Polymorphisms in Toll-Like Receptor 4 Underlie Susceptibility to Tumor Induction by the Mouse Polyomavirus

Abstract

PERA/Ei (PE) mice are susceptible to tumor induction by polyomavirus (Py), while C57BR/cdJ (BR) mice are resistant. Antigen-presenting cells from BR mice respond to the virus with interleukin-12 (IL-12) and those from PE mice with IL-10. These polarized cytokine responses underlie the development of effective antitumor immunity in BR mice and the lack thereof in PE mice. An ex vivo cytokine production assay using spleen cells from infected [PE × BR] F2 mice together with a genome-wide SNP (single-nucleotide polymorphism)-based QTL (quantitative trait locus) analysis was used to map the determinant of cytokine production to a region of chromosome 4 carrying the Toll-like receptor 4 (TLR4) gene. Genotyping of infected F2 mice showed concordance of TLR4 allele-specific DNA sequences with the cytokine profile. Cytokine responses elicited by Py are MyD88 dependent. Bacterial lipopolysaccharide (LPS), a known TLR4 ligand, induced the same polarized responses as the virus in these host strains. Spleen cells from C3H/HeJ and C57BL/10ScNJ LPS-nonresponsive mice challenged in vitro with Py showed an impaired IL-12 response but were unaffected in IL-10 production. TLR4s of strains PE and BR differ by 3 amino acid substitutions, 2 in the extracellular domain and 1 in the intracellular domain. cDNAs encoding the TLR4s signaled equally to an NF-κB reporter in 293 cells in a ligand-independent manner. When introduced into TLR2/TLR4 double-knockout macrophages, the TLR4 cDNA from BR mice conferred a robust IL-12 response to Py and no IL-10 response. The TLR4 cDNA from PE mice failed to confer a response with either cytokine. These results establish TLR4 as a key mediator of the cytokine response governing susceptibility to tumor induction by Py.

INTRODUCTION

Mouse polyomavirus (Py) is the oldest member and among the best studied of the polyomavirus group. Distinct advantages in this experimental system derive from the genetics of the mouse as the natural host. Inbred strains of mice vary greatly in terms of their susceptibilities to tumor development and to acutely lethal infection by the virus (4, 5, 8). Studies of host immune responses have uncovered mechanisms of susceptibility and resistance operating at multiple levels involving aspects of innate and adaptive immune systems (26). In mice carrying the H2k major histocompatibility complex (MHC) haplotype, the endogenous mouse mammary tumor virus superantigen Mtv-7 sag confers susceptibility by eliminating essential precursors of virus-specific T lymphocytes (16). C57BR/cdJ (BR) mice, which carry H2k but lack Mtv-7, mount effective T cell responses against the virus and are resistant (15, 16, 34). Other H2k strains, such as PERA/Ei (PE), while free of endogenous mammary tumor virus superantigens (33), are nevertheless susceptible based on their innate immune response (31).

The innate responses of PE and BR mice to Py have been compared (31). Antigen-presenting cells (APCs) from PE mice respond to the virus with the type 2 cytokine interleukin-10 (IL-10) and those from BR mice with the type 1 cytokine IL-12. B cells contribute along with macrophages and dendritic cells to these responses (32). Although PE mice mount a transient virus-specific cytotoxic-T cell response, this gives way under the influence of IL-10 to a failure to sustain adaptive immunity to the tumors (31). [PE × BR] F1 mice develop tumors, although the tumors develop at later times and are fewer than in PE mice. Tumor susceptibility is inherited as a codominant or quantitative trait governed apparently by a single gene (33).

Recombinant IL-12 protects F1 mice completely and PE mice partially from the development of Py-induced tumors (31). Results of further investigations pointed to the importance of a Toll-like receptor (TLR) in mediating different cytokine responses (32). Polyomavirus-like particles elicit the same host-specific cytokine responses as those elicited by infectious virus. In contrast, individual capsomeres corresponding to single-pentamer assemblies of the major capsid protein VP1 do not (32). These findings suggest recognition by a TLR and the importance of highly repetitive structural motifs or pathogen-associated molecular patterns (PAMPs) in the fully assembled virus capsid. An interesting parallel may be drawn between these findings and those regarding the induction of T cell-independent type 2 antibody (Ab) responses, which are elicited by whole virus but not by capsomeres (26, 27).

