ABSTRACT
Nontyphoidal serovars of Salmonella enterica are pathogenic bacteria that are common causes of food poisoning. Whereas Salmonella mechanisms of host cell invasion, inflammation, and pathogenesis are mostly well established, a new possible mechanism of immune evasion is being uncovered. Programmed death ligand 1 (PD-L1) is an immunosuppressive membrane protein that binds to activated T cells via their PD-1 receptor and thereby halts their activation. PD-L1 expression plays an essential role in the immunological tolerance of self-antigens but is also exploited for immune evasion by pathogen-infected cells and cancer cells. Here, we show for the first time that Salmonella infection of intestinal epithelial cells causes the induction of PD-L1. The increased expression of PD-L1 through Salmonella infection was seen in both human and rat intestinal epithelial cell lines. We determined that cellular invasion by the bacteria is necessary for PD-L1 induction, potentially indicating that Salmonella strains are delivering mediators from inside the host cell that trigger the increased PD-L1 expression. Using knockout mutants, we determined that this effect largely originates from the Salmonella pathogenicity island 2. We also show for the first time in any cell type that Salmonella combined with gamma interferon (IFN-γ) causes a synergistic induction of PD-L1. Finally, we show that Salmonella plus IFN-γ induction of PD-L1 decreased the cytokine production of activated T cells. Understanding Salmonella immune evasion strategies could generate new therapeutic targets and help to manipulate PD-L1 expression in other diseases.
INTRODUCTION
Programmed death ligand 1 (PD-L1) is a hallmark transmembrane immunoregulator protein that can be expressed constitutively or upregulated during inflammation on hematopoietic cells, such as B cells, T cells, and dendritic cells, as well as nonhematopoietic cells, such as muscle, liver, heart, and intestinal epithelial cells (IECs) (1–3). PD-L1 binds to programmed death 1 (PD-1) on activated T cells and inhibits T cell receptor activation pathways, thus decreasing inflammatory cytokine production, T-cell proliferation, and survival (4). The absence or inhibition of PD-L1 can lead to autoimmune or inflammatory diseases, respectively (1, 5). When expressed on nonhematopoietic tissues, PD-L1 aids in peripheral tolerance to self-antigens (5, 6). It has been shown that PD-L1 plays a major role in controlling inflammation and peripheral tolerance in the gastrointestinal tract (5–8). PD-L1 can also stimulate the generation of T regulatory cells, which further contributes to the tolerogenic state within the gut (9).
The gastrointestinal tract plays a vital role in regulating immune responses and is home to the majority of microorganisms found in the human body. The intestinal lumen consists of hundreds of bacterial species that contribute to host digestion, nutrient availability, and the tolerogenic immune environment of the gut. In a healthy host, commensal bacteria represent the majority of the lumenal microbiota. However, upon consumption of contaminated food or water, pathogenic bacteria may enter this tightly regulated environment. Salmonella enterica serovar Typhimurium is one such pathogenic bacterium that causes a typhoid-like disease in mice or acute gastroenteritis in humans (10). Although not normally fatal in humans, Salmonella induces fever, severe diarrhea, and abdominal cramping (11).
The epithelial intestinal barrier is crucial in helping to control inflammatory responses and contributes to mucosal tolerance (12). Critical to Salmonella's pathogenicity is invasion of the intestinal epithelia (13). Invasion is carried out by the structural and effector proteins of a type 3 secretion system (T3SS) encoded within the Salmonella pathogenicity island 1 (SPI-1) and expressed under the control of the transcription factor hilD (14, 15). Once individual Salmonella bacteria successfully invade host cells, a shift in pH and limiting nutrients signal to the bacteria the change in environment (16–18). Consequently, Salmonella downregulates SPI-1 and induces SPI-2, a T3SS whose gene products facilitate survival in this unique niche. The effectors encoded by SPI-2 facilitate intracellular survival of Salmonella by preventing the host cell's lysosome from fusing with the Salmonella-containing vacuole (16, 19), thus maintaining a suitable reservoir for Salmonella. In addition, recent studies have elucidated immunomodulatory functions of SPI-2 effectors (reviewed in references 20 and 21), which serve to manipulate the host cells and thereby prevent an immune response that would otherwise combat Salmonella intracellular survival.
