IRF4 rearrangement may predict favorable prognosis in children and young adults with primary head and neck large B‐cell lymphoma

Abstract Purpose Large B‐cell lymphoma with IRF4 rearrangement (LBCL, IRF4+) has been recently recognized as a specific entity that is frequently associated with young age and favorable prognosis. However, whether the good outcome of the disease is due to IRF4+ or other factors remains obscure. We thus analyzed 100 young patients with primary head and neck LBCL to see the clinicopathologic correlates of IRF4+. Methods The histopathology, immunophenotype, IRF4 status of the tumors, and clinical data were reviewed. Results Twenty‐one tumors were diagnosed as LBCL, IRF4+, which were more frequently associated with a follicular growth pattern, medium‐sized blastoid cytology, germinal center B‐cell‐like, and CD5+ phenotype, compared with IRF4− ones. While most of the patients received chemotherapy with or without radiation, eight IRF4+ patients received mere surgical resection of the tumor and exhibited excellent outcome. IRF4+ cases featured a significantly higher complete remission rate, and better survivals compared with IRF4− ones. Multivariate analysis confirmed IRF4+ correlates with a better survival. Conclusion Our work confirmed the unique clinicopathologic features of LBCL, IRF4+, and disclosed for the first time the independent favorable prognostic impact of IRF4+. These findings may further unravel the heterogeneity of LBCL occurring in youth, and aid in risk stratification and tailoring the therapeutic strategy.


| INTRODUCTION
Large B-cell lymphoma with IRF4 rearrangement (LBCL, IRF4+) was recently recognized as a specific entity characterized genetically by IRF4 translocation (mostly with IG genes), 1,2 and the disease was introduced by the revised 4th edition of the World Health Organization (WHO) Classification of Hematopoietic and Lymphoid Neoplasms as a new provisional entity. 3 This type of lymphoma affects more frequently children under 18 years of age, typically with a limited stage involving the head and neck region, such as the Waldeyer ring, and is characterized by a favorable prognosis. 4 In 2011, the translocations of IGH, IGL, and IGK with IRF4 were identified by Salaverria et al. in 20 of 427 (5%) LBCL patients. Those lymphomas were overwhelmingly of the germinal center B-cell (GCB) subtype, and featured high-level expression of IRF4/MUM1 protein, but lacked the t (14;18)/BCL2 gene translocation. It was found such lymphomas possessed a distinctive gene expression profile. 1 Further work with targeted gene sequencing and copy number array revealed additional aberrations, including frequent IRF4 and NF-kB-related gene mutations, overexpression of the downstream genes of the NF-kB pathway, losses of 17p13 and gains of chromosome 7, 11q12.3-q25. 5 LBCL, IRF4+ hence represents a previously unrecognized subset of B-cell lymphomas with distinctive clinicopathologic characteristics and molecular alterations.
Most pediatric germinal center-derived B-cell non-Hodgkin lymphomas, such as follicular lymphoma or diffuse LBCL (DLBCL), feature a relatively favorable outcome. Given the facts that most LBCL, IRF4+ patients are children and young adults with a GCB-derived tumor, and the probability of IRF4 rearrangement has been reported to decrease significantly with aging, 1,6 one may wonder whether the better outcome of LBCL, IRF4+ is attributable or related to the young age or the molecular changes such as IRF4 rearrangement. In other words, whether IRF4 rearrangement is an independent prognostic factor of LBCL in children and young adults remains unclear.
We thus conduct the current retrospective case-control study to evaluate the clinicopathologic correlates of IRF4 rearrangement in young patients with a primary head and neck LBCL.

| Patients
Patients with DLBCL diagnosed between 2015 and 2020 were selected from the files of the Department of Pathology, Fudan University Shanghai Cancer Center. All of the selected cases fulfilled the following criteria: (1) patients with a Stage I-II disease of primary DLBCL of the head and neck region at the age ≤35 years, (2) transformed DLBCL from other low-grade B-cell lymphomas were excluded. Altogether 106 cases were collected, among which 100 with follow-up information and available materials were employed for pathologic review. Clinical data, including sex, age, Ann Arbor stage, B symptoms, treatment, and the outcome were collected either from medical records or by telephone inquiries. The use of human tissue samples was approved by the Ethics Committee at Fudan University Shanghai Cancer Center.

