Absence of myeloperoxidase and CD8 positive cells in colorectal cancer infiltrates identifies patients with severe prognosis

Abstract

Colorectal cancer (CRC) infiltration by cells expressing myeloperoxidase (MPO) or CD8 positive T lymphocytes has been shown to be independently associated with favorable prognosis. We explored the relationship occurring between CD8+ and MPO+ cell CRC infiltration, its impact on clinical-pathological features and its prognostic significance in a tissue microarray (TMA) including 1,162 CRC. We observed that CRC showing high MPO+ cell infiltration are characterized by a prognosis as favorable as that of cancers with high CD8+ T cell infiltration. However, MPO+ and CD8+ CRC infiltrating cells did not synergize in determining a more favorable outcome, as compared with cancers showing MPO^high/CD8^low or MPO^low/CD8^high infiltrates. Most importantly, we identified a subgroup of CRC with MPO^low/CD8^low tumor infiltration characterized by a particularly severe prognosis. Intriguingly, although MPO+ and CD8+ cells did not co-localize in CRC infiltrates, an increased expression of TIA-1 and granzyme-B was detectable in T cells infiltrating CRC with high MPO+ cell density.

Abbreviations

CRC

colorectal cancer

IHC

immunohistochemistry

MMR

mismatch repair

MPO

myeloperoxidase

NK

natural killer

PTL

peritumoral lymphocytes

TMA

tissue microarray.

Introduction

Despite advanced multimodal treatment regimens, CRC is still the fourth most common cause of cancer-related death worldwide in both genders.¹ Besides specific mutations^2,3 and genomic and epigenomic instability, ^4,5 the interaction of malignant cells with the tumor microenvironment has been demonstrated to play critical roles in cancer development and progression. ^6,7

Infiltration by immunocompetent cells and cytokine and chemokine gene expression significantly influence CRC outcome.^8-11 Indeed, the analysis of CRC “immune contexture”¹² has been suggested to outperform the prognostic significance of tumor node metastasis (TNM) staging and might contribute to personalized treatment decisions.^13,14 In particular, CRC infiltration by CD8+ T cells has repeatedly been shown to be associated with favorable clinical outcome.^8,10,15,16

On the other hand, the role of innate immune system has not been studied in comparable detail. Controversial data have been reported regarding CRC infiltration by natural killer (NK) cells^17-19 and macrophages.^20-23 Recently, we have observed that CRC infiltration by cells expressing MPO, an enzyme typically released by activated granulocytes, represents an independent favorable prognostic factor, as emerging from the analysis of a large cohort of patients.²⁴

Experimental models have suggested that antitumor effects of granulocytes are largely dependent on the interaction with T cells.^25-27 Furthermore, most recently, early stage lung cancer-associated neutrophils have been shown to be able to stimulate T cell responses in humans.²⁸

Prompted by these reports, in this study we have explored the relationship occurring between CD8+ and MPO+ cell infiltration in CRC, its impact on clinical-pathological features and its prognostic significance.

Materials and Methods

Tissue microarray construction

Construction of the TMA has previously been described.¹⁶ Briefly, 1,162 unselected, non-consecutive, primary CRCs, sampled between 1985 and 1998, were collected from the tissue biobank of the Institute of Pathology, University Hospital Basel, performing translational research with the approval of the Ethics Committee Beider Basel (EKBB), in compliance with ethical standards and patient confidentiality. Samples were formalin-fixed and paraffin-embedded CRC tissue blocks were prepared. Tissue cylinder of a 0.6 mm diameter were punched from morphologically representative areas of each donor block and brought into one recipient paraffin block (30 × 25 mm), using a semi-automated tissue arrayer. Each punch was made from the center of the tumor so that each TMA spot consisted of at least 50% tumor cells.

