IL22BP Mediates the Antitumor Effects of Lymphotoxin Against Colorectal Tumors in Mice and Humans

Abstract

Background & Aims

Unregulated activity of interleukin (IL) 22 promotes intestinal tumorigenesis in mice. IL22 binds the antagonist IL22 subunit alpha 2 (IL22RA2, also called IL22BP). We studied whether alterations in IL22BP contribute to colorectal carcinogenesis in humans and mice.

Methods

We obtained tumor and nontumor tissues from patients with colorectal cancer (CRC) and measured levels of cytokines by quantitative polymerase chain reaction, flow cytometry, and immunohistochemistry. We measured levels of Il22bp messenger RNA in colon tissues from wild-type, Tnf^–/–, Lta^–/–, and Ltb^–/– mice. Mice were given azoxymethane and dextran sodium sulfate to induce colitis and associated cancer or intracecal injections of MC38 tumor cells. Some mice were given inhibitors of lymphotoxin beta receptor (LTBR). Intestine tissues were analyzed by single-cell sequencing to identify cell sources of lymphotoxin. We performed immunohistochemistry analysis of colon tissue microarrays from patients with CRC (1475 tissue cores, contained tumor and nontumor tissues) and correlated levels of IL22BP with patient survival times.

Results

Levels of IL22BP were decreased in human colorectal tumors, compared with nontumor tissues, and correlated with levels of lymphotoxin. LTBR signaling was required for expression of IL22BP in colon tissues of mice. Wild-type mice given LTBR inhibitors had an increased tumor burden in both models, but LTBR inhibitors did not increase tumor growth in Il22bp^–/– mice. Lymphotoxin directly induced expression of IL22BP in cultured human monocyte–derived dendritic cells via activation of nuclear factor κB. Reduced levels of IL22BP in colorectal tumor tissues were associated with shorter survival times of patients with CRC.

Conclusions

Lymphotoxin signaling regulates expression of IL22BP in colon; levels of IL22BP are reduced in human colorectal tumors, associated with shorter survival times. LTBR signaling regulates expression of IL22BP in colon tumors in mice and cultured human dendritic cells. Patients with colorectal tumors that express low levels of IL22BP might benefit from treatment with an IL22 antagonist.

What You Need to Know.

Background and Context

Unregulated activity of interleukin 22 (IL22) promotes intestinal tumorigenesis in mice. IL22 binds the antagonist IL22 subunit alpha 2 (IL22RA2, also called IL22BP).

New Findings

Lymphotoxin signaling regulates expression of IL22BP in colon; levels of IL22BP are reduced in human colorectal tumors and associated with shorter survival times. LTBR signaling regulates expression of IL22BP in colon tumors in mice and in cultured human dendritic cells.

Limitations

This study was performed using human tissue samples and mice; further studies are needed in humans.

Impact

Patients with colorectal tumors that express low levels of IL22BP might benefit from treatment with an IL22 antagonist.

Interleukin (IL) 22 is produced by immune cells at mucosal surfaces.^{1, 2, 3, 4} In the intestine, IL22 promotes intestinal stem cell–mediated epithelial regeneration and thus intestinal integrity, and it can also protect intestinal stem cells against genotoxic stress.⁵ However, uncontrolled and prolonged IL22 activity promotes intestinal tumorigenesis, as shown in mouse models by others and by us.^{6, 7, 8, 9, 10, 11, 12} These data are further supported by human studies, which showed that IL22 is highly expressed in human colon cancer and correlates with a chemoresistant state.^13,14

On the basis of these data, IL22 has been proposed as a potential therapeutic target in colorectal cancer (CRC). However, using a systemic blockade of IL22 as a treatment strategy might have serious adverse effects because IL22 also has beneficial effects, as described above. Thus, it will be important to identify a biomarker that would allow the selection of patients who might benefit from an IL22 blockade because the activity is uncontrolled and prolonged. Interestingly, there is a soluble IL22 receptor, IL22 binding protein (IL22BP, IL22RA2), which is produced by dendritic cells (DCs), T cells, and eosinophils in humans.^7,15,16 It binds to IL22 and limits its bioavailability by prohibiting the binding of IL22 to the membrane-bound IL22R1.^17,18 Accordingly, murine studies showed that IL22BP plays a protective role in CRC, whereas IL22 can promote carcinogenesis.^7,8 However, the source and the significance of IL22BP in human CRC is unclear. Furthermore, the pathway inducing IL22BP is currently unknown. Finally, whether IL22BP might serve as a biomarker to predict the outcome in CRC and to select patients who might benefit from a blockade of IL22 is unclear.

Here, we found that IL22BP is down-regulated in human CRC, thereby indicating an impaired control of IL22 in this malignancy. In line with these data, we showed that low IL22BP expression is an independent risk factor for lower survival in patients with CRC. Furthermore, we found that IL22BP production in DCs is regulated by lymphotoxin (LT) LTα1β2 via the noncanonical nuclear factor (NF) κB pathway. Finally, using CRC mouse models, we proved that IL22BP bridges the antitumorigenic effect of LT.

Methods

Isolation of Hematopoietic Cells From Murine and Human Intestine

Hematopoietic cells were isolated from macroscopically healthy human colonic mucosa, CRC tissue, or murine colon. Human tissues were obtained freshly after surgical removal of tumors from patients diagnosed with CRC. After removal of the Peyer’s patches and the adventitial fat, the murine colon was cut longitudinally. Prepared samples were washed with phosphate-buffered saline (PBS). For isolation of intraepithelial lymphocytes (IELs), the intestinal tissue was incubated in Hank’s balanced salt solution containing 1 mmol/L dithioerythritol, followed by a dissociation step using 1.3 mmol/L EDTA for 20 minutes at 37°C, respectively. To isolate lamina propria lymphocytes, the tissue was further cut in small pieces and minced with a scalpel. The remaining tissue was incubated for 45 minutes at 37°C on a shaking incubator in Hank’s balanced salt solution (with Ca²⁺ and Mg²⁺) with collagenase (1 mg/mL) and DNase I (10 U/mL), and supernatant was collected. Leukocytes were further enriched by Percoll gradient centrifugation (GE Healthcare, Princeton, NJ). If not stated otherwise, IEL and lamina propria lymphocytes were collected and pooled.