Salmonella may have several mechanisms to escape host immune detection, but most recently it has been shown to do so by increasing the PD-L1 expression of infected B cells to limit CD8 T cell responses (22, 23). These findings corroborate previous literature demonstrating that Salmonella-driven PD-L1 expression contributes to the killing of activated CD4 T cells soon after animal infection (24). Similar PD-L1 induction and immune suppression occur in other viral or bacterial infections, such as Helicobacter pylori infection of gastric epithelial cells (25), indicating that it is a common and successful immune evasion strategy. Our objective was to determine whether Salmonella caused an increase of PD-L1 in IECs, and if so, the effects of PD-L1 induction on T cell activation.
RESULTS
Salmonella induces PD-L1 in IECs.It is known that Salmonella induces PD-L1 in cells of the immune system (22–24, 26). Since this pathogen encounters IECs at an early stage of infection, we sought to determine whether Salmonella can also induce PD-L1 in this important cell type. In order to investigate changes in expression of PD-L1 on IECs, we used the well-established IEC colorectal adenocarcinoma cell lines, Caco-2 and HT-29. Basal expression of PD-L1 in Caco-2 and HT-29 cells was found to be low (data not shown), making these cell lines excellent models to study PD-L1 production in human IECs, provided the pathway components are expressed.
Caco-2 and HT-29 enterocytes are sometimes cultured together to recapitulate intestinal characteristics, including tight-junction formation from Caco-2 cells and mucous secretion from HT-29 cells. IEC-6 cells are cells isolated from rat intestinal epithelium that are also widely used for enterocyte research. Using these IECs, we compared the abilities of several intestinal bacteria to induce PD-L1 expression, as measured with quantitative PCR (qPCR) 24 h after initial exposure (Fig. 1). The Gram-negative Escherichia coli and Gram-positive Lactobacillus gasseri were chosen as representative commensal bacteria that enterocytes regularly encounter. E. coli and L. gasseri inoculation elicited no change of basal PD-L1 expression in any cell type. In contrast, the pathogenic Salmonella bacteria greatly induced PD-L1 mRNA expression. This effect was not unique to human IECs, since similar results were demonstrated in rat IECs (Fig. 1D). Salmonella increased PD-L1 expression from 5- to 100-fold, depending on the cell type. The largest induction occurred in HT-29 cells (approximately 80-fold compared to nontreated), whereas Caco-2 and IEC-6 cells demonstrated lesser but significant induction ranging from 4- to 12-fold. PD-L1 induction was independent of Gram stain classification, as neither E. coli nor L. gasseri had an effect. In order to minimize variability of responses from multiple cell types, we chose to further the investigation of Salmonella-mediated PD-L1 induction in Caco-2 cells.
Salmonella increased PD-L1 mRNA expression in human and rat intestinal epithelial cells. Intestinal epithelial cells were incubated with the commensal bacterium Escherichia coli, Nissle strain (Nis), or Lactobacillus gasseri (LaB) or the pathogenic bacterium Salmonella enterica serovar Typhimurium (ST) for 1 h before bacterial removal and gentamicin addition. Intestinal epithelial cells were cultured for a further 24 h, after which the RNA was isolated, quantified via qPCR, and normalized to GAPDH. PD-L1 expression was measured in a 3:1 mixture of Caco-2:HT-29 cells (A), HT-29 cells alone (B), Caco-2 cells alone (C), or rat IEC-6 cells (D). One-way analysis of variance (ANOVA) with Dunnett's posttest was performed to compare each mean to nontreated cells (NT) in every cell type. Graphs are representative of averages of three individual preparations.
To corroborate the mRNA transcript increase, we looked for differences in Caco-2 PD-L1 surface protein expression after Salmonella treatment. We first generated a novel Salmonella strain that constitutively expressed blue fluorescent protein (BFP) (see Table S2 in the supplemental material) to identify infected Caco-2 cells. At 24 h after Salmonella inoculation, the cells were trypsinized, stained for surface PD-L1, and analyzed via flow cytometry. The PD-L1 mean fluorescent intensity was significantly increased (by about 60%) in Salmonella-infected cells compared to uninfected cells from the same well (Fig. 2). Nontreated Caco-2 cells were used as a negative control, and gamma interferon (IFN-γ) was used as a positive control since it is known to increase expression of PD-L1 via IFN regulatory factors found in the PD-L1 promoter region (27).
