Genomic heterogeneity in primary cutaneous follicular center lymphomas reveals different clinicopathological subgroups

Abstract

Primary cutaneous follicle center lymphomas (PCFCLs) are indolent B‐cell neoplasms limited to the skin and effectively managed with local therapies. Distinguishing PCFCL from systemic follicular lymphoma (sFL) with cutaneous involvement (FL_CI) is challenging due to overlapping features. We performed an integrated pathological and genetic analysis of skin samples of 24 PCFCL and 10 FL_CI, which showed subtle pathological changes (decreased BCL2 and CD10, increased CD23 expression in PCFCL), but markedly distinct mutational landscapes. FL_CI exhibited recurrent mutations in chromatin‐modifying genes (CREBBP 90%, KMT2D 90%, and EP300 30%), closely resembling sFL. In contrast, PCFCL displayed a more heterogeneous profile, with mutations affecting B‐cell development, cell adhesion, and immune evasion. These molecular alterations identified three distinct PCFCL subgroups. Group 1 (44%) harbored mutations in immune evasion genes (IRF8, FAS, CIITA, and B2M) and was associated with CD10 negativity, diffuse architecture, and localization in non‐photoexposed areas. Group 2 (20%) showed activating STAT6 and CREBBP mutations, consistent CD23 and CD10 positivity, and exclusive presentation in sun‐exposed sites. Group 3 (20%) shared a similar clinicopathological profile to Group 2 but was defined by TNFRSF14 mutations. Clonal origin and mutational evolution were studied in five patients, one with synchronous PCFCL and four with sequential samples from two PCFCL and 2 FL_CI. This analysis supports a model of PCFCL oncogenesis driven by circulating progenitors following complex evolutionary patterns, including convergent evolution and greater clonal diversity at relapse. Overall, our study refines the mechanisms driving PCFCL pathogenesis, while providing a framework for PCFCL differential diagnosis and clinical management.

INTRODUCTION

Primary cutaneous follicle center lymphoma (PCFCL) is the most common type of primary cutaneous B‐cell lymphoma. It is a low‐grade, indolent lymphoma that originates from follicle center B cells and primarily affects the skin without evidence of systemic disease at diagnosis. It typically presents as solitary or grouped plaques, nodules, or tumors, most commonly on the scalp, forehead, or trunk. ¹ One of the main diagnostic challenges in PCFCL is distinguishing it from secondary cutaneous involvement by systemic follicular lymphoma (FL_CI). PCFCL, unlike systemic follicular lymphoma (sFL), has a more favorable prognosis, responds to local therapy, and rarely progresses to systemic involvement. ¹ Traditionally, immunohistochemical strong expression of BCL2 and CD10, as well as the presence of the t(14;18)/IGH::BCL2 translocation, has been considered supportive of secondary involvement. However, none of these features have been established as unequivocal criteria, highlighting that, although rare, PCFCL may harbor BCL2 rearrangement. ² , ³ Moreover, 5%–10% of PCFCL cases may exhibit dissemination during follow‐up, ⁴ requiring systemic and more aggressive therapeutic regimens. It remains unclear whether these cases represent PCFCL with subsequent secondary dissemination, or if they are FL_CI that initially presented below the threshold of detection, as imaging studies and bone marrow examinations sometimes fail to detect subtle systemic disease. ⁵ A better understanding of PCFCL and FL_CI pathogenesis can also assist in the identification of these aggressive cases using molecular assays. Genomic studies in these neoplasms are limited,6, 7, 8, 9, 10 but they suggest that PCFCL has a distinct mutational profile from nodal FL,11, 12, 13 with fewer chromatin‐modifying gene mutations in skin‐restricted cases. However, the molecular features of PCFCL remain insufficiently characterized due to the scarcity of samples and the complexity of diagnostic assessment.

