Biopsies of 118 patients receiving an allogeneic SCT at the University Hospital Regensburg, Germany, from 2008 to 2016 were included in the study. All patients gave informed consent for the GI-tract biopsies. Patient characteristics are given in Table 1. The study was approved by the local ethical review board of the University of Regensburg (approval numbers: 02/220, 09/059, and 17-619-101).

Gastro-intestinal biopsies were obtained either in the course of screening study in asymptomatic patients (median 30 days, range 13 - 2,345 days), because of clinical symptoms indicative of de novo onset (median 51 days, range 14 - 479 days), or during persistence or recurrence of acute GI-GvHD (median 130 days, range 37 - 538 days). Biopsies were histologically graded for the acute GI-GvHD according to the Lerner’s grading system (16). There were 56 patients who had no histological signs of acute GI-GvHD, 36 patients had GvHD grade 1, 11 patients had GvHD grade 2, 9 and 6 patients had GvHD grade 3 and 4, respectively.

Following retrieval, biopsies were immediately transferred to 500 µl RNA protect (Qiagen) and were stored at - 80°C until RNA extraction. RNA extraction, including DNA digestion, was performed with RNeasy Mini Kit (Qiagen) as per the manufacturer’s instructions. RNA concentration and purity were monitored with NanoDrop and Bioanalyser, respectively. One microgram of RNA was transcribed to cDNA using moloney murine leukemia virus (MMLV) reverse transcriptase (Promega) according to the manufacturer’s instructions. qPCR was performed on a Mastercycler Ep Realplex (Eppendorf) for GPR43 and GPR41 and on HT 7900 Real-Time PCR system (Applied Biosystems) for IL22 using QuantiFast SYBR Green PCR Kit (Qiagen). Genes of interest were normalized to the 18S rRNA reference gene.

Following gene-specific primer pairs were used:

IL22, forward: 5’-AGC-CCT-ATA-TCA-CCA-ACC-GC-3’, reverse: 5’-TCT-CCC-CAA-TGA-GAC-GAA-CG-3’; GPR43, forward: 5’- GTA-GCT-AAC-ACA-AGT-CCA-GTC-CT -3’, reverse: 5-CTA-GGT-GTT-GCT-TTG-AAG-CTT-GT -3’; GPR41, forward: 5’- GCC-AAC-TGC-ACT-AGG-TCT-GGA-GAG-3’, reverse: 5’-CTT-GCC-CAC-GAA-GAC-CAC-CA-3’; 18S rRNA, forward: 5’-ACC-GAT-TGG-ATG-GTT-TAG-TGA-G-3’, reverse: 5’-CCT-ACG-GAA-ACC-TTG-TTA-CGA-C-3’.

Formalin-fixed paraffin embedded (FFPE) biopsies were cut approximately 3 µm thick and were immersed in xylene and descending alcohol line. Antigen retrieval was performed with citrate buffer (pH 7.2) at 350 watts for 32 minutes in a microwave. Subsequently, the biopsy was blocked with 20% bovine serum albumin (BSA) for 20 minutes, rinsed three times with PBS followed by incubation with anti-IL-22 (mouse monoclonal, clone 2D5, Merck Millipore) and anti-CD3 (rabbit monoclonal, clone SP7, Thermo Scientific) primary antibodies at 1:50 dilution in 1% BSA for 1 hour. Primary antibodies were labeled with Alexa Flour (AF) 488 (for IL-22) and AF594 (for CD3) and secondary antibodies for another hour. Nuclei were counterstained with DAPI. Images were taken with a Zeiss epifluorescence microscope.

Data were analyzed in SPSS version 26 (IBM) and R version 4.1.2 (The R Foundation for Statistical Computing). Test of normality was performed using Shapiro-Wilk test. Normally distributed data were analyzed with t-test or one way ANOVA. Non-normally-distributed data were analyzed with Mann-Whitney U test or Kruskal-Wallis test. Correlations were assessed by using Pearson or Spearman-rho correlation coefficient for normal and non-normally distributed data, respectively. Kaplan-Meier method was used to calculate the probability of TRM. To analyze the impact of IL22 on TRM, uni- and multivariable competing risk Cox regression models were used accounting for relapse related mortality as competing risk event. Time to event was defined from day of biopsy to day of death or to last day of patient being confirmed to be alive (censored cases). The multivariable model was adjusted for the additional covariates GI-GvHD, age, steroids, stage of disease, donor type, broad spectrum antibiotics and conditioning. Since biopsies were taken at different time points after SCT, we also added time from SCT to biopsy as potential confounder to the model. Hazard Ratios (HR) and corresponding 95%-confidence intervals (95%-CI) are reported as effect estimates. All competing Cox regression analyses were evaluated using the CumIncidence function (17) and the factor2ind function (18) of the cmprsk_2.2-11 package in R version 4.1.2. Multivariable competing risk regression analyses was done in 110 of 118 patients since data on antibiotics at the time of biopsy were missing on 8 patients. A p-value < 0.05 was considered statistically significant for all analyses. Data were depicted using Graph Pad Prism, version 7.00 (GraphPad Software).