Tumor Induction

Mice were injected intraperitoneally with azoxymethane (AOM) (Sigma-Aldrich, St Louis, MO) at a dose of 7.5 mg/kg body weight. After 5 days, mice received 2.5% dextran sodium sulfate (DSS) (MP Biomedicals, Irvine, CA; molecular weight, 36,000–50,000 Da) in drinking water for 5 days, followed by 16 days of regular water. This cycle was repeated twice.¹⁹ Mice were given a blocking Fc-fusion protein against LT beta receptor (LTBR) (aLTBR) or PBS on days 5, 19, 33, 47, 61, and 75 of the experiment. The mice were euthanized on day 80 of the experiment.

Tissue Microarray

A preexisting colon cancer prognosis tissue microarray (TMA) was used in this study.^20,21 The TMA consisted of 1475 tissue cores distributed across 3 different paraffin blocks containing 522, 458, and 495 tissue cores. These numbers include a standard control area with 62 non–colon cancer tissue spots (normal tissues of other organs) that are used for immunohistochemistry staining quality controls. The patients underwent CRC surgery between 1988 and 1996 at the Department of Surgery of the University Hospital of Basel. A single 0.6-mm tissue core per tumor was arrayed. All colon cancer tissue spots represent primary tumors from 1413 patients with a median age of 71 years (range, 30–96 years). After cutting the paraffin blocks and immunohistochemical staining of IL22BP, 1110 of the 1413 patients could undergo further analysis—the remaining tissue cores were either lost or their quality did not allow proper analysis. Raw survival data were either obtained from the cancer registry or collected from the patients’ attending physicians. The mean follow-up time was 46 months (range, 1–152 months). Formalin-fixed (buffered neutral aqueous 4% solution), paraffin-embedded tumor material was used. The pathologic stage, tumor diameter, and nodal status were obtained from the primary pathology reports. All slides from all tumors were reviewed by 2 pathologists to define the histologic grade and the histologic tumor type. Details on the composition are given in Table 1. IL22BP-specific antibody from R&D Systems (MAB 10871; Minneapolis, MN) was used at a 1:350 concentration. To visualize IL22BP, the EnVision Kit (Dako Agilent Technologies, Santa Clara, CA) was used. The sections were counterstained with hematoxylin. IL22BP presence was then determined in a blinded fashion, and the results were sent to a separate scientist who performed the data analysis. All human studies were approved by the local ethical committee (Ethik-Kommission der Ärztekammer Hamburg).

Table 1.

Patient Characteristics in the Validation Data Set

	Patients, n	IL22BP–, %	IL22BP+, %
Total	1110	84.30	15.70
pT stage
pT1	41	87.8	12.2
pT2	154	77.3	22.7
pT3	719	84.6	15.4
pT4	180	88.3	11.7
pN stage
pN0	550	83.3	16.7
pN1	294	85.0	15.0
pN2	232	86.2	13.8
Grading
G1	21	95.2	4.8
G2	928	84.1	16.0
G3	145	82.8	17.2
Location
Anal	8	75.0	25.0
Appendix	1	100.0	0.0
Ascendens	120	83.3	16.7
Caecum	171	84.2	15.8
Descendens	50	92.0	8.0
Rectosigmoid	73	84.9	15.1
Rectum	377	81.4	18.6
Sigmoid	207	85.0	15.0
Transversum	96	88.5	11.5
Histological type
Adeno	993	84.2	15.8
Mucinous	1	0.0	100.0
Medullary	100	85.0	15.0
Signet cell	7	85.7	14.3
Others	5	60.0	40.0
Vascular invasion
Absent	590	81.0	19.0
Present	504	87.9	12.1
Age, y, mean ± SD	70.2 ± 0.35	69.1 ± 0.9	70.4 ± 0.4

SD, standard deviation.

Human Cohort Used for Cytokine Expression Analysis

Human tissues were obtained freshly after surgical removal of tumors from patients diagnosed with CRC. Both the healthy colonic mucosa and parts of the tumor were collected and frozen until the RNA isolation procedure. In this study, we analyzed the expression levels of cytokines in 99 patients with CRC who had surgery between 2010 and 2015. Of those patients, 92 were used for the survival analysis in the discovery data set. No survival data were available for the remaining 7 patients. Ages and tumor stages are summarized in Supplementary Table 2. All human studies were approved by the local ethical committee (Ethik-Kommission der Ärztekammer Hamburg).

Statistical Analysis

Statistical analysis was performed with GraphPad Prism Software (GraphPad Software, San Diego, CA). For comparison of groups, the nonparametric 2-sided Mann–Whitney test was used. Bonferroni correction was used to counteract the problem in the case of multiple comparisons. For time-dependent weight loss data, a repeated-measures analysis of variance (ANOVA) to assess the significance of the main effects and an experimental group × time interaction was used. The significance level alpha was set to .05. The Pearson correlation was used for correlative analyses. The significance level was set to .05. Additionally, Anne Daubmann, a trained statistician, who is one of the coauthors of this publication, conducted multivariate survival analyses using multiple Cox models with SAS, version 9.4 (SAS Institute, Cary, NC). The independent variables were the proteins (IL22BP, LT-α [LTA], LT-β [LTB]), age, sex, TNM stage, and grading. All other variables were kept constant for the purpose of the multivariate analysis. The messenger RNA (mRNA) expression of the cytokines was transformed by using a base 2 logarithm. For an easier interpretation, the estimators of the cytokines were back transformed. A check of proportional hazards assumption using Schoenfeld residuals showed a violation for age at operation. Therefore, we used a multiple extended Cox model with a time-dependent effect for age.

Results

Interleukin 22 Binding Protein Is Down-regulated in Colorectal Cancer and Correlates With Lymphotoxin Expression

We initially analyzed the expression of IL22, IL22R1, and IL22BP in healthy mucosa and tumor tissue from patients with CRC. In line with previous reports,^9,22 we found increased levels of IL22 and the related cytokine IL17A in tumors compared to healthy tissue control samples (Figure 1A and Supplementary Figures 1 and 2). Interestingly, although the levels of IL22R1 were not different between the groups, we observed significantly decreased IL22BP levels in CRC, suggesting a dysregulation of the IL22-IL22BP axis in intestinal malignancies, which may result in high IL22 activity (Figure 1A and Supplementary Figure 2). We confirmed these findings at the protein level by using flow cytometry (Figure 1B–E and Supplementary Figure 3). Furthermore, we showed that DCs and eosinophils are the major source of IL22BP, but we found only minor expression in T cells in the human intestine in CRC (Figure 1D–F). Of note, IL22BP expression was reduced in DCs, eosinophils, and CD4⁺ T cells in tumors compared to the healthy adjacent mucosa of patients with CRC (Figure 1D–F). However, the total cell number of DCs, eosinophils, and CD4⁺ T cells in tumors compared to the healthy adjacent mucosa of patients with CRC was not reduced (Supplementary Figure 4). These data indicate that the reduced IL22BP expression in CRC is due to an impaired production from these cells.