In this study, we expand the clinicopathologic and molecular understanding of PCFCL. The cohort size, together with comprehensive and well‐characterized clinical, pathological, and molecular data, has allowed us to address some of the biological mechanisms underlying PCFCL pathogenesis. Moreover, in this work, we provide rational insights to better differentiate PCFCL from FL_CI—an essential step toward improving clinical management.

METHODS

Patients and clinicopathological evaluation

A total of 34 patients diagnosed with PCFCL or FL_CI, with available formalin‐fixed paraffin‐embedded (FFPE) and frozen tissue (OCT) material from Hospital Clinic of Barcelona (Spain) during the period 2007–2023, were included in the study. The study was designed in line with the Declaration of Helsinki and was approved by the Hospital Clinic of Barcelona Review Board (HCB/2018/0553). Informed consent of all patients was obtained, according to the Institutional Review Board of Hospital Clinic de Barcelona. The sample set comprised 25 biopsies of PCFCL from 24 patients, including two synchronous lesions from the same patient, and 10 biopsies from 10 patients with FL_CI; all corresponding to time point 1. In addition, recurrent samples were available for a subset of cases, including two time‐point‐2 samples from PCFCL cases, and two time‐point‐2 samples and one time‐point‐3 sample from two FL_CI cases. All the cases included underwent a comprehensive diagnostic reassessment, including a detailed review of the original histopathology as well as the patients' clinical records and complementary studies. The diagnosis of PCFCL and FL_CI was established following the criteria of the 2018 update of World Health Organization‐EORTC classification for primary cutaneous lymphomas, ¹ the 5th edition of the World Health Organization Classification of Hematolymphoid Tumours ¹⁴ and the International Consensus Classification of Mature lymphoid Neoplasms (2022). ¹⁵ PCFCLs were diagnosed in patients presenting with lesions confined to the skin, without evidence of extracutaneous disease at baseline evaluation. In contrast, secondary cutaneous involvement was defined in patients with a previous diagnosis of sFL who subsequently developed skin lesions or in those who presented with extracutaneous involvement at the time of diagnosis. We systematically collected comprehensive clinical and biological data to assess overall health, included the skin location and timing of initial and recurrent lesions to identify relapse patterns, and documented all treatments, including both skin‐directed and systemic therapies. Disease extent and systemic involvement were assessed using bone marrow biopsies and positron emission tomography (PET)/computed tomography (CT) or CT imaging. In seven patients with PCFCL, bone marrow biopsy was not performed at the time of diagnosis, as these cases were diagnosed in earlier years and staged in accordance with the recommendations then in effect. Under the 2007 ISCL/EORTC guidelines, ¹⁶ bone marrow examination was not mandatory for indolent primary cutaneous lymphomas such as PCFCL unless clinical, laboratory, or imaging findings suggested extracutaneous involvement. In these cases, the clinical presentation and histopathologic features were consistent with a primary cutaneous origin, and no clinical, laboratory, or imaging findings indicated systemic involvement neither at diagnosis nor during follow‐up (Table 1). Longitudinal follow‐up enabled monitoring of disease progression and clinical outcomes. BCL2 translocations and pathological evaluation, including morphology and immunohistochemistry, were performed in FFPE samples following standard protocols (Supporting Information S1: Supplementary Methods).

Table 1.

Clinical characteristics of the 34 patients included in the study (24 primary cutaneous follicle center lymphoma, PCFCL, defined as skin‐limited disease with clinicopathological features supporting a cutaneous origin, and 10 systemic follicular lymphoma with cutaneous involvement, FL_CI, defined as systemic cases subsequently developing skin lesions).

