Gastrointestinal biopsies were obtained and analyzed from a total of 90 adult patients receiving an HSCT between October 2009 and November 2013. Patient characteristics are summarized in Table 1 . All patients gave informed consent, the biopsy studies and scientific analyses were approved by the local ethical review board (approval no 02/220 and 09/059). All studies were performed under the regulations of Helsinki. Biopsies were either obtained in the course of a screening study in asymptomatic, clinically acute GI-GvHD-free patients (median 62 days after HSCT, n=57) or because of clinical symptoms indicative of de novo onset (median 78 days after HSCT, n=24) or persistence/recurrence of acute GI-GvHD (median 175 days after HSCT, n=9). Biopsies were obtained through upper or lower GI endoscopy. 33 biopsies were obtained from the small intestine and 57 biopsies from the large intestine. Clinical grading was performed according to the Glucksberg’s criteria (26) and the histological grading was performed according to the Lerner’s criteria (27).

Total RNA was extracted from intestinal biopsies using the RNeasy Mini Kit (QIAGEN) according to the manufacturer’s recommendation. In brief, tissues were centrifuged at 4000 rpm for five minutes and were transferred to 700 µl RLT buffer (Qiagen) supplemented with 1% β-mercaptoethanol. Tissues were then sonicated for five seconds, 3-5 times, till the complete homogenization was ensured. The homogenates were filtered through QIAshredder column by centrifugation at room temperature (RT) for 3 min at 13000xg before proceeding to RNA extraction. To remove potential DNA contaminations, on-column DNA digestion with the RNase-free DNase Set (Qiagen) was implemented according to the protocol. RNA concentration was measured with ND-1000 NanoDrop Spectrophotometer (Thermo Fisher Scientific). RNA integrity quality was controlled using Agilent Bioanalyzer (Böblingen) as per the manufacturer’s instructions. RNA was stored at -80 °C until further use.

Total RNA was reverse transcribed into complementary DNA (cDNA) using Moloney murine leukemia virus reverse transcriptase (M-MLV RT) (Promega) enzyme. Random decamers (Promega) were used to prime cDNA synthesis. The volume of 200 ng (or 1 µg when stated) of total RNA was adjusted to 13 μl with nuclease-free ddH₂O and mixed with 1 μl Random Decamers (Promega) and 1 μl dNTPs (10 mM) on ice. Secondary structures of RNA were dissolved by 5 min incubation in thermocycler at 65°C followed by immediate incubation on ice for 1 min. After mixing with 4 μl M-MLV Buffer (5x;Promega) the samples were incubated at 42°C for 2 min. Reverse transcription started upon addition of 1 μl RT enzyme (50 min, 42°C) and was stopped by heat inactivation of the enzyme (15 min, 70°C). cDNA samples were stored at -20°C until further use.

The expression of VDR was determined by the 48.48 Fluidigm Dynamic Arrary Integrated Fluidic Circuits (IFC) together with reference genes GAPDH and HPRT as explained elsewhere (28). The Dynamic Array™ chip was primed by injecting a control line fluid into each accumulator on opposite sides of the chip. Following this, 5 μl of each assay was loaded into the respective inlets on the left side of the chip and 5 μl of each sample was loaded into the respective inlets on the right side of the chip. The samples and assays were mixed onto the chip, in the center of the array by the IFC Controller. The chip was then run using the BioMark Gene Expression Data Collection software, according to the given parameters for the 48.48 dynamic arrays.

qPCR and melting curve analysis were performed by running the following temperature program: 2400 s at 70°C and 30 s at 60°C, followed by a hot start for 60 s at 95°C, 40 PCR cycles of 5 s at 96°C for denaturation and 20 s at 60°C for annealing and elongation. The melting curve analysis consisted of 3 s at 60°C followed by heating up to 95°C with a ramp rate of 1°C/3 s. All samples were normalized to the GAPDH, and their relative expression was calculated. Following are the primer sequences: VDR, Sense: GGCTACCACTTTTACATGGTCA, Antisense: TCATAATCCCGGCCTCTCTCT; GAPDH, Sense: TTCGACAGTCAGCCGCATC, Antisense: GCCCAATACGACCAAATCCGT. Anti-microbial peptides (AMPs) were cycled (including primer sequences) and analyzed as previously described (29).

