Good prognosis for follicular lymphoma with estrogen receptor α‐positive follicular dendritic cells

Abstract Follicular lymphoma (FL) has a meshwork of follicular dendritic cells (FDCs). We previously demonstrated the presence of estrogen receptor alpha (ERα)+CD23+ FDCs in grades 1‐2 FL. The significance of FDCs as a prognostic factor in FL remains unknown. The current study aimed to compare clinicopathological features, including prognosis, between FL with and without ERα+ FDCs. This study evaluated the clinicopathological significance of ERα expression in 70 FL patients by immunostaining. The presence of ERα mRNA on FDCs from 5 FL patients was confirmed by CD21/ERα double staining (immunohistochemistry and in situ hybridization). We defined patients with frequent ERα expression as the ERαhigh group and those with infrequent ERα expression as the ERαlow group. Thirty‐two patients were assigned to the ERαhigh group (45.7%), and 38 patients were assigned to the ERαlow group (54.3%). Both overall survival (OS) and progression‐free survival (PFS) were significantly better in the ERαhigh group than in the ERαlow group (OS, log‐rank, P = .0465; PFS, log‐rank, P = .0336). Moreover, high ERα expression on FDCs was an independent prognostic factor for OS in both the univariate ([hazard ratio] HR, 0.163; P = .0260) and multivariate (HR, 0.050; P = .0188) analyses and for PFS in both the univariate (HR, 0.232; P = .0213) and multivariate (HR, 0.084; P = .0243) analyses. ERα mRNA expression was detected in CD21+ FDCs within the neoplastic follicles of FL patients. In conclusion, a neoplastic follicular microenvironment with ERα‐positive FDCs might affect the grade and presence of the follicular pattern of FL and improve patient prognosis.


| INTRODUCTION
Follicular lymphoma (FL) is a germinal center (GC)-derived lymphoma 1,2 that is frequently followed by an indolent clinical course. 3,4 As prognostic factors, the Follicular Lymphoma International Prognostic Index (FLIPI) 5 and FLIPI2 6 are commonly used. Recently, m7-FLIPI, 7 which includes the mutation status of 7 genes, and progression of disease within 2 years (POD24 or POD2), 8 which is defined as relapse or progression of FL within 24 months (2 years) after diagnosis, have been used as prognostic factors. In addition, various proteins of neoplastic cells themselves, such as CD5, 9 GNA13, 10 and FOXP-1, 11 are considered prognostic factors. In the microenvironment, it is unclear whether tumor-associated macrophages and programmed cell death-1 + cell infiltration are associated with prognosis. 12,13 Concerning follicular dendritic cells (FDCs), a tight CD21 + FDC meshwork frequently promotes transformation into large B-cell lymphoma, 3 although the extent of Ki-M4p + or the CD23 + FDC meshwork is not associated with treatment outcome or survival time. 14 We previously demonstrated that the number of ERα + cells was positively correlated with the width of the CD23 + FDC meshwork in both nonneoplastic GC and neoplastic follicle, and that estrogen receptor alpha (ERα) + CD23 + FDCs supported the neoplastic follicular microenvironment of grades 1-2 (G1-2) FL, but not G3 FL, suggesting the possibility of the usefulness of antiestrogen therapy against FL. 15 This study first compared clinicopathological features and prognosis between FL patients with more and less frequent ERα + FDCs and investigated the significance of ERα + FDCs in the FL microenvironment.

| Patients and samples
We investigated 70 tissue samples from FL patients before treatment.
Pathological diagnoses were determined at Yamagata University Hospital and Yonezawa City Hospital in Japan and Harbin Medical University Cancer Hospital in China between 2003 and 2018 using the rituximab-containing regimen. Specific FL variants and subtypes, including testicular FL, in situ FL, duodenal-type FL, pediatric-type FL, and primary cutaneous follicle center lymphoma, were excluded from this study. The FL specimens were classified as G1-2 (n = 35), G3A (n = 22), or G3B (n = 13) in accordance with the WHO Classification Revised Fourth Edition. 16 The histological pattern of these cases was classified as follicular pattern (n = 52) and another pattern (n = 18). 16 Tissues were fixed in 10% neutral-buffered formalin for 6 to 12 hours at room temperature, embedded in paraffin (formalin-fixed paraffinembedded; FFPE), and used for hematoxylin-eosin staining, immunostaining (IHC), and in situ hybridization (ISH