Surface expression of PD-L1 was increased after Salmonella treatment. Flow cytometry was performed on BFP-containing Salmonella-treated Caco-2 cells; infected cells are represented as BFP+, and uninfected cells from the same well are represented as BFP−. At 24 h after bacterial exposure, the Caco-2 cells were stained for PD-L1. A negative control of nontreated cells (NT) and a positive control of 5 ng/ml IFN-γ-treated cells are included for reference. A paired Student t test was performed between the BFP+ and BFP− cells only. The histogram depicts representative averages from three individual preparations.
We performed preliminary in vivo infections of five mice to identify whether PD-L1 is expressed on epithelial cells in the presence of Salmonella (Fig. 3). After analysis of at least 10 images from the infected ceca, we qualitatively observed strong PD-L1 expression on epithelial cells near the Salmonella. PD-L1 is also basally expressed on some mouse intestinal cells, and this may depend on the type of intestinal epithelial cell or its microenvironment. The PD-L1 staining was usually brighter on epithelial tissue closest to the Salmonella than in areas nonadjacent to the bacteria (Fig. 3). Unfortunately, due to limitations of properly stained tissues, we were unable to confidently quantify the PD-L1 staining patterns, and we can therefore only claim this as a preliminary observation that needs further validation. For the scope of this work, we aimed to address mechanistic and functional questions of this PD-L1 upregulation in a controlled, in vitro setting.
PD-L1 expression is present in regions of dense Salmonella infection. Five 7-week-old female C57BL/6 mice were administered oral antibiotics. After 24 h, they received 109 CFU orally of a spontaneous streptomycin-resistant isolate of Salmonella. At 48 h after inoculation, the mice were euthanized, and ceca were harvested. Slides were stained for Salmonella (green), PD-L1 (red), and nuclei (blue). (A) Confocal microscopy was performed, and a representative image was selected. (B) Enlargement of the white box from the image in panel A.
Mechanism of Salmonella-mediated PD-L1 induction.We chose to continue mechanistic investigations into Salmonella PD-L1 induction using qPCR rather than protein analysis, since qPCR is a more sensitive assay. We hypothesized that Salmonella invasion was required for the induction of PD-L1 from enterocytes, due to the lack of induction from the noninvasive E. coli and L. gasseri shown in Fig. 1. To characterize the relationship between numbers of invasive bacteria and PD-L1 induction, we first inoculated Caco-2 cells with various doses of Salmonella ranging from 106 to 108 CFU/well. At 24 h postexposure, we simultaneously harvested half of each well for RNA and half for bacterial invasion assays. Figure 4A displays a positive correlation between PD-L1 expression and intracellular CFU/ml. The exponential curve was the best-fitting and most logical trend line. Such induction dynamics could be due to Caco-2 cells reaching their limit of internalized bacteria but not yet reaching their PD-L1 induction threshold. One explanation for this positive correlation could be due to initial signaling from the various doses of Salmonella inocula. To rule this out, we chose to test Salmonella mutant strains with various abilities to invade. ΔhilE and ΔhilD mutants have, respectively, increased and decreased ability to invade cells compared to wild-type Salmonella (15, 28, 29). At 24 h after we administered the same numbers of bacteria, we compared the PD-L1 induction versus the intracellular CFU/ml of wild-type (WT), ΔhilE, or ΔhilD Salmonella strains (Fig. 4B). Again, the level of invasion positively correlated with PD-L1 induction, thus demonstrating that PD-L1 production induced by Salmonella is dependent upon the extent of cellular invasion.
Salmonella dose and invasion strength positively correlated with PD-L1 induction. (A) Various doses of Salmonella were administered to Caco-2 cells for 1 h before external bacteria were washed away, gentamicin was added, and Caco-2 cells were cultured for a further 24 h. (B) Equal amounts of wild type (WT) and Salmonella mutants that demonstrate decreased and increased cellular invasion (ΔhilD and ΔhilE mutants, respectively) were administered to Caco-2 cells for 1 h. After 24 h, Caco-2 cells were trypsinized, and each well was divided in half for either RNA isolation or a bacterial invasion assay. PD-L1 induction was plotted against intracellular CFU/ml.
