Discovery of rs9984273 in clinical trials investigating intravenous IFN β-1a for the treatment of ARDS

We have performed two significant clinical trials investigating the use of intravenous IFN β-1a for the treatment of ARDS [18, 24]. The rationale for performing these trials was in the ability of intravenous IFN β to increase the expression of 5’-ectonucleotidase, CD73 on lung endothelium. CD73 is a key protective molecule and the rate limiting enzyme in converting pro-inflammatory adenosine monophosphate (AMP) into anti-inflammatory adenosine [25]. The first open-label Phase I/II study investigating intravenous IFN β-1a for ARDS produced remarkable results, showing ARDS mortality of only 8% [24]. The trial was small, consisting of only 37 ARDS patients in the active arm, but it led us to perform a substantially larger placebo controlled trial of 296 ARDS patients [18]. In this randomized clinical trial, intravenous IFN β showed no benefit over placebo in the entire study population. However, further laboratory analyses revealed that glucocorticoids block IFN signaling and the upregulation of CD73 in the lung endothelium and increased mortality in the active arm of the study [5]. Patients (n = 66) who did not receive glucocorticoids with IFN β had 28-day mortality of 10.6%, while patients (n = 78) who did receive glucocorticoids with IFN β had 28-day mortality of 39.7%. The deleterious effect of glucocorticoids administered together with intravenous IFN β was evident despite adjusting to disease severity and the likelihood of receiving glucocorticoid treatment in propensity matched multivariable regression analyses [5].

Further, pre-defined targeted resequencing and genetic analyses of entire gene regions were performed for genes encoding CD73 (NT5E) and IFN α/β receptor alpha and beta chains (IFNAR1 and IFNAR2). In the initial analyses, the SNP associations with the IFN β-1a treatment response were inspected for all the SNPs in the sequenced regions and for the statistically significant associations (P < 0.05) with mortality was inspected. Out of these analyses, the minor C allele found with sequencing (rs9984273) was shown to significantly associate with decreased mortality in the IFN β treatment arm when compared to those patients having the major allele T. Patients with ARDS carrying the minor C allele had day-28 mortality of only 10.9% (vs 31.0% without minor C allele) when treated with IFN β. Figure 1A, 1B shows the survival curves up to day 360 for active and placebo arm by SNP status (n = 202 with the SNP status available). As can be seen, similar phenomenon was not witnessed in the placebo arm. When testing for the interaction using multivariate Cox model adjusted for Acute Physiology and Chronic Health Evaluation (APACHE) II score and concomitant glucocorticoid use, the interaction between the treatment group and SNP status was significant (p = 0.046). Furthermore, day 28 mortality was analyzed using multivariate logistic regression in the full population (n = 296). In this analysis, concomitant use of glucocorticoids with IFN β-1a was associated with increased mortality (OR 3.30; 95% CI 1.79–6.08; P < 0.001), while the presence of the minor allele C in rs9984273 was associated with lower mortality at day 28 (OR 0.31; 95% CI 0.13–0.72; P = 0.006) (Fig. 1C). There was no significant fluctuation in the effect of the polymorphism on day 28 mortality between countries (p = 0.7462 for polymorphism country interaction, the forest plot in Additional file 1: Fig. S1). Thus, the use of glucocorticoids and the presence of the minor allele in this SNP associate to the outcome of intravenous IFN β treatment in patients with ARDS.

Fig. 1

The effect of the rs9984273 genotype on mortality and cytokine levels in the INTEREST trial investigating intravenous IFN β-1a for ARDS treatment. A genetic sample was available from 202/296 (104/144 in the IFN β-1a treatment arm and 98/152 in the placebo arm) of the study subjects. Number of the subjects at risk is indicated on the x- axis. A The rs9984273 minor allele C (CC or CT) was significantly associated with survival in patients compared to those who were homozygous with the major allele (TT). B A similar association was not observed in the placebo arm. C Further adjusted logistic regression analysis for Day 28 mortality encompassing the entire Phase III study population (both active and placebo arms) revealed that the use of glucocorticoids had the strongest association with mortality. D The logarithmic change in serum IFN γ and IL-6 levels according to the genetic background of rs9984273 in patients of the INTEREST trial with ARDS of pulmonary origin. Patients with the minor allele (CT or CC) had more rapid decrease in serum IFN γ and IL-6 levels toward normal when glucocorticoids were administered than the patients homozygous with the major allele (TT) in rs9984273. Numbers of the patients at each time point are indicated

