Prognostic significance of HLA class I and II expression in patients with diffuse large B cell lymphoma treated with standard chemoimmunotherapy

Abstract

Loss of tumor cell human leukocyte antigen (HLA) is an immune escape mechanism for malignancies. However, the effect of low HLA class I or class II expression in diffuse large B cell lymphoma (DLBCL) treated with chemoimmunotherapy with the monoclonal antibody rituximab is largely unknown. We retrospectively analyzed samples and other data from 144 patients with DLBCL who were newly diagnosed in our institution and treated with standard R-CHOP therapy. We used antibodies against pan-HLA class I and pan-HLA class II molecules to assess HLA expression and its effect on prognosis. In a multivariate analysis, loss of HLA class II expression was a significantly independent adverse factor for progression-free survival (PFS; hazard ratio 2.3; 95 % confidence interval 1.2–4.6; P = 0.01). Although HLA class I loss of expression did not correlate with prognosis, the combination of HLA class I⁺ with either low peripheral lymphocyte count or CD3⁺ lymphocyte count was an adverse prognostic factor for PFS. Loss of HLA class II is an International Prognostic Index (IPI)-independent adverse factor for PFS in patients with DLBCL treated with standard therapy. However, in contrast to other solid cancers, HLA class I loss was not solely a prognostic factor in DLBCL.

Electronic supplementary material

The online version of this article (doi:10.1007/s00262-016-1883-9) contains supplementary material, which is available to authorized users.

Introduction

Diffuse large B cell lymphoma (DLBCL) is a heterogeneous disease. Gene expression analysis has identified four prognostic signatures for DLBCL: proliferation, human leukocyte antigen (HLA) class II, lymph nodes, and germinal center differentiation [1]. The importance of HLA class II and lymph node signatures suggests that host response to lymphoma cells is a crucial survival determinant [2].

In immune response, cytotoxic T cells react to antigens presented by HLA class I molecules, which are expressed on virtually all nucleated cells; HLA class II molecules are expressed on only a few types of antigen-presenting cells such as B cells, macrophages, and dendritic cells. Class I and II molecules provoke helper T cell responses, which assist the cytotoxic T cell responses. Therefore, loss of HLA in malignancy cells may directly promote evasion of from immune surveillance.

In DLBCL patients treated with chemotherapy without rituximab (RTX), loss of lymphoma cell HLA-DR (an HLA class II molecule) was reported to be a significant adverse prognostic factor [3–6]. However, few studies have investigated the predictive effect of HLA-DR [7] or pan-HLA class II loss in patients treated with RTX-based regimens, and limited data are available on the prognostic impact of HLA class I expression or loss in DLBCL.

Standard treatment for DLBCL is chemoimmunotherapy with the monoclonal antibody RTX and the chemotherapeutic agents of doxorubicin, cyclophosphamide, vincristine, and prednisolone (R-CHOP).

In the current study, we investigated HLA class II expression and its predictive effect in patients with DLBCL when uniformly treated with R-CHOP. We also evaluated a newly developed antibody that recognizes HLA class I molecules and is suitable for immunostaining of formalin-fixed tissue [8].

Patients and methods

Patients

We retrospectively analyzed data from 144 patients with DLBCL who were initially diagnosed and treated with R-CHOP therapy in our institute between January 2003 and December 2010. All diagnoses were confirmed following review by hematopathologists in our institute using the morphological and immunohistochemistry criteria defined in the World Health Organization classification, 4th edition [9]. Patients were selected based on the availability of their clinical information and histological material for new immunohistochemistry analysis in this study. We excluded cases of primary central nervous system (CNS) or testicular lymphoma, which require different treatment strategy. Data were collected from medical records and our institutional database. The following baseline clinical characteristics were collected: sex, age, disease stage, presence of bulky disease, B symptoms, performance status, number of extranodal lesions, lactate dehydrogenase (LDH) level, peripheral absolute lymphocyte count (ALC), and peripheral CD3⁺ T cell (CD3) count.

Definitions

All patients were staged according to the Ann Arbor system [10] and were also assigned International Prognostic Index (IPI) scores [11]. Overall survival (OS) was measured as the time from the initial diagnosis until death from any cause. Progression-free survival (PFS) was the time from initial diagnosis until disease progression/relapse or death from any cause.