Figure 1.

IL22BP is down-regulated in CRC and correlates with LT. (A) Relative mRNA expression of IL22, IL22RA1, and IL22BP in tumors and macroscopically healthy mucosa surgically removed from patients with CRC (n = 99). (B, C) Representative FACS plots and corresponding statistics showing the production of IL17A, IL22, interferon gamma, and TNF-α by CD4⁺ T cells isolated from healthy colon and tumor tissue (n = 12). (D, E) Representative FACS plots (from the same patients) and corresponding statistics showing the production of IL22BP by CD45^– cells, CD4⁺ T cells (CD4⁺ TC), DCs and eosinophils (Eos) isolated from healthy colon and tumor tissue (n = 8 patients). (F) Relative IL22BP mRNA expression by CD4⁺ TC, DCs, and eosinophils isolated and FACS sorted from healthy colon and tumor tissue (n = 5). (G) Correlation matrix of several genes in normal tissue and tumor from patients with CRC (statistically significant results are marked with an asterisk). (H) Correlation of LTA and LTB expression with IL22BP in patients with CRC. (I) Relative mRNA expression of LTA and LTB in tumors and macroscopically healthy mucosa surgically removed from CRC patients (n = 99). Data are presented as mean ± standard error of the mean. ∗P < .05, ∗∗P < .01, and ∗∗∗P < .001, as assessed by Wilcoxon matched pairs test (A, C, and E) and Mann-Whitney U test (F). The Pearson correlation was used for correlative analysis. If not stated otherwise, P > .05 is considered nonsignificant. FACS, fluorescence-activated cell sorting.

We next hypothesized that this imbalance in the IL22-IL22BP axis is caused by an altered cytokine environment within the tumor. We therefore correlated the expression of several cytokines with IL22BP. We selected cytokines that have been shown to be important for the regulation of IL22 or IL22BP. We previously published that tumor necrosis factor alpha (TNF-α) regulates IL22BP in T cells in inflammatory bowel disease (IBD).¹⁶ In addition, we measured the TNF superfamily members LTA and LTB. IL18 was also included because it is known to negatively regulate IL22BP expression in DCs in mice.⁷ Furthermore, IL17A was measured because it is commonly coexpressed with IL22 and has been shown to influence the progression of intestinal malignancies.²² Additionally, IL23 and transforming growth factor beta were chosen as known regulators of IL22.²³ IL22RA1 was examined because it is the receptor of IL22. In line with our previous findings in IBD,¹⁶ we found a positive correlation between the expression of TNF-α and IL22BP. However, this correlation was even stronger for other TNF superfamily members, namely, LTA and LTB (Figure 1G and H and Supplementary Figure 5).

In conclusion, we showed that IL22BP expression is down-regulated in human CRC. Furthermore, IL22BP correlated with the expression of TNF superfamily members, namely TNF-α, LTA, and LTB.

Interleukin 22 Binding Protein Expression Is Regulated by Lymphotoxin

We next aimed to study whether TNF-α, LTA, and LTB might regulate IL22BP expression. To this end, we first measured Il22bp mRNA levels under steady state conditions in the colon of wild type (WT), Tnf^–/–, Lta^–/–, and Ltb^–/– mice. Interestingly, we found a strong down-regulation of Il22bp in the colon of Lta^–/– and Ltb^–/– mice but not of Tnf^–/– mice (Figure 2A). Of note, we reported previously that TNF-α drives CD4⁺ T-cell–derived IL22BP during IBD most likely in an indirect manner.16 However, DCs represent the main source of IL22BP in the colon under steady state conditions, whereas CD4⁺ T cells show only very low IL22BP expression.^15,16 Thus, the total amount of Il22bp was not altered in the colon of Tnf^–/– mice. Taken together, these data suggest a link between LT and IL22BP.

Figure 2.

IL22BP expression is regulated by LT. (A) Relative mRNA expression of Il22bp in the colon of WT, Tnf^–/–, Lta^–/–, and Ltb^–/–mice. (B) Relative mRNA expression of Il22bp in the colon of mice injected intraperitoneally with a Fc-Fusion protein blocking LTBR (aLTBR) or agonistic (ACH6) LTBR antibody. (C) Heatmap showing the fold-change expression values after blocking LTBR (aLTBR) or administration of agonistic (ACH6) LTBR antibodies compared to control mice. LT controls colitis development by regulating IL22BP. (D) Representative colonoscopy pictures, (E) endoscopic colitis score and weight loss of WT and Il22bp^–/– mice given PBS or agonistic antibodies against LTBR (ACH6). (F) The expression of Il22, Il22bp, and Ltb during the course of DSS colitis (n = 4 for each group and timepoint). Data are presented as mean ± standard error of the mean. ∗P < .05, ∗∗P < .01, and ∗∗∗P < .001, as assessed by 1-way ANOVA with Bonferroni post hoc tests (A, B) or repeated-measures ANOVA (E, F). If not stated otherwise, P > .05 is considered nonsignificant (ns).

Because Lta^–/– and Ltb^–/– mice have an underdeveloped immune system,²⁴ we aimed to verify that IL22BP down-regulation is not merely the result of a broadly impaired mucosal immunity. We therefore administered either agonistic antibodies (ACH6) or an antagonistic Fc-Fusion protein (aLTBR) against the LTBR to WT mice. We found that LTBR blocking significantly lowered Il22bp expression in the colon, mesenteric lymph nodes, and spleen. However, the agonistic antibody had no significant effect (Figure 2B and C and Supplementary Figure 6). Importantly, the levels of other cytokines that have been previously shown to regulate IL22BP, such as TNF-α and IL18,^7,16 were not affected by the administration of aLTBR, therefore arguing against a secondary effect mediated by those molecules (Figure 2C and Supplementary Figure 6).