Members of the TLR family of pattern recognition receptors play a critical role in controlling the outcome of microbial infections. The primary factor that determines which TLRs participate in any given infection is the ligand which an individual receptor detects. As such, TLRs that are important for controlling bacterial infections typically detect bacterial cell surface components, such as lipopolysaccharide (LPS) (detected by TLR4) or flagellin (detected by TLR5). Similarly, TLRs that are important for controlling viral infections typically detect nucleic acid structures that are commonly found in viral genomes, such as double-stranded RNA (TLR3) or unmethylated CpG DNA (TLR9). Py is internalized along an endosomal pathway leading to the endoplasmic reticulum, where uncoating occurs (29). This pathway of cell entry and the fact that empty virus-like particles induce the same host-specific cytokine responses as those induced by the virus rule against a requirement for viral DNA (CpG) interacting with TLR9. Definitive evidence implicating a specific TLR in the response to Py infection is lacking. The present investigation was undertaken to identify the genetic and immunological determinants of cytokine responses which underlie susceptibility to tumor induction by Py in the PE × BR host system.

MATERIALS AND METHODS

Mouse strains and virus inoculations for the ex vivo cytokine assay.

PERA/Ei (PE) and C57BR/cdJ (BR) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). All mice were bred and maintained in a specific-pathogen-free (SPF) animal facility at Harvard Medical School. Newborn mice were inoculated intraperitoneally (i.p.) with ∼50 μl of a virus suspension containing 10⁶ PFU of the PTA strain of polyomavirus. Infected mice were transferred and held in a dedicated infected-animal facility. Infected mice were sacrificed at 7 days postinoculation, and spleen cells were cultured for 24 h. Peritoneal exudate cells (PECs) from 3- to 4-week-old PE and BR mice were collected in cold phosphate-buffered saline (PBS). Culture supernatants were assayed for IL-10 and IL-12 by double-sandwich enzyme-linked immunosorbent assay (ELISAs). Protocols for animal studies were reviewed and approved by the Harvard Medical Area Standing Committee on Animals.

Virus purification and infections of cultured cells for assays of cytokine production.

The “high-tumor” strain PTA of polyomavirus (10) was grown on primary mouse kidney cells and purified by cesium chloride density gradient centrifugation to remove any possible LPS contamination. Assurances that virus preparations were free of LPS came from the pathogen-free status of ICR mice (Harlan Laboratories) used for primary kidney cultures and from the findings that crude lysates of uninfected cultures as well as portions of the density gradients above the virus band were negative in tests of cytokine production. Macrophage cell lines established from wild-type and MyD88-knockout (KO) mice on a C57BL/6 background were described previously (17, 25). Infections were carried out at a multiplicity of infection of 2 to 5 PFU per cell. Culture supernatants were assayed for IL-10 and IL-12 by ELISAs at 48 h postinfection.

Generation of TLR4 cDNA clones.

Total RNAs were extracted from spleens of PE and BR mice with QIAzol lysis reagent (Qiagen) and chloroform and cleaned up with an RNeasy minikit (Qiagen). One microgram of each RNA preparation was used for cDNA library production with a QuantiTect reverse transcription kit (Qiagen). The TLR4 gene was amplified with High Fidelity Platinum PCR SuperMix (Invitrogen) and a primer set (mTLR4 F [5′-CAGGATGATGCCTCCCTGGCTCCTG] and mTLR4 R [5′-CTCCTCAGGTCCAAGTTGCCGTTTC]). The amplified gene was cloned into pcDNA3-TOPO (Invitrogen). Clones were isolated from Escherichia coli transformants and verified by sequencing. The clones were further used for luciferase analyses and as templates for generating pLenti viral clones (see below). A set of primers (mTLR4 Lenti F [5′-TAGGATCCGATGATGCCTCCCTGGCTCCTGGCTA] and mTLR4 Lenti R [5′-TAGGGCCCTCAGGTCCAAGTTGCCGTTTCTTGTTC]) were used to amplify the TLR4 gene from TLR4-pCDNA3 clones. The PCR products were inserted into BamHI and ApaI sites in the plasmid pLenti6/V5-D-TOPO (Invitrogen). The Lenti-cDNA clones were isolated from transformed bacterial cells and used for the production of lentivirus.

