transcripts are category 1, and RV⁺ transcripts are category 2. Induction of IFN occurs through pattern recognition receptors (16, 17) that undergo transcriptional auto-amplification (Fig. 2B, category 3). Signal transduction following RV infection results in initial induction of transcripts via IRFs (category 4), and other transcription factors that are considered (18) IRF3 dependent (category 5), IRF3 augmented (induced by IRF3 and/or other transcription factors, category 6), or IRF3 independent (including NF-κB–dependent transcripts, category 7) (9, 16, 17, 19, 20–22). This transduction culminates in the transcription of type I IFN genes (Ifn-α4 and Ifn-β primarily, and Ifn-α5 by receptor-mediated amplification, category 8) (21). The levels of these 81 transcripts were measured simultaneously in 976 single cells collected from 11 RV-infected suckling mice (739 cells, including 120 RV⁺ cells) and 5 uninfected control mice (237 cells), resulting in more than 79,000 Ct measurements (Dataset S1). Consistent intra-assay Ct values were obtained across single cells (Fig. 2C, Fig. S2, and Table S2), and a distinctly multimodal signal distribution (Fig. 2D) was observed for certain transcripts, indicating the presence of subpopulations of cells.

Hierarchical Clustering of Single IECs Based on Their Antiviral Response to RV Infection.

To visualize and organize the cellular diversity presented at the single-cell level, individual cells were grouped into “uninfected” (cells derived from mice not exposed to RV), “RV⁺” (cells from infected mice containing at least one RV transcript), and “bystanders” (cells derived from infected mice not containing any RV transcripts) (Fig. S3). The phrase “cells from infected animals (CIAs)” refers to both RV⁺ and bystander cells. Single cells within the three groups—RV⁺, bystander, and uninfected—were then clustered by algorithmic assembly based on their levels of transcripts from category 1 (Fig. 3A and Fig. S3). The resulting heat map clusters each category based on aggregated IEC-specific gene expression levels, a phenotype generally reflecting their location along the villous. Cells from uninfected mice exhibited significant heterogeneity in levels of Lct, Anpep/CD13, Esa/Tacstd1, Ace2, and Krt20 (Fig. 3A and Table S3) at a 5% level of significance by both-sided nonparametric test with false discovery rate (FDR) correction (SI Materials and Methods, Statistical Methods). Each cell was classified as either enterocyte^Hi or enterocyte^Low based on the overall levels of these transcripts. In contrast to this heterogeneity, category 1 transcripts Tff3 (Fig. S5), Top2a, Mki67, Muc2, and Birc5 were expressed at similar levels in both enterocyte populations (Fig. 3A and Table S3). Of note, uninfected enterocyte^Hi cells had significantly elevated basal expression of type I IFN transcripts and other antiviral genes compared with uninfected enterocyte^Low cells (P < 0.05) (SI Materials and Methods, Statistical Methods). Elevated type I IFN transcription in the enterocyte^Hi cells correlated with high levels of Irf1 and Irf7, but not Irf3, transcripts. Increased basal expression of type I IFN has been noted in differentiated epithelial cells (23) and proposed to promote efficient IFN induction following pathogen detection (“revving-up” model) (24). Interestingly, in RV⁺ CIAs, RV transcripts were found at significantly higher levels in the enterocyte^Hi cells (P < 0.0001, Mann–Whitney test, Fig. S4).

Fig. 3.

Single-cell–based hierarchical reconstruction of innate immune response signaling states. (A) Hierarchical clustering of single cells grouped at left based on virus-derived replication states and at right (horizontal dotted lines) on category 1 transcripts, groups as shown in B. Assay columns from left to right are in the order shown in Table S2. (B) Antiviral states in different subpopulations following RV infection of IECs. NF-κB–dependent transcripts from category 7 are indicated by asterisks. Statistically significant changes (P < 0.05) relative to population-matched uninfected cells in the positive (blue) or negative (red) direction are indicated. Numbers refer to the transcript categories in Fig. 2B.