Treatment policy

Patients with early stage (stage I-II) disease were treated with three to six cycles of CHOP chemotherapy and three to eight cycles of RTX, followed by radiation therapy if required. Patients with advanced-stage disease were treated with six to eight cycles of R-CHOP, followed by radiation therapy for patients with bulky disease only. Patients who failed to respond to primary therapy or relapsed after primary therapy received salvage chemotherapy; if chemo-sensitive and aged younger than 65 years, patients received high-dose chemotherapy with autologous stem cell transplantation.

Immunohistochemistry

Immunohistochemistry was performed on formalin-fixed paraffin-embedded tissues and was carried out using a panel of monoclonal antibodies as previously reported [12]. Briefly, for antigen retrieval, 4-μm sections were cut from each paraffin block, deparaffinized, and autoclaved at 121 °C in citrate buffer, pH 6.0, for 10 min. Antibodies against the following antigens were used: CD3 (PS1, 25×; Novocastra, Newcastle upon Tyne, UK), CD5 (4C7, 100×; Novocastra), CD10 (56C6, 100×; Novocastra), CD20 (L26, 200×; Dako, Glostrup, Denmark), Bcl-2 (124, 100×; Dako), Bcl-6 (PG-B6p, 20×; Dako), MUM1 (MUM1p, 200×; Dako), and Ki-67 (MIB-1, 100×; Dako). In addition, pan-HLA class II (CR3/43, 200×; DAKO) and pan-HLA class I (EMR8-5, 800×; Hokuto, Japan, and Abcam, Cambridge, UK) were used in this study. The pan-HLA class I antibody was recently developed and can recognize HLA-A, HLA-B, and HLA-Cw [8]. To validate its recognition of pan-HLA class I expression, we used an anti-β2-microglobulin antibody (D8P1H, 1:1000, Cell Signaling Technology, Danvers, MA, USA). Cross-tabulation was undertaken to confirm that HLA class I-positive cases were also positive for anti-β2-microglobulin and that HLA class I-negative cases were also negative for anti-β2-microglobulin (Supplementary Table 1).

An autostainer was used for staining with the standard polymer method (Dako Autostainer Plus). All samples were judged to be positive for CD20 and negative for CD3. Immunoreactivity for CD5, CD10, Bcl-2, Bcl-6, and MUM1 was judged positive if ≥30 % of tumor cells were stained, according to our previous study [12]. For Ki-67, ≥80 % was defined as a high labeling index. We selected the area in each slide with the greatest staining, manually counted 500 nuclei, and calculated the percentage of positive cells. To classify each case as having either a germinal center B (GCB) phenotype or a non-GCB phenotype, we used a panel of three antigens (CD10, Bcl-6, and MUM1) and the algorithm reported by Hans et al. [13].

HLA-DR was judged to be positive or negative according to a previous study [14], using an internal positive control for reactive small lymphocytes or accessory cells. The percentages of HLA class I⁺ tumor cells among all tumor cells were semi-quantified as <5, 5–25, 25–50, 50–75, 75–95, or >95 %. We analyzed the associations between prognoses and each cutoff point, as no significant cutoff point had been reported previously.

Fluorescence in situ hybridization (FISH) analysis

To detect allele loss from the HLA lesion, we developed FISH probes to cover a long HLA lesion. We performed double-target FISH using in-house HLA BAC probes for the loci of HLA-A (RP11-836G11) and HLA-B (RP11-35J4) and a centromere 6 probe (Abbott, Abbott Park, IL, USA). HLA-A and HLA-B probes were labeled using the Histology FISH Accessory kit (Dako) as previously reported [15]. Nuclear staining was performed using 4′,6-diamidino-2-phenylindole (DAPI). We visualized stained sections using a four-color fluorescence microscope (BIOREVO; Keyence Corporation, Osaka, Japan).

Statistical analysis

Categorical data were analyzed using Fisher’s exact test (two-sided). The Mann–Whitney U and t tests were applied, where appropriate, to assess median or mean differences between groups. OS and PFS were calculated using the Kaplan–Meier method. Surviving patients were censored at the last follow-up visit. Risk factors for OS and PFS were evaluated using Cox proportional hazard models. Statistical analyses were performed using the SPSS version 11.0 statistical package (SPSS Inc, Tokyo, Japan). P < 0.05 was considered significant. Cross-tabulation analysis was performed using Fisher’s exact test.