Production of lentiviral stocks.

PE- and BR-TLR4-pLenti clones were mixed with ViraPower packaging mix (Invitrogen) and transfected into 293FT cells with Lipofectamine 2000 (Invitrogen). The viral supernatants were harvested after 48 to 72 h, spun at 3,000 rpm for 10 min, and filtered with 0.45-μm Millex syringe filters (Millipore). Virus was concentrated by mixing 4 volumes of viral supernatant with 1 volume of 5× polyethylene glycol (PEG) precipitation solution (40% polyethylene glycol, 0.4 M NaCl, and 2 mM EDTA), stored overnight at 4°C, and centrifuged at 3,000 rpm for 30 min. Pelleted virus was resuspended in PBS buffer and stored at −80°C.

Luciferase assay.

Human embryonic 293 kidney cells were cotransfected with pBIIX with an NF-κB promoter driving firefly luciferase (12) and pRL-TK (Promega) and either pcDNA3, pcDNA3-PE TLR4, or pcDNA3-BR TLR4. At 24 h posttransfection, cells were infected with polyomavirus and incubated for 16 h. Cells were washed twice with cold PBS and lysed in 200 μl of passive lysis buffer (Promega, Madison, WI). Twenty microliters was subjected to a luciferase assay. Firefly luciferase activities were normalized to those of Renilla luciferase by using the Promega Dual-Luciferase reporter assay system.

Lentivirus infections and cytokine assays.

Immortalized bone marrow-derived macrophages from TLR2/TLR4 double-knockout mice (BEI Resources) were infected with lentivirus expressing either PE TLR4 or BR TLR4 in Dulbecco's modified Eagle's medium (DMEM)–10% fetal bovine serum with 10 mM HEPES and 10 μg/ml Polybrene. After 16 h, cells were infected with cesium chloride density gradient-purified polyomavirus and incubated for an additional 24 h. A two-site capture ELISA was performed to detect cytokines in the culture supernatant. Pairs of matched monoclonal Abs specific to IL-10 and IL-12 and the relevant recombinant cytokine proteins as standards were all purchased from BD Pharmingen. TLR4 expression in lentivirus-infected cells was measured by fluorescence-activated cell sorter (FACS) analysis with phycoerythrin-conjugated anti-mouse TLR4 from eBioscience.

Genome-wide SNP array, QTL mapping, and TLR4 genotyping.

DNA from parental and 114 F2 mice were extracted and genotyped at the Partners Healthcare Center for Personalized Genetic Medicine (PCPGM) core at Harvard Medical School using a whole-genome single-nucleotide polymorphism (SNP) panel with Sequenom iPLEX, as described previously (20). Of 394 SNPs in the panel, 167 were informative. The F2 phenotypic data based on cytokine secretion along with the genotypic data for the 167 informative SNPs were imported into J/QTL (http://churchill.jax.org/software/jqtl.shtml), a graphical interface to R/QTL, for quantitative trait locus (QTL) mapping (7). QTLs were identified by using a binary-phenotype model with the expectation maximization (EM) algorithm through the “run one QTL genome scan” function. P values were adjusted genome wide (18) by using 100,000 permutations. A 1.5-log of odds (LOD) support interval (19) expanded to the nearest marker was calculated for the region surrounding the significant locus found on chromosome 4. To determine the genotype at the TLR4 gene, genomic sequences between bp 66501650 and 66502337 on chromosome 4 (corresponding to mRNA sequences between bp 1666 and 2353 [GenBank accession number NM_021297]) were amplified. This region includes 4 SNPs that distinguish the BR and PE parental alleles: rs13489095, rs27883164, rs13498478, and rs13489097. Sequences were aligned by using Sequencher 5.0. All 4 SNPs were concordant with respect to genotype for each sample tested.

RESULTS

Cytokine responses to Py are MyD88 dependent.

MyD88 functions as a key adaptor protein in signaling through TLRs (1). As a further test for the involvement of a TLR, macrophages from wild-type or MyD88-knockout mice were exposed to virus, and levels of secreted IL-12 and IL-10 were measured at 48 h postinfection (Table 1). Macrophages from wild-type mice showed a clear IL-12 response and no IL-10 response. Macrophages from MyD88-knockout mice failed to show a detectible IL-12 response but responded instead with IL-10. These results are consistent with the B6 strain origin of the macrophage cell lines, as B6 mice, like BR mice, are highly resistant to tumor induction (15) and are known to be IL-12 responders to Py (P. Velupillai, unpublished observation). These results establish that the elicitation of cytokine responses by Py is MyD88 dependent, in line with previously reported findings implicating TLR involvement (32).