	PCFCL (n = 24)	FL_CI (n = 10)
Sex
Male	19 (79%)	7 (70%)
Female	5 (21%)	3 (30%)
Mean age at diagnosis (years; range)	56 (26–75)	60 (42–65)
Bone marrow biopsy involvement at diagnosis	0 (0%)	10 (100%)
Performed	17 (71%)	10 (100%)
Not performed	7(29%)	0 (0%)
Extracutaneous PET/CT uptake at diagnosis	0 (0%)	5 (50%)
Performed	21 (87.5%)	5 (50%)
Not performed	3 (12.5%)	5 (50%)
Relapse of the disease
No relapse	8 (34%)	3 (30%)
Relapse	16 (66%)	7 (70%)
Same location	12 (50%)	0 (0%)
Other location	4 (17%)	7 (70%)
Mortality due to lymphoma	0 (0%)	2 (20%)
Time of follow‐up (years; median)	14 (5–27)	15 (7–31)
Treatment
Topical corticosteroids	8 (33%)	0 (0%)
Intralesional corticosteroids	11 (45.8%)	0 (0%)
Surgical excision	6 (25%)	0 (0%)
Local radiotherapy	10 (41.6%)	4 (40%)
Intralesional rituximab	3 (12.5%)	0 (0%)
Intravenous rituximab	9 (37.5%)	8 (80%)
R‐CHOP	1 (4.2%)	6 (60%)
R‐bendamustine	0 (0%)	3 (30%)
Biopsy site	n  = 25	n  = 10
Head	11 (44%)	1 (10%)
Thorax	1 (4%)	2 (20%)
Abdomen	2 (8%)	1 (10%)
Back	7 (28%)	5 (50%)
Upper extremity	2 (8%)	0 (0%)
Lower extremity	2 (8%)	1 (10%)

Note: One PCFCL patient (C050) had two synchronous tumors. Several patients received more than one treatment.

Abbreviations: PET/CT, positron emission tomography–computed tomography; R‐bendamustine, combination of rituximab and bendamustine; R‐CHOP, combination of rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone.

Mutational and B‐cell receptor clonality analysis

DNA from 40 fresh skin frozen samples from 34 patients was extracted using the QIAmp DNA Mini Kit (Qiagen). We designed a custom targeted NGS panel of 136 genes recurrently altered in B‐cell lymphoma (Supporting Information S3: Table 1). The design, library preparation, sequencing, and bioinformatic analysis were performed as previously published, with minimal modifications (Supporting Information S1: Supplementary Methods). ¹³ Copy number alterations (CNAs) from were analyzed using the tool “cpCaller” (https://github.com/nadeulab/cpCaller, Supporting Information S1: Supplementary Methods). We also analyzed seven matched lymph node samples from FL_CI patients preserved in FFPE (Supporting Information S1: Supplementary Methods). In addition, mutational profile data from sFL cases at diagnosis (n = 54), previously performed and published at our center using a targeted NGS panel, were included for comparative analyses. ¹³

Immunoglobulin gene rearrangement analysis

Immunoglobulin gene rearrangements in the longitudinal cohort (2 FL_CI and 2 PCFCL) and two synchronous samples from the same PCFCL patient were studied using a custom capture‐based NGS panel that covers the IGH V(D)J regions IGK and IGL VJ, and class switch regions. ¹⁷ The IGHV gene rearrangement of extra 11 samples (3 FL_CI and 8 PCFCL cases) was analyzed using the SOPHiA DDM™ Community CLL Clonality Solution. In both cases, IgCaller was used to perform analysis. ¹⁸

Statistical analysis

Quantitative variables were compared among groups by means of the Wilcoxon non‐parametric test. Fisher's exact test was used to compare categorical variables. Statistical significance was defined as a P value < 0.05.

RESULTS

Clinicopathological features

We included 24 PCFCL patients and 10 FL_CI patients (Table 1). The PCFCL group comprised 19 males (79%) and 5 females (21%), while the FL_CI group included 7 males (70%) and 3 females (30%). The mean age at diagnosis was 56 years for the PCFCL group and 60 years for the FL_CI group. Lymphadenopathies or masses were not observed at diagnosis in PCFCL cases. Two PCFCL (8%) showed extracutaneous systemic spread, 5 and 6 years after diagnosis (patient C021 and C035, respectively), whereas the rest of the PCFCL (92%) remained restricted to the skin during a median follow‐up of 14 years (range 5–27 years). In our series, all FL_CI patients developed cutaneous involvement following the diagnosis of sFL. The mortality rate for PCFCL and FL_CI patients was 0% and 20%, respectively.