Results

Patient, disease, and treatment characteristics

The main clinical and treatment characteristics of patients are shown in Table 1. Median age of patients was 62 years (range: 24–88 years). Numbers of patients for each IPI score were as follows: IPI score was score 0–1, n = 78; score 2, n = 33; score 3, n = 20; and score 4–5, n = 13. Seventy-two (50 %) patients received standard R-CHOP therapy without radiation therapy, and the remaining 72 patients received standard R-CHOP therapy followed by radiation therapy. The median follow-up period was 58 months (range 7.4–112.4 months).

Table 1.

Patient, disease, and treatment characteristics

	Total
	n = 144	%
Age, median (range)	62 (24–88)
Sex, male/female	73/71	51/49
IPI factors
Age >60 years	82	57
Elevated LDH	63	44
PS ≥2	17	12
Stage III–IV	48	33
Number of extranodal lesion ≥2	23	16
IPI
0–2	111	77
3–5	33	23
B symptoms	8	6
Bulky disease	11	8
Peripheral ALC, median (range)	1400 (200–3900)
Peripheral CD3+ lymphocyte count*, median (range)	996 (111–2254)
Initial treatment
R-CHOP followed by RT	72	50
R-CHOP without RT	72	50

ALC absolute lymphocyte count, IPI International Prognostic Index, LDH lactate dehydrogenase, PS performance status, RT radiation therapy

* n = 136

Immunohistochemistry

Representative staining results for HLA class I and HLA class II expression are shown in Fig. 1. HLA class II staining was clearly divided into positive (n = 115, 80 %) and negative (n = 29, 20 %) samples. No samples exhibited the cytoplasmic and Golgi staining pattern observed in Hodgkin lymphoma [16, 17]. Of the 144 patients, 62 (43 %) had samples that were at least 95 % HLA class I⁺; other cases showed heterogeneous down-regulation of HLA class I. Semi-quantified HLA class I⁺ stains were <5 %, n = 11; 5–25 %, n = 22; 25–50 %, n = 7; 50–75 %, n = 17; 75–95 %, n = 25; and >95 %, n = 62. Considering the majority of the cases were positive at a level of >95 %, and no cutoff level was determined by the prognosis, we arbitrary used the 95 % level as the level to determine positivity for further analysis. The ALC or CD3 count also did not affect prognosis. Other results for the 144 samples included CD10⁺: 40 (28 %); Bcl-6⁺: 96(67 %); MUM1⁺: 71(49 %); CD5⁺: 14 (10 %); Bcl-2⁺: 97 (67 %); GCB phenotype: 61 (42 %) and non-GCB phenotype: 83 (58 %).

Fig. 1.

Immunohistochemical staining for HLA class I and HLA class II. HLA class I was positive in all tumor cell membranes (a), whereas it was negative in most tumor cells (b). HLA class II was positive in all tumor cell membranes (c), whereas it was negative in most tumor cells (d). In the HLA⁻ panels (b, d), some small lymphocytes, which are HLA class I⁺ or HLA class II⁺, are visible as internal controls

Positive HLA class I expression correlated with positive for β2-microglobulin expression. In cases negative for β2-microglobulin, staining was weak for HLA class I staining in 24 cases and strong in two cases. In cases positive for β2-microglobulin, staining was weak for HLA class I in 20 cases and strong in 10 cases. In cases strongly positive for β2-microglobulin, staining was weak for HLA class I in 13 cases and strong in 36 cases. Fisher’s exact test for count data showed strong correlation; the P value was <0.01.

FISH analysis successfully detected signals in 44 of 144 samples; no allele loss was detected when compared with the centromere probe (data not shown).