Table 1.

Cytokine responses to Py by macrophages from MyD88-knockout and wild-type mice^a

Cytokine	Infection	Mean cytokine concn (pg/ml) ± SD
		MyD88 KO	Wild type
IL-12	−	<25	<25
	+	<25	426 ± 8
IL-10	−	<25	<25
	+	114 ± 7	<25

^aNumbers represent cytokine concentrations (pg/ml) in culture supernatants at 48 h postinfection. Shown are averages of data from two independent experiments with triplicate determinations ± SD.

A single gene governs the difference in cytokine responses between PE and BR mice.

An ex vivo cytokine production assay was used as a first step toward identifying the gene(s) governing the specificity of the cytokine responses. We equated the “cytokine response gene” with the “tumor susceptibility gene” to map the PE × BR cross. The use of this assay is justified in lieu of tumor studies based on its rapidity, simplicity, and quantitative endpoints as well as on biological findings indicating the importance of polarized cytokine responses in the development of adaptive tumor immunity in this system (31, 32). IL-10 dominates the response to Py in infected PE mice, and IL-12 dominates in infected BR mice. F1 mice show a transient IL-12 response that rapidly gives way to an IL-10-dominated response (31).

One hundred fourteen newborn F2 mice were infected. Spleen cells taken 1 week after infection were cultured for 24 h, and culture supernatants were assayed for IL-10 and IL-12 by ELISAs. IL-12 and IL-10 concentrations were combined into log₂-transformed IL-12/IL-10 ratios for each mouse, and the values were used to plot the distribution of responses (Fig. 1). Cutoffs appear around −2 (corresponding to an IL-12/IL-10 ratio of 0.25) and +2 (corresponding to an IL-12/IL-10 ratio of 4.0), yielding three distinct groups, labeled PE-like, F1-like, and BR-like. These ranges found in F2 mice are consistent with values found previously for parental and F1 strains (31). The observed distribution of 27 PE-like, 61 F1-like, and 26 BR-like responders is in good agreement with the expected 1:2:1 ratio for a single gene (χ² = 0.5789) governing the cytokine response, with the allele carried by PE mice acting in a codominant manner, as found previously for tumor susceptibility (33).

Fig 1.

Distribution of F2 mice by cytokine response. The numbers of F2 mice with a given cytokine ratio are displayed against the log₂-transformed ratio (see the text).

The cytokine response gene is mapped to an interval on chromosome 4.

To map the gene governing the cytokine responses, a genome-wide scan was carried out on DNAs from 15 PE-like responders and 17 BR-like responders (Fig. 1). A panel of 394 SNPs distributed across the 19 autosomes was used. F1 mice generated from both parental directions gave the same results, indicating no involvement of the X chromosome. QTL software was used to calculate LOD scores based on 167 informative SNPs in the panel (Fig. 2A). A single prominent peak was seen on chromosome 4, centered on SNP rs4224562. The association with this locus was highly significant, with an LOD score of 10.21 (P value of <0.0001). A genotype plot for this region of chromosome 4 (Fig. 2B) shows that all 17 BR-like responders were homozygous for the BR allele of SNP rs4224562. The 15 PE-like responders were either homozygous or heterozygous for the PE allele of this SNP, consistent with the dominance of the IL-10 response in PE mice. These results suggest a single gene on chromosome 4, the dominant allele of which is carried by strain PE and governs the IL-10 response. The 1.5-LOD support interval (19) defines a region from Mbp 65.483 to 96.092 on chromosome 4 which is likely to contain the cytokine response gene. The TLR4 gene is located in this interval at Mbp 66.489 to 66.591. Since TLR4-mediated signal transduction requires MyD88, we considered this receptor a logical candidate gene governing cytokine responses and tumor susceptibility in the PE × BR cross.

Fig 2.