| Histological review
Formalin-fixed, paraffin-embedded (FFPE) specimens were used for the morphological and ancillary studies. Routine hematoxylin and eosin-stained sections were reviewed by three of the authors (XNJ, TX, and XQL) and diagnosis was rendered according to the updated WHO classification. 3 Consensus was reached on uncertain cases after discussion.

| Statistics
Overall survival (OS) was calculated from the date of diagnosis to the date of death or the last follow-up.
Progression-free survival (PFS) was calculated from the date of diagnosis to the date of disease progression, relapse, death from any cause, or the last follow-up. The survival rates of DLBCL patients were estimated by the life table method. OS and PFS curves were generated by using the Kaplan-Meier method and the log-rank test. Multivariate analyses were performed using a logistic regression model for response and Cox regression models for survival. All variables with a value of p < 0.1 in univariate analysis were submitted for multivariate model analysis. The results were considered statistically significant if a p value <0.05. Statistical analysis was carried out by using GraphPad Prism (8.0, GraphPad Software).

| Patient characteristics and pathologic findings
The clinical and pathologic characteristics of the 100 young patients with a Stage I-II disease of primary DLBCL of the head and neck region are summarized in Table 1. IRF4 rearrangement was detected in 21 patients. The median age of IRF4+ cases was 17 years, with a male to female ratio of 2:1. The most common site of involvement was cervical lymph nodes (61.9%), followed by Waldeyer ring/tonsils (38.1%). All of the IRF4+ cases featured an IPI score of 0-2, and all except one case were demonstrated to have Stage I disease. No obvious differences in the distribution of age, sex, tumor location, staging, or IPI score were observed between the IRF4+ cases and those lacking IRF4 rearrangement (IRF4−).
Histologically, the vast majority of the cases displayed changes compatible with a diagnosis of DLBCL. Of note, nine cases each of both cohorts displayed a concurrent follicular component, that is, follicular LBCL, in addition to the diffuse neoplastic infiltrate. Compared with IRF4− cases, IRF4+ tumors exhibited more frequently a follicular growth pattern (42.9% vs. 11.4%) (p = 0.0020). In the current study, we defined a follicular pattern by the presence of more than one low-power field (2.4 mm 2 ) comprising compact, or sometimes, confluent neoplastic follicles ( Figure 1A). Follicular colonization, typically, replacement of the peripheral outer zone of germinal centers with lymphoma cells was seen in two cases ( Figure 1B,C). Scattered tingible body macrophages were present, at least focally, in 17 cases with IRF4+ tumors, with a conspicuous starry sky pattern reminiscent of Burkitt lymphoma being noted in only one case ( Figure 1D). Cytologically, 61.9% of all IRF4+ tumors were composed predominantly of atypical centroblasts to medium-sized blastoid cells (usually with less distinct nucleoli) ( Figure 1E), with the remaining ones being composed predominantly of centroblasts and centrocytes. In contrast, IRF4− tumors were basically composed of mixed cell population comprising immunoblasts to centrocytes or centroblasts and centrocytes ( Figure 1F) (p < 0.0001).
All cases stained positively for BCL6, whereas CD10 and IRF4/MUM1 positivity was detected in 60 and 56 cases, respectively. For the IRF4+ cohort, the staining intensity for IRF4/MUM1 was principally strong (Figure 2A), although a heterogeneous staining pattern was also demonstrated in some cases. With regard to the cell-of-origin subtyping, most LBCL, IRF4+ cases (95.2%) featured a GCB subtype ( Figure 2B), whereas the IRF4− ones more frequently (40.5%) displayed a non-GCB phenotype (p = 0.001). Of note, 23.8% of IRF4+ tumors were CD5-positive ( Figure 2C), and the incidence was much higher than that of IRF4− tumors (1.3%) (p = 0.0010). In general, CD5-positive LBCL, IRF4+ cases demonstrated similar clinicopathologic characteristics with CD5negative LBCL, IRF4− ones, although the former tended to be more frequently associated with younger patients, Waldeyer ring lesions, and medium-sized blastoid cytology (p < 0.0001) (Table S1). In addition, one case each of the IRF4+ tumors were diffusely positive for CD23 and CD30 staining, respectively. High-level expression of BCL2 or MYC was detected in nearly one-third of IRF4+ cases each; all of those cases, however, lacked BCL2 or MYC gene rearrangements ( Figure 2D-F). Neither IRF4+ nor IRF4− cases were EBER-positive.

