Immunogenic cell death by neoadjuvant oxaliplatin and radiation protects against metastatic failure in high-risk rectal cancer

Eligible patients had histologically verified rectal adenocarcinoma that was considered high risk by magnetic resonance imaging (MRI): T2 cases that presented tumor threatening the anal levator muscles, T3 cases that had mesorectal fascia margin of less than 3 mm, T4 cases (organ-infiltrating tumor), or cases that had involved pelvic cavity lymph nodes (N1-2 disease). The full eligibility criteria and evaluation and follow-up procedures have been detailed previously [25]. The present study subpopulation of 50 patients was selected because of the completeness of biobank plasma samples and clinical data.

This single-arm study was conducted to evaluate the safety and efficacy of the intensified neoadjuvant therapy [27]; the exploratory end points of this report were encouraged by findings from analysis of the safety and efficacy data. The induction NACT was given as 2 cycles of the Nordic FLOX regimen (oxaliplatin 85 mg/m² on day 1 and bolus fluorouracil 500 mg/m² and folinic acid 100 mg on days 1 and 2 every second week). The sequential CRT consisted of radiation delivered 5 days per week to a total dose of 50 Gy in 25 fractions to the tumor bed and 48 Gy in 23 fractions to regional lymph nodes, based on 2-/3-dimensional conformal planning models, with concomitant weekly oxaliplatin 50 mg/m² and capecitabine 825 mg/m² twice daily on days of radiotherapy. The neoadjuvant schedule was continuously adjusted according to treatment toxicity from each of the therapeutic agents, as detailed previously [27], principally by reducing the oxaliplatin dose scheduling to avoid compromising radiation delivery, as protraction of the total CRT time theoretically might permit tumor cell repopulation and thereby the survival and dissemination of therapy-resistant cell clones. Radical excision of the residual tumor was planned within 8 weeks after completion of the neoadjuvant treatment. Patients did not proceed to further therapy. Routine blood samples were collected within the standard patient follow-up. Tumor mutational KRAS status was determined for 39 (78%) of the present cases [28]. In addition to the routine diagnostic MRI performed at baseline and 4 weeks after CRT completion, 42 cases (84%) were also assessed after completion of NACT (post-NACT). The tumor boundary was manually contoured by the study radiologist on all tumor-containing axial T₂-weighted images [29] at baseline and post-NACT. Tumor volume (V) change was calculated as ∆V_NACT = [(V_NACT − V_baseline)/V_baseline] × 100. The histologic scoring of the resected tumor specimens (ypTN stage) was recorded as treatment surrogate end point. One patient had disease progression in the pelvic cavity during the neoadjuvant treatment and, therefore, proceeded to palliative surgery; hence, histologic tumor response was not available. The patients included in this report were enrolled from 5 October 2005 through 2 December 2009 and final censoring was done on 8 August 2013 for recording of distant metastasis-free survival (DMFS) and OS.

Patients had plasma sampled over 9 weeks: at baseline, post-NACT, and following CRT completion (post-CRT). The samples were stored at − 80 °C until analysis undertaken in April 2017 (i.e., after 86–138 months of storage). HMGB1 was analyzed with the Human HMGB1® ELISA Kit (Shinto, Kanagawa, Japan), following the manufacturer’s manual, after 1:2 sample dilution and using the mean value of two technical replicates for further analyses. The post-NACT plasma sample was missing for one patient, and this case was accordingly left out from some analyses.

The HMGB1 absolute values were ln-transformed to achieve normality in estimate analyses and expressed as mean ± SEM. Associations between HMGB1 and various patient and disease factors were determined by independent samples t-test, one-way ANOVA, and Pearson correlation test, as appropriate. Alterations in HMGB1 during the neoadjuvant treatment were analyzed by paired samples t-test and described by profile plots with mean ± SEM. DMFS was calculated from the time of study enrollment to the day of metastatic progression, death of any cause, or end of follow-up (a maximum of 5 years after the date of surgery or at final censoring), whichever occurred first. OS was measured from the date of enrollment to death of any cause or final censoring. Associations between HMGB1 and DMFS and OS were analyzed with univariable and multivariable Cox proportional hazards models, and results were presented as hazard ratio (HR) with 95% confidence interval (CI). Due to the limited number of patients and events, potentially predictive factors other than age and sex were omitted from the multivariable models. The Kaplan–Meier method was used to estimate DMFS differences among groups with various CRT oxaliplatin doses, with the log-rank test to examine any statistical significance. All tests were two-sided and P-values less than 0.05 were considered statistically significant. Analyses were carried out using STATA version 15. GraphPad Prism version 7.0 was used for illustrations.

The 21 women and 29 men with median age of 56.5 years (range 30–73) presented with T2 (10%), T3 (58%), or T4 (32%) disease, the majority (82%) with involved lymph nodes and tumor wild-type KRAS status (67% of the 39 analyzed cases; Table 1). As illustrated in Fig. 1, the median plasma HMGB1 showed a modest increase from the absolute value of 1.13 ng/ml (range 0.23–5.25) at baseline through 1.34 ng/ml (range 0.23–5.67) post-NACT and 1.57 ng/ml (range 0.28–4.03) post-CRT, yet with a significant group difference between the baseline and post-CRT values of ln 0.23 ± 0.10 ng/ml (P = 0.029). Baseline HMGB1 was not associated with patient or tumor characteristics (Table 1).