The cytokine response gene maps to an interval on chromosome 4. (A) Genome-wide QTL scan of 15 PE-like and 17 BR-like F2 mice determined from the experiment depicted in Fig. 1. (B) Genotype plot in the region of linkage on chromosome 4 for the 32 mice described above for panel A. Blue indicates homozygous PE, red indicates homozygous BR, and green indicates heterozygous mice. The asterisk indicates the SNP corresponding to the LOD peak in panel A (see the text).

Sequencing identifies polymorphisms in the TLR4s of PE and BR mice, and these polymorphisms are linked to cytokine responses in F2 mice.

To confirm the expected presence of polymorphisms in the TLR4 genes between PE and BR mice, cDNAs were cloned from total spleen RNAs. The entire coding sequences and 3′-untranslated regions were sequenced (Fig. 3). Three single-nucleotide substitutions gave rise to 3 amino acid substitutions in exon 3: 2 in the ectodomain and 1 in the cytoplasmic domain. Two synonymous substitutions were also found in exon 3. The 3′-untranslated regions showed two SNPs along with an expansion of a dinucleotide repeat [(GT)n] in strain PE, as indicated.

Fig 3.

Polymorphisms in the TLR4 gene between PE and BR mice. cDNAs were cloned from spleens and sequenced. Three nonsynonymous changes with amino acid substitutions are indicated. Two synonymous changes in Exon 3 were also noted: AAC in BR mice and AAT in PE mice at cDNA nucleotide position 1746 and AAC in BR mice and AAT in PE mice at position 2178 (see the text). TM, transmembrane sequence; UTR, untranslated region.

To test the TLR4 gene as the determinant of the cytokine response to Py in the PE × BR cross, four of the SNPs embedded in the TLR4 locus were utilized (see Materials and Methods). A single 687-bp amplicon covering these SNPs was used to genotype F2 mice. The 21 parental-like responders (Fig. 1) not previously included in the genome-wide scan (Fig. 2) were tested. These included 9 BR-like and 12 PE-like responders. The former responders were all homozygous for the BR allele SNPs. The latter group of PE-like responders included 5 that were homozygous for the PE allele SNPs and 7 that were heterozygous. Full concordance was seen across the four TLR4 gene-specific SNPs in each of the 21 mice. These results are in full agreement with the dominance of the PE (IL-10) response and with the TLR4 gene as the determinant of cytokine production.

LPS and Py elicit the same polarized cytokine responses.

To functionally test the TLR4 gene as the candidate gene governing responses to Py, comparisons were made with LPS, a known TLR4 ligand. LPS, together with LPS binding protein (30), CD14, and MD2 (28, 35), acts through TLR4 (6) to activate a complex regulatory network of transcription factors, resulting in increased NF-κB transcription (2). Peritoneal exudate cells (PECs) are most likely the first cells to encounter the virus following i.p. inoculation, the route used in studies of tumor induction by the virus (33). PECs from naive PE and BR mice were exposed to LPS at 100 ng/ml for 40 h. LPS induced the same polarized host-specific cytokine responses as those induced by the virus, with IL-10 predominating in PE mice and IL-12 predominating in BR mice. Concentrations in culture supernatants from PE mice were 168 ± 8 pg/ml of IL-12 and 707 ± 16 pg/ml of IL-10, and those in supernatants from BR mice were 805 ± 18 pg/ml of IL-12 and 275 ± 25 pg/ml of IL-10 (averages of data from 3 animals per group ± standard deviations [SD]). Similar results were found with spleen cells (not shown). These results using LPS as a known TLR4-specific PAMP indicate intrinsic differences in TLR4s between the parental strains and show that Py and LPS elicit the same skewed cytokine responses in these strains.

To extend the analysis comparing Py and LPS, mice known to be altered in their LPS responses were tested for their responses to the virus. Two pairs of related strains are available based on previous work defining certain alleles of the TLR4 gene as the determinant of responsiveness to LPS. C57BL/10ScSnJ mice are normally responsive, while the closely related strain C57BL/10ScCr carries a large deletion including the TLR4 locus and is hyporesponsive (21, 22). Similarly, strain C3H/HeJ carrying a substitution in the cytoplasmic domain of TLR4 (P712H) confers hyporesponsiveness to LPS, while the closely related strain C3H/HeSnJ is normally responsive (23). Spleen cells from uninfected mice of these strains were challenged in vitro with purified Py, and cytokine responses were measured (Table 2). In both pairs, the IL-12 responses were 6-fold higher in spleen cells from the LPS-responsive strain than in those from the LPS-nonresponsive strain. TLR4-null or mutant mice are thus impaired in their responses to Py. IL-10 responses were the same in LPS-responsive and -nonresponsive mice, indicating that the induction of this cytokine by Py occurs via a TLR4-independent pathway.