| Treatment and outcome
The patients received diverse treatments, including surgical excision, immunochemotherapy, with regimens most commonly of R-CHOP (rituximab, cyclophosphamide, hydroxydaunomycin, oncovin, and prednisone) and occasionally of R-CVP (rituximab, cyclophosphamide, vincristine, and prednisone), excisional surgery followed by chemotherapy, or sometimes combined chemotherapy and radiotherapy with or without a prior tumor mass excision. Involved site radiation therapy was administered to 10 patients in 1 month after the last cycle of chemotherapy, with a prescribed dose of 30 Gy (Table 1). While the vast majority of LBCL, IRF4− cases (94.9%) received chemotherapy or combined treatment, a remarkable proportion (38%) of LBCL, IRF4+ cases received surgical excision only and were not submitted for further chemoradiotherapy (p < 0.0001).
As for the outcome to the treatment (  achieved CR with a significantly higher CR rate (100%) than the LBCL, IRF4− ones (88.6%) (p = 0.0030). In addition, more advanced stage and a non-GCB phenotype also had an adverse impact on the response rate. In the logistic regression, the only variable showing an independent adverse influence on the CR rate was a higher stage (II), with a risk ratio (RR) of 6.9% and 95% confidence interval (CI) of 1.2-43.9 (p = 0.047). After a median follow-up of 60.5 months, the 5-year OS and PFS in the global series were 86% and 84%, respectively. IRF4+ patients had a better OS (100% vs. 82.3%) (p < 0.0001) and PFS (100% vs. 79.7%) (p < 0.0001) than IRF4− patients ( Figure 3A,B). Univariate analysis also showed a more advanced stage (II) had an adverse prognostic impact on OS (55.6% vs. 92.7%) (p < 0.0010) ( Figure 3C) and PFS (50.0% vs. 91.5%) (p < 0.0010) ( Figure 3D), whereas the non-GCB phenotype only correlated with a poorer PFS (p = 0. 0075) instead of OS ( Figure 3E,F). There was no statistical difference with respect to response and survival in relation to anatomic site of involvement, follicular growth pattern, or expression status of CD5. In the multivariate analysis, IRF4+ turned out to be the only parameter that could predict a better OS (RR, 3.7; 95% CI, 1.0-13.4) (p = 0.0499) and PFS (RR, 3.6; 95% CI, 1.1-11.9) (p = 0.0328), whereas a more advanced stage was only associated with an unfavorable PFS (RR, 7.2; 95% CI, 1.4-37.8) (p < 0.0001).