| Evaluation of ERα IHC
IHC was performed as previously described, 17 and a specific antibody for ERα (EP1; rabbit IgG, DAKO, Agilent Technologies, Santa Clara, California) was used. IHC was performed using an Autostainer Link 48 system (Agilent Technologies). ERα reactivity was estimated as previously described. 15 Briefly, cells positive for ERα were counted in five neoplastic follicles for each case. In FL specimens with diffuse proliferation, positive reactions were counted in five high-power fields indicated that the majority of ERα + FDCs simultaneously expressed CD23. 15 Therefore, we estimated ERα/HPF as a semi-quantitative marker of the width of ERα + CD23 + FDC meshwork in this study.

| IHC of other proteins
To confirm the immunophenotype of FL, IHC was performed using Technologies). The reactivity of CD10, CD20, BCL2, BCL6, and MUM1 was considered positive if more than 30% of the neoplastic cells were positive. The area of the neoplastic follicle was determined from images and estimated as the gross area by ImageJ as previously described. 15,18,19 The CD23 + FDC pattern was divided into 2 groups: none/dim or focal/marginal 4 /diffuse.

| Double staining (IHC and ISH) of serial sections
To identify ERα-expressing cells, double staining (IHC/ISH) was performed using the FFPE serial sections of the 5 FL patients with only follicular pattern and the 5 FL patients with diffuse pattern, and one uterine endometrioid carcinoma patient (as a positive control) as previously described, 20 with minor modifications. Briefly, in one section, CD21 immunostaining was performed using an Autostainer Link 48 system (DAKO, Agilent Technologies) and visualized with 3,3 0 -diaminobenzidine. For the ISH of other sections, digoxigenin-labelled riboprobes encoding human ERα (human estrogen receptor 1, 854 bp, nucleotides 3001-3845) and an ISH kit (Genostaff, Tokyo, Japan) were used. Probes were hybridized for 24 hours at 60 C, incubated with anti-digoxigenin-alkaline phosphatase (Roche Diagnostics, Indianapolis, Indiana) for 1 hour at room temperature, and visualized with BCIP/ NBT Substrate System (DAKO, Agilent Technologies) for 24 hours at room temperature. The visual images for CD21 IHC and ERα ISH were overlaid by using ImageJ. 18,19 Sense probes were used as negative controls. Nuclear staining was performed with Kernechtrot.

| Statistical analysis
To compare the clinicopathological characteristics of FL, the χ 2 test, Fisher's exact test or the Mann-Whitney test was used. The endpoint of overall survival (OS) was defined as the time from diagnosis until allcause death, and the endpoint of progression-free survival (PFS) was defined as the time from diagnosis until relapse due to FL. 9,10 Kaplan-Meier curves of OS and PFS were drawn and compared by the log-rank test. To propose prognostic factors, univariate and multivariate Cox proportional hazards regression models were used. Statistical analyses were performed using JMP version 14 (SAS Institute, Tokyo, Japan). Differences with P values <.05 were considered significant in each analysis.

| IHC and ISH analyses of ERα
In the IHC analysis, ERα expression was detected mainly in FDCs in neoplastic follicles, as previously reported. 15 OS, PFS, and ERα/HPF were compared between the ERα high-expression (ERα high ) and ERα low-expression (ERα low ) groups. As a result, most of the significant differences in OS and PFS were observed when FL patients were divided into 2 groups: ≥ 3 ERα/HPF and <3 ERα/HPF. Therefore, patients with ≥3 ERα/HPF were as assigned to the ERα high group, and the other patients were assigned to the ERα low group (<3 ERα/HPF).
Thirty-two patients (32/70, 45.7%) were assigned to the ERα high group, and 38 patients (38/70, 54.3%) were assigned to the ERα low group. ERα/HPF were 10.2 ± 6.45 in the ERα high group (Figure 1a) and 0.20 ± 0.54 in the ERα low group (P < .0001). In sequential IHC/ISH, the expression of ERα mRNA was detected in FDCs within the neoplastic follicles of 5 FL patients (Figure 1b,c). ERα mRNA was not detected in diffuse proliferation area of 5 FL patients ( Figure 1d) and CD21 + FDC meshwork hardly/did not exist in this area.