The polymorphism affected differently the outcome of men and women in the INTEREST trial. Majority of the patients (70%) in the trial were male. 42.4% of men and 51.4% of women had CC/CT. Mortality of women was 11.8% for TT and 8.3% for CC/CT patients; mortality of men was 28.9% for TT patients and 12.5% for CC/CT patients. Thus, TT increases risks of death only in men, OR 2.85, p = 0.028, while it is not seen among the women, OR 1.47, p = 0.63 for an unknown reason.

Contribution of the SNP rs9984273 polymorphism was further investigated by a retrospective analysis, in which we performed focused genetic analyses of IFNAR2 on the earlier Phase I/II study population. Phase I/II and III study populations were comparable according to demographics and disease severity; however, we identified one major difference. In the Phase I/II study 40% of the study subjects receiving IFN β received overlapping glucocorticoids, but 73% of these patients had the rs9984273 the minor C allele. In contrast, in the phase III study 54% of the patients received glucocorticoids, but only 46% of the subjects receiving overlapping glucocorticoids with IFN β-1a carried the minor C allele. Hence, the rs9984273 C allele was highly enriched in the Phase I/II study population compared to the Phase III study, and more patients received glucocorticoids in the Phase III than in the Phase I/II study, thus suggesting a possible explanation for the difference between mortality in the studies.

Rs9984273 associates with IFN γ and IL-6 production in patients receiving glucocorticoid treatment

The SNP rs9984273 locates in the 3’ untranslated region (3’-UTR) of the IFNAR2 gene (chr21:33,262,760 (GRCh38)). To investigate potential binding sites of GR in the region, we utilized position weight matrices compiled from individual genomic sites from the Transfac database (Transfac matrix table, Release 2020.3) or derived from the Encode ChIP-seq data for GR [21]. The Transfac-based results showed that the SNP resides in a predicted transcription factor binding site of GR (Transfac matrix similarity score 0.864, chr21:33,262,746–33,262,761). Furthermore, the SNP was predicted to alter a GR binding site on the basis of the ChIP-seq data (GR_known3 from [21]).

Following our in silico evaluations, we next wanted to determine, whether rs9984273 was associated with the immune status of the patients when glucocorticoids were administered. The Phase III ARDS trial (INTEREST) was optimal to analyze this effect as it had cytokine profiling of the patients for the first 14 days. IFN γ and IL-6 were chosen as indicators of immune activation as they are also associated with poor outcome in ARDS and COVID-19 [24, 26]. Only those patients with ARDS of pulmonary origin (pneumonia or pulmonary sepsis) were chosen for these analyses to best resemble ARDS of COVID origin and adjust for a heterogenic etiology of an all-comer ARDS trial, e.g., excluding aspiration. We found high interindividual variations in IFN γ and IL-6 levels and the values were not statistically significantly different between the patients homozygous with the major T allele and those with the minor allele C (CC or CT) in rs9984273 at the beginning, day 0 (IFN γ: 19.2 ± 38 pg/ml for CC/CT and 24 ± 40.5 pg/ml for TT; IL-6: 75.9 ± 275.3 pg/ml for CC/CT and 300.6 ± 1178.5 pg/ml for TT), but after day 7 IFN γ and IL-6 levels of the patients with the CC/CT genotype start to decrease back to normal faster than in TT patients (Fig. 1D). We hypothesize that administering patients with glucocorticoids is not harmful, if they possess at least one copy of the minor allele despite receiving simultaneous IFN β; however, in the patients under glucocorticoid treatment, the major allele TT is associated with higher levels of IFN γ and IL-6 (RMANCOVA, P < 0.0001), and the two anti-inflammatory agents, glucocorticoids and IFN β, become harmful. IL-6 and IFN γ values are presented separately according to patients’ mortality status on Day 28 in order to account of lost values due to death (Additional file 1: Fig. S2). Cytokine associations remain similar according to the genetic background despite clinical outcome.