Clinicopathological features

Clinicopathological data for the patients included in the TMA were collected retrospectively in a non-stratified and non-matched manner.¹⁶ Annotation included patient age and gender, tumor location (n = 1,150), pT and pN stage (n = 1136 and n = 1,118, respectively), grade (n = 1,138), histologic subtype (n = 1,162), vascular invasion (n = 1,136), border configuration (n = 1,135), presence of peritumoral lymphocytic (PTL) inflammation at the invasive tumor front (n = 1,137), and overall survival (n = 1,162). Tumor border configuration and PTL inflammation were evaluated according to Jass using the original H&E slides of the resection specimens corresponding to each TMA punch.²⁹ Based on mismatch repair (MMR) status, defined according to MLH1, MSH2, and MSH6 expression³⁰ the TMA included 975 MMR-proficient tumors and 187 MMR-deficient tumors.

Median event-free follow-up time was of 68 mo. Data on local recurrence and presence of distant metastases were available for 412 and 418 patients, respectively (Table 1).

Table 1.

Association of MPO+ and CD8+ cell infiltration with clinical-pathological features in 1,162 CRC

		MPO low CD8 low		MPO high CD8 low		MPO low CD8 high		MPO high CD8 high
Characteristics		n = 790	(68 %)	n = 117	(10 %)	n = 206	(18 %)	n = 49	(4 %)	p-value*
Age	years (mean)	70		71		69		72		0.656
Gender	Female	396	(50.1)	64	(54.7)	122	(59.2)	25	(51.0)	0.047
	Male	394	(49.9)	53	(45.3)	84	(40.8)	24	(49.0)
Tumor location	Left-sided	516	(65.3)	83	(70.9)	124	(60.2)	25	(51.0)	0.28
	Right-sided	264	(33.4)	33	(28.2)	81	(39.3)	24	(49.0)
Histologic subtype	Mucinous	66	(8.4)	11	(9.4)	11	(5.3)	3	(6.1)	0.395
	Non-mucinous	724	(91.6)	106	(90.6)	195	(94.7)	46	(93.9)
pT stage	pT1–2	124	(15.7)	33	(28.2)	57	(27.7)	12	(24.5)	<0.0001
	pT3–4	649	(82.2)	83	(70.9)	141	(68.4)	37	(75.5)
pN stage	pN0	376	(47.6)	54	(46.2)	133	(64.6)	40	(81.6)	<0.0001
	pN1–2	383	(50.4)	59	(50.4)	64	(31.1)	9	(18.4)
Tumor grade	G1–2	686	(86.8)	101	(86.3)	170	(82.5)	36	(73.5)	0.12
	G3	89	(11.3)	14	(12.0)	28	(13.6)	12	(24.5)
Vascular invasion	Absent	550	(69.6)	85	(72.6)	155	(75.2)	40	(81.6)	0.023
	Present	225	(28.5)	30	(25.6)	43	(20.9)	8	(16.3)
Tumor border	Pushing	261	(33.0)	40	(34.2)	100	(48.5)	29	(51.2)	<0.0001
	Infiltrating	513	(64.9)	75	(64.1)	98	(47.6)	19	(38.8)
PTL inflammation	Absent	633	(80.1)	94	(80.3)	149	(72.3)	29	(59.2)	0.018
	Present	143	(18.1)	21	(17.9)	49	(23.8)	19	(38.8)
Local recurrence	Absent	144	(18.2)	33	(28.2)	50	(24.3)	15	(30.6)	0.0003
	Present	131	(16.6)	18	(15.4)	18	(8.7)	3	(6.1)
Distant metastasis	Absent	219	(27.7)	43	(36.8)	62	(30.1)	19	(38.8)	0.011
	Present	60	(7.6)	8	(6.8)	7	(3.4)	0	(0.0)
Microsatellite stability	Deficient	120	(15.2)	17	(14.5)	34	(16.5)	16	(32.7)	0.256
	Proficient	670	(84.8)	100	(85.5)	172	(83.5)	33	(67.3)
5-year survival rate	(95%CI)	51.3%	(47.6–55.0)	69%	(60.4–78.9)	71%	(64.7–78.0)	64.3	(51.2–80.8)	<0.0001

^*Correlation between the MPO^low / CD8^low subgroup with the other 3 subgroups. Percentage may not add to 100% due to missing values of same variables.