Regarding treatments, PCFCL patients received skin‐directed therapies, including topical corticosteroids, intralesional corticosteroids, surgical excision of lesions, local radiotherapy, intralesional rituximab, and in selected cases with multifocal lesions, intravenous rituximab. In contrast, the FL_CI group underwent more intensive treatments, primarily chemotherapy according to the R‐CHOP regimen (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone). Relapse was observed in the majority of cases in both groups (relapse‐free cases: PCFCL, 8/24; FL_CI, 3/10). PCFCL relapses occurred at the same original location in most cases, whereas all FL_CI skin relapses occurred far from the original biopsy locations (P < 0.005) (Supporting Information S2: Figure 1A). The median time to relapse for PCFCL cases confined to the skin was 24 months (n = 13; range, 5–97 months), while PCFCL cases with subsequent systemic involvement and FL_CI cases relapsed earlier, with a median of 9 months (n = 2; range, 5–36 months) for PCFCL, and a median of 15 months (n = 9; range, 2–73 months) for FL_CI. Lastly, regarding the tumor location of the skin biopsies, PCFCL biopsies were more frequent from the head (11/25, 44%), whereas FL_CI tumors frequently affected the back (5/10, 50%). Remarkably, a significant association was observed between PCFCL cases and localization in sun‐exposed areas, defined as the head, neck, and arms—anatomical sites that are chronically exposed to sunlight throughout most of the year in Spain, where the patients resided (P = 0.043) (Table 1, Supporting Information S2: Figure 1B,C).

Histopathological examination of all specimens revealed a dermal lymphoid infiltrate with sparing of the epidermis, including the 25 PCFCL skin biopsies obtained from 24 patients (patient C050 had two synchronous tumors at diagnosis, one from the face and one from the arm) and the 10 skin biopsies from 10 patients with FL‐CI. Among the PCFCL samples, the most common growth patterns were diffuse (11/25, 44%) and mixed (with both diffuse and nodular components present) (9/25; 36%). The most frequent cytomorphology was medium to large cell predominance (9/25; 36%), followed by small centrocyte predominance (6/25; 24%) or small centrocyte with large‐cell aggregates (7/25; 28%). A minority of cases (3/25; 12%) showed alternative cytological features, such as spindle‐ or clear‐cell. By contrast, FL_CI samples showed a mixed (4/10; 40%), nodular (3/10; 30%), or diffuse (3/10; 30%) pattern (Supporting Information S2: Figure 2A), with high small centrocyte predominance (6/10; 60%).

We performed a panel of IHC staining to study the phenotypical and pathological characteristics of both diseases (of note, due to occasional non‐evaluable immunostaining, slight variations in case numbers are observed across markers). The neoplastic cells in all cases displayed expression of CD20 and CD79a when tested, as well as BCL6. Only one case of FL_CI was negative for BCL6 but showed clear positivity for CD10. BCL2 expression was found in all evaluable FL_CI cases (one case was considered non‐evaluable) and in 11/22 (50%) PCFCL (P = 0.01; Figure 1A). CD10 expression was significantly higher in FL_CI (9/10; 90% vs. 8/24; 33%; P < 0.05; Figure 1A). CD23 was negative in the majority of FL_CI cases (9/10; 90%), while it was expressed—either focally or weakly—in nearly half of the PCFCL cases (11/23; 48%; P = 0.054; Figure 1A). In both groups, the majority of cases exhibited a T‐cell infiltrate that did not exceed the number of B cells (PCFCL: 19/25, 76%; FL_CI: 6/10, 60%) (Supporting Information S2: Figure 2B). The proliferative index (Ki‐67) was ≥30% in the majority of PCFCL cases (18/25; 72%) and in half of the FL_CI cases (5/10, 50%), indicating a higher proliferative rate in PCFCL (Supporting Information S2: Figure 2C). BCL2 rearrangements, assessed by FISH, were detected in the majority of FL_CI cases (8/9; 88%) but only in two cases from the PCFCL cohort (C041 and C061; 2/11; 18.2%; Supporting Information S2: Figure 2D). Importantly, both PCFCL cases demonstrated clinicopathological findings fully consistent with PCFCL, supported by negative bone marrow biopsies and the absence of extracutaneous involvement on PET/CT imaging. Among PCFCL, only case C061 exhibited a BCL2 copy number gain, whereas BCL2 gains in the FL_CI cohort were observed in cases C058 and C063.