Intestinal tissue of patients with the diagnosis of no/mild acute GI-GvHD (Lerner Score 0-1) and severe acute GI-GvHD (Lerner Score 2-4) were used for this study. An in-situ immunohistochemical characterization and quantification (analogue to IRS-Score) of VDR were performed. For the in-situ characterization, standard routine diagnostic procedures and antibodies (VDR: Clone D2K6W, Cell Signaling) were used. All immunohistochemical stains were performed on tissue sections, prepared from-formalin-fixed paraffin-embedded tissue blocks. Tissues were sectioned into 2–3 µm slices. All the tissues were fixed in 4% neutral buffered formalin. Immunohistochemical staining was conducted using a Roche Ventana Benchmark Ultra automated slide stainer (Ventana Medical Systems, Roche, France). Immunofluorescence was performed as described previously (30). Briefly, after antigen retrieval, biopsies were incubated with rabbit monoclonal VDR antibody (dilution 1:100) for one hour. After washing three times with PBS, biopsies were again incubated with anti-rabbit Alexa Fluor (AF) 594 (dilution 1:100) for one hour in dark. Nuclei were counterstained with DAPI, samples were sealed with mounting media. VDR signals were observed using a Zeiss epifluorescence microscope.

Biopsies were quantified in a semi-quantitative fashion. The mucosa vitality was scored based on the presence of intact epithelium. The percentage of VDR expressing cells was scored with respect to available mucosal membrane in the biopsies. The staining score was calculated as a product of multiplication between positive cells proportion score (0-4) and the staining intensity score (0-3).

Until May 2012, patients received low dose Vitamin D3, i.e. 1000 to 5000 IU/d Colecalciferol. From May 2012 onwards, the patients received high dose 20.000 IU/ml Colecalciferol upon admission to the hospital followed by daily administration of 10.000 IU/ml Vitamin D3.

Data analysis was performed in SPSS v26 (IBM Corp., Armonk, NY, USA). Test of normality was performed using the Shapiro-Wilk test. Since data were observed to be non-normally distributed, Mann-Whitney or Kruskal Wallis tests were performed and Spearman correlation coefficients were computed. Box plots were used to visualize continuous data and dot plots for semi-quantitative data. For multivariable analysis, patients were classified into two groups based on receiver-operator curve (ROC) and Youden index for best sensitivity and specificity. Area-under-the curve (AUC) was reported. Time to event was defined from day of biopsy harvest to the day of death or the last day of the patient being confirmed to be alive (censored cases). The last follow up of each patient was as recent as of January 2022. Kaplan-Meier curve was generated and log-rank test was used analyze the association of VDR with the risk of TRM. To further analyze the association of VDR expression groups with TRM, Cox regression models were applied. The multivariable Cox model was adjusted for additional covariates GI-acute GI-GvHD, age, steroids, stage of disease, and donor type. Hazard Ratios (HR) and corresponding 95%-confidence intervals (95% CI) are reported as effect estimates. To separate transplant-related hazard from relapse-related hazard, competing risk analyses were evaluated using the CumIncidence function (31).

First, we were interested in analyzing the expression of VDR mRNA in terms of histological acute GI-GvHD. Patients were classified into no/mild Lerner grade 0-1 and high Lerner grade 2-4 GI-GvHD. Interestingly, VDR mRNA expression was significantly downregulated in higher grade Lerner acute GI-GvHD patients when compared to lower grade Lerner acute GI-GvHD patients (p = 0.007, Figure 1A ). Additionally, we also observed that patients who were supplemented with low dose vitamin D3 (1000-5000 IU/day) in our center had significantly lower intestinal VDR gene expression when compared to patients who were supplemented with high dose vitamin D3 (10000-20000 IU/day) within first 100 days (data not shown). As it is well-known that epithelial cells express the VDR, it was necessary to clarify if the VDR loss was solely due to epithelial denudation or if VDR loss could also be seen in absence of severe epithelial damage. For this purpose, we classified patients according to the clinical status of acute GI-GvHD. Clinically, GI-GvHD free patients were classified as the “screening” group, patients with a start of aGvHD were classified as the “onset” group, and the patients with progressive aGvHD were classified as the “ongoing” group. We observed that onset patients and ongoing patients displayed significantly lower VDR expression as compared to screening patients (p = 0.010 for onset vs. screening, p = 0.034 for ongoing vs. screening (not significant after multiple testing by Bonferroni/Sidak correction, p = 0.08), Figure 1B ). It is important to note that aGvHD onset patients started to show GvHD did not however show severe epithelial damage. The reduction of VDR mRNA in onset patient samples therefore implied that VDR loss was evident in GvHD even when the epithelium was intact. In conclusion, VDR mRNA expression is significantly downregulated in histological and clinical GvHD.