Single-cell–based clustering allowed us to derive dimensional enterocyte^Hi- and enterocyte^Low-specific “signaling states” for CIA populations (RV⁺ and bystanders), RV⁺ alone, and bystander cells alone (Fig. 3B). A specific state is defined as a significant (P < 0.05) perturbation in transcription levels compared with control cells from an identically clustered population from uninfected mice (Fig. 3B, Fig. S5, Fig. S6 and Tables S4–S9). The IEC innate response to RV involved both redundant and unique patterns of transcriptional activation in RV⁺ and bystander CIAs. Induction of Rsad2 (viperin), Isg15, Isg20, Ifit1/Isg56, and Ifit2/Isg54 constituted a conserved “core” signature of the host response found in all IECs from CIAs. Other “signatures” specific to different combinations of RV⁺, bystander, enterocyte^Hi, and enterocyte^Low states also were visualized (Fig. 3B). For example, Irf7 was induced in enterocyte^Low cells and Mx2 in enterocyte^Hi populations. In contrast, Lgp2 was induced in both RV⁺ and bystander enterocyte^Low cells but not in enterocyte^Hi populations.

This “bottom-up” reconstruction from single cells revealed signaling heterogeneity in local cellular niches. In all CIAs, RV infection either did not induce type I IFN transcription (in the enterocyte^Low population) or caused a significant decline in steady-state transcript levels (in the rarer enterocyte^Hi population), compared with uninfected control mice. These IFN effects occurred in both RV⁺ and bystander cells, indicating that murine RV infection of the suckling mouse mediates both direct and indirect inhibitory effects on IFN transcription in IECs.

Type I IFN Induction Occurs in Nonepithelial Hematopoietic Cells by a Replication-Independent, IFN-αβγ Receptor/STAT1–Dependent Mechanism.

To determine where in the intestine RV-induced type I IFN transcription occurs, we bulk-sorted ESA⁺CD45⁻ (epithelial, ∼91% purity), ESA⁻CD45⁺ (hematopoietic, ∼96% purity), and ESA⁻CD45⁻ (stromal, ∼99% purity) cells from uninfected and EW-infected suckling mice (Fig. 4A), and analyzed purified RNA by qRT-PCR (Fig. 4B). Certain ISGs, including the IRF3-dependent transcripts Rsad2 and Ifi203, were significantly up-regulated in epithelial and hematopoietic intestinal cells, but only weakly in stromal cells. However, induction of both Ifn-α4 and Ifn-β transcripts occurred overwhelmingly in the intestinal hematopoietic cell compartment and not in epithelial or stromal cells, consistent with our single-cell analysis of CIAs and supporting the notion that a major proportion of the intestinal type I IFN is induced outside the primary RV replication site in IECs (Fig. 4B).

Fig. 4.

Rotavirus induction of intestinal type I IFN responses occurs in hematopoietic cells via replication-independent, IFN receptor–dependent, and STAT1-dependent signaling. (A) Suckling mice were infected with EW and at 16 hpi, intestinal cell populations were purified by flow cytometry based on the epithelial cell marker Esa and the hematopoietic cell marker CD45 into the numbered populations. (B) Total RNA from cell populations was then analyzed by qRT-PCR. (C) Mice were infected with RRV, the monoreassortant D6-2, or EW, and total intestinal RNA was analyzed 16 hpi for viral loads. (D) wt or IFN-αβγ-R–deficient mice were infected with the RV strains indicated, and total intestinal RNA was analyzed by qRT-PCR at 16 hpi. Data shown are mean ±SD and representative of at least two experiments (n = 3–10 mice in each group per experiment).

Because RV replicates predominantly in IECs rather than in hematopoietic cells, we suspected that the type I IFN induction in hematopoietic cells was not highly dependent on robust viral replication. To correlate IFN induction with viral replication, we examined the ability of RRV, a heterologous simian RV that is replication restricted in the mouse, to activate an IFN response (Fig. 4C). To rule out viral entry-dependent requirements for RRV replication, we also infected mice with D6/2, a monoreassortant that encodes the RRV major entry protein (VP4) on a murine EW genetic background, and examined the levels of RV, type I IFN, and ISG transcripts in bulk intestinal lysates (Fig. 4C). Despite differences in VP4 genes, both D6/2 and EW replicate to significantly higher levels in the intestine than RRV (∼10⁵ to 10⁶ times). Thus, the intestinal replication restriction of RRV is unlikely the result of a cell entry deficiency. In contrast to its distinct heterologous replication defective phenotype, RRV induced expression of type I IFN and ISG transcripts to a similar extent as D6/2 and EW (Fig. 4D). This result implies that the early innate immune response to RV is substantially independent of viral replication in the intestine.