Table 2.

Cytokine responses to Py by APCs from LPS-responsive and LPS-nonresponsive mice^a

Mouse strain (LPS responsiveness)	Mean cytokine concn (pg/ml) ± SD
	IL-10		IL-12
	Control	Virus	Control	Virus
C3H/HeJ (−)	50 ± 5	750 ± 55	50 ± 9	250 ± 45
C3H/HeSnJ (+)	150 ± 18	900 ± 85	150 ± 20	1500 ± 120
C57BL/10ScNJ (−)	100 ± 12	850 ± 70	50 ± 6	300 ± 28
C57BL/10ScSnJ (+)	100 ± 15	800 ± 75	100 ± 15	1800 ± 170

^aSpleen cells from uninfected mice of each strain were exposed to Py. Culture supernatants were tested for IL-10 and IL-12 secretion after 72 h. Shown are averages of determinations from three mice of each strain ± SD.

Functional tests for differences intrinsic to TLR4s from PE and BR mice.

To test directly if the difference in the cytokine responses to Py in PE and BR mice is intrinsic to their TLR4 genes, vectors expressing the complete coding regions of the TLR4s were established and introduced into 293 cells. The cDNAs were transfected along with an NF-κB–luciferase reporter as an indicator of signaling through TLR4. Transfected cells were incubated for 24 h and then either challenged with virus or mock infected and incubated for another 16 h. FACS analysis was used to monitor the expression of the TLRs. Both PE and BR TLR4s induced the activation of NF-κB (Fig. 4) in uninfected cells. Py challenge had little or no effect on the degree of activation, although it did result in a roughly 2-fold enhancement of expression of the TLRs, as measured by the mean fluorescence index. The ability of the TLR4s to signal to NF-κB in 293 cells in the absence of virus is due most likely to overexpression driving ligand-independent dimerization. While confirming at least a partial functionality of the parental TLR4s, these results fail to discriminate between them in terms of their responses to the virus.

Fig 4.

NF-κB activation in 293 cells by TLR4 from PE and BR mice. Mean fluorescence indices (MFI) are given as a measure of levels of TLR4 expression. Error bars represent standard deviations of data from 3 independent determinations (see the text).

We addressed the functions of PE and BR TLR4s in a physiologically relevant cell type by introducing the cDNAs into established bone marrow-derived macrophages from TLR2/TLR4 double-KO mice. The TLR4 cDNAs were inserted into a lentiviral vector and transduced into macrophages (Table 3). The expression of the TLRs was confirmed by FACS analysis, with mean fluorescence indices of 7,425 ± 29 for PE TLR4, 3,510 ± 30 for BR TLR4, and 812 ± 26 for the empty vector control. Macrophages expressing BR TLR4 showed a strong IL-12 response and no IL-10 response when challenged with Py. Cells expressing PE TLR4 showed no detectible response with respect to the secretion of either IL-10 or IL-12. These cells were nevertheless able to respond to treatment with an anti-CD40 antibody, which led to increases in the levels of both cytokines.

Table 3.

Cytokine responses to Py by macrophages from TLR2/TLR4 double-knockout mice expressing PE or BR TLR4^a

Treatment	Py infection	Mean cytokine level (pg/ml) ± SD
		IL-10	IL-12
Lentivirus
PE TLR4	+	<25	<25
BR TLR4	+	<25	2,402 ± 185
Empty vector	+	<25	<25
Anti-CD40 antibody
PE TLR4	+	808 ± 78	1,138 ± 120
BR TLR4	+	814 ± 64	1,222 ± 95

^aValues represent the averages of data from two independent experiments with triplicate measurements in each experiment. Controls without lentiviral or Py infection gave <25 pg/ml of both cytokines.