Immunohistochemistry

Standard indirect immunoperoxidase procedures were used for immunohistochemistry (IHC; ABC-Elite, Vector Laboratories, Burlingame, CA). Briefly, slides were dewaxed and rehydrated in distilled water. Endogenous peroxidase activity was blocked using 0.5% H₂O₂. Sections were first incubated with 10% normal goat serum (DakoCytomation, Carpinteria, CA) for 20 min and then with primary antibody at room temperature. Primary antibodies used were specific for MPO (clone 59A5 Novocastra, Newcastle, UK), CD8 (clone C8/144B, DakoCytomation, Switzerland), Granzyme B (clone NCL⁻L-GRAN-B, Novocastra, Newcastle, UK), and TIA-1 (clone IM 2550, Immunotech, Marseille, France). Subsequently, sections were incubated with peroxidase-labeled secondary antibody (DakoCytomation) for 30 min at room temperature. To visualize target antigens, sections were immersed in 3-amino-9-ethylcarbazole plus substrate-chromogen (DakoCytomation) for 30 min, and counterstained with Gill's hematoxylin.

MPO+, CD8+, granzyme B+, and TIA-1+ tumor infiltrating cells were counted for each punch (approximately one high power [20x] field). All immunohistochemical reading was performed by trained research fellows (R.D. or G.S.) and data were independently validated by two additional investigators (L.To. and C.H.).^16,24

Immunofluorescence studies

For immunofluorescence staining, human colorectal cancer sections were blocked with 5% goat serum diluted in phosphate-buffered saline (PBS) containing 0.3% Triton X-100 for one hour at room temperature and then incubated with rabbit polyclonal anti-myeloperoxidase (Ventana Medical Systems, 1:50) and mouse monoclonal anti-CD8 (BD Biosciences, 1:100) reagents for one hour at 37°C. Slides were then washed with PBS and incubated for one hour at RT with goat anti-mouse and anti-rabbit Ig, Alexa Fluor 488- or 546-conjugated antibodies (Invitrogen, 1:1000). During the last 10 min of incubation, 4′, 6-diamidino-2-phenylindole (DAPI, Invitrogen, 1:500) was added for nuclear counterstaining. Sections were analyzed under a fluorescence microscope (Olympus BX61, Olympus Switzerland) and images were captured with 10x and 20x magnification using a digital camera and AnalySIS software (Soft Imaging System GmbH).

Statistical analysis

Threshold values used to classify CRC with low or high MPO+ or CD8+ infiltration were obtained, as previously reported, by means of receiver operating characteristic (ROC) curves and regression trees.^16,24 Subsequently, specific cut-off values were set at 60 cells/TMA-punch for MPO+ and at 10 cells/TMA-punch for CD8+ infiltration.

Four subgroups characterized by combined, dichotomized, MPO+, and CD8+ cell infiltration (MPO^low/CD8^low, MPO^high/CD8^low, MPO^low/CD8^high, and MPO^high/CD8^high) were identified. Chi-Square or Fisher's exact tests were used to determine the correlation between MPO+ and CD8+ infiltration and clinicopathological features. Survival data were analyzed by the Kaplan–Meier method and compared by log rank test. To prove the stability and reproducibility of the data obtained, we repeated the log rank test in 100 randomly selected subsets of 700 patients each. The assumption of proportional hazards was verified for both markers by analyzing the correlation of Schoenfeld residuals and the ranks of individual failure times. Any missing clinicopathological information was assumed to be missing at random. Subsequently, MPO+ and CD8+ cell infiltration data were entered into multivariate Cox regression analysis and hazard ratios (HR) and 95% confidence intervals (CI) were used to determine prognostic effects on survival time. Statistical analyses were performed using R software (Version 2.15.2, www.r-project.org).

Results

Patient characteristics

The cohort under investigation included 1,162 patients with CRC with a median age of 71 years (range 30–96), as previously described.^16,24 According to the TNM-classification, most of the tumors were of pT3/pT4 stage (n = 910), with lower numbers of pT1/pT2 (n = 226) cancers. Among the malignancies under evaluation, 603 (53%) were pN0, whereas 515 were pN1–2. While 993 (87%) tumors were of grade 1–2 and 705 (62%) showed infiltrating tumor border configuration, in the majority of cases (n = 830: 73%) vascular invasion was absent. MMR-proficient tumors accounted for a large majority of the cancers under investigation (n = 975: 83%). Within the follow-up period, median disease-specific survival was 68 mo (range 0–152), while mean 5-year-survival was 56 mo (95% confidence interval: 54–59 mo) (Table 1).