| DISCUSSION
DLBCL represents a large group of biologically heterogeneous neoplasms, among which have been recognized some specific entities or subtypes with characteristic clinicopathologic and molecular features, such as plasmablastic lymphoma, EBV-positive DLBCL and so on, whereas the remaining ones are provisionally labeled as DLBCL, not otherwise specified (NOS), in the WHO classification. DLBCL may affect any site of the body, including the head and neck region. Most head and neck DLBCL cases actually represent systemic disease with head and neck involvement, while about 40% of DLBCL cases are localized to head and neck sites, that is, primary head and neck lymphomas, with both nodal and extranodal involvement. Any extranodal site may be involved, including Waldeyer ring/ palatine tonsil, palate, mandible, maxilla, and tongue. [8][9][10] It has been recognized that the prognosis of primary head and neck DLBCL is mainly related to the subtypes, clinical stage, and International Prognostic Index, instead of sites of involvement. 11 Specific DLBCL subtypes, such as high grade B-cell lymphoma, usually feature their distinctive morphologic, immunophenotypic, and molecular genetic features, and are often associated with specific therapeutic strategies.
LBCL, IRF4+ is a rare disease affecting most frequently children and young adults, with a decreasing relative incidence with age among all LBCLs. 1,5,12 It typically involves the head and neck region such as Waldeyer ring or cervical lymph nodes, and consistently expresses IRF4/ MUM1 due to an IRF4:IG juxtaposition. Although IRF4 rearrangement is a defining feature of this special type of lymphoma with relatively favorable prognosis, it should be noted such a molecular alteration is not specific for the entity. Since it has been found that IRF4 rearrangement can occur in some other aggressive B-cell lymphomas as T A B L E 2 Univariate analysis of clinicopathologic parameters that may influence therapeutic response and survival. well, especially those which are associated with MYC or BCL2 rearrangement, mainly in adults. 5,13 We thus selected a large cohort of DLBCL cases occurring in youth with relatively early-stage disease confined to the head and neck region, to compare the clinicopathologic features of LBCL, IRF4+ and IRF4− ones, and to observe the prognostic implication of IRF4 rearrangement in these patients. The LBCL, IRF4+ cases in the current study present most frequently with cervical lymphadenopathy or tonsillar masses. A slight male predominance, although not statistically significant, is noticed in the IRF4+ patients compared with IRF4− ones. Histologically, the presence of a follicular growth pattern may be highly suggestive of a potential diagnosis of LBCL, IRF4+, which is much less frequently observed in IRF4− cases. The cytologic features of IRF4+ tumors differ from those of IRF4− ones, too, since the characteristic medium-sized blastoid cells are rather commonly seen in IRF4+ tumors, but are generally lacking in the latter ones. In addition, the cell-of-origin subtype of the tumors seems critical for the diagnosis of LBCL, IRF4+, as all but one case feature a GCB phenotype.