Figure 1.

Molecular and pathological characteristics of primary cutaneous follicular center lymphomas (PCFCL) and systemic follicular lymphomas with cutaneous involvement (FL_CI). (A) Oncoprint displaying recurrent mutated genes (≥2 tumors) in the cohort, including 25 PCFCL and 10 FL_CI. Baseline features include BCL2, CD10, and CD23 positivity by immunohistochemistry. Two synchronous tumors from the same patient are indicated with an asterisk (*). (B) Comparative bar plots showing mutation frequencies in 35 highly recurrent genes in PCFCL (left) and FL_CI (right) (≥3/25 in PCFCL and/or ≥2/10 in FL_CI).

Genomic landscape of cutaneous follicular lymphoma reveals divergent oncogenic mechanisms in PCFCL

Custom targeted sequencing detected a total of 284 mutations in the initial biopsies (Supporting Information S3: Table 2). Cases of both PCFCL and FL_CI had a similar number of mutations (eight mutations per sample in both groups, Supporting Information S2: Figure 2E). A total of 58 genes were found to be mutated in more than two samples (Figure 1A). The most recurrently affected genes were directly related to FL pathogenesis, including frequent alterations in the chromatin modifiers CREBBP (18/35, 51%), KMT2D (10/35, 28%), EP300 (6/35, 17%), and TNFRSF14 (10/35, 28%).

A closer comparison between the two subtypes was performed by focusing on genes with recurrence rates higher than 10% in either subtype, resulting in a set of 35 genes (Figure 1B). Top recurrently mutated genes in FL_CI involved chromatin regulators (CREBBP, KMT2D, EP300, and EZH2). KMT2D mutations were nearly exclusive to FL_CI. Multi‐hit mutations in CREBBP (2/10, 20%) and KMT2D (7/10, 70%) occurred in FL_CI but were rare or absent in PCFCL (CREBBP: 1/25, 4%; KMT2D: 0/25, 0%). All CREBBP mutations were affecting the KAT domain except two mutations in two different PCFCL cases (p.Pro1948fs and p.Leu2409Ile). Indeed, only two genes were recurrently mutated specifically in FL_CI: UBE2A (3/10) and BCL2 (2/10). BCL2 mutations are consistent with aberrant somatic hypermutation associated with BCL2 translocation, previously described in FL. On the contrary, in PCFCL, we observed a higher number of genes associated with immune evasion (IRF8, FAS, CIITA, and B2M) and JAK/STAT signaling (STAT6, PTPRD). As expected, recurrent mutations in PCFCL were concentrated in known hotspots—such as the KAT11 domain of CREBBP, the DNA‐binding site of STAT6, the FAS death domain, and both termini of IRF8 (Supporting Information S2: Figure 3). We found fourteen genes that were exclusively mutated in PCFCL with varying recurrence rates (STAT6, PTPRD, ITPKB, REL, SETD2, B2M, ETS1, GNA13, BCOR, CIITA, FOXO1, TMSB4X, TP53BP1, and ZFP36L1) (Figure 1B).

Pathway enrichment analysis showed that mutations affecting chromatin regulators were increased in FL_CI (Figure 2, annotated in Supporting Information S3: Table 1, Supporting Information S2: Figure 4), with a median of three chromatin regulator genes mutated per sample (P = 0.003). On the other hand, PCFCL showed increased mutations in several pathways, such as adhesion and migration genes (P = 0.005), B‐cell differentiation (P = 0.014), and immune escape (P = 0.03). We also identified a trend toward increased mutations in the JAK/STAT pathway in PCFCL samples (P = 0.08). No significant differences were detected in other pathways, such as proliferation, B‐cell receptor (BCR) pathway, or other cell signaling pathways.