Because the agonistic LTBR antibody in steady state did not affect IL22BP levels, we hypothesized that this effect might be related to the already high amount of IL22BP and LT during homeostasis. To test this hypothesis, we analyzed the effect of the agonistic antibody in a murine model of acute colitis, when IL22BP is down-regulated. Mice were given DSS in drinking water for 5 days, leading to acute colonic tissue damage and inflammation, which we monitored using mouse colonoscopy and body weight. In this model, L22BP expression in the colon decreases during the inflammatory phase.⁷ First, we confirmed an IL22BP down-regulation and an IL22 up-regulation at the peak of disease (Figure 2F). Second, we observed that the expression curve of Ltb mimicked that of Il22bp, suggesting that LT might regulate IL22BP in DSS colitis (Figure 2F). Finally, we proved that agonizing LTBR resulted in a partial rescue of IL22BP expression. In line with that, we also saw significantly augmented disease in WT mice receiving the agonistic antibody (Figure 2D–F). Notably, this difference in disease outcome was IL22BP dependent because no effect was observed in Il22bp^–/– mice (Figure 2D and E). These findings suggest that LT down-regulation plays a critical role in tissue regeneration. Its down-regulation results in lower IL22BP levels, which facilitate the IL22-driven intestinal regenerative programs. Next, we assessed the effect of LTBR blockade in DSS colitis. However, blockade of the LTBR did not affect colitis severity (Supplementary Figure 7). These data are in line with our previous results showing that IL22BP is down-regulated in DSS colitis. The strong down-regulation of IL22BP in this setting might explain why Il22bp^–/– mice do not have an altered disease severity.⁷

Next, we aimed to test whether LTBR signaling would also influence IL22 expression. Of note, previous reports have shown a regulation of IL22 via an LTBR-IL23 axis in Citrobacter rodentium infection in mice.^25,26 For this reason, we also used the C rodentium infection model to study the role of LT signaling on the IL22-IL22BP axis. We orally gavaged the mice with C rodentium and performed intraperitoneal injections of either ACH6 antibody or PBS during the infection. We killed the mice after 8 days and could then confirm an up-regulation of Il23 and Il22 in the colon of mice that were given agonistic LTBR antibodies (Supplementary Figure 8). Of note, Il22bp expression was also induced by the LTBR agonistic antibody, in a similar way to the effects observed in the DSS colitis experiment (Figure 2). However, we saw no effect of LTBR agonizing or blockade on IL22 levels in steady state conditions and during DSS colitis (Figure 2F and Supplementary Figure 7).

In conclusion, we observed that LT regulates IL22BP expression in the colon both in the steady state and in an inflammatory setting.

Interleukin 22 Binding Protein Mediates the Antitumorigenic Effects of Lymphotoxin in Mice

Next, we aimed to determine whether the LT-IL22BP axis plays a role in intestinal tumorigenesis. To this end, we blocked LTBR in WT and Il22bp^–/– mice during the course of a colitis-associated colon cancer model.²⁷ We used the well-established AOM/DSS model in which the injection of a mutagen (AOM) is followed by repeated cycles of intestinal inflammation induced by DSS. After 3 cycles of DSS administration, the tumor burden in mice was assessed by using mouse colonoscopy. In line with our previous findings, under the steady state conditions, we noticed that blocking LTBR resulted in lower Il22bp levels in tumor tissue (Figure 3C). Furthermore, antagonizing LTBR promoted tumor number and overall tumor load in WT mice while it had no effect in Il22bp^–/– mice, indicating that the antitumorigenic effect of LT in this model is IL22BP dependent (Figure 3A and B). Of note, these results cannot be attributed to different colitis severity, because LTBR blockade did not affect colitis severity (Supplementary Figure 7). Next, we confirmed the antitumorigenic effect of LT in a second colon cancer model. In this model, mice underwent laparotomy and were injected with luciferase-expressing MC38 colon cancer cells into the cecum of WT mice. After the injection, we tracked the tumor growth by measuring the bioluminescence of the cecal tumors. Of note, the effects of LT in this model were also dependent on the presence of IL22BP as shown in Il22bp-deficient mice (Figure 3D and E). Collectively, our results indicate that that the antitumorigenic effect of LT in colon cancer in mouse models is IL22BP dependent.

Figure 3.

IL22BP bridges the antitumorigenic effects of LT in mice. (A) Representative pictures of tumors (scale bar = 2 mm) and (B) the corresponding statistics showing tumor number and tumor score/load (based on endoscopic diagnosis) at indicated timepoints are shown. WT and Il22bp^–/– mice were given PBS or blocking Fc-Fusion protein against LTBR (aLTBR). (C) Relative mRNA expression of Il22 and Il22bp in tumors after LTBR blockade compared to PBS controls (n = 4). (D) Representative bioluminescence pictures and (E) corresponding statistics of tumors developed after the intracecal injection of luciferase-expressing MC38 cells. Data are presented as mean ± standard error of the mean. ∗P < .05, ∗∗P < .01, and ∗∗∗P <.001, as assessed by 1-way ANOVA with Bonferroni post hoc tests (B and E) and Mann-Whitney U test (C). If not stated otherwise, P > .05 is considered nonsignificant.

Because we identified LT to be the key regulator of IL22BP in CRC, we next aimed to determine the source of LT in this context. To this end, we performed single-cell RNA sequencing of immune cells infiltrating human CRC (Figure 4A–C). LTB was primarily expressed by CD4⁺ T cells, B cells, and natural killer T cells (Figure 4B and C and Supplementary Figure 9). LTA transcripts were detected in only a few cells. Therefore, we could not accurately assess the cellular source of LTA based on single-cell sequencing. To identify the functionally relevant source of LT for the regulation of IL22BP, we measured the expression of IL22BP in the colon of Rorc^CreLta^flox/flox mice (in which all cells that have or had expressed Rorc, including T cells, do not produce LTA) and Cd19^CreLta^flox/flox mice (in which B cells cannot produce LTA). Interestingly, IL22BP expression was reduced in Rorc^CreLta^flox/flox mice, whereas we found only an incremental and nonsignificant reduction in Cd19^CreLta^flox/flox mice (Figure 4D). These data indicate that T cell–derived LT, and not B cell-derived LT, plays a key role in regulating IL22BP expression in the colon. Rorc Cre deletes in all T cells,²⁸ and it is also expressed by subsets of innate lymphoid cells (ILCs) (ILC-3 and lymphoid tissue inducer cells). Therefore, these data do not preclude an additional effect of ILC-derived LT. Of note, we observed high expression of LT in human tumor-infiltrating T cells and lineage-negative cells (which include ILC-3 und lymphoid tissue inducer cells) by reverse-transcription polymerase chain reaction from sorted cells (Figure 4E)

Figure 4.