In addition to the necessity of entering the host cell, we wanted to uncover whether a known effector protein(s) could be responsible for initiating signaling that resulted in PD-L1 induction. Srinivasan et al. found SPI-2 had some involvement in PD-L1 induction of expanding responder CD4 T cells in Salmonella-infected mice (24). Therefore, we compared an SPI-2 knockdown mutant (ΔssrAB; deletion of an essential SPI-2 transcriptional regulatory control element; 5 × 107 CFU/well) to various WT Salmonella inocula (106 to 108 CFU/well). The ΔssrAB mutant demonstrated less PD-L1 induction than the WT when tested at the same concentration of intracellular bacteria (Fig. 5A). Therefore, one or multiple genes from SPI-2 are involved in PD-L1 induction of IECs. To further identify a single bacterial component important in PD-L1 induction, seven SPI-2-encoded effector mutants were created and tested (ΔspvC, ΔsspH1, Δslrp, ΔsseJ, ΔsseL, ΔspvD, and ΔsrfA; see Table S2 in the supplemental material) based on their suggested effector functions (20, 21, 30). Each strain was inoculated at four doses (1 × 107, 3 × 107, 5 × 107, and 8 × 107 CFU/well) to generate respective intracellular bacteria versus PD-L1 expression curves (Fig. 5B). The two effector mutants with the least amount of PD-L1 induction were the ΔspvD and ΔsseL mutants, whereas the remaining mutants had no change or showed greater PD-L1 induction than the WT slope. None of the seven mutants, however, prevented PD-L1 induction with an efficacy equal to that of the ΔssrAB mutant, indicating that either a different effector or a combination of effectors is responsible for PD-L1 induction.
Salmonella pathogenicity island 2 is involved in Salmonella-induced PD-L1 induction. (A) Various doses of WT Salmonella or the ΔssrAB mutant were administered to Caco-2 cells. (B) Salmonella mutants were inoculated onto Caco-2 cells at various doses. At 24 h postinoculation, the Caco-2 cells were trypsinized, and half of the cells were used for an invasion assay, while the other half were isolated for RNA. PD-L1 induction was plotted against intracellular CFU/ml. Exponential curves were fitted to each strain.
Overall, we interpreted these results as demonstrating that invasion of cells is a necessary precursor to SPI-2 expression and effector secretion. Indeed, Salmonella relies on a combination of environmental signals to sense its environment and expresses SPI-2 in response to those cues that are present within the host cell (namely, altered nutrient availability, cation concentration, and acidity). In the case of the ΔhilD mutant, which fails to invade the epithelia, extracellular bacteria are not exposed to the signals that stimulate SPI-2 induction. For this reason, ΔhilD mutants elicit the same epithelial response as SPI-2 mutants. Conversely, the greater invasion observed in a ΔhilE mutant results in a greater number of intracellular bacteria expressing SPI-2. As a result, it stands to reason that the effect of SPI-1 perturbation on PD-L1 induction is not due to SPI-1-encoded effectors. Rather, SPI-1 expression and bacterial invasion of epithelia are necessary precursors to SPI-2 induction, with the necessity of individual SPI-2 effectors being well supported by Fig. 5 and the published data of others (24).
Salmonella and IFN-γ synergistically induce PD-L1.IFN-γ is a proinflammatory cytokine secreted from T cells during early stages of Salmonella infection (31, 32); however, in our isolated cultures of IECs, IFN-γ is not included, since there are no immune cells present to produce it. IFN-γ can induce PD-L1 expression (27) and may do so after it is produced during Salmonella infection. Therefore, it is physiologically relevant to investigate IFN-γ addition to our in vitro infection model. We compared PD-L1 expression of IECs after Salmonella infection alone or combined with IFN-γ treatment (Fig. 6A and B). In these experiments, PD-L1 expression was increased about 5- and 25-fold in Salmonella-treated HT-29 cells and Caco-2 cells, respectively. PD-L1 expression was increased about 1,500- and 110-fold in IFN-γ-treated HT-29 cells and Caco-2 cells, respectively. Astoundingly, PD-L1 expression was increased about 6,000- and 4,400-fold in dual Salmonella plus IFN-γ-treated HT-29 cells and Caco-2 cells, respectively. In both cell types, this dual treatment was identified as a synergistic induction of PD-L1 (see the synergy calculations in Fig. S1 in the supplemental material).
Salmonella and IFN-γ synergistically increase PD-L1 expression and increase IFN-γ receptor expression. HT-29 cells (A and C) and Caco-2 cells (B and D) were either not treated (NT) or treated with Salmonella (ST) for 1 h. The bacteria were then washed away, gentamicin was added, and the cells were incubated in media without IFN-γ or with 1 ng/ml IFN-γ (A and C) or 5 ng/ml IFN-γ (B and D) for 24 h. Synergy calculations were performed on PD-L1 induction (see Fig. S1 in the supplemental material) for panels A and B. One-way ANOVA with Holm-Sidak's multiple-comparison test was performed for panels C and D. The graphs reflect representative averages from three individual preparations.