Clinicopathological features associated with MPO+ and CD8+ CRC infiltration

Representative pictures of low and high MPO+ and CD8+ cell infiltration in CRC are shown in Figure 1. Based on regression tree analysis, cut-off values of 60 and 10, respectively, were selected to identify CRC with low or high MPO+ or CD8+ cell infiltration.^16,24 Data related to the four subgroups identified by MPO+ and CD8+ cell density (MPO^low/CD8^low, MPO^high/CD8^low, MPO^low/CD8^high, and MPO^high/CD8^high), as evaluated for each clinicopathological feature under investigation, are reported as absolute numbers and percentages in Table 1. Dropouts were due to loss of punches during TMA staining preparation or missing information, and usually accounted for <5% of data.

Figure 1.

High and low CRC infiltration by MPO+ and CD8+ cells. CRC samples were stained with MPO and CD8+ specific reagents. Tumor punches are representative of low (A and B) and high (C and D) density of CRC infiltration by MPO+ (A and C) and CD8+ (B and D) cells, respectively. Magnification: 20×.

Tumors with MPO^low/CD8^low infiltration were characterized by a significantly (p < 0.01) higher pT stage, as compared to CRC displaying MPO^high/CD8^low, MPO^low/CD8^high, and MPO^high/CD8^high cell density. In contrast, no significant differences were observed among tumors showing MPO^high/CD8^low, MPO^low/CD8^high, and MPO^high/CD8^high infiltrates. These data suggest an impact of MPO+ cell infiltration “per se” on primary cancer size, irrespective of CD8+ T cell infiltrate.

On the other hand, effects of CD8+ T cell infiltration were clearly detectable on pN stage. Indeed, CRC with MPO^high/CD8^low infiltrate were characterized by a significantly more frequent pN1–2 stage, as compared with tumors with MPO^low/CD8^high infiltrates (p = 0.0006). Cancers with MPO^low/CD8^low infiltrate did not significantly differ from those with MPO^high/CD8^low infiltrate, nor did those with MPO^high/CD8^high infiltrates significantly differ from those with MPO^low/CD8^high infiltration.

Incidence of local recurrences and distant metastases, as assessed in a subgroup of CRC (see above), was significantly more frequent in CRC with MPO^low/CD8^low infiltration than in the other cohorts under investigation. These did not significantly differ from each other. On the other hand, PTL inflammation was found to be more frequently detectable only in CRC with MPO^high/CD8^high infiltration, as compared to all other subgroups.

Tumor grade and vascular invasion did not significantly (p < 0.01) differ in CRC with different immune cells infiltration profiles.

The nature of tumor border (pushing vs. infiltrating) has been reported to impact on CRC prognosis. ³¹ Indeed, infiltrating tumor border, associated with poor survival, was detected with significantly (p = 0.006) higher frequency in CRC with MPO^high/CD8^low than in tumors with MPO^low/CD8^high infiltrate, thus suggesting effects related to CD8+ lymphocyte infiltration. Accordingly, no significant differences were observed between tumors with MPO^low/CD8^low or MPO^high/CD8^low infiltrates and MPO^low/CD8^high or MPO^high/CD8^high infiltration.

Prognostic significance of MPO^low/CD8^low, MPO^high/CD8^low, MPO^low/CD8^high, and MPO^high/CD8^high CRC infiltration

Survival rates at 5 years were significantly different depending on the nature of immune infiltrate (Table 1). Most importantly, Kaplan–Meier plots clearly indicated that prognosis was significantly more severe in patients bearing cancers with MPO^low/CD8^low infiltration (Fig. 2). In contrast, comparable long-term survival could be observed in patients bearing tumors characterized by MPO^high/CD8^low, MPO^low/CD8^high, and MPO^high/CD8^high infiltrates. Multivariate analysis confirmed these data since MPO^low/CD8^low cell density in CRC was independently associated with unfavorable prognosis, as compared to the other three subgroups after adjusting for several known prognostic factors such as age, sex, T stage, N stage, tumor grade, vascular invasion, tumor border configuration, and microsatellite stability (Table 2).