T cells regulate IL22BP via LT. (A) Cell clusters identified using single-cell sequencing in 2 patients with CRC. (B, C) LTB expression by each cell cluster. (D) Expression of Il22ra2 in the healthy colon of Rorc^CreLta^wt/wt, Rorc^CreLta^flox/flox, Cd19^CreLta^wt/wt, and Cd19^CreLta^flox/flox mice under steady state conditions. (E) Expression of LTA and LTB in cells sorted from the tumors of patients with CRC. The “Rest” population compromises CD45⁺ cells that were neither lineage-negative nor CD3⁺ or CD11c⁺. Data are presented as mean ± standard error of the mean. ∗P < .05, ∗∗P < .01, and ∗∗∗P < .001, as assessed by Mann-Whitney U test. If not stated otherwise, P > .05 is considered nonsignificant.

Lymphotoxin Controls the Expression of Interleukin 22 Binding Protein via the Noncanonical Nuclear Factor κB Pathway

We then aimed to identify whether the effect of LT on IL22BP expression was mediated directly or indirectly by LTBR. To this end, we used an assay based on human monocyte-derived DCs (MDDCs), which produce high levels of IL22BP.^7,29 Indeed, exposure of MDDCs to LTα1β2 or an agonistic antibody against the human LTBR resulted in higher IL22BP levels both on mRNA and on protein levels (Figure 5A–C), suggesting a direct effect of LTBR signaling on IL22BP production. Because LTBR signaling activates both the canonical and noncanonical NF-κB pathways,³⁰ we hypothesized that NF-κB signaling might be a mechanism by which LT induces IL22BP. We could confirm an activation of RelA (part of the canonical NF-κB pathway) and RelB (noncanonical NF-κB pathway) after exposure of MDDCs to LTα1β2 (Figure 5D–F and Supplementary Figure 10). Based on the positive correlation between NF-κB activation and IL22BP production, we hypothesized that LT might regulate IL22BP by activating NF-κB. To test this hypothesis, we performed a short hairpin RNA (shRNA)–mediated knockdown of RelA and RelB (Supplementary Figure 11). We found that the knockdown of RelB resulted in lower levels of IL22BP, whereas the knockdown of RelA had no significant effect (Figure 5G–I). These results complement our previous findings, which showed no effect of TNF-α on the expression level of IL22BP in DCs,¹⁶ because TNF-α signaling activates the canonical, but not the noncanonical, NF-κB pathway.³¹

Figure 5.

LT controls the expression of IL22BP via the noncanonical NF-κB pathway. (A, B) Representative FACS plots and corresponding statistics showing IL22BP production by MDDCs given recombinant human LTα1β2 or an agonistic antibody against the human LTBR (CBE11) for 48 hours (n = 4). (C) The corresponding IL22BP expression measured using RT-PCR (n = 5). (D) Representative Western blot and (E) the corresponding statistics showing the activation of RelA and RelB after exposure of MDDCs to LTα1β2 (n = 3; biological replicates). Results are normalized to the 0 hour timepoint. (F) Western blot analysis showing the activation of NF-κB signaling and IL22BP during the differentiation of MDDCs. (G, H) Representative FACS plots and corresponding statistics showing IL22BP production by MDDCs after the knockdown of RelA or RelB (2 different shRNAs used for each) compared to scrambled shRNA as control (n = 3). (I) Western blot showing the efficiency of the shRNA-mediated knockdown of RelA and RelB. Data are presented as mean ± standard error of the mean ∗P < .05, ∗∗P < .01, and ∗∗∗P < .001, as assessed by 1-way ANOVA with Bonferroni post hoc tests. If not stated otherwise, P > .05 is considered nonsignificant. Ctrl, control; PBMC, peripheral blood mononuclear cell; RT-PCR, reverse-transcription polymerase chain reaction.

Thus, we conclude that IL22BP is induced by LTα1β2 via the noncanonical NF-κB pathway in human DCs.

High Interleukin 22 Binding Protein Levels Are Associated With a Favorable Outcome in Patients With Colorectal Cancer

Finally, we aimed to identify the clinical importance of the reduced LT and, thus, IL22BP levels in human tumor tissues. For this purpose, we compared the survival of 92 patients, including 46 patients with high and 46 patients with low IL22BP expression levels (Figure 6A). The median IL22BP expression level was used as the cutoff. Using a multivariate analysis, we found that patients with high IL22BP mRNA levels had a significantly better prognosis (discovery data set) (Supplementary Table S2) (Figure 6B and D). One caveat of this analysis was that patients with lower IL22BP expression had a slightly more advanced tumor stage (Figure 6C). Specifically, samples with low expression levels of IL22BP were overrepresented in patients with lymph node involvement or distant metastases. Therefore, we determined IL22BP levels in a larger second cohort consisting of 1110 patients with CRC (validation data set) (Table 1). To this end, we performed immunohistochemical staining of IL22BP in a paraffin-embedded tissue microarray (Fig 6A and E). Based on the staining, the patients were divided into 2 groups; in 1 group (176 patients), IL22BP protein was detected, and in the other (934 patients) it was largely undetectable (Figure 6E). The results obtained with a multivariate analysis confirmed our initial findings showing that patients with high IL22BP expression in the tumor had a better clinical outcome (Fig 6F–H).

Figure 6.