Rs9984273 contributes to the expression of IFNAR together with rs2236757

After the discovery of rs9984273, and recognizing its role in the outcome of the previous clinical ARDS studies, we then investigated, whether this polymorphism contributes to the expression of the IFNAR (composed of IFNAR1 and IFNAR2 subunits) by immunohistochemistry using antibodies against IFNAR subunits of the receptor in surgical lung specimens from 14 patients. In this material, five samples had the major allele TT, 8 samples CT, and 1 CC for rs9984273. The primary antibody was titrated to the level, at which the alveolar epithelium and vasculature still remained strongly positive in a subset of the samples. The scoring was done without any knowledge of the genotype. The results clearly show that all high expressors of IFNAR were CT and overall the CT/CC group had statistically significantly higher IFNAR level than the TT group. However, four out of nine samples from the CC/CT group show medium level of expression, indicating that other factors are involved in regulating IFNAR expression. Therefore, we decided to analyze another SNP, rs2236757 in IFNAR2 that has been found to contribute to disease severity of COVID-19 [27]. Eighty percent (4/5) of the highest IFNAR expressors had both rs2236757 A and rs9984273 C alleles, whereas three out of four medium expressors had rs2236757 G with rs9984273 C (Fig. 2B). Further, the A allele of rs2236757 alone was not enough for the high IFNAR expression as 80% (4/5) of low IFNAR expressors with rs9984273 TT had rs2236757 A. In summary, 8/9 of medium/low expressors did not have both rs2236757 A and rs9984273 C alleles (p = 0.005, Fisher’s exact test). Thus, A and C alleles of these SNPs together significantly contribute to the high expression of IFNAR in our lung specimens.

Fig. 2

Rs9984273 is associated with the expression of IFNAR and regulation of MX1 response. A Illustrative examples of IFNAR expression in lung specimens from patients with a CT (left panel) and a TT (right panel) genotype. Arrows indicate the alveolar epithelium and arrowheads the vessels. Stained negative controls are shown in the insets. B Combined expression intensity scores of rs9984273 (CC/CT and TT) and rs2236757 (AG, AA and GG) in lung specimens (n = 9 and 5, respectively). The single CC sample was pooled with CT samples for statistics between CC/CT and TT expressors. Color codes for rs2236757: AG blue, GG black, AA empty. Statistics: Mann–Whitney U-test, two-tailed. C MX1 response. Peripheral blood mononuclear cells (PBMC) were isolated from healthy donors having CC/CT (n = 8) or TT (n = 5) and interferon response was measured with (dark blue) or without (light blue) hydrocortisone (HC) by MX1 upregulation. The graphs show fold change (RQ) compared to the untreated sample (placebo without HC). Significance was calculated using unpaired t test with Welch’s correction

We also analyzed the number of CD73 positive vessels and expression level of CD73 in lung specimens incubated for 4 days with IFN β. There were no differences in the number of CD73 positive vessels/mm2 (2.6 ± 1.4 for CC/CT and 2.5 ± 0.8 for TT), but the staining intensity of CD73 was significantly higher in samples of CC/CT than TT patients (3.2 ± 0.4 and 2.2 ± 0.8, respectively; p = 0.04). Moreover, we collected peripheral blood mononuclear cells from CC/CT and TT individuals and measured MX1 (a key mediator of the interferon-induced antiviral response against a wide range of viruses) after IFN β treatment. CC/CT individuals have significantly better response than TT individuals to IFN β measured as MX1 increase suggesting better antiviral response (Fig. 2C).

CT and TT patients display different features in IFN β-induced STAT1 and STAT2 signaling and responsiveness to glucocorticoids

As the immune status and IFNAR expression are at least partially under the control of rs9984273 and the formation of the signal transducer and activator of transcription 1/2 and interferon regulatory factor 9 (STAT1/STAT2/IRF9) complex and its translocation into the nucleus is required to trigger type I IFN responsive genes, we next examined STAT1, phosphorylated STAT1 (pSTAT1), STAT2 and pSTAT2 expression in the lung specimens cultured in the presence of IFN β with or without hydrocortisone (HC) [4, 5]. We found that TT patients tended to have lower total STAT1 expression and statistically significantly less pSTAT1 than the CT patients and HC inhibited its nuclear location (Fig. 3A, B). In contrast, TT patients had significantly more STAT2 than the CT patients and HC did not decrease its nuclear expression like in CT patients (Fig. 3C, D), while no statistical differences in pSTAT2 were observed (Fig. 3 E). Examples of pSTAT1 and 2 staining are shown in Additional file 1: Fig. S3. For the STAT-analyses, we were limited being able to only include three out of five CT patients to the combined data shown in Fig. 3, as two sample donors were already on corticosteroids at the time of taking the lung specimen. In these two specimens, already on glucocorticoids, IFN β could not induce efficient nuclear localization of STAT1 and STAT2, further demonstrating the clinical role of glucocorticoids in this phenomenon.