Figure 2.

Effects of MPO+ and CD8+ tumor infiltration on overall survival in patients with CRC. Kaplan–Meier overall survival curves were designed according to MPO+ and CD8+ tumor infiltration in patients with CRC (n = 1131 ). Cut-off values established by regression tree analysis, as previously reported,^16,24 were 60 cells/punch for MPO and 10 cells/punch for CD8 cell infiltration. Cumulative effects of tumor infiltration by MPO+ and CD8+ cells were also explored. A: Kaplan–Meier overall survival curves of tumors with low MPO+ and low CD8+ cell infiltration (MPO CD8 - line) in comparison to all other three groups (MPO^high/CD8^low, MPO^low/CD8^high, and MPO^high/CD8^high, continuous line) (p < 0.0001). B: MPO CD8+ – line refers to tumors with MPO^low/CD8^low infiltration (284 deaths observed in 767 patients). MPO CD8 −+ line refers to tumors with MPO^low/CD8^high infiltrates (63 deaths observed in 200 patients). MPO CD8 ++ line refers to tumors with MPO^high/CD8^high infiltration (15 deaths observed in 48 patients) and MPO CD8 +− line refers to CRC with MPO^high/CD8^low cell infiltration (35 deaths observed in 116 patients) (p < 0.0001).

Table 2.

Multivariate Hazard Cox regression survival analysis

	HR (+95% CI)	p-values
MPOlow/CD8low vs. all other combinations	0.35 + 0.34	0.002
Age	1.03 + 0.005	<0.0001
Gender	0.68 + 0.09	<0.0001
pT stage	1.81 + 0.08	<0.0001
Tumor grade	1.27 + 0.13	0.056
pN stage	1.91 + 0.06	<0.0001
Vascular invasion	1.46 + 0.11	0.0004
Tumor border configuration	1.46 + 0.12	0.0016
Microsatellite stability (deficient vs. proficient)	1.80 + 0.15	<0.0001

Multivariate analyses showing Hazard Ratios and p-value for all CRC (n = 1062 less than 1420 due to missing values) conferred by CD8+ and MPO density, age, sex, tumor size, tumor grade, vascular invasion, tumor border configuration, and microsatellite stability.

Expression of activation markers in CD8+ T cells infiltrating CRC with high MPO+ density

To evaluate interactions potentially occurring between MPO+ and CD8+ cells infiltrating CRC, we analyzed the expression of CD8+ T cell markers, as related to MPO+ infiltration in CRC. We observed that in the presence of high MPO+ cell infiltrate, CRC infiltrating CD8+ cells expressed significantly more frequently granzyme B and TIA-1,¹⁶ as compared with cells from CRC with low MPO+ infiltration (Fig. 3A). However, immunofluorescence analysis of CRC sections from surgically excised tumors (n = 5) clearly indicated that MPO+ and CD8+ CRC infiltrating cells did not co-localize (Fig. 3B).

Figure 3.

Granzyme B and TIA+ cell infiltration of CRC is increased in tumors with high MPO+ cell infiltration. Absolute numbers of granzyme B+ (A) and TIA-1+ (B) CRC infiltrating cells, as identified by IHC in CRC TMA punches displaying MPO^low/CD8^low, MPO^high/CD8^low, MPO^low/CD8^high or MPO^high/CD8^high cell infiltration. (C) Shows a representative immunofluorescence staining of CRC infiltrating MPO+ and CD8+ cells. Magnification 20x, scale bar: 50 µm.

Discussion

There is a resurgent interest in the scientific community regarding the role of granulocytes in tumor immunobiology.^28,32,33 Indeed, tumor infiltration by granulocytes has usually been associated with poor prognosis.²⁵ However, experimental models in the past have suggested an antitumor role elicited through the cross-talk between granulocytes and T cells.^26,27 More recently, granulocyte polarization, similarly to macrophages, has been described³³ and the ability of granulocytes maturing within tumor tissues to promote lymphocyte activation has been reported.²⁸ Furthermore, CRC cells have been shown by us and others to produce factors promoting granulocyte chemoattraction and survival.^34,35 Finally, we have observed that MPO+ cell infiltration of CRC represents an independent marker of favorable prognosis.²⁴ Importantly, MPO+ cells appear to represent a subset of CD66b+ CRC infiltrating cells.²⁴

Our data provide novel insights into the prognostic relevance of the interaction between innate and adaptive immune system in CRC microenvironment, and raise a number of points of potentially high-clinical impact.