Prognostic factors for PFS and OS

The nine immunohistochemical markers [12] and three clinical factors shown in Table 2 were assessed for their effects on PFS. Univariate analyses revealed that five factors—loss of HLA class II, Bcl-2 positivity, high IPI score (3–5), low ALC/monocyte count, and c-MYC and Bcl-2 coexpressions, were significant adverse factors. As shown in the Kaplan–Meier curves for HLA class II⁺ and HLA class II⁻ PFS (Fig. 2a, b), the HLA class II⁻ group had significantly shorter PFS (log-rank tests, P = 0.03). Multivariate analysis of three among these five factors, that were found to be significant in the univariate analyses, showed that loss of HLA class II and significantly affected PFS (hazard ratio [HR] 2.3; 95 % confidence interval [CI] 1.2–4.6; P = 0.01) as did Bcl-2 positivity and high IPI score. However, loss of HLA class I did not correlate with PFS for any tested cutoff values evaluated (Fig. 2a, b).

Table 2.

Univariate and multivariate analysis of progression-free survival

Variable	Univariate analysis			Multivariate analysis
	HR	95 % CI	P	HR	95 % CI	P
HLA class I
Positive	1
Negative	0.96	0.5–1.8	0.9
HLA class II
Positive	1			1
Negative	2.0	1.1–3.9	0.03	2.3	1.2–4.6	0.01
Bcl-2
Negative	1			1
Positive	2.9	1.3–6.5	0.01	2.7	1.2–6.1	0.02
CD5
Negative	1
Positive	0.9	0.3–2.7	0.9
CD10
Negative	1
Positive	1.1	0.6–2.2	0.8
Bcl-6
Negative	1
Positive	0.5	0.3–1.01	0.06
c-MYC
Negative	1
Positive	1.05	0.5–2.02	0.9
c-MYC+Bcl-2
Without co-expression	1
Co-expression	1.958	1.05–3.67	0.04
MUM1
Negative	1
Positive	1.2	0.6–2.2	0.6
Cell of origin
GCB	1
Non-GCB	1.5	0.8–2.7	0.2
Ki-67
<80 %	1
≥80 %	1.4	0.7–2.6	0.3
IPI
0–2	1			1
3–5	2.7	1.4–5.0	0.002	2.6	1.4–4.9	0.003
Peripheral ALC
>1000	1
≤1000	1.3	0.6–2.6	0.6
ALC/monocyte
>1000, ≤630*	1
Rest	1.88	1.02–3.47	0.04
Peripheral CD3+ lymphocyte count**
>700	1
≤700	1.2	0.6–2.4	0.6

ALC absolute lymphocyte count, CI confidence interval, GCB germinal center B, HLA human leukocyte antigen, HR hazard ratio, IPI International Prognostic Index

* Cases with peripheral ALC > 1000/μl and monocyte count ≤630/μl. ** n = 136

Fig. 2.

Kaplan–Meier curves for progression-free survival (PFS) by HLA positivity (i.e., >95 % of tumor cells were stained in IHC) (a, b) and by HLA class I positivity (i.e., >95 % of tumor cells were stained in IHC) and absolute lymphocyte count (ALC) (c) or CD3⁺ lymphocyte count (d). a The HLA class II⁻ group had significantly shorter PFS than the HLA class II⁺ group (P = 0.03). b The HLA class I⁺ and HLA class I⁻ groups did not significantly differ in PFS (P = 0.9). Loss of HLA class I did not correlate with PFS for any tested cutoff values evaluated. c Among HLA class I⁺ patients, those with low ALC (≤1000/μl) had significantly shorter PFS than those with high ALC (>1000/μl; P = 0.002). However, among HLA class I⁻ patients, PFS did not significantly differ between those with low and high ALC (P = 0.1). d Similarly, among HLA class I⁺ patients, those with low CD3⁺ lymphocyte counts (≤700/μl) had significantly shorter PFS than those with high counts (>700/μl; P = 0.02); whereas among HLA class I⁻ patients, PFS did not significantly differ between those with low and high CD3⁺ lymphocyte counts (P = 0.2)

The same nine immunohistochemical markers and three clinical factors shown in Table 2 were also assessed for their effects on OS. No significant adverse factors were found.

