Predictive immunohistochemical features for tumour response to chemoradiotherapy in rectal cancer

Background Reduced expression of cluster of differentiation (CD) 133 and cyclo‐oxygenase (COX) 2, and increased density of CD8+ tumour‐infiltrating lymphocytes, are associated with a favourable tumour response to preoperative chemoradiotherapy (CRT). This study aimed to evaluate these markers in relation to tumour response after preoperative CRT in two rectal cancer cohorts. Methods Patients with low rectal cancer who underwent radical resection and preoperative short‐term CRT in 2001–2007 (retrospective cohort) and long‐term CRT in 2011–2017 (prospective cohort) were analysed. Pretreatment biopsies were stained immunohistochemically using antibodies to determine CD133 and COX‐2 expression, and increased CD8+ density. Outcome measures were tumour regression grade (TRG), tumour downstaging and survival. Results For 95 patients in the retrospective cohort, the incidence of TRG 3–4 was 67 per cent when two or three immunohistochemistry (IHC) features were present, but only 20 per cent when there were fewer features (P < 0·001). The incidence of tumour downstaging was higher in patients with at least two IHC features (43 versus 22 per cent with fewer features; P = 0·029). The 49 patients in the prospective cohort had similar rates to those in the retrospective cohort (TRG 3–4: 76 per cent for two or more IHC features versus 25 per cent with fewer features, P < 0·001; tumour downstaging: 57 versus 25 per cent respectively, P = 0·022). Local recurrence‐free survival rates in patients with more or fewer IHC features were similar in the retrospective and prospective cohort (P = 0·058 and P = 0·387 respectively). Conclusion Assessment of CD133, COX‐2 and CD8 could be useful in predicting a good response to preoperative CRT in patients with lower rectal cancer undergoing neoadjuvant therapy. Further studies are needed to validate the results in larger cohorts and investigate a survival benefit.


Introduction
Preoperative chemoradiotherapy (CRT) is currently the standard for locally advanced rectal cancer, and aims to lead to tumour regression, downstaging 1 -3 and increased resectability 1 -4 . However, neoadjuvant CRT has been associated with postoperative complications including anastomotic leakage and worse anal sphincter function following surgery 5 -7 . Thus, the investigation of features associated with patient responsiveness is essential to avoid unnecessary treatment. and radiotherapy. However, COX-2 promotes the radioresistance of cancer cells via p38/mitogen-activated protein kinase-mediated cellular antiapoptosis 14 , and selective COX-2 inhibitors reportedly increase the susceptibility of tumours to radiation by inhibiting DNA repair processes 15 . In addition, COX-2 is a powerful angiogenesis-inducible factor 16 , and induces radioresistance in cancer cells efficiently by increasing blood supply. Further, CD8+ TILs have been reported to affect prognosis positively 17 , possibly indicating that the density of CD8+ TILs is a crucial parameter for determining immunocompetence. Some studies 18,19 have also demonstrated that radiotherapy and chemotherapy are more efficient in immunocompetent conditions.
With this background, the research hypothesis investigated in the present study was that increased density of CD8+ TILs and reduced expression of CD133 and COX-2 may predict tumour response to preoperative CRT. The study aimed to assess these features in a retrospective cohort treated with short-term CRT and a prospective cohort that received long-term CRT.

Retrospective cohort
The study was approved by the internal review board at the National Defense Medical College; all patients consented to the collection and study of specimens.
Details of all consecutive patients with stage II-IV rectal cancer undergoing preoperative CRT followed by total mesorectal excision between September 2001 and October 2007 at the National Defense Medical College Hospital, a general hospital affiliated to the medical college in Japan, were reviewed and included.
Preoperative CRT was indicated when the distal margin of the tumour was located below the peritoneal reflection, with a preoperative diagnosis of cT3-4 status, obtained using digital examination, colonoscopy, barium enema and MRI. CT was used to determine the extent of extrapelvic tumour spread. A short-axis value of 5 mm was used as the cut-off point for determining lymph node metastasis: 5 mm or more and less than 5 mm were regarded as metastasis-positive and metastasis-negative respectively. Tumour size was estimated from lateral X-ray images taken during barium enema in the pretreatment stage. During this period, patients were treated using short-term preoperative CRT (20 Gy (5 daily doses of 4 Gy) and tegafur-uracil 400 mg/day for 7 days throughout the period of irradiation), followed by total mesorectal excision. In all patients, two opposing fields were used to treat the entire treatment area, which included the anal canal, primary tumour, mesorectal and presacral lymph nodes, and lymph nodes along the internal iliac vessels, those up to the upper border of the fifth lumbar vertebra, and those at the obturator foramina. Data were obtained from medical records and analysed retrospectively.