To further investigate this phenomenon, we sought to determine what could be occurring in the mammalian cell postinfection to cause this synergy. Interestingly, we found that IFN-γ receptor expression was significantly increased with dual treatment of Salmonella infection and IFN-γ (Fig. 6C and D). This likely causes the cells to become even more responsive to IFN-γ and explains the synergistic upregulation of PD-L1.
Functional analysis of PD-L1 induction.We wanted to evaluate whether the induction of PD-L1 was sufficient to obtain a functional decrease in T cell activation. We first established that activated Jurkat T cells demonstrated a time-dependent increase of the PD-L1 receptor, PD-1, when stimulated with anti-CD3, anti-CD28 beads (see Fig. S2 in the supplemental material). Therefore, T cells were activated with anti-CD3, anti-CD28 beads while, in parallel, Caco-2 cells were activated with Salmonella plus IFN-γ as before. We chose the dual stimulation of Caco-2 cells for functional experiments due to the physiological relevance to an in vivo infection. In order to control for extraneous signaling and restrict the assessment to the functionality of surface PD-L1, we used transduced Caco-2 cells that express either scrambled short hairpin RNA (shRNA) (control) or PD-L1 shRNA (suppression of PD-L1 induction). At 24 h after activation, T cells were washed and resuspended in serum-free medium to minimize the cytokine presence for the downstream enzyme-linked immunosorbent assay (ELISA). Next, the activated T cells were cultured with the transduced, pretreated, and washed Caco-2 cells (Fig. 7A). After 24 h of coculture, T cells were collected for qPCR analysis (Fig. 7B and C), and IFN-γ was measured from the supernatant via ELISA (Fig. 7D). Proinflammatory cytokine transcripts interleukin-2 (IL-2) and IFN-γ were both increased about 2-fold from T cells cocultured with PD-L1 shRNA Caco-2 cells over controls. Thus, suppression of PD-L1 from the shRNA in Caco-2 cells caused increased cytokine production from T cells. In summary, Salmonella infection plus IFN-γ decreased T cell activation through the induction of PD-L1.
Salmonella plus IFN-γ-induced PD-L1 expression decreased activated T cell cytokine production. (A) Experimental setup. Caco-2 cells that express scrambled shRNA or PD-L1 shRNA were infected and treated with IFN-γ as described in Fig. 6. After 24 h, the Caco-2 cells were washed twice in PBS, and preactivated and washed Jurkat T cells were added to the Caco-2 plate in serum-free medium (SFM). After 24 h of coculture, the medium was collected and centrifuged. (B and C) RNA was extracted from the cell pellet for qPCR assays, and cytokines were graphed as the fold change over T cells cocultured with nontreated scrambled shRNA Caco-2 cells. (D) Soluble IFN-γ was measured from supernatants via ELISA. Data were analyzed with a one-tailed Mann-Whitney test, and the P values are indicated on each graph. Each histogram depicts representative averages from at least three individual preparations.
DISCUSSION
We have shown in an isolated in vitro setting that PD-L1 is induced by Salmonella and is dependent on Salmonella invasion and effector gene(s) from SPI-2 (Fig. 8). Figures 4 and 5 illustrate PD-L1 induction, not as a function of Salmonella inocula, but rather as a function of viable intracellular bacteria recovered at the experimental endpoint. Thus, changes in bacterial viability as a consequence of the SPI-2 mutations are controlled for in these assays. As a result, PD-L1 induction is a function of the specific mutation, rather than a result of altered bacterial viability. Of the SPI-2 genes tested, the most influential candidates were sseL and spvD, albeit their induction of PD-L1 was still higher than the complete knockdown of SPI-2 achieved by the ΔssrAB mutant. Nevertheless, these effectors would be worth pursuing as agents of PD-L1 induction. Although ssrAB is considered to be the master regulator of SPI-2 (21), it is possible that other genes under the control of ssrAB outside of this pathogenicity island could be involved. In addition, we have shown for the first time that IFN-γ increased PD-L1 expression synergistically with Salmonella invasion. Importantly, the increased protein expression of PD-L1 was found to be functional, in that it decreased activated T cell IL-2 and IFN-γ production. It would be interesting to pursue other signaling changes that occur within activated T cells exposed to Salmonella-infected IECs.