Figure 2.

Oncogenic pathways target by mutations in primary cutaneous follicle center lymphoma (PCFCL) and systemic follicular lymphoma with cutaneous involvement (FL_CI). Box plots showing the number of mutations per sample in specific pathways across both subtypes of cutaneous FL. P‐values were calculated using the Wilcoxon rank‐sum test.

Distinct mutation patterns and elevated genomic heterogeneity distinguish PCFCL from FL_CI and systemic FL

To explore the relationship between the mutated genes in PCFCL and FL_CI with those observed in sFL, we compared the 35 most frequently mutated genes identified in PCFCL and FL_CI tumors with the mutational profile of sFL at diagnosis, recently characterized in our center (Figure 1B, Supporting Information S3: Table 3). ¹³ This comparison showed that the genes most frequently mutated in PCFCL (>15% of cases) were rarely mutated in sFL (<10%), except for chromatin regulators (CREBBP, EP300, and EZH2), TNFRSF14, and IRF8 (Figure 3A). Notably, sFL samples exhibited a striking enrichment of KMT2D mutations (72% in sFL vs. 8% in PCFCL, P = 0.0001) and BCL2 mutations (28% in sFL vs. 0% in PCFCL, P = 0.0018). In contrast, genes recurrently mutated in FL_CI showed similar mutation frequencies with non‐significant trends for RRAGC (30% vs. 7% in sFL, P = 0.07), POU2F2 (20% vs. 2% in sFL, P = 0.06), and for ATM and SPEN (both 20% vs. 4% in sFL, P = 0.11) (Figure 3B).

Figure 3.

Mutational landscapes across systemic folicular lymphoma (sFL), primary cutaneous follicle center lymphoma (PCFCL), and systemic follicular lymphoma with cutaneous involvement (FL_CI) subtypes. (A) Bar plots comparing mutation frequencies in 35 highly recurrent genes between PCFCL (left) and sFL (right). UBE2A, ZFP36L1, and TP53BP1 were not included in the panel sequencing study of systemic FL. (B) Bar plots comparing mutation frequencies in 35 highly recurrent genes between FL_CI (left) and sFL (right). UBE2A, ZFP36L1, and TP53BP1 were not included in the panel sequencing study of systemic FL. (C) Principal component analysis of the mutational profiles of PCFCL (n = 25), FL_CI (n = 10), and sFL (n = 54). (D) Principal Component 1 values compared across lymphoma subtypes (PCFCL, FL_CI, and systemic FL). P‐values calculated using the Wilcoxon rank‐sum test. (E) Usage frequency of IGHV gene subgroups in two independent series of systemic FL and our cohorts of PCFCL (n = 11) and FL_CI (n = 5).

To systematically characterize the mutational patterns distinguishing these entities, we performed principal component analysis on the mutation profiles across PCFCL, FL_CI, and sFL samples. PCFCL cases clustered separately from FL_CI and sFL cases (Figure 3C). Comparison of PC1 scores using t‐test confirmed significant differences between PCFCL and sFL (P = 0.0001), while FL_CI and sFL did not differ significantly (Figure 3D), suggesting distinct biology and oncogenic pathways in PCFCL tumors. Interestingly, PCFCL showed higher dispersion in the PC1 scores, suggesting a more heterogeneous mutational landscape than the detected in both FL_CI and sFL cases.

