A type I interferon footprint in pre-operative biopsies is an independent biomarker that in combination with CD8+ T cell quantification can improve the prediction of response to neoadjuvant treatment of rectal adenocarcinoma

The study addresses the need for predictive biomarkers to identify rectal cancer patients who will respond to neoadjuvant treatment. While surgery is central to curative treatment, neoadjuvant therapy reduces local recurrence and improves resectability, yet only about 20% achieve a clinical complete response and treatment carries risks and side effects. Prior work indicates that TIL density and the Immunoscore may have prognostic value, and CD8+ TILs have been suggested as predictors of complete response, although findings are inconsistent. Mechanistically, cytotoxic T lymphocyte (CTL) activity (e.g., via granzyme B) and type I interferon (IFN) signaling (e.g., cGAS-STING pathway, radiotherapy-induced IFN) may enhance antitumor immunity. MxA is a type I IFN-stimulated gene product that can serve as a specific footprint of type I IFN exposure. The study hypothesizes that quantifying T cell subsets and a type I IFN footprint (MxA) in pre-operative biopsies can improve prediction of response to neoadjuvant therapy in rectal adenocarcinoma.

Design and cohort: Retrospective study of patients with rectal cancer treated at Sahlgrenska University Hospital/Östra (Gothenburg, Sweden). From 1774 patients (2007–2019) identified in the Swedish Colorectal Cancer Registry, 130 were included based on receipt of neoadjuvant therapy and an interval >14 days to surgical resection permitting tumor regression assessment. Demographic and clinical variables (e.g., T stage, neoadjuvant treatment type) were retrieved. Specimens and staining: Pre-operative biopsies were FFPE and sectioned at 4 μm on Superfrost Plus slides. After deparaffinization (xylene, graded ethanol) and heat treatment, multiplex immunofluorescence (mIF) was performed (antibodies and TSA reagents per Table S1). H&E staining was performed on adjacent sections for pathology review and to delineate invasive tumor areas. Image analysis and cell identification: A pathologist marked invasive tumor regions on H&E which were mapped to mIF sections. Slides were scanned (TissueFAXS) and analyzed in StrataQuest (v7.0.1.189). Fluorescent intensity thresholds for markers were set to define cell types; validation was performed by manual counts in randomly selected regions from 10 samples by two investigators, comparing mean manual counts to software counts. CD3 staining intensity variability was addressed by stratifying sections into weak, intermediate, and strong groups; CD3+ cutoff values were adjusted accordingly. γδTCR+ cells were quantified and subtracted from CD3+ to define conventional T cells. Tumor compartments annotated included tumor epithelium (PanCK+), tumor stroma (non-epithelial area within tumor border), and a 50 μm stromal rim around epithelium. Outcome (response): Tumor regression grade (TRG) was determined on surgical specimens per AJCC criteria (0–3; 0 = pathologic complete response, pCR; 3 = no response) by a pathologist, with senior review for uncertainties. Analysts of pre-operative biopsies were blinded to TRG until statistical analysis. Quantified markers: Densities (cells/mm2) of CD3+, CD8+, CD8+GrzB+ CTLs, γδTCR+ T cells in whole tumor and compartments. MxA+ (type I IFN footprint) cells were quantified in tumor stroma (MxA+ PanCK− cells). Median densities reported included conventional CD3+ and CD3+CD8+ T cells (medians approximately 1065 and 184 cells/mm2, respectively; ranges provided in text). Scoring/stratification models: Biopsy-adapted Immunoscore (ISB) computed as mean ranked percentile of CD3+ and CD8+ densities, categorized into low (<25), intermediate (25–70), and high (>70). For MxA, similar percentile categorization was applied based on stromal MxA density. A combined heat-map approach classified patients by ranked percentiles of CD8 (whole tumor) and MxA (stroma) into six qualitative categories: both high (dark red), high+intermediate (red), high+low (light red), both intermediate (white), intermediate+low (light blue), both low (dark blue). An additional combined IS of CD8 and MxA (IS MxA+CD8+) was also evaluated. Statistics: Group comparisons used Wilcoxon rank-sum tests (p<0.05 significant). Associations used Pearson correlation and unilateral linear-by-linear association tests for ordinal data. Logistic regression used glm() in R (v4.2.1). Analyses performed in GraphPad Prism v9.0.2 and R. Sensitivity, specificity, and diagnostic odds ratio (DOR) were computed for models and single markers. Clinical covariates (T stage, type of neoadjuvant treatment: short-course radiotherapy [SCRT], chemoradiotherapy [CRT], chemotherapy [CT]) were considered in multivariable models. Cohort summary: n=130; female 33%; mean age 63 (25–88). Clinical stage: T1–2 5%, T3 48%, T4 45%; N0 27%, N1–2 72%; M0 85%, M1 15%. Neoadjuvant treatment: SCRT 44%, CRT 52%, CT 4%. TRG: 0 (10%), 1 (12%), 2 (45%), 3 (33%).