High IL22BP levels are associated with a favorable outcome in patients with CRC. (A) Graphical representation of the experimental and statistical setup in the 2 independent patient cohorts. (B) Kaplan-Meier curve comparing the survival of patients with high and low relative IL22BP mRNA expression (divided by the median IL22BP expression value). Patients who did not survive the first 30 days after surgery were excluded from the analysis. The statistical significance was determined by the log-rank test. (C) TNM stage of patients with high or low IL22BP mRNA expression levels. (D) Results of the multivariate analysis showing the effects of doubling LTB or IL22BP expression and other selected parameters on patient survival. (E) Representative pictures of immunohistochemical staining of paraffin-embedded CRC tissue from a pre-existing tissue microarray. (F) Kaplan-Meier curve comparing the survival of patients in which IL22BP was detectable (IL22BP positive) or absent (IL22BP negative) based on immunohistochemical staining. The statistical significance was determined by the log-rank test. (G) TNM stage of patients in whom IL22BP was detectable or undetectable. (H) Results of the multivariate analysis showing the effects of IL22BP and other selected parameters on patient survival. CI, confidence interval; expr., expression; HR, hazard ratio; IHC, immunohistochemistry; neg. negative; pos., positive.

Finally, we performed a similar analysis for other gastrointestinal/mucosal tumors using The Cancer Genome Atlas.³² Interestingly, high IL22BP expression was associated with a significantly better outcome in CRC but not in the other malignancies analyzed, suggesting that this effect may be specific for CRC (Supplementary Figure 12).

In conclusion, low IL22BP expression is an independent risk factor for lower survival, suggesting that IL22BP is an important prognostic biomarker in patients with CRC.

Discussion

Tumor progression does not depend only on the intrinsic genetic changes in cancer cells but also on the immunologic tumor microenvironment. In fact, in situ analysis has shown that immune cell infiltration affects the prognostic outcome of patients with tumors.^{33, 34, 35, 36, 37, 38, 39, 40, 41} Even though the immune system as a whole is able to generate antitumor responses, there is evidence suggesting that certain immune cells and cytokines actually promote tumor progression. Therefore, it is crucial to understand the cellular and molecular pathways underlying this dichotomy. Furthermore, given that we are in the age of emerging immunotherapy, we need to learn exactly which molecular pathways are involved in directing the immune system toward either pro- or antitumorigenic responses. To this end, this study focused on the expression and regulation of the antitumorigenic factor IL22BP, which controls the protumorigenic effects of the cytokine IL22.

Interestingly, we found that IL22BP is down-regulated in human CRC. Likewise, LTA and LTB were also down-regulated. Furthermore, there was a clear correlation between LTA and LTB and IL22BP. Based on these data, we hypothesized that LTα1β2 might regulate IL22BP. Indeed, using mouse models and assays containing human cells, we proved that LTα1β2 regulates IL22BP via the noncanonical NF-κB pathway.

In addition to DCs, eosinophils produce IL22BP in the human colon in the steady state.^15,16 In line with these data, we showed that DCs and eosinophils are the main source of IL22BP in human CRC. However, CD4⁺ T cells showed only an incremental IL22BP expression in CRC. Interestingly, in patients with IBD, CD4⁺ T cells produce significant amounts of IL22BP. This suggests that the micro-milieu, which is very different between CRC and IBD, might differentially regulate IL22BP expression by CD4⁺ T cells.

Both pro- and antitumorigenic effects of LT have been described in humans.^{42, 43, 44} Previous studies identified not only direct effects of LT on cancer cells but also indirect effects of LT on the tumor microenvironment.⁴⁴ For instance, LT signaling plays a critical role in the development of tumor-associated tertiary lymphoid structures, which in turn, are associated with a positive prognosis in CRC.^45,46 The mechanisms behind the dual functions of LT remain unclear. However, the present study identifies a mechanism by which LT orchestrates the immunologic microenvironment in CRC in a protective manner. By activating the noncanonical NF-κB pathway, LT promotes IL22BP expression that subsequently antagonizes the protumorigenic effects of the cytokine IL22. In mouse CRC models, the antitumorigenic effects of LT were completely dependent on IL22BP. Moreover, we showed that LTα1β2 regulates IL22BP via the noncanonical NF-κB pathway. Interestingly, we previously reported that IL18, which also activates NF-κB, suppresses IL22BP expression by DCs during intestinal tissue injury.⁷ These apparently contradictory data might be explained by our results showing that the noncanonical NF- κB pathway was critical for the LT-mediated IL22BP induction in DCs. On the contrary, IL18 signals primarily via the canonical NF- κB pathway.⁴⁷

Finally, NF-kb2–deficient mice have reduced disease severity compared to WT mice in the DSS-colitis model, and consequently, they also develop a reduced tumor load in the colitis-associated CRC model.⁴⁸ Interestingly, our data indicate that blockade of LTBR reduces IL22BP production and consequently leads to a higher tumor burden in mice. Notably, the induction of IL22BP by LT is dependent on RelB and, therefore, on the noncanonical NF- κB activation. In contrast to the aforementioned study, blockade of the LTBR did not change colitis severity. Of note, total knockout mice were used by Burkitt et al.⁴⁸ Taken together, these results point toward a different role of noncanonical NF- κB activation in different cell types during colitis and cancer. Therefore, further studies using conditional knockout mice will be critical to decipher these signaling pathways in the context of intestinal carcinogenesis.

Interestingly, our data obtained from the analysis of human CRC samples indicate that an imbalance of the LT-IL22BP axis influences the prognosis of patients with CRC. Studying CRC in 2 independent patient cohorts, we showed that low IL22BP levels were linked to a poor prognosis, whereas high IL22BP levels showed significantly better survival. One caveat of this analysis was that patients with higher IL22BP expression in the discovery data set had a slightly less advanced tumor stage. To overcome this, we determined IL22BP levels in a larger second cohort with 1110 patients. However, we had to use different quantification methods in the discovery (IL22BP mRNA expression) and validation data sets (IL22BP IHC staining). Accordingly, the cutoff methods used were also different. Despite these technical differences, we could validate our finding that IL22BP expression represents an independent risk factor for a poor prognosis in CRC. Importantly, the results of the discovery data set suggest a possible link between low levels of IL22BP and lymph node involvement or distant metastases. It remains unclear whether there is a causal link underlying this finding. It is possible that both metastasis-dependent and metastasis-independent effects of IL22BP influence CRC progression and, thus, patient survival. Further studies are needed to test this hypothesis.

Taken together, our data show that IL22BP mediates the antitumor effects of LT. Moreover, we found that IL22BP expression in the tumor can serve as a biomarker to identify patients with a worse clinical outcome. Of note, IL22 also has beneficial effects, and thus it is crucial to identify patients who might benefit most from an IL22 blockade. Indeed, our data indicate that these patients could be selected based on the IL22BP expression level.