First, CRC showing high MPO+ cell infiltration are characterized by a prognosis as favorable as that of cancers with high CD8+ T cell infiltration. This effect is also detectable in cancers with low CD8+ infiltrate and is reflected by a lower pT stage as compared with CRC with low MPO+ and CD8+ cell infiltration. Remarkably, Rao et al. have reported that CD66b+ cell infiltration in CRC is associated with severe prognosis in a cohort of 229 patients from southern China.³⁶ This discrepancy could be attributed to the use, as marker of CRC infiltrating cells, of CD66b instead of MPO. Indeed, MPO appears to be expressed in a subset of CD66b+ CRC infiltrating cells in our cohort,²⁴ as previously observed by others in a different cancer type.³⁷ Alternatively, different genetic backgrounds in the cohorts of patients under investigation, e.g. from East Asia or Western Europe, could also play a role. Moreover, importantly, differences in gut microbiomes of patients from disparate geographic areas might also be involved. ^38,39

Second, MPO+ and CD8+ CRC infiltrating cells do not appear to synergize in determining a more favorable outcome, as compared with cancers showing MPO^high/CD8^low or MPO^low/CD8^high infiltrates. It is tempting to speculate that CRC infiltration by these cell types might reflect different phases of immune response to CRC. While innate immune system MPO+ cells might contribute to the control of tumor outgrowth and progression in earlier stages of the disease, adaptive immune response, elicited by CD8+ T cells, might be required in more advanced phases. Accordingly, concomitant CD3+ T cell infiltration and HLA class I expression by tumor cells have recently been identified as favorable prognostic markers in resected liver metastases deriving from CRCs.⁴⁰

Third, our data contribute to the identification of a subgroup of CRC with a particularly severe prognosis, characterized by MPO^low/CD8^low tumor infiltration. Probably, patients bearing these cancers could be eligible for adjuvant treatments following surgery, irrespective of conventional TNM staging.

Fourth, irrespective of their independent prognostic significance, a crosstalk between MPO+ and CD8+ CRC infiltrating cells appears to be likely to occur, as suggested by the increased expression of TIA-1 and granzyme-B in T cells infiltrating CRC with high MPO+ cell density. Importantly, the expression of both these markers by CRC infiltrating cells has repeatedly been reported to be associated with an improved prognosis.¹⁶ However, we could clearly observe that MPO+ and CD8+ cells do not co-localize while infiltrating CRC. These data, in agreement with immunohistochemical and clinical data, may suggest that CRC infiltration by MPO+ and CD8+ cells do not obviously depend on each other, although they might concomitantly occur. Thus, they might reflect different phases of immune responses induced by both tumor cells and microbial products from the gut lumen, including cytokine and chemokine production, promoting immune cells recruitment and activation.

A major limitation of our study is represented by its retrospective character. However, this analysis of well-characterized cohorts of patients paves the way for the currently ongoing planning of more ambitious prospective investigations. On the other hand, the study population refers to CRC surgically treated between 1985 and 1998, thus prior to a widespread use of neoadjuvant treatment regimens in rectal cancer. Therefore, while our results may not be fully representative of the current clinical treatment, they are more likely to faithfully mirror CRC immunobiology, in the absence of chemo-irradiation treatments.

Molecular mechanisms underlying the antitumor effects of MPO+ CRC infiltrating cells remain largely undefined and further research is clearly warranted to clarify them. Nevertheless, our data contribute to the identification of myeloid cells as key players in antitumor immune responses.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

Financial support for this study was provided by the SNF Grants Nr. PP00P3–133699, Nr. 31003A-122235 and Nr. 320030–120320, the Italian Association for Cancer Research (AIRC) IG Grant Nr. 10555, the Lazio Regional Agency for Transplantation and Related Diseases, and the Dr. Hans Altschueler Stiftung.