Prospective cohort
The study was registered at the University Hospital Medical Education Network clinical trial registry (study ID: UMIN000011993 and UMIN000013486) and received approval from the internal review board. Signed informed consent was obtained from all patients before enrolment. All consecutive patients with stage II-III rectal cancer undergoing long-term preoperative CRT between July 2011 and April 2017, followed by surgery, were included. The neoadjuvant approach was changed because of a temporary increase in reports of adverse events associated with short-term preoperative CRT regimens 5,20,21 . Patients were treated with long-term preoperative CRT (45 Gy (25 daily doses of 1⋅8 Gy), and S-1 and irinotecan), followed by total mesorectal excision. Preoperative CRT was indicated for patients with cT3-4 rectal cancer where the distal margin was located below the peritoneal reflection.
Pretreatment assessment procedures were similar to those used in the retrospective study; however, in the prospective cohort patients with stage IV tumours were excluded. S-1 was administered orally on days 1-5, 8-12, 22-26 and 29-33, based on body surface area (BSA): BSA below 1⋅25 m 2 , 80 mg/day; BSA 1⋅25 to less than 1⋅5 m 2 , 100 mg/day; and BSA 1⋅5 m 2 or above, 120 mg/day. Irinotecan was administered as a continuous intravenous infusion on days 1, 8, 22 and 29 22 . A four-field box technique was used, and the treatment field of radiotherapy was as described previously 23 . The superior margin of the typical irradiation field was set at the level between the fifth lumbar and first sacral vertebra. The inferior margin was set at 3-4 cm below the inferior edge of the primary lesion, as defined principally by a line to the inferior margin of the ischial tuberosity. Lateral margins were 1 cm lateral to the lesser pelvis cavity. The anterior margin was defined as the posterior margin of the pubic symphysis, and the posterior margin was defined as the centre of the sacral bone, as observed from the lateral view. Patient data were collected prospectively, and survival analyses were performed in 2019.

Outcome measures
Outcome measures were tumour regression grade (TRG), tumour downstaging and survival. TRG was assessed semiquantitatively using haematoxylin and eosin-stained slides, as described previously 25 . Briefly, TRG was categorized as follows: TRG 0, no regression; TRG 1, dominant tumour mass with obvious fibrosis in 25 per cent or less of the tumour; TRG 2, dominant tumour mass with obvious fibrosis in 26-50 per cent of the tumour; TRG 3, dominant fibrosis outgrowing the tumour (more than 50 per cent); and TRG 4, only a fibrotic mass with no viable cancer cells. Tumour downstaging was defined as pathological findings of ypT0-2, because all cancers had been estimated to be cT3-4 according to inclusion criteria. Local recurrence-free survival was defined as the time from surgery to local recurrence in patients with rectal cancer, whereas relapse-free survival was defined as the time to the first relapse or death from any cause 26 .
All patients received regular follow-up at the outpatient clinic. Physical examination and blood testing, including testing for carcinoembryonic antigen and carbohydrate antigen 19-9 levels, were performed every 3 months. Contrast CT was done every 6 months. If patients did not attend the clinic, follow-up was by telephone interviews once a year.

Statistical analysis
Categorical variables were compared using the χ 2 or Fisher's exact test, whereas unpaired t tests were used to compare normally distributed continuous variables. The association between pretreatment parameters (immunohistochemistry (IHC) and clinical/pathological variables) and TRG was tested using logistic regression, with a forced entry procedure to determine the hazard ratio (HR) and 95 per cent confidence interval. Cut-off values for clinical/pathological variables (age, tumour size and distance to anal verge) were based on receiver operating characteristic (ROC) curve analysis of the probability of TRG 3 or 4. Survival probabilities were calculated using the Kaplan-Meier method, with comparisons made using the log rank test.
All statistical analyses were performed using the JMP ® 12 software (SAS Institute, Cary, North Carolina, USA). P < 0⋅050 was considered statistically significant.

Results
All clinical/pathological and IHC features for the two cohorts are shown in Table 1.

Retrospective cohort
Of 101 patients reviewed, six were excluded owing to an insufficient volume of archival paraffin-embedded tissue blocks from pretreatment biopsy specimens for IHC staining, leaving for data analysis 95 consecutive patients.
When the number of IHC features of the three markers was assessed (Table 4), retrospective data indicated that the incidence of TRG 3-4 was 67 per cent (20 of 30) in patients with two or three features, but only 20 per cent (13 of 65) in patients with no or one feature (P < 0⋅001) (positive   Values in parentheses are 95 per cent confidence intervals. *Cut-off determined by receiver operating characteristic (ROC) curve analysis. †Well or moderately differentiated tubular adenocarcinoma versus moderately differentiated or mucinous tubular adenocarcinoma, or signet ring cell carcinoma. CD, cluster of differentiation; COX, cyclo-oxygenase; TIL, tumour-infiltrating lymphocyte; CEA, carcinoembryonic antigen; CA, carbohydrate antigen. ‡Logistic multivariable analysis of variables with P < 0⋅050 in univariable analysis. Similarly, prospective data demonstrated that the incidence of TRG 3-4 was 76 per cent (16 of 21) in patients with two or three markers, but only 25 per cent (7 of 28) in those with no or one factor (P < 0⋅001) (PPV, 76 per cent; NPV, 75 per cent). The rate of tumour downstaging was also higher in patients with more features (12 of 21, 57 per cent) than in those with fewer (7 of 28, 25 per cent) (P = 0⋅022) ( Table 4).