The conclusion that early intestinal IFN responses depend on activation of hematopoietic cells by RV outside the replicating epithelial compartment is supported further by the observation that EW-infected suckling mice lacking IFN-αβγ-R are profoundly defective in their ability to induce type I IFN (∼700×) and to a lesser extent, ISG (less than ∼10×) transcription (Fig. 4D). These results indicate that within absorptive villous IECs, which are capable of synthesizing type I IFN transcripts and are the primary target of RV infection and pathogenesis, the establishment of an antiviral state by locally secreted IFNs is inhibited by, and does not protect against, homologous RV replication.

To gain insight into the mechanism by which this inhibition occurs, we infected mice lacking STAT1 (defective in types I, II, and III IFN-mediated feedback signaling) with RRV or EW and compared the induction of antiviral transcripts (Fig. 5). Infection with RRV or EW resulted in similar induction of Lgp2, as well as of the IRF3-responsive transcripts Rsad2 (viperin) and Isg15 (Fig. 5A). In contrast, RRV induced significantly higher levels of the NF-κB–dependent transcripts A20 and IκB-α compared with EW (P < 0.005, Welch’s t test adjusted for multiplicity). RRV also induced higher levels of types I, II, and III IFNs compared with EW in the absence of STAT1-mediated feedback amplification, indicating that heterologous RV cannot effectively regulate IFN induction in cells that encounter virus (Fig. 5B). Further analysis of total intestinal lysates by immunoblotting demonstrated accumulation of IκB-α following infection with EW but not RRV (Fig. 5C). Finally, to examine the relative contribution of types I, II, and III IFNs in restricting replication of EW and RRV in the intestine, we infected wt, IFN-αβγR^−/−, and STAT1^−/− suckling mice with either strain, and compared levels of RV gene 11 transcripts (Fig. 5D). The replication of EW was increased modestly by the combined absence of types I and II IFN receptors (∼25×), and there was no subsequent increase in STAT1^−/− mice. In contrast, replication of RRV increased substantially in IFN-αβγR^−/− mice (∼250×), with a further substantial enhancement in STAT1^−/− mice (∼20×). These findings indicate that early during RV infection, although type III IFNs may incrementally restrict the replication of heterologous RRV, they are not particularly effective in restricting homologous murine RV replication in the suckling mouse.

Fig. 5.

The induction of IFNs by heterologous RV correlates with its inability to inhibit NF-κB. (A–C) STAT1-deficient suckling mice were infected with EW or RRV, and at 16 hpi total intestinal lysates were analyzed by qRT-PCR (A and B) or immunoblotting (C). (D) wt, IFN-αβγ-R–deficient, or STAT1-deficient mice were infected with the RV strains indicated, and total intestinal RNA was analyzed by qRT-PCR at 16 hpi. Data shown are mean ±SD and representative of at least two experiments (n = 3–5 mice in each group per experiment for RNA analysis and n = 2 pooled mice per sample per experiment for protein analysis).

Discussion

Multiparametric analyses of antiviral responses generally are performed using pools of cells, resulting in “averaged” per cell measures. This approach is useful but has inherent drawbacks. Signal averaging obscures the heterogeneity in signaling states of individual cells; this is a limitation especially when the cell population consists of both infected and bystander cells. Here, we measured multiple early antiviral IFN response parameters in bulk small intestine, intestinal cell population subsets, and isolated single intestinal villous epithelial cells during acute infection with RV. Although this population of IECs is derived from the same stem cell and would be considered homogenous based on surface marker expression, we found considerable heterogeneity in response to virus. RV⁺ and bystander cells from infected mice displayed unique and shared signaling responses, and within each group, subsets of cells with unique antiviral responses were delineated. Individual enterocytes within the group may vary with regard to gene expression because each differs from the others with regard to physical position along the crypt–villus and proximal–distal axes, and distance from important regulatory nonepithelial cell types (e.g., lymphoid aggregates, dendritic cells). Thus, different enterocytes have different immediate microenvironments, which may lead to differences in gene expression, including innate immune responses.