To evaluate the protein expression of VDR, we performed immunofluorescent staining of VDR protein expression in the gut biopsies of patients and found that higher grade Lerner patients (grade 2-3) had reduced VDR expression as compared to lower Lerner patients (grade 0-1) ( Figure 2A ). Results were confirmed by single antibody immunohistochemistry (IHC) in gut biopsies of patients (lower grade Lerner 0-1, n = 15, higher grade Lerner 2-4, n=18) after HSCT. To ensure that protein differences were not influenced by GvHD- induced intestinal damage, biopsies were scored for mucosa vitality based on intact epithelium. Mucosa vitality was severely compromised in higher grade Lerner GvHD patients compared to lower grade Lerner GvHD patients (low Lerner: mean = 96%, median = 100%, min/max = 50% - 100%; high Lerner: mean = 61%, median = 60%, min/max = 10% – 95%, p= 5x10^-6). Therefore, we made a cut-off of at least 85% mucosa vitality to ensure intact epithelium (lower grade Lerner, n=14; higher grade Lerner, n=8). An exemplary image of VDR staining from small intestine and large intestine is shown in Figures 2C, D . The number and intensity of VDR⁺ enterocytes in small intestine and VDR⁺ colonocytes in large intestine were strongly reduced in higher grade Lerner GvHD patients despite an intact epithelium. Of interest, we found that an overall epithelial VDR expression was significantly reduced in higher grade Lerner GvHD patients compared to lower grade Lerner GvHD patients (p = 0.03, Figure 2E ). Furthermore, the reactivity score that evaluates the staining intensity within VDR⁺ cells was significantly reduced in higher grade Lerner GvHD (p = 0.02, Figure 2F ). In summary, we observed that low epithelial VDR expression associated with acute GI-GvHD.

To examine the impact of VDR expression on prognosis after HSCT, we categorized patients into the groups who died as a consequence of TRM and who did not display TRM. Of note, patients dying of infection, acute GI-GvHD or toxicity were compounded as TRM group. We found a significant downregulation of VDR mRNA expression in the patient group who exhibited TRM compared to patients without TRM (median survival time in no TRM group= 88 months, median survival time in TRM group = 11 months, p = 4x10^-4, Figure 3A ). In order to deal with varying length of observation times, a defined 1-year TRM from the biopsy retrieval was additionally analyzed. Again, we observed a strong reduction of VDR mRNA in the TRM group when compared to the non-TRM group (median survival time in no TRM group = 12 months, median survival time in TRM group = 3 months, p = 0.018, Supplementary Figure 1A ). To further address this association, we classified patients based on high and low VDR expression. The point within the ROC curve with highest sensitivity and specificity was chosen for dichotomization of VDR mRNA expression. The AUC was 0.714 (95% CI = 0.604 – 0.824, p = 4x10^-4). The Kaplan-Meier survival curve revealed a significantly higher probability of TRM in patients with low VDR expression (median survival time in VDR low group = 14 months, median survival time in VDR high group = 67 months, log-rank p = 1x10^-5, Supplementary Figure 1B ). In the context of HSCT, the relapse related mortality (RRM) is a competing event for TRM. In our patient cohort, 13 patients died of relapse. Therefore, we performed competing risk analyses to account for RRM as a competing event. We observed a significant association of low VDR expression with the probability of TRM (p = 1.6x10^-6, Figure 3B ). In contrast, no significant association of VDR expression with relapse was observed (p = 0.654, Figure 3B ).

To determine the potency of VDR as an independent risk factor for TRM, a Cox regression model was performed. In the unadjusted model, low VDR expression and higher grade Lerner acute GI-GvHD (grade 2-4) were associated with the risk of TRM ( Table 2 ). After adjustment for classical risk factors, multivariate Cox regression showed low VDR expression (p = 0.001, hazard ratio 4.14, 95% CI 1.78 – 9.63) and higher grade Lerner acute GI-GvHD (grade 2-4) (p = 0.001, hazard ratio 2.65, 95% CI 1.33 – 5.31) as an independent risk factor for TRM ( Table 2 ). Classical risk factors such as age, disease status, donor type were not found to be significantly associated. When we looked at 1-year outcome from the time of biopsy, we again observed low VDR and severe acute GI-GvHD as a risk factor for TRM ( Supplementary Table 1 ). Although steroid treatment seemed to suppress VDR expression in univariate and multivariate analyses (data not shown), Cox regression analyses demonstrated that steroid treatment is not a risk factor for TRM. Taken together, the data indicates that low VDR expression predicts outcome after HSCT.

Anti-microbial peptides (AMPs) are known downstream targets of the VitD3-VDR pathway. Paneth cells express AMPs and are known to be expressed abundantly in the small intestine (29). Therefore, we investigated the correlation between VDR and AMPs gene expression in small intestinal biopsies in our patient cohort. We found a significant correlation of VDR expression with defensins: DEFA5, DEFA6 genes ( Figure 4A ). Within the correlation plot, it can be visualized that lower grade Lerner patients show high VDR and AMPs (blue circles) and higher grade Lerner patients show low VDR and AMPs (red circles) gene expression. To further substantiate the relation between VDR and AMPs, we tested the AMPs expression comparing patients with low and high VDR expression. As expected, we found low AMPS in “VDR low” patients while “VDR high” patients showed significantly high AMPs (DEFA5: p = 0.005, DEFA6: p = 0.026, Figure 4B ).