Fig. 3

Patients homozygous for the major allele of rs9984273 (TT) have higher STAT2 expression in the lungs than the heterozygous patients with the minor allele (CT) and glucocorticoids do not inhibit its nuclear expression in TT patients. A STAT1 expression in the lung after a 4-day culture in the presence of IFN β with or without hydrocortisone (HC). B pSTAT1 expression in the same specimens as in A. C Example photomicrographs showing higher STAT2 expression in a TT patient than in a CT patient and the effect of HC on its nuclear translocation. Most STAT2 remains in the cytoplasm of the CT patients, whereas nuclear expression is prominent in the TT patient. Indicated insets are shown in the bottom row. D Combined results of all patients noting that two CT samples are excluded in the data as the patients were already under glucocorticoid treatment at the time of sample acquisition. E pSTAT2 expression in the same samples as in A, B and D. Ns, not significant; *P < 0.05; **P < 0.01; and ***P < 0.001. Statistics: two-way ANOVA with Šídák's multiple comparisons test

Rs9984273 is associated with COVID-19 hospitalization

Adhering to our recent clinical and laboratory findings, and the outbreak of the SARS-CoV2, we also explored the role of rs9984273 in COVID-19 through publicly available databases. After all, without a doubt both endogenous and exogenous type I IFNs and steroids play a fundamental role in disease progression and survival concerning COVID-19. Recent genome-wide analyses of COVID-19 patients have revealed that SNP rs2236757 in IFNAR2 has a significant impact on disease severity [27], but the functional role of these findings is yet to be explained and the contribution to signaling remains to be revealed. Until now, the importance of rs9984273 has not been recognized. Therefore, using the COVID-19 Host Genetics Initiative database, we explored whether there was an association between rs9984273 and disease severity (Table 1). We discovered that the presence of the minor allele of rs9984273 was associated with less hospitalization for COVID-19 (OR 0.96; 95%CI 0.93–1.00; P = 0.035), when comparing hospitalized to non-hospitalized COVID-19 patients. Hospitalized patients with COVID-19 also differed from the general public in the rs9984273 genotype distribution (OR 0.96; 95%CI 0.93–0.98; P = 0.001). Comparing the patient group with the most severe disease (very severe confirmed respiratory COVID-19) to the general public, the rs9984273 polymorphism showed a statistically significant risk association with the minor allele being associated with a lower risk of severe disease (OR 0.93; 95%CI 0.89–0.98; P = 0.006). The number of cohorts involved in the individual comparisons varied from 3 to 60 with no sign of heterogeneity between the cohorts (P > 0.3).

Table 1 Associations of rs9984273 C allele* to disease severity of COVID-19 in meta-analyses

We also tested, whether carrying the minor alleles of rs9984273 C and rs2236757 A, correlated with mortality in our ARDS cohort. We found that patients possessing the minor alleles for both genotypes have significantly decreased mortality than those homozygous with the major allele for both genotypes. The lower mortality among the minor allele carriers (n = 51) compared to non-carriers (n = 41) was seen at days 28, 90, 180, and 360, with 28-day mortality of 10% vs. 27% (P = 0.03), 90-day mortality of 16% vs. 37% (P = 0.02), 180-day mortality of 18% vs. 39% (P = 0.02), and 360-day mortality of 20% vs. 39% (P = 0.04), respectively. Although rs2236757 in IFNAR2 is associated with severe COVID-19 infection and low IFNAR expression to life-threatening disease [27], rs2236757 and rs9984273 are only in a mild linkage disequilibrium with (r2 = 0.19, D’ = 0.99) in the European population. Nevertheless, they may still significantly cooperate and regulate the expression of IFNAR as four out of five patients possessing both minor alleles had high levels of IFNAR expression.