DISCUSSION

PE mice are fully susceptible and BR mice are completely resistant to tumor induction by Py. Evidence presented previously indicated that this striking difference depends on a single gene (33) operating at the level of innate immune responses, specifically in the polarized cytokine responses elicited by the virus from APCs of the two strains (31, 32). IL-10 produced by PE mice and IL-12 produced by BR mice are key cytokines in regulating Th2 and Th1 responses, respectively. IL-10 is known to downregulate Th1 cytokines (11), consistent with the dominance of the type 2 response and susceptibility to tumors in [PE × BR] F1 mice. Here we have used an ex vivo cytokine production assay as a proxy for tumor susceptibility. The cytokine response by APCs was used to phenotype Py-infected [PE × BR] F2 mice. By use of an SNP array and QTL analysis, the genetic determinant(s) of the response was mapped to a single locus on chromosome 4. The TLR4 gene resides in this region of chromosome 4 and emerged as a logical candidate based on previously reported evidence for the involvement of a TLR in determining the specificity of cytokine responses (32). The sequencing of parental TLR4 genes confirmed the presence of multiple SNPs. The genotyping of F2 mice revealed a complete concordance of parental TLR4 allele-specific DNA sequences with the cytokine responses of F2 mice.

Cytokine responses elicited by Py are MyD88 dependent. Evidence specifically in support of the TLR4 gene comes from results comparing responses to Py with the responses to LPS, a known TLR4 ligand. LPS elicited the same skewed responses as those elicited by Py in APCs from PE and BR mice. An asymmetry emerged in results for Py and splenocytes from LPS-responsive and -nonresponsive mice. IL-12 responses to Py were impaired in cells from LPS-nonresponsive mice, while IL-10 responses were unaffected. These results support TLR4 as the mediator of the IL-12 response to Py and indicate that the IL-10 response is elicited by the virus through a different pathway.

Both PE and BR TLR4s are able to signal to an NF-κB–luciferase reporter in 293 cells, indicating that the recruitment of TIRAP and MyD88 to the TIR domain of TLR4 and further downstream signaling proceed normally under conditions of overexpression. However, due to the facts that signaling was not stimulated by virus in 293 cells and the TRIF-mediated signaling pathway downstream of TLR4 is largely absent in these cells (J. Kagan, unpublished results), we sought a more relevant cell culture system in which the functions of PE and BR TLR4s could be compared. A macrophage cell line from a TLR2/TLR4 double-knockout mouse and lentiviral vectors were used to assess functional differences intrinsic to the TLR4s. TLR4 from BR mice conferred a robust IL-12 response to Py and no detectible IL-10 response in these cells. The IL-12 response that leads to effective tumor immunity in Py-infected BR mice is therefore dictated by TLR4. TLR4 from PE mice was unable to confer a response to Py involving either cytokine. The failure of PE TLR4 to confer an IL-10 response in these cells is consistent with results for splenocytes from LPS-nonresponsive mice and indicates that Py induces IL-10 through a TLR4-independent pathway. The latter pathway may depend on TLR2, which is absent in these macrophages. Further work will be required to understand the basis of the IL-10 response to Py in PE mice.

Taken together, the results of genetic and immunological approaches have led to the somewhat surprising conclusion that TLR4, the prototypical antibacterial TLR, plays an important role in initiating tumor immunity following infection by an oncogenic virus. It will be interesting to determine the mechanism of activation of TLR4 by Py in BR mice and compare it to the well-studied pathway of activation by LPS in further studies. Further investigations will be required to determine whether the functional differences between PE and BR TLR4s result from amino acid substitutions in one or both of the sites in the ectodomain potentially affecting the recognition and binding of the Py capsid or in the cytoplasmic domain possibly affecting binding and signaling through one or more adaptor protein complexes. The latter may include the TIRAP/MyD88 complex, which mediates signaling from TLR4 at the plasma membrane or, following internalization, the TRAM/TRIF pathway (14, 28) or both.

The human polyomaviruses are a rapidly expanding group (3, 13, 24). Although widespread in human populations and generally associated with silent persistent infections, under rare circumstances, these viruses cause severe disease. Immunological risk factors are likely involved, as in the case of Merkel cell carcinoma, a highly malignant form of cancer associated with the Merkel cell polyomavirus (9). Little is known about how the human innate immune system recognizes these viruses, how immune responses in general regulate the establishment of these largely silent infections, or how a genetic predisposition or a breakdown in immune responses may result in the development of disease. Further studies along the lines presented here may serve as a useful guide for studies of the human polyomaviruses and their roles in disease.