Salmonella induces functional PD-L1. A schematic of Salmonella shows the invasion of intestinal epithelial cells (IECs), and Salmonella residing in a Salmonella-containing vesicle (SCV). Invasion and Salmonella pathogenicity island 2 (SPI-2) were found to influence IECs to induce PD-L1 expression through an unidentified host cell pathway. The addition of IFN-γ synergistically enhanced the Salmonella-driven PD-L1 induction. The resulting increase in PD-L1 surface protein was shown to decrease, but not inhibit, the cytokine production of activated T cells.
The in vivo infection system is highly intricate, and the multifactorial signaling cascades could all affect PD-L1 induction. First, there are many subsets and differentiation states of cells within the intestinal epithelium that could regulate PD-L1 in different manners. Also, differences in local IFN-γ concentrations would affect PD-L1 expression via synergy. Alternatively, other unknown variables could affect PD-L1 expression. These factors highlight the advantage of using the in vitro infection model as was done for this work. Further in vivo experimentation with multiple time points, mutant bacteria, and an IFN-γ signal blockade is warranted to answer these questions and show the causation of PD-L1 induction.
The functional effects of Salmonella-induced PD-L1 expression on the course of infection and disease symptoms are of utmost interest. Salmonella infection still causes robust gastroenteritis, and the effect of PD-L1 upregulation is to slightly dampen (but not fully inhibit) the overwhelming inflammation and cytokine production seen during infection. The transient PD-L1 upregulation is certainly not able to halt gastroenteritis but rather is most likely one of many avenues Salmonella uses in an attempt to stall the immune system. These questions could be answered with a PD-L1 blockade during infection. It is possible that disease symptoms and severity would increase without the control of PD-L1, causing increased lethality of the host. Alternatively, a lack of PD-L1 signaling could allow T and other immune cells to clear the infection faster. There is also a less likely possibility that PD-L1 plays a redundant role during Salmonella infection, and with the numerous pro- and anti-inflammatory signaling molecules seen during infection, the quantity of PD-L1 may not robustly affect the outcome of disease progression.
Our findings of Salmonella-induced PD-L1 induction are corroborated by previous reports of in vivo Salmonella infection or Salmonella-based therapies (22–24, 26). However, it was not determined whether the cells in each of these studies were directly infected by Salmonella or if the PD-L1 induction was an indirect effect through cytokines. The authors of these in vivo studies fail to recognize IFN-γ as a source of PD-L1 induction. It is likely that the increased PD-L1 expression of these studies was due to (i) local IFN-γ or other cytokine production after Salmonella infection (31, 32) and (ii) the synergy of Salmonella internalization plus IFN-γ, as we show here (Fig. 6). The occurrence of immune suppression through Salmonella-mediated PD-L1 induction on several cell types is potentially powerful and could be molded into a useful therapy in downregulating undesired immune activation.
PD-L1 regulation and inhibition are being actively investigated and used in clinical stages of inflammatory diseases, such as cancer (33), colitis (34), and bacterial (23) and viral (35) infections. Therefore, novel means of PD-L1 manipulation and a better understanding of its regulation would be advantageous for novel therapeutic strategies (36). In a nonobese diabetic mouse model, systemic Salmonella administration prevented the onset of diabetes, and the authors showed that this prevention was caused by an induction of PD-L1 on dendritic cells (26). Salmonella has also been proposed as a cancer therapy for the targeted killing of tumor cells due to general activation of the immune system, direct killing, competition of nutrients, and tumor-specific immune activation when tumor antigens are delivered with Salmonella (37, 38). Interestingly, it was found that administration of an anti-PD-1 antibody combined with Salmonella delivering tumor-peptide was more effective at tumor eradication than the Salmonella or anti-PD-1 therapy alone (39). Most likely, the addition of the anti-PD-1 eliminated the Salmonella-induced PD-L1-mediated immune evasion of the surviving tumor cells. To further the therapeutic potential of Salmonella in cancer, we show that using a SPI-2 knockdown strain of Salmonella could decrease the PD-L1 induction normally seen after internalization, thereby potentially increasing the benefits of Salmonella anticancer therapy.
MATERIALS AND METHODS
Cell cultures and reagents.Caco-2, HT-29, IEC-6, and Jurkat T cells (clone E6-1) (American Type Culture Collection [ATCC], Manassas, VA) were maintained according to recommended conditions and media unless otherwise specified. Media and fetal bovine serum (FBS) were purchased from Life Technologies (Carlsbad, CA).