Comparing characteristics of HLA class II⁺ and HLA class II⁻ groups

We examined the differences in background characteristics between the HLA class II⁺ and HLA class II⁻ groups (Table 3). The HLA class II⁻ group tended to have fewer HLA class I⁺ cases (P = 0.06) and more cases with involved extranodal lesions (P = 0.06) than the HLA class II⁺ group: this observation was not statistically significant, however. Furthermore, we found no significant differences between the background characteristics of the HLA class II⁺ and HLA class II⁻ groups, including the median peripheral blood ALC and CD3⁺ lymphocyte count.

Table 3.

Comparison of characteristics between HLA class II positive and negative cases

	HLA class II				P
	Positive		Negative
	n = 115	%	n = 29	%
Age, median (range)	62 (24–88)		64 (41–80)		0.5
Sex, male/female	58/57		15/14
IPI factors
Age >60 years	58	50	18	62	0.3
Elevated LDH	49	43	14	48	0.5
PS ≥2	11	10	6	21	0.1
Stage III–IV	37	32	11	38	0.7
Number of extranodal lesion ≥2	18	16	5	17	0.8
IPI score
0–2	89	77	22	76	0.8
3–5	26	23	7	24
Nodal/extranodal	62/52	54/46	10/19	34/66	0.06
B symptoms	7	6	1	3	1
Bulky disease	7	6	4	14	0.2
Response to initial treatment
CR or PR	107	93	25	86	0.3
Refractory	8	7	4	14
Immunophenotype
CD5 positive	12	10	2	7	0.7
CD10 positive	30	26	10	34	0.6
Bcl-2 positive	78	68	19	66	0.8
Bcl-6 positive	78	68	18	62	0.7
MUM1 positive	59	51	12	41	0.4
GCB/non-GCB	48/67	42/58	13/16	45/55	0.8
Ki-67 >80	36	31	12	41	0.4
HLA class I positive	64	56	10	34	0.06
Immunological state
Peripheral ALC, median (range)	1400 (200–3900)	1400 (210–3600)	0.8
Peripheral CD3+lymphocyte count, median (range)	984* (129–2254)	920** (111–1965)	0.6

ALC absolute lymphocyte count, CR complete remission, GCB germinal center B, HLA human leukocyte antigen, IPI, International Prognostic Index, LDH lactate dehydrogenase, PR partial remission, PS performance status

* n = 108, ** n = 28

Impact of ALC and CD3 count

We found no effect of ALC or CD3 counts on PFS. We carried out a subgroup analysis in the HLA class I⁺ and HLA class I⁻ groups. In the HLA class I⁺ group, low ALC (≤1000/μl) was a significant adverse factor for PFS (P = 0.002; Fig. 2c) when the cutoff value of HLA class I positivity was defined as >95 %. Low ALC did not affect PFS in the HLA class I⁻ group (P = 0.1, Fig. 2c).

Similarly, low CD3 count (≤700/μl) was a significant adverse factor for PFS (P = 0.02, Fig. 2c) in the HLA class I⁺ group but not in the HLA class I⁻ group (P = 0.2, Fig. 2d).

To evaluate the prognosis of subgroups based on the combination of HLA class I positivity and ALC, multivariate analysis was performed. When IPI score, HLA class II positivity, and Bcl-2 positivity were included in the multivariate analysis of the HLA class I⁺ group, patients with low ALC had a significantly higher PFS hazard ratio than HLA class I⁺ patients with higher ALC or HLA class I⁻ patients (HR 2.3; 95 % CI 1.04–5.0; P = 0.04) (Table 4). Similarly, the HLA class I⁺ group with low CD3 count (≤700/μl) had a significantly shorter PFS and higher PFS HR than cases with higher CD3 counts or the HLA class I⁻ group (HR 2.4; 95 % CI 1.1–5.2; P = 0.03).

Table 4.