Although RV infection resulted in the activation of several antiviral transcripts in specific groups of single cells (Fig. 3B), we found that these did not include the type I IFNs themselves. In a small subpopulation of IECs (enterocyte^Hi cells), RV infection mediated a significant decrease in type I IFN transcription, both in RV⁺ and bystander cells. In most IECs (enterocyte^Low cells), virus infection did not lead to an induction of type I IFN transcripts despite a robust induction of IRF3-dependent genes. Interestingly, levels of NF-κB–dependent Peli1 and A20/Tnfaip3 transcripts and the NF-κB RelA/p65 subunit transcript itself were significantly lower in bystander cells than IECs from uninfected mice. These reductions were not observed in RV⁺ cells, but NF-κB activation was not observed (Fig. 3B). NF-κB–dependent activation reportedly is essential for IFN synthesis following RV infection (6) as well as for basal IFN synthesis in uninfected revved-up cells (25), so the observed block in NF-κB–dependent activation likely plays an important role in the inability of RV-infected enterocytes to initiate IFN transcription. Other signaling changes also may contribute to the profound block in type I IFN induction in IECs following homologous RV infection. Infected enterocyte^Low cells had higher levels of Lgp2 transcripts: Lgp2 inhibits IFN synthesis in RV-infected IECs in vitro (16). Similarly, transcripts encoding Ifit1/ISG56 (an inhibitor of the IFN promoter), Rsad2 (viperin), and Isg15 (negative regulators of the IFN response) (26–28) were up-regulated in all CIA cells (Fig. 3B). The specific identities of RV-induced factors that mediate the suppressive changes in bystander villous IECs remain to be determined.

RV induced the expression of several IRF3-responsive transcripts in both bystanders and RV⁺ cells. In the absence of type I IFN induction, it seems likely that these transcripts were regulated, at least in the bystander CIAs, by type II (e.g., Igtp, Irf9) and/or III IFNs (e.g., category 5 and 6 transcripts) (22, 23), or by type I IFN secreted from hematopoietic cells. In contrast to type I IFNs, type III IFNs can be induced by either IRF3 or NF-κB activation alone and result in ISG expression that essentially is indistinguishable from that induced by type I IFN (22, 29). Thus, RV-activated IRF3 signaling in IECs might be sufficient to induce vSIG, type III IFN, and downstream ISG expression. We previously demonstrated that degradation of IRF3 by the RV NSP1 protein occurs more efficiently following IRF3 activation and at a step following the phosphorylation of a cluster of amino acids at the carboxyl terminus of IRF3 (9). Thus, one interesting possibility is that the IRF3-mediated transcription observed in single IECs represents the activation of this pathway before efficient NSP1-mediated degradation of IRF3 at later times.

Homologous (and heterologous) RV infection of suckling mice elicited a type I IFN response in bulk intestinal tissue lysates that overwhelmingly originated in hematopoietic cells (Fig. 4B). Induction did not require substantial viral replication as similar levels of IFNs were observed following infection with simian and murine RV strains (Fig. 4 C and D). Studies of primary human circulating plasmacytoid dendritic cells demonstrated that psoralen-inactivated RRV efficiently induces IFN (30). Further studies are under way to identify the exact nature of the specific IFN transcriptionally active hematopoietic cells and the mechanism initiating the RV signaling event in that cell. Although this IFN response does not appear to be highly effective in controlling homologous murine RV replication, it seems to play a critical role in restricting the replication of some heterologous RV strains in the gut and elsewhere (Fig. 5D) (4).

Our studies demonstrate that villous IECs are capable of transcribing type I IFN and IFN-stimulated genes (Fig. 3), indirectly providing evidence that RV actively counters both type I IFN synthesis and the effects of IFN-induced factors in IECs. It seems likely that the virus directly blocks the NF-κB signaling pathway. Previous studies using IL28Rα^−/− and STAT1^−/− mice provide conflicting evidence as to whether deletion of type III IFN signaling significantly enhances homologous RV intestinal replication and/or fecal shedding in suckling mice (10, 31). We observed that EW replication at 16 hpi in the intestine of suckling mice was enhanced only modestly in the absence of types I and II IFNs, and no further increase was observed in STAT1-deficient mice (Fig. 5D). In contrast, the replication of the simian RRV was found to be substantially enhanced by the absence of types I and II IFN receptors, and incrementally by the absence of STAT1 (presumably reflecting the additive antiviral effect of type III IFN on early simian RV replication) (Fig. 5D). However, whether type III IFNs play a role in the duration of RV replication and/or viral clearance will require additional kinetic studies. At the single-cell level, we measured only type I IFN levels; thus, the exact role of type III IFN, if any, in regulating homologous RV replication in IECs requires further evaluation. However, several type I/III IFN-dependent ISGs (Fig. 3B) were induced efficiently within individual RV⁺ IECs during acute infection with homologous RV despite its robust replication phenotype. Hence, depletion of type III IFN would be anticipated to confer only an incremental advantage to homologous RV replication.