Bacteria.The bacteria used in this study are described in Table S1 in the supplemental material. Escherichia coli Nissle 1917 was isolated and cultured from a commercial probiotic as previously described (40). Lactobacillus gasseri 33323 and Salmonella enterica serovar Typhimurium 14028s were purchased from the ATCC. For Salmonella mutants made in this study, the red recombinase method was utilized to construct mutants of Salmonella, as described previously (41) using the primers listed in Table S2 in the supplemental material. The E. coli and Salmonella strains were maintained in Luria-Bertani (LB) media. L. gasseri was cultured and maintained in MRS broth (BD Biosciences, Franklin Lakes, NJ).
For fluorescent analysis, Salmonella was engineered to express constitutively the Tag-Blue fluorescent protein (BFP; Evrogen). The open reading frame (ORF) of BFP was codon optimized by GenScript and inserted in the EcoRV site of the pUC57Kan vector, preceded by a constitutive promoter, and followed by a BglII restriction site (the sequence is provided in Table S2 in the supplemental material). Using this BglII site and the EcoRI site already present on pUC57Kan, the chloramphenicol acetyltransferase (CAT) gene, with flanking flip recombinase sites, was cloned from plasmid pKD3 directly downstream of the BFP gene. The constitutive promoter, BFP ORF, and the downstream CAT gene were then moved together into Salmonella Typhimurium, replacing the phoN gene using the red recombinase method (41). The final construction was verified by PCR before fluorescent bacteria were used in flow cytometry experiments.
Bacterium-IEC cocultures.In preparation for an experiment, 20,000 IECs (single or 3:1 Caco-2:HT-29 for mixed cultures) were seeded in a 24-well plate, fed every 2 days while gradually changing cells to serum-free medium (Dulbecco modified Eagle medium, 1× insulin/transferrin/selenium, 1× nonessential amino acids, and 1× antibiotic-antimycotic [Thermo Fisher Scientific]), and maintained for 1 week after reaching confluence (usually 14 days total). At this time, IECs were cultured in antibiotic-free, serum-free media overnight. In addition, overnight cultures of bacteria were grown in 5 ml of the indicated broth. The next morning, the bacteria were diluted 1:5 in fresh broth and grown for another 1 to 2 h, the absorbance was measured, and the cultures were standardized to an optical density at 600 nm of 1.0. Next, 1 ml of culture was centrifuged at 10,000 × g for 1 min and resuspended in 1 ml of IEC serum-free media. Unless otherwise specified, the bacteria were then added to cells at a 10% volume (50 μl/well in a 24-well plate). For Salmonella, this equates to a multiplicity of infection of about 50. After bacterial addition, IEC plates were centrifuged at 800 × g for 10 min at room temperature, followed by incubation in a humidified incubator (37°C and 5% CO2) for 1 h. The IECs were then washed twice with 1× phosphate-buffered saline (PBS) with gentamicin, followed by incubation with serum-free media (plus 150 μg/ml gentamicin) for 1 h. Finally, the medium was replaced with serum-free media (with 15 μg/ml gentamicin) for 24 h unless otherwise specified. Where applicable, 1 to 5 ng/ml human IFN-γ (ProSci, Poway, CA) was added to Caco-2 cells as indicated in the figure legends.
Salmonella infection of mice and analysis.All work was carried out in accordance with the International Guiding Principles for Biomedical Research Involving Animals. Five 7-week-old female C57BL/6 mice were administered 20 mg of streptomycin sulfate orally. After 24 h, they received 109 CFU orally of a spontaneous streptomycin-resistant isolate of Salmonella Typhimurium 14028s. At 48 h after inoculation, the mice were euthanized, and the ceca were preserved in 10% formalin until processing. Microscope slides were stained for Salmonella (green; Thermo Fisher Scientific, Salmonella common core antibody, MA518259) PD-L1 expression (red; Cell Signaling, rabbit anti-mouse PD-L1 D5V3B), and DNA (blue; DAPI [4′,6′-diamidino-2-phenylindole]). Confocal microscopy was performed on at least 10 slides from the five animals. Epithelial tissue was qualitatively compared between regions of Salmonella presence or absence within the same mouse.