Multivariate analysis of progression-free survival

Variable	HR	95 % CI	P
HLA class I
Negative or positive with high ALC	1
Positive with low ALC	2.3	1.04–5.0	0.04
HLA class II
Positive	1
Negative	2.3	1.2–4.5	0.01
IPI
0–2	1
3–5	2.5	1.3–4.8	0.004
Bcl-2
Negative	1
Positive	2.4	1.1–5.6	0.03

ALC absolute lymphocyte count, CI confidence interval, HR hazard ratio, IPI International Prognostic Index

Discussion

The immune systems can recognize tumor cells and modulate their growth and progression. Indeed, T lymphocytes that can specifically recognize tumor-associated antigens (TAA) are commonly found in cancer patients [18, 19]. Among these TAA-specific T lymphocytes, cytotoxic CD8⁺ T cells recognize tumor cells by the interaction between the HLA class I-TAA complex on tumor cells and the T cell receptor on T lymphocytes, whereas cytotoxic CD4⁺ T cells recognize tumor cells by the interaction between HLA class II-TAA complex on tumor cells and the T cell receptor on T lymphocytes. Therefore, loss of HLA expression in malignant cells directly permits escape from immune surveillance and can thus be an adverse prognostic factor. In this study, we investigated the loss of HLA class II and HLA class I expression from DLBCL cells and evaluated the effect of this loss on prognosis of patients treated with R-CHOP therapy.

Our findings demonstrated that HLA class II loss is an IPI-independent adverse factor for PFS in DLBCL treated with R-CHOP therapy. For DLBCL located in immune-privileged sites, such as CNS lymphoma, or testicular lymphoma, HLA class II loss frequently leads to large genomic deletions of the HLA class II gene cluster [20]. In DLBCL types not associated with immune-privileged sites, a common mechanism of HLA class II loss is transcriptional repression of the HLA class II gene cluster, including the master transcription factor, class II transactivator (CIITA) [21]. HLA class II affects cellular immune response by presenting peptides to CD4 cells and enhancing CD8⁺ T cell stimulation. Specifically, fewer tumor-infiltrating CD8⁺ T cells were detected in HLA class II⁻ samples than in HLA class II⁺ samples [3]. Moreover, cytotoxic CD4⁺ T cells capable of directly recognizing and killing tumor cells presenting surface HLA class II through T cell receptor interactions have recently been reported [22, 23]. Hence, HLA class II loss is a mechanism by which tumor cells can escape immune surveillance during or after chemotherapy.

In the pre-RTX era, loss of HLA-DR or HLA class II was reportedly an IPI-independent adverse prognostic factor [3–6]. However, with RTX treatment, prognostic results for HLA-DR or HLA class II loss have been inconclusive; although gene expression profiling analysis showed HLA-DR loss to be an adverse factor [7], immunohistochemical analysis did not [24]. Our results support the earlier gene profiling analysis, and our immunohistochemical analysis shows that HLA class II loss remains an IPI-independent adverse factor for PFS in DLBCL treated with RTX.

We evaluated anti-HLA class I molecule expression for the first time in cases of malignant lymphoma treated with R-CHOP therapy. Cases that were positive for the HLA class I molecule correlated with those cases with positive for β2-microglobulin, thus validating the antibody use. In addition, we examined hemizygous HLA gene loss that cannot be detected by the antibodies using FISH analysis. The cases identified had not lost the allele and in our cohorts that did not contain CNS and testicular lymphomas; hemizygous HLA was not a common mechanism for loss of HLA expression. However, we cannot rule out the possibility that these cases included the mutated gene of β2-microglobulin gene as reported previously [25].

Furthermore, we may have overlooked the copy number neutral loss of heterozygosity on HLA genes, a limitation of our study. Microsatellite analysis or by SNP-analysis can reveal the loss of heterozygosity (LOH), although DNA quality and quantity of the formalin-fixed paraffin-embedded tissue (FFPE) samples hampered the analysis. Moreover, we did not have access to the necessary germ line tissue that is necessary as a control, as these highly polymorphic regions do not allow for use of pooled DNA or human genetic variation data.

We found that HLA class I expression level was not a prognostic marker for patients were treated with R-CHOP therapy, a finding that contrasts with the previous reports showing HLA class I loss to be an adverse factor in renal [26], bladder [26], and esophageal cancers [27, 28]. This conflict can be partly explained by the triggering of direct killing or antibody-dependent cellular cytotoxicity (ADCC) function by NK cells or during RTX therapy to compensate for the negative effects of HLA class I loss. NK cells recognize HLA class I⁻ tumor cells as “missing self,” in contrast to recognition via HLA class I by cytotoxic T cell (CTL). Therefore, CTL and NK cells complement each other in immune surveillance. Moreover, DLBCL cells lacking HLA class I expression can reportedly escape from CTL recognition but are vulnerable to NK cell attack [25]. HLA class I down-regulation in DLBCL cells is also associated with sensitivity to RTX-mediated ADCC by NK cells [29]. Taken together, HLA class I⁻ DLBCL cells can escape from CTL recognition but become more vulnerable or sensitive to ADCC by NK cells, thus overcoming the negative effects of HLA class I loss with RTX treatment.