Supplementary Methods

Animals

Il22^–/– and Il22bp^–/– mice are described elsewhere.^1,2 Tnf^–/–, Lta^–/–, and Ltb^–/– mice were obtained from the Jackson Laboratory (Bar Harbor, ME). Conditional Lta- and Ltb-knockout mice were kindly provided by Andrey A. Kruglov.^3,4 Age- and sex-matched knockout mice and cohoused, in-house–bred C57/BL6 WT mice 8 to 18 weeks of age were used for all experiments. All animals were cared for in accordance with the Institutional Animal Care and Use Committee of Yale University or the institutional review board Behörde für Soziales, Familie, Gesundheit und Verbraucherschutz (Hamburg, Germany).

Flow Cytometry

Human fluorochrome-conjugated antibodies anti-CD45, anti-CD4, anti-CD3, anti-CD11c, anti-Siglec-8, anti-IL17A, anti-IFNγ, anti-TNFα, and anti-IL22 and mouse antibodies anti-CD45, anti-CD11c, anti-CD3, anti-CD4, and anti-Siglec-F were purchased from BioLegend (San Diego, CA). Additional anti-human CD3 and anti-human CD4 were purchased from BD Biosciences (San Jose, CA). Anti-human IL-22BP antibody (clone 87554) and IgG2B isotype control were purchased from R&D Systems (Minneapolis, Mn). To identify dead cells, 7-amino-actinomycin D staining (BioLegend) was performed. For the identification of lineage-negative cells, the following fluorochrome-conjugated antibodies were used: anti-CD1a, anti-CD5, anti-CD8, anti-CD11c, anti-CD14, anti-CD19, anti-CD34, anti-CD56, anti-CD123, anti- αβTCR, anti-FceRI-a, anti-γδTCR, anti-CD16, anti-CD94 (all purchased from BioLegend), anti-CD4 (purchased from BD Biosciences), and anti-CD303 (purchased from Miltenyi Biotec, Bergisch Gladbach, Germany). For extracellular staining, isolated hematopoietic cells from colon and tumor tissue nodes were incubated for 20 minutes at 4°C. Intracellular cytokine staining was performed after the extracellular staining. First, the cells were diluted in 100 μL 4% formaldehyde for 20 minutes at room temperature. After centrifugation, the cells were resuspended for 4 minutes in 0.1% NP-40 to permeabilize the cell membranes. After 4 minutes, the cells were washed with FACS buffer and centrifuged (400g, 5 minutes, 4°C). The cells were then stained with the respective antibodies. The staining was performed in the dark for 1 hour at 4°C. Afterward, the cells were washed, centrifuged, and resuspended in FACS buffer for acquisition. A modified staining protocol was used for staining IL22BP. The staining of IL22BP was performed without ex vivo restimulation. After staining the cells intracellularly for 1 hour with the I22BP antibody, the cells were centrifuged and resuspended in 5 mL FACS buffer. After overnight incubation, the cells were centrifuged and washed again with FACS buffer before being resuspended in 300 μL FACS buffer for the acquisition. Cells were sorted on a FACS Aria II or acquired on a LSRII Fortessa flow cytometer (BD), respectively. Data were analyzed with FlowJo software (FlowJo, Ashland, OR).

Endoscopic Procedures

Colonoscopy was performed in a blinded fashion for colitis and tumor monitoring using the Coloview system (Karl Storz, Tuttlingen, Germany) as previously described.⁵ Briefly, colitis scoring was based on granularity of mucosal surface, stool consistence, vascular pattern, translucency of the colon, and fibrin visible (0–3 points for each). Tumor sizes were graded from 1 to 5. Tumors observed during endoscopy were counted to obtain the total number of tumors per animal. The total tumor score per mouse was calculated as the sum of all tumor sizes.

Lentiviral Transfer of Short Hairpin RNAs Into Monocyte-Derived Dendritic Cells

Lentiviral vectors expressing shRNAs under control of the human U6 promoter (MISSION pLKO.1-puro) directed against RelA (TRCN0000014686, TRCN0000014687) and RelB (TRCN0000014713, TRCN0000014716) as well as a nontargeted control shRNA (SHC002) were obtained from Sigma-Aldrich (Munich, Germany). The production of lentiviral particles has been described in detail earlier,⁶ and protocols are available online (http://www.LentiGO-Vectors.de). To allow transduction of MDDCs with HIV-1–derived lentiviral vectors, 1 × 10⁶ mononuclear cells in 1 mL medium per well of a 24-well plate were treated 1 day after isolation with 30 μL VSV-G pseudotyped simian immunodeficiency virus (SIV)–derived vector-like particles. The vector-like particles were produced by using the plasmid pSIV3⁺ (kindly provided by Jeremy Luban, University of Massachusetts Medical School) and were used to transfer the vpx protein into the cells.⁷ Subsequent addition of 100 μL VSV-G pseudotyped, HIV-1–derived lentiviral particles led to the stable integration of shRNAs and the puromycin resistance gene. This protocol resulted in transduction rates of up to 95%, as determined by flow cytometry using the control vector LeGO-G2-Puro⁺,⁸ expressing eGFP and puromycin resistance, having a titer of 5 × 10⁷/mL as measured on 293T cells. The selection of successfully transduced cells with 1 μg/mL puromycin in the culture medium was started 2 days after transduction.

RNA Analysis

Total RNA was extracted from tissue and cells of colon, lymph nodes, liver, and spleen using Trizol Reagent (Invitrogen, Waltham MA). The High Capacity cDNA Synthesis Kit (Applied Biosystems, Foster City, CA) was used for cDNA synthesis. Primers and probes were purchased from Applied Biosystems. Human primers and probes including reference: IL22 (Hs01574154_m1), IL22RA1 (Hs00222035_m1), IL22RA2 (Hs00364814_m1), IL17A (Hs00174383_m1), IL23 (Hs00900828_g1), IL18 (Hs01038788_m1), TNF (Hs01113624_g1), LTA (Hs04188773_g1) , LTB (Hs00242739_,1) , LTBR (Hs01101194_m1), and HPRT1 (Hs02800695_m1). Mouse primers and probes including reference: Hprt (Mm01545399_m1), Il22 (Mm00444241_m1), Il22ra2 (Mm01192969_m1), Lta (Mm00440228_gH), Ltb (mm00434774_g1), Il23 (mm00518984_m1), Tnf (Mm00443258_m1), and Il18 (Mm00434226_m1). Real-time polymerase chain reaction (PCR) was performed using the Kapa Probe Fast qPCR Master Mix (Kapa Biosystems, Charlestown, MA) on the StepOne Plus system (Applied Biosystems). For both human and mouse, relative expression was normalized to HPRT and calculated using the 2^-ΔΔCt method.