Bacterial invasion assay.At the time of harvest, Caco-2 cells were washed in PBS, trypsinized, and split in half for RNA isolation and bacterial invasion assessment. Caco-2 cells were gently lysed with 0.1% Triton X-100 (Sigma) at 37°C for 15 min. Next, the lysates were serially diluted, plated onto LB agar plates, and grown overnight at 37°C, and the colonies were counted.
qPCR.At the time of harvest, IECs were either frozen in RNAlater (Thermo Fisher Scientific) or immediately processed for RNA isolation using RNeasy kits and on-column DNase (Qiagen) according to the manufacturer's protocol. RNA was quantified with a NanoDrop 1000 (Thermo Scientific), and 100 ng of each sample was used to synthesize cDNA using iScript cDNA synthesis kit (Bio-Rad). Finally, cDNA was combined with 70 nM primers (see Table S3 in the supplemental material) and SsoAdvanced SYBR green Supermix (Bio-Rad), and samples were run in technical triplicates and analyzed against a standard curve on every plate using CFX Connect Real-Time PCR detection system (Bio-Rad). Samples were normalized to GAPDH (glyceraldehyde-3-phosphate dehydrogenase) reactions run on the same plate. Where indicated, samples were further normalized to nontreated groups to show the fold difference in expression.
Flow cytometry.BFP Salmonella strains were used in all flow cytometry experiments to minimize cell processing and nonspecific staining for internalized bacteria. At the time of harvest, Caco-2 cells were trypsinized, washed in 1% bovine serum albumin (BSA) in PBS, blocked in 20% normal mouse serum for 20 min, stained with anti-human PD-L1–phycoerythrin (BD Biosciences) for 40 min, washed twice, and resuspended in 1% BSA in PBS for immediate analysis on a FACSAria (BD Biosciences). Data were further analyzed with FCS Express 5 (De Novo Software, Glendale, CA).
Lentiviral transduction.Lentiviral plasmid constructs containing scrambled or shRNA against PD-L1 (CCATCAAGTCCTGAGTGGTAAGACCACCA) were purchased from Origene (Rockville, MD) and transfected into HEK293T cells (ATCC). In accordance with Origene instructions, lentivirus was extracted from HEK293T cell culture supernatant and distributed onto Caco-2 cells with 6 μg/ml Polybrene (Santa Cruz Biotechnology, Santa Cruz, CA). Positively transduced Caco-2 cells were selected and maintained with 5 μg/ml puromycin dihydrochloride (Santa Cruz Biotechnology).
T cell activation assay.A total of 3 × 106 Jurkat T cells were activated in 5 ml of RPMI–5% FBS media with 150 μl of washed anti-CD3, anti-CD28 Dynabeads (Thermo Fisher) as recommended by the manufacturer's protocol. In a separate 48-well plate, lentiviral transduced Caco-2 cells were activated with Salmonella plus 5 ng/ml IFN-γ as described above. After 24 h, the Caco-2 cells were washed twice in PBS. Activated T cells were washed in PBS, centrifuged, and resuspended in 7 ml of serum-free medium. Next, 0.25 ml of activated T cells (∼1.5 × 105 cells) was distributed to each well of Caco-2 cells (half of plate). After 24 h, the culture medium was collected and centrifuged. The supernatant was applied in technical duplicates onto a human IFN-γ ELISA plate (Thermo Scientific), and the ELISA was performed as directed by the manufacturer. The T cell pellets were harvested for RNA and subsequent qPCR assays.
Statistical analysis.Data were analyzed and graphed with either Microsoft Excel or GraphPad Prism (La Jolla, CA), and the statistical analysis details are indicated in the figure legends (*, P < 0.05 if not otherwise indicated). All graphs show the data from one representative experiment of at least three experiments performed. Within each histogram, biological triplicates are averaged, and error bars indicate standard errors of the mean. Synergy calculations were performed in Microsoft Excel testing for Bliss independence (42, 43).
ACKNOWLEDGMENTS
This study was supported by the Defense Threat Reduction Agency HDTRA1-13-1-0037 and NIH DP2:1DP20D007155-01. This material is based on work that is supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture, under award 2016-67012-25184. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture.
FOOTNOTES
- Received 18 September 2017.
- Returned for modification 14 October 2017.
- Accepted 4 February 2018.
- Accepted manuscript posted online 12 February 2018.
Supplemental material for this article may be found at https://doi.org/10.1128/IAI.00674-17.
- Copyright © 2018 American Society for Microbiology.