Whether low ALC together with low CD3 count is a poor prognostic factor remains unclear. A Japanese study in DLBCL patients showed that low ALC was significantly associated with worse prognosis among patients treated with CHOP alone, but not among those treated with R-CHOP [30]. ALC only affects non-germinal center-type DLBCL [31]. Because both lymphocytes and T cells recognize HLA class I, we predicted and confirmed that low ALC would affect prognosis when combined with HLA class I expression. In this case, neither ALC nor HLA class I expression was a prognostic factor; however, if combined, low ALC with high HLA class I expression was significantly associated with worse prognosis. Our study showed that low ALC or CD3 counts were significant adverse prognostic factors only for HLA class I⁺ cases. Our study also showed the HLA class I⁺/low-ALC group had a significantly higher PFS HR than patients with high ALC, negative HLA class I, or both. These results suggest that HLA class I⁺ cases, which are less sensitive to direct attack or antibody-dependent cytotoxic ADCC by NK cells, are more dependent on T cell immune surveillance.

Recently, the impact of the lymphocyte/monocyte ratio at diagnosis [32] or during therapy [33] for DLBCL has been recognized as a more reliable prognostic marker. We confirmed that the low lymphocyte/monocyte ratio is a poor prognostic marker at diagnosis. Monocyte/macrophages are potent antigen-presenting cells (another type of HLA class II + cell), again highlighting the importance of HLA class II molecules.

The present study was not intended to detect the most significant prognostic factor, but to address the impact of HLA class I and class II molecules on prognosis. Confounding factors are present, and a single antibody such as Bcl-2 is not sufficient to fully characterize the heterogeneity of DLBCL biology [34]. We found that IPI remains the most significant factor for prognosis. Each parameter such as stage I-II disease alone did not affect the prognosis. This finding validates the use of IPI; however, it should be noted that IPI was not associated with class I or class II expression as shown in Table 1.

We could not identify any factors that significantly affected OS. Although we cannot speculate on the reason for this, the paucity of events for analysis and/or response and survival after relapse involving heterogeneous treatment should be considered.

Major limitations of our study include its retrospective and single-institution design. Our results require confirmation in prospective multicenter trials. The finding that both HLA class I expression and absolute lymphocyte and CD3 count in the peripheral blood correlated with prognosis should be regarded as an exploratory result.

Another limitation is the possibility that we might have overlooked copy number neutral LOH or point mutations in the HLA loci. Analysis of the highly polymorphic region requires germ line DNA, an approach which requires further investigation.

In conclusion, our study demonstrated that HLA class II loss is an IPI-independent adverse factor for PFS in DLBCL treated with R-CHOP therapy. However, HLA class I loss detected by immunostaining was not a significant prognostic factor. Our study also showed that the HLA class I⁺/low ALC group had significantly poorer prognosis than patients with high ALC and negative HLA class I, or both. These results suggest that lymphoma–host immune balance may be important for prognosis prediction or immunotherapy strategy choice.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ADCC

Antibody-dependent cellular cytotoxicity

ALC

Absolute lymphocyte count

CIITA

Class II transactivator

CTL

Cytotoxic T cell

DLBCL

Diffuse large B cell lymphoma

FFPE

Formalin-fixed paraffin embedded

FISH

Fluorescent in situ hybridization

HLA

Human leukocyte antigen

HR

Hazard ratio

IHC

Immunohistochemistry

LOH

Loss of heterozygosity

OS

Overall survival

PFS

Progression-free survival

R-CHOP

RTX plus doxorubicin, cyclophosphamide, vincristine, and prednisolone

RTX

Rituximab

TAA

Tumor-associated antigen

Compliance with ethical standards

Conflict of interest

The authors have no conflicts of interest to declare. This research was approved by the Institutional Review Board of the National Cancer Center (2014-185).