To date, we have been unable to identify an in vitro replication system that recapitulates the strikingly different phenotypes of heterologous simian and homologous murine RV replication in the suckling mouse intestine. Here, we found that both homologous and heterologous RV activation of type I IFN in bulk intestinal lysates depends on signal amplification via IFN receptors and STAT-1 and is predominantly hematopoietic cell derived. In wt mice, replication of RV in the intestinal epithelium thus resembles a situation in which exogenous IFN is added, and homologous RV can successfully overcome such an antiviral state. Our findings also help explain the puzzling observations that exogenously added type I/II IFNs fail to suppress homologous RV replication in IECs of suckling mice and that the inhibition of types I, II, and III IFN receptor-mediated (i.e., STAT1) signaling does not significantly enhance murine RV replication in suckling mice (4, 10, 32, 33). We identified unique subsets of RV⁺ and bystander IEC responses associated with the high replication phenotype of homologous infection. Remarkably, the “remote” inhibition of NF-κB transcription by RV in bystander IECs via unknown factors is similar to recently identified strategies encoded by commensal and pathogenic gut bacteria to influence intestinal homeostasis and disease (34). Absence of NF-κB–mediated transcription also was observed within RV⁺ cells, and at the bulk level this correlated with an increase in IκB-α protein (Fig. 5C) and a lack of NF-κB–dependent transcriptional induction (Fig. 5A). The accumulation of IκB-α protein in RV-infected cells has been shown to be a consequence of NSP1-mediated degradation of β-TrCP. In contrast, we found that simian RV infection did not result in IκB-α protein accumulation—presumably a result of NF-κB pathway activation because the absence of IκB-α was accompanied by NF-κB–dependent transcriptional induction (Fig. 5). Finally, important heterogeneity in the in vivo antiviral response to RV was identified using a unique single-cell analytic approach to resolve averaged bulk measures. RVs are one of several important intestinal pathogens, including noroviruses, astroviruses, and shigella. Each of these pathogens induces specific changes in the intestinal cell it invades, causing ramifications on the overall host immune and inflammatory response. We predict that similar analytic strategies to compare global and localized host signaling following such infections or live vaccinations at other mucosal sites (35), such as the respiratory epithelium, also might reveal the existence of cell type–specific anti-innate immunity strategies in vivo.

Materials and Methods

Viruses and Infection.

The in vivo propagation and infectivity of virulent wt murine RV EW was described previously (5). Three- to 5-d-old suckling 129Sv mice were orally inoculated with 10⁵ 50% diarrhea dose (DD₅₀) of EW, 10⁵ pfu of D6/2, or 10⁷ pfu of RRV, or were mock infected. Mice lacking IFN-αβγ receptors or STAT1 were infected as described previously (4). Sixteen hours post infection, mice were killed and intestinal tissues collected for enterocyte isolation and bulk analysis. Mice were maintained in the Veterinary Medical Unit of Palo Alto Veterans Affairs Health Care System. The Stanford Institutional Animal Care Committee approved all animal studies.

Tissue Preparation.

Tissues were processed as described in SI Materials and Methods.

Antibody Staining and Flow Cytometry.

Cells were stained using antibodies and flow cytometry performed as described in detail in SI Materials and Methods.

Bulk Analysis of Tissue by RT-PCR and Immunoblots.

Proximal small intestines or double-sorted ESA⁺CD45⁻, ESA⁻CD45⁺, and ESA⁻CD45⁻ cells were collected into TRIzol (Invitrogen) for RNA purification. RT-PCR and immunoblotting were carried out as described in SI Materials and Methods.

Gene Expression Analysis in Single Cells.

Single cells (purity >95%) were double-sorted into individual wells of 96-well plates containing 5 μL of lysis buffer (CellsDirect qRT-PCR kit, Invitrogen) and RNase inhibitor (Invitrogen). After osmotic lysis, reverse transcriptase (RT), Taq polymerase (Platinum Taq, Invitrogen), and 0.1× Taqman assay mix were loaded into each well. RT (15 min—50 °C, 2 min—95 °C) was followed by preamplification for 20 cycles of 15 s—95 °C, 4 min—60 °C. Diluted total RNA from uninfected and infected small intestines was run in parallel on each chip. Preamplified cDNA from each well and Taqman qPCR mix (Applied Biosystems) were inserted into sample inlets, and assays were inserted into the assay inlets of a 96.96 Fluidigm chip (Fluidigm) with a Hamilton STARlet robot. The chip was loaded in a chip loader (Nanoflex, Fluidigm) and transferred to a reader (Biomark, Fluidigm) for thermocycling and fluorescent quantification.

Statistical Analyses.

See SI Materials and Methods for a detailed description of statistical methodology.