Table 3 Predictive values for three markers used in pretreatment biopsy specimens for tumour regression grade and downstaging
Univariable analysis of data from the prospective cohort indicated that two or three IHC markers (P < 0⋅001) and pretreatment tumour size of 46⋅0 mm or less (P = 0⋅015) were significantly associated with a good tumour response (TRG 3-4), whereas other pretreatment parameters failed to show significance. In multivariable analysis of these two variables, both the IHC factor (HR 9⋅13, 95 per cent c.i. 2⋅44 to 41⋅00; P < 0⋅001) and pretreatment tumour size (HR 4⋅38, 1⋅04 to 22⋅50; P = 0⋅044) were independently associated with a good tumour response.

Survival analysis
Of 87 patients in the retrospective cohort who had an R0 resection, 26 died a median of 47⋅0 (range 2⋅1-149⋅5) months after surgery. The median duration of follow-up for the remaining 61 patients was 78⋅2 (29⋅0-191⋅8) months. The difference between 5-year local recurrence-free survival rate in patients with two or three features (100 per cent) and those with no or one feature (88 per cent) was of borderline significance (P = 0⋅058). In addition, the 5-year relapse-free survival rate did not differ in these two subgroups (80 versus 68 per cent respectively; P = 0⋅501). Among the 44 patients with an R0 resection in the prospective cohort, four died a median of 48⋅2 (range 12⋅3-66⋅3) months after surgery. The median duration of follow-up for the remaining 40 patients was 38⋅8 (21⋅3-87⋅4) months. Neither the 5-year local recurrence-free survival rate (100 per cent for patients with 2 or 3 features versus 96 per cent for those with 0 or 1 feature; P = 0⋅387) nor the relapse-free survival rate (100 versus 76 per cent respectively; P = 0⋅085) was significantly different.

Discussion
In recent years, several studies have evaluated the potential of molecular biomarkers to predict tumour response to CRT. A review of the literature, including several studies reporting on gene expression profiles associated with tumour response to CRT, concluded that there was little consistency with respect to the selected genes for determining CRT sensitivity 27 . This variation could be ascribed to the use of different methodologies. Some studies 28 -33 also examined genetic/epigenetic changes or protein expression levels, although research is still at an early stage. Nevertheless, large-scale validation studies of predictive markers are necessary before incorporating such methodologies into future clinical practice. Although a number of promising predictive classifiers have been proposed, successful validation has not yet been achieved.
In the present study, findings from the retrospective cohort indicated that expression of CD133 and COX-2, and the density of CD8+ TILs were significant predictors of tumour response to preoperative CRT. However, the prospective study, although limited by sample size, disclosed excellent predictive value for the three markers, except for CD133 immunoreactivity. However, the interval between preoperative CRT and surgery was longer in the long-term preoperative CRT cohort, and this may have had a negative influence on the radioresistant property of CD133+ cancer cells.
The present study also showed that the increased density of CD8+ TILs is a prominent predictor of tumour downstaging. A previous study 9 demonstrated that a high density of CD8+ TILs in pretreatment biopsy specimens correlated strongly with enhanced CD8+ lymphocyte aggregation at the tumour margin after preoperative CRT. This could suggest that CRT-induced CD8+ lymphocytes may respond to cancer cells located at the invasive front, resulting in tumour downstaging. Conversely, increased CD8+ lymphocyte aggregation at the tumour margin after CRT was rarely observed in cancers with a low density of CD8+ TILs in pretreatment biopsy specimens 9 , which prevented downstaging. Patients with rectal cancer who did not have preoperative CRT also showed a positive association between the density of CD8+ TILs in biopsy specimens and the level of CD8+ lymphocyte aggregation at the tumour margin. However, the level of CD8+ lymphocyte aggregation at the tumour margin in patients who had preoperative CRT was strikingly higher than that in those who had surgery alone 9 . These data suggest that patients with a high density of CD8+ TILs in pretreatment biopsy specimens demonstrate tumour downstaging as a benefit of preoperative CRT, owing to enhanced immunoreactivity after preoperative CRT at the invasive tumour margin.
Overall, the combined analysis of the three markers could be a powerful tool for identifying the chemoradiosensitivity of patients with lower rectal cancer, consistent with the research hypothesis. Of note, a previous report 8,9 also demonstrated excellent interobserver agreement for evaluating CD133 and COX-2 immunostaining, and CD8+ TIL density: 85⋅4 per cent (κ = 0⋅68), 92⋅9 per cent (κ = 0⋅76) and 89⋅9 per cent (κ = 0⋅79) respectively. This study has some limitations. First, the preoperative CRT regimens adopted and the intervals between preoperative CRT and surgery differed between the prospective and retrospective studies. Second, the characteristics of biopsy specimens may not represent those of whole tumours.
A prospective trial involving patients with only two or three positive findings for treatment using preoperative CRT (study ID: UMIN000026306) is currently ongoing, to obtain more robust findings. The results of this trial will be necessary to determine the clinical usefulness of these proposed IHC markers.