| Relationship between the frequency of ERα expression and the clinicopathologic features of FL
The comparison of each clinicopathological feature between the ERα high and ERα low groups is shown in Table 1. Regarding the histological grade, there was more G1-2 FL in the ERα high group than in the ERα low group and more G3 FL in the ERα low group than in the ERα high group (P < .0001). Moreover, the ERα high group had a higher frequency of the follicular proliferative pattern than the ERα low group (P < .0001). Immunohistochemically, the ERα high group had a higher F I G U R E 1 Immunohistochemistry (IHC) of the estrogen receptor alpha (ERα) protein and in situ hybridization (ISH) of ERα mRNA in follicular lymphoma (FL). IHC shows that ERα is expressed on follicular dendritic cells (FDCs) in a neoplastic follicle of G1-2 FL (a). The area in the red dashed frame is 0.1 mm 2 . The number of ERα-positive cells in the frame is 23. Therefore, the number of ERα-positive cells/highpower field (HPF; ×40 magnification, 0.159 mm 2 ) was 36.57, and formalin-fixed paraffin-embedded tissue sections from FL patients were used for double staining of CD21 (IHC) and ERα (ISH). The left panel (b) shows hybridization with an antisense riboprobe, and the right panel (c) shows hybridization with a sense riboprobe (control) on serial sections. The expression of ERα mRNA was detected in FDCs within the neoplastic follicle of FL patients. CD21 + FDCs were stained with 3,3 0 -diaminobenzidine (brown), and ERα mRNA + cells were stained with NBT-BCIP (dark blue); double stained CD21 + /ERα mRNA + FDCs are shown as arrows in (b). The expression of ERα mRNA was not detected in diffuse proliferation area of FL patients, although CD21 + FDCs was slightly detected (d; Arrowhead). Nuclear staining was performed with Kernechtrot. Bars, 50 μm frequency of the CD23 + FDC meshwork than the ERα low group (P < .0001). In the initial therapy, there was a more watchful wait in the ERα high group than in the ERα low group (P = .0077). However, other features, such as age, sex, FLIPI, t(14;18)(q32;q21), the expression of other markers (CD10, BCL2, BCL6, and MUM1), the initial therapy regimen, and the rate of complete response to initial therapy were not significantly different between the ERα high and ERα low groups.

| Comparison of survival between the ERα high and ERα low groups
Kaplan-Meier curves of OS and PFS are shown in Figure 2. The ERα high group had a significantly better prognosis for both OS and PFS than the ERα low group (OS, log-rank, P = .0465; PFS, log-rank, P = .0336).

T A B L E 1
ERα expression on follicular dendritic cell (FDC) in follicular lymphoma (70 cases)

| DISCUSSION
There was a higher frequency of G1-2 FL, the follicular pattern and CD23 + FDCs in the ERα high group than in the ERα low group in this study ( Table 1). The FDC immunophenotype of G1 FL resembles that of the light zone (LZ). 21 We previously described that ERα + CD23 + FDCs were distributed both in the LZ of GC and in the neoplastic follicle from G1-2 FL, unlike G3 FL, 15 indicating that the microenvironment of the ERα high group is similar to the LZ of GC supported by ERα + CD23 + FDCs.
To the best of our knowledge, this study was the first to reveal an association between high ERα expression and a good prognosis in FL. Established prognostic factors, including FLIPI 5 and FLIPI2, 6 are used to judge the pre-treatment status. Two additional prognostic factors were recently reported: m7-FLIPI 7 is used to determine the pretreatment status, and minimal residual disease 22  and could be either a good or poor prognostic factor of transformation. 3,23 These results suggest that the relation between FDCs and prognosis remains ambiguous. However, we first described frequent cancers. 25 Furthermore, tamoxifen is also considered a G proteincoupled estrogen receptor (GPER) agonist and suppresses the proliferation of Jurkat cells, a T-ALL cell line expressing GPER. 26 GPER promotes the survival of mantle cell lymphoma cells. 27