First, we were interested in whether GvHD patients show distinct IL22 mRNA expression compared to patients who did not develop acute GI-GvHD. When we classified patients according to Lerner grades, IL22 expression did not significantly differ between patients without GvHD or with mild to severe GvHD (Supplementary Figure 1A). When comparing screening biopsies of patients without clinical GI-GvHD to those derived at clinical onset, significant upregulation of IL22 was found in the latter (p = 0.043, Supplementary Figure 1B). Since IL-22 administration has a positive impact on survival in murine GvHD (11, 12), we tested the hypothesis that IL22 expression might have an impact on transplant-related mortality (TRM) and therefore separated the patients in regard to both, GvHD and TRM. In our cohort, patients experiencing TRM died mainly of infection, GvHD, or toxicity. The precise cause of death is listed in Table 2. Most importantly, within the patient group suffering from acute, histologically proven GI-GvHD, those with TRM showed significantly lower IL22 mRNA expression compared to those without TRM (p = 0.007, Figure 1A). Similarly, clinically defined onset patients tended to show lower IL22 expression in the TRM group as compared to the non-TRM group (p = 0.060, Figure 1B). Interestingly, in the patient group without TRM, histological GvHD was associated with a significantly higher IL22 expression compared to patients without histological GvHD (p = 0.048, Figure 1A). From a clinical perspective, in the patient group without TRM, IL22 expression was pronounced at the onset of GvHD compared to screening biopsies (p = 0.022, Figure 1B). In conclusion, IL22 is significantly upregulated in GvHD patients not dying from TRM.

When we examined the impact of IL22 on prognosis after allogeneic SCT, including patients with and without GvHD, we observed significantly lower IL22 expression in the TRM group compared to the non-TRM group (p = 0.010, Figure 2A). To further address this association, we dichotomized patients into high and low expression of IL22 based on ROC curve analysis and a Youden index of 0.6435. Kaplan-Meier analysis revealed a significantly higher probability of TRM in patients with low IL22 expression (p = 0.004, Supplementary Figure 2). In addition to the 36 patients dying by TRM, 31 patients died of relapse (relapse related mortality, RRM) as a competing event. Moreover, observation times, as measured from biopsy retrieval until the occurrence of the event (such as TRM or relapse) or the last follow up, varied. Therefore, a time-dependent competing risk analysis was performed revealing a significant association of low IL22 with the probability of TRM (p = 0.005, Figure 2B). In contrast, no significant association of IL22 and RRM was observed (p = 0.334, Figure 2B). Time-dependent Cox regression analysis revealed low IL22 (p = 0.007, hazard ratio 2.72, 95% CI 1.32 to 5.61) and acute GI-GvHD (p = 0.010, hazard ratio 3.12, 95% CI 1.32 to 7.39) as independent risk factors for TRM (Table 3). Established risk factors such as age, steroids, stage of disease, donor type, broad spectrum antibiotics or conditioning were not significantly associated. Furthermore, no influence of the time between the recent transplantation and biopsy retrieval was found. In summary, low expression of IL22 in intestinal biopsies associates with an increase in transplant-related mortality after allogeneic stem cell transplantation.

The application of broad-spectrum antibiotics (Abx) within 7 days before obtaining biopsies was used as an indirect indicator of microbiota damage (19) with 60% of patients receiving carbapenems (meropenem), 35% of patients receiving piperacillin/tazobactam and only 5% of patients receiving cephalosporins (ceftazidim). Since the lower GI-tract harbors the majority of microbiota (20), we analyzed IL22 status in lower GI-tract (n = 86) with respect to use of Abx treatment and observed a significant lower IL22 mRNA in patients receiving Abx compared to patients not treated with Abx (p = 0.019, Figure 3). When we included upper GI tract biopsies in our analyses, we still observed marked suppression of IL22 in antibiotic group with a trend towards significance (no Abx: median = 1.88, range = 0 - 898; Abx: median = 1.21, range = 0 - 85.8, p = 0.085). In conclusion, reduction of commensals as implicated by the use of Abx negatively affected IL22 expression in the lower GI in our cohort.

To identify IL-22 cellular source in patient biopsies, we performed immunofluorescent staining using IL-22- and CD3-specific antibodies in sigmoid colon biopsies (Supplementary Figure 3). IL-22⁺ cells were detected both within CD3⁺ and CD3^- cells. Microbiota-derived SCFA were shown to increase IL-22 expression in CD3^- ILC3 via GPR43 and both CD3^- ILCs and CD3⁺ T cells via GPR41 in murine models (14, 15). To further explore the role of microbiota on IL22 expression in allogeneic SCT patients, we examined the association of IL22 with SCFA receptors GPR43 and GPR41 in the intestinal biopsies of the lower GI tract (Figure 4). Indeed, a significant positive correlation between IL22 and GPRs was observed.