Infection With Citrobacter rodentium

WT or Il22bp^–/– mice were orally gavaged with C rodentium (10⁹ colony-forming units/200 μL). Right after the injection and after 6 days, mice were intraperitoneally injected with either ACH6 antibody or PBS. On day 8 after infection, mice were killed, and colon and cecum were isolated for analysis.

Preparation of Cancer Cell Lines for Intracecal Injection

MC38 cells were cultured at 37°C in 5% CO₂/95% air using Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum, 2 mmol/L l-glutamine, 1 mmol/L pyruvate, 100 U/mL penicillin, and 100 mg/mL streptomycin. Cell lines were tested for Mycoplasma species (by PCR). Cells were transduced with lentivirus CAG.Luc.eGFP by the standard procedure, and stable clones were generated by puromycin selection. For in vivo injections, cells were harvested using trypsin, incubated with Trypan blue, counted as described elsewhere, and injected into the cecum of anesthetized mice. Only 95% viable cells were used in vivo.

Orthotopic Cecum Colon Cancer Model

The cecum of anesthetized mice (8 weeks, n = 10) was exteriorized through an abdominal laparotomy. 1 × 10⁶ MC38 CAG.Luc.eGFP cells were injected into the cecal wall between the mucosa and the muscularis externa layers by using a 30-gauge needle.⁹, ¹⁰, ¹¹ A proper implantation into the cecum was confirmed at day 0 by a localized and unique bioluminescent signal into the abdominal cavity. Mice successfully injected were imaged by bioluminescence imaging 10 days later from the ventral view.

Bioluminescence Imaging

Cells and mice were serially imaged on an in vivo imaging system, and data were analyzed using Living Image, version 4.3 (Perkin-Elmer, Waltham, MA), after the addition of 300 μg/mL d-luciferin to culture media or delivery of 1 mg intravenous d-luciferin.⁹

Injection of Antibodies Against Lymphotoxin Beta Receptor

In the indicated experiments, mice were intraperitoneally injected with 75 μg of a blocking Fc-fusion protein against the LTBR (called aLTBR throughout the article) or agonistic LTBR antibodies (ACH6) as described previously.^3,12

Analysis of The Cancer Genome Atlas Data

Data from The Cancer Genome Atlas¹³ were used to validate the findings of our 2 patient cohorts. Data were analyzed using GEPIA, a web server for cancer gene expression profiling.¹⁴ The data sets for colon and rectal adenocarcinoma were pooled for the analysis. Patients were divided into those with high and low IL22BP expression levels. The median expression was used as a cutoff.

Western Blot

To analyze the NF-kB pathway and IL-22BP activation, total cell lysates of indicated cell populations were separated in a 10% sodium dodecyl sulphate polyacrylamide gel electrophoresis assay, transferred to polyvinylidene difluoride membranes (Merck Millipore), probed with different primary antibodies (Cell Signaling Technology, Danvers, MA, and R&D Systems) (Supplementary Table 1), subsequently incubated with the appropriate horseradish peroxidase–conjugated secondary antibodies (Dako), and finally visualized with the enhanced chemiluminescence substrate (Merck Millipore, Billerica, MA).

Dextran Sodium Sulfate Colitis

For acute DSS colitis induction, mice received 2.5% DSS in drinking water for 5 days, followed by 7 days of regular water. The mice were treated with ACH6, aLTBR, or PBS (injected intraperitoneally) on days 1 and 8 of the experiment. According to the animal protocol, mice were killed if they lost more than 20% of their initial body mass.

Single-Cell Sequencing

The sorted cellular suspension was loaded on a 10× Genomics (San Jose, CA) Chromium instrument to generate single-cell gel beads in emulsion (GEMs). Single-cell RNA-sequencing libraries were prepared as described by the 10× Genomics Single Cell 3′ v2 Reagent Kit user guide and using the following reagent kits: Chromium Single Cell 3′ Library & Gel Bead Kit v2 (PN-120237), Chromium Single Cell A Chip Kit (PN-120236), and Chromium i7 Multiplex Kit (PN-120262). Briefly, after reverse-transcription, GEMs were broken, and the single-strand complementary DNA (cDNA) was cleaned up with DynaBeads MyOne Silane Beads (Thermo Fisher Scientific; P/N 37002D) and SPRIselect Reagent Kit (Beckman Coulter, Pasadena, CA; P/N B23318). Purified cDNA was amplified for 12 cycles before being cleaned up using SPRIselect beads. cDNA libraries were prepared by using the reagents in the GemCode Single-Cell 3′ Library Kit, and 15 PCR cycles were used for the samples’ index PCR. Single-cell libraries were sequenced on an Illumina (San Diego, CA) HiSeq 4000 machine in 150–base pair paired-end mode. After demultiplexing, we used the Cellranger pipeline (v. 3.0.1, 10× Genomics) to process fastq files, aligning the reads to the GRCh38 human genome draft. The resulting barcode-read count matrices were processed in R (R Foundation for Statistical Computing, Vienna, Austria) with the Seurat package (v. 2.3.4). Briefly, we removed low-quality and suspected doublet cells, resulting in 3860 and 787 cells for tumor and healthy tissue samples, respectively. After normalization and scaling, we used the shared nearest neighbor algorithm in Seurat to detect clusters and visualized the data using t distributed stochastic neighbor embedding.

Generation of Monocyte-Derived Dendritic Cells

Peripheral blood mononuclear cells were isolated from freshly collected blood using density gradient centrifugation. Monocytes were obtained using magnetic-activating cell sorting with anti-CD14 magnetic beads purchased from Miltenyi Biotec. The monocytes were then plated at a density of 1 × 10⁶ in RPMI media supplemented with 10% fetal bovine serum and human IL4 and granulocyte-macrophage colony-stimulating factor (both purchased from R&D Systems) at a final concentration of 25 ng/mL. Media were changed after 3 and 6 days, and cells were collected after 3, 6, and 8 days for further analysis.