- T cell densities vs response: No significant differences in densities of conventional CD3+ or CD8+ T cells, CD8+GrzB+ CTLs, or γδTCR+ T cells between complete responders (TRG 0) and non-responders (TRG 3), or TRG 0 vs pooled TRG 1–3, across whole tumor, tumor epithelium, stroma, or 50 μm peritumoral rim. A non-significant trend was observed for increased CD8+GrzB+ CTLs in responders. - Type I IFN footprint: Significantly higher densities of MxA+ stromal cells were observed in TRG 0 compared with TRG 3 and with pooled TRG 1–3, indicating an association between a stromal type I IFN response and pCR. MxA+ density did not correlate with densities of T cell subsets, including CTLs. - Logistic regression: Models using MxA or CD8 densities alone did not reach significance for predicting tumor regression. The combined heat-map model (CD8 or MxA high) showed a significant association with complete tumor regression (p<0.01). When adding clinical covariates (T stage, neoadjuvant treatment type) in multivariable analysis, the association remained but was attenuated (p<0.01); clinical covariates were not significant, possibly due to limited power. - Stratification patterns: Among TRG 0 patients, all but one were in the high percentile of either CD8 or MxA; none were low in both, and 23.1% were high in both. In TRG 3, only 2.3% were high in both, and approximately 10% were low in both. - Diagnostic performance (Table 2): • CD3: sensitivity 0.38, specificity 0.71, DOR 6.48. • CD8: sensitivity 0.54, specificity 0.73, DOR 16.53. • MxA: sensitivity 0.62, specificity 0.74, DOR 27.52. • ISB (CD3+CD8): sensitivity 0.46, specificity 0.74, DOR 10.53. • IS MxA+CD8+: sensitivity 0.23, specificity 0.76, DOR 2.67. • Heat-map (MxA or CD8 high): sensitivity 0.77, specificity 0.64, DOR 83.33. Overall, stromal MxA density is an independent biomarker associated with pCR, and combining CD8 density (whole tumor) with stromal MxA via a heat-map approach improves prediction compared with ISB or single markers.

The study sought to refine prediction of response to neoadjuvant therapy in rectal adenocarcinoma using pre-operative biopsy biomarkers. While traditional TIL measures and ISB provide prognostic information, they yielded limited predictive power for pCR in this cohort. Incorporating a type I IFN footprint, measured by stromal MxA expression, identified a significant association with pCR independent of T cell density. The lack of correlation between MxA and TIL densities suggests that type I IFN-driven biology contributes information orthogonal to T cell enumeration, potentially reflecting enhanced antigen presentation and CTL priming (e.g., via cGAS-STING and type I IFN signaling). A combined qualitative heat-map stratification leveraging equal weighting of CD8 (whole tumor) and MxA (stroma) improved sensitivity and diagnostic odds ratio, indicating better discrimination of responders than ISB or single-parameter models. Mechanistically, tumors with both abundant CTLs and an active type I IFN milieu may be most permissive to effective antitumor responses after neoadjuvant therapy, whereas either parameter alone may suffice in some patients. These findings align with the concept that type I IFN augments the cancer-immunity cycle and CTL effector function, yet highlight that simple TIL counts alone may not capture the full predictive landscape.

Quantifying a type I IFN footprint (stromal MxA+) in pre-operative rectal cancer biopsies serves as an independent biomarker associated with pathological complete response after neoadjuvant therapy. When combined with total CD8+ T cell density using a straightforward heat-map stratification, predictive performance improves beyond ISB and single-marker approaches, achieving higher sensitivity and diagnostic odds. This two-parameter, biopsy-based method could aid in pre-treatment identification of patients likely to achieve pCR, informing watch-and-wait strategies or treatment tailoring. Future work should validate the heat-map model in independent cohorts, standardize mIF quantification, investigate the cellular sources and subtypes of type I IFN within the tumor microenvironment, and explore γδ T cell subset contributions.

- Single-center retrospective cohort with modest sample size (n=130), limiting statistical power and generalizability. - Lack of external validation for the heat-map stratification approach. - mIF-based quantification may introduce variability (e.g., CD3 intensity variation) compared with chromogenic IHC. - Logistic regression models for single markers (MxA, CD8) did not reach significance; multivariable models may be underpowered. - MxA is an indirect surrogate of type I IFN exposure and does not identify cytokine subtypes or producing cell populations. - γδ T cell analyses lacked subset resolution, which may obscure opposing functional roles. - Potential intracohort dependency of percentile-based scoring may limit transferability without standardization.