Complete remission
In contrast, 40% of IRF4− cases are of non-GCB subtype. None of the tested cases in our series carries chromosomal translocations involving the MYC or BCL2 gene, which is also in consistent with that documented in the literature. 1-5 LBCL, IRF4+, therefore, seems to harbor low genomic complexity. It is noteworthy that a small fraction of LBCL, IRF4+ cases may also harbor BCL6 rearrangement, most of which are young adult patients with Waldeyer ring, cervical lymph node, or bowel involvement. 1,5,[13][14][15] On the other hand, some LBCL cases with concurrent IRF4 and BCL2 rearrangement have been recognized as well, which frequently affects other sites of older adults, and is often morphologically atypical, and may thus represent DLBCL, NOS, rather than typical LBCL, IRF4+. 13 The differential diagnosis of LBCL, IRF4+ includes first pediatric-type follicular lymphoma (PTFL), which also affects children and adolescence predominantly, and present similarly with a follicular lymphoid proliferation composed of large neoplastic cells. PTFLs, however, are principally nodal lesions and rarely involve extranodal sites. [16][17][18] Histologically, PTFL features purely follicular architecture without diffuse infiltration. High-level expression of IRF4/MUM1 is rarely seen. Most importantly, PTFL uniformly lacks a IRF4 translocation. [19][20][21] It has been recently reported that LBCL, IRF4+ frequently harbors mutations of NF-kB-related genes such as CARD11. Intriguingly, CARD11 alterations are present exclusively in cases with a diffuse growth pattern, while the MAP2K1 gene mutations are predominantly present in PTFL, 22 suggesting that the molecular profiles may potentially impact the morphological changes of these tumors. Conventional follicular lymphoma, Grade 3B (FL3B), or follicular LBCL, needs to be distinguished from LBCL, IRF4+, too, due to the similar morphological findings and a frequent IRF4/ MUM1-positive phenotype, but conventional FL3B occurs more commonly in adults, and lacks the IRF4 translocation. 23 Given the fact that LBCL, IRF4+ may express CD5, distinction with CD5-positive DLBCL should be also taken into consideration; the later more frequently affects adults, and usually lacks follicular components and IRF4 rearrangement. 24 Exceptional LBCL, IRF4+ cases with CD5 or CD23 expression need to be distinguished from prolymphocytic progression or large cell transformation of chronic lymphocytic leukemia/small lymphocytic lymphoma. The latter, however, is basically a disease affecting elderly population, and LEF1 positivity and lack of IRF4 rearrangement may aid in differential diagnosis as well. 25 It has been indicated that LBCL, IRF4+ is a peculiar lymphoma type associated with relatively favorable prognosis. 3 However, it remains unclear whether the better outcome of the disease is related to the genetic changes, that is, IRF4 rearrangement, or other clinical and pathologic features of this tumor, such as the young age of the patients, relatively early-stage disease at presentation, and a predominant GCB-derived tumor phenotype, which are all well-known factors associated with a better prognosis. We hence design the current study to compare the clinicopathologic characteristics of LBCL, IRF4+ and LBCL, IRF4− that occur in the head and neck region of young patients with limited stage disease. The results indicate that IRF4+ cases feature better initial response to the treatment and better survival during follow-up. To ascertain whether the cell-of-origin, that is, GCB and non-GCB subtypes, will influence the biological behavior, we further conducted subgroup analysis (Table S2), and found that, even within GCB-subtype cases, IRF4 rearrangement retains to be associated with a better outcome. Further multivariate analysis has demonstrated that both clinical stage and IRF4 status are independent prognostic factors, that is, either the presence of IRF4 rearrangement or an earlier stage (Stage I) correlates with a relatively favorable outcome in those DLBCLs arising from the head and neck region. Of note, while Stage I correlates with a better PFS only, the rearrangement of IRF4 correlates not only with a superior PFS, but also a better OS. Therefore, it seems reasonable and necessary to enroll both clinical staging and the detection for IFR4 rearrangement into the prognosis-estimating system for those young patients with limited stage disease of DLBCL involving the head and neck region. Twenty-five percent of IRF4+ cases in our series were CD5-positive LBCLs; such an incidence is almost the same as that documented in the literature. CD5 can be expressed in 5%-10% of DLBCL, NOS, and CD5-positivity is believed to be associated with an inferior outcome of this tumor. 26 However, in our LBCL, IRF4+ case series, CD5-positivity seems to have no effects on the behavior of this tumor.
As documented in the literature, most of the LBCL, IRF4+ patients have been administered systemic treatment, including chemotherapy, with or without radiation, and the cure rate is usually high. 1,5,27 However, there is limited existing evidence to support treatment deescalation. [27][28][29][30] For example, whether those patients with lymphoma lesions completely removed by surgery still need complementary systemic treatment remains unclear. Nearly 40% of all IRF4+ patients in our series received mere surgical resection, without additional chemotherapy or radiation, all of whom turned to be cured later. These patients were all characterized by localized lesions (Stage I and no bulky disease), and other clinicopathologic characteristics comparable with those patients who had received systemic treatment (Table S3). Based on these findings, a more conservative therapeutic strategy of complete resection followed by watchful wait might be adequate as well as safe for those with solitary, resectable lesions, which may certainly aid in avoiding possible overtreatment.
In conclusion, rearrangement of the IRF4 gene not only defines a group of LBCL with distinct clinicopathologic features, but also is of practical implication for prognosis prediction and therapeutic strategy decision. We have compared for the first time the pathologic and prognostic differences between IRF4+ and IRF4− primary head and neck DLBCLs, and found that IRF4 rearrangement may represent an independent prognostic factor indicative of better outcome. These findings further unravel the heterogeneity of DLBCL arising in young patients, which may provide additional parameters for risk stratification and individualized treatment.