Analysis of tumor gene expression data sets has shown that mutant KRAS status is associated with reduced infiltration of cytotoxic T cells and suppression of the adaptive IFN-γ response in CRC [30], and a possible mechanism for mutant KRAS-driven CRC immune tolerance was recently reported [31]. In this context, the modest neoadjuvant HMGB1 alterations within the entire patient group segregated into distinct patterns when patients were categorized into those harboring tumors with mutant (N = 13) or wild-type (N = 26) KRAS status (the upper panel of Fig. 2). The wild-type group revealed a rise of ln 0.40 ± 0.13 ng/ml (P = 0.006) from baseline to NACT completion, whereas the mutant group had a slight decline. During the succeeding CRT, plasma HMGB1 did not essentially change in either of the groups.

Similarly, HMGB1 profile plots over the neoadjuvant treatment course were compared for patients who over the total follow-up period did (N = 13) or did not (N = 37) experience a DMFS event (the middle panel of Fig. 2). The group with the aimed study outcome (durable freedom from distant recurrence) had an increase in HMGB1 of ln 0.27 ± 0.12 ng/ml (P = 0.036) during NACT, after which the level was consolidated. In contrast, the patient group with metastatic failure had a non-significant decline initially before plasma HMGB1 reverted during the sequential CRT.

For OS (the lower panel of Fig. 2), the small group of patients who died during the follow-up period (N = 7) had a neoadjuvant HMGB1 profile pattern mirroring that of a positive DMFS event, with decline during NACT before reverting during CRT, both alterations at the margin of statistical significance (P = 0.053 and P = 0.049, respectively). Likewise, individuals alive at censoring (N = 43) displayed a pattern closely resembling that of patients without DMFS event. Acknowledging the small number when specifically examining those alive with metastatic disease at censoring (N = 6), it was still notable that the post-NACT HMGB1 was not maintained during the sequential CRT. Besides those pointed at, none of the described alterations within a group was statistically significant.

Since the plasma HMGB1 profiles suggested that alterations during NACT in particular (termed ∆HMGB1) might be indicative of disease outcome, we investigated whether ∆HMGB1 might correlate with early treatment effects (Table 2). This factor correlated moderately well with the change at the same time point of monocyte count (N = 46, R = 0.30, P = 0.040), indicating that an early systemic immune response might have been invoked. However, ∆HMGB1 was not associated with ∆V_NACT (the initial tumor volume response, N = 42) or ypTN response of the surgical specimen (N = 47–48), implying that a biological connection between ICD from the induction NACT and MRI-assessed and histologic tumor effects did not exist.

At censoring, the median time to a DMFS event was 11.0 months (range 3.3–26.4) with 31.4 months (range 8.0–52.6) to death for the deceased (all from the metastatic disease), while the median follow-up time for participants still alive was 77.3 months (range 45.0–94.4). No difference in DMFS or OS was observed between the patient groups with mutant or wild-type KRAS tumor status (not shown). As may be expected from the profile plots, ∆HMGB1 distinguished between patient groups with and without DMFS and OS events (P = 0.041 and P = 0.005, respectively) and metastatic disease leading to death or not (P = 0.048; Table 2). Moreover, as shown by Table 3, the higher the ∆HMGB1, the lower were the risks of metastatic failure and death. Age and sex, which might have affected the long-term outcome, were included in the Cox regression models, but ∆HMGB1 remained an independent predictor of DMFS (HR 0.26, 95% CI 0.11–0.62, P = 0.002) and OS (HR 0.14, 95% CI 0.04–0.51, P = 0.003); other prognostic factors (e.g, the status of extramural venous invasion at baseline MRI or involved lymph nodes in the resected specimen) were omitted in the multivariable models. Post-estimation tests of the proportional hazard assumption were not significant, supporting the validity of the models.

To tentatively clarify whether the ICD induction during NACT would be indicative of the disease outcome in its own capacity, the patients’ treatment status during the sequential CRT was specifically examined. The study subjects were first categorized into those who during CRT received the total oxaliplatin dose that was planned in the protocol (90–100%, N = 18) and those who had a modest reduction (70–89% of the planned dose, N = 8) or a substantial reduction (< 70% of the planned dose, N = 24) because of treatment toxicity. Whereas none of patients (0 of 8) with the modest oxaliplatin dose reduction experienced a DMFS event, 21% of individuals (5 of 24) with a substantial reduction did; of note, 44% of patients (8 of 18) who received the full-planned oxaliplatin dose later had metastatic failure. Hence, the last-mentioned group had considerably worse outcome compared to patients who had the oxaliplatin dose reduced during CRT (P = 0.020; Fig. 3). With respect to the radiotherapy, all of the 50 cases in the present analysis received the total dose of 50 Gy without interruption in radiation delivery, which is critical for the elimination of clonogenic cells within the source of disease dissemination (the radiation target volume).