Lastly, when available, we analyzed the rearranged IGHV gene sequences from both groups (n = 11 PCFCL and n = 5 FL_CI) and compared the IGHV gene family usage with that reported in two studies of systemic FL (Berget et al. [n = 104] and Garcia‐Álvarez et al. 2019 [n = 138]). ¹⁹ , ²⁰ Interestingly, FL_CI showed a distribution similar to that observed in systemic FL, with three of five samples (60%) using IGHV3 genes (Figure 3E). In contrast, PCFCL displayed a marked bias toward IGHV1 gene usage (7/11, 63%), which was significantly higher than in both FL systemic cohorts and in our FL_CI (Fisher's exact test: P = 0.0033 vs. Berget et al.; P = 0.0001 vs. Garcia‐Álvarez et al.; P = 0.033 vs. FL_CI). Notably, two IGHV1 genes (IGHV1‐46 [2/11] and IGHV1‐69 [3/11]) and one IGHV3 gene (IGHV3‐7 [2/11]) recurred among PCFCL cases, whereas IGHV1‐46 and IGHV1‐69 are reported to represent <5% of cases in systemic FL cohorts.

Identification of three PCFCL subgroups with unique genomic, pathological, and clinical profiles

We observed that genes related to immune escape mechanisms were frequently mutated in PCFCL when compared to FL_CI. Mutations in IRF8, FAS, CIITA, and/or B2M were found in 44% PCFCL (11/25) but only in 2/10 FL_CI (20%). Moreover, 6 out of the 11 (55%) tumors harbored recurrent mutations in two or more genes (Figure 4A), with no recurrence observed in FL_CI cases. Interestingly, PCFCL mutations in immune escape genes showed a mutually exclusive pattern with mutations in STAT6 or TNFRSF14, accounting in total to 10/25 cases (40%) (Mutual exclusivity by Fisher test: P = 0.011). Specifically, 20% of cases (5/25) had STAT6 mutations, while 20% had TNFRSF14 mutations, without mutations in immunosuppression genes and/or STAT6. These profiles identify three subgroups of PCFCL, each characterized by particular genetic patterns targeting different oncogenic mechanisms and correlating with distinct clinicopathological features. Four cases lacked mutations in any of these genes and remained unclassified.

Figure 4.

Genomic and clinicopathological features classify primary cutaneous follicle center lymphoma (PCFCL) into three distinct groups. (A) Oncoprint showing alterations in genes associated with immune escape (IRF8, FAS, CIITA, and B2M), STAT6, and TNFRSF14. Baseline annotations include CD10, CD23, and BCL2 positivity (IHC), T‐cell infiltration, growth pattern (diffuse vs. follicular/mixed), cytomorphology, photoexposure, and relapse behavior, as well as copy number alterations in REL and TNFRSF14 genes. (B–E) Bar plots showing differences in CD10 expression (B), CD23 expression (C), growth pattern (D), and T‐cell infiltration (E) in the three groups of PCFCL. (F) Frequency of CREBBP mutations in the PCFCL subgroup. (G) Tumor location comparison between Group 1 (immune escape‐related, n = 11) and joined Groups 2 and 3 (n = 10). (H, I) Different distribution in photoexposed areas (H) and recurrence behavior (I) of the different PCFCL groups.

Analysis of two of the most recurrent CNAs described in PCFCL ²¹ —TNFRSF14 deletions and REL amplifications—also revealed marked differences among the three groups. TNFRSF14 deletions were identified in 4 of 25 PCFCL cases (16%) and were mutually exclusive with TNFSF14 mutations. Notably, all TNFRSF14 deletions occurred in Group 1 cases (4/11; 36%). Conversely, gains or high‐level gains (>3 gene copies) were detected in 8 of 25 cases, again predominantly in Group 1 (6/11; 54%), whereas only one case from Groups 2 and 3 showed a REL gain. Taken together, these findings suggest that these novel groups could differ in their underlying oncogenic mechanisms.

Pathological inspection of these three groups (Supporting Information S2: Figure 5) revealed that tumors in Group 1 showed lower CD10 expression compared to Group 2 (P = 0.035, Figure 4B), while all Group 2 tumors were CD23‐positive (P = 0.03, Figure 4B,C), consistent with STAT6‐driven activation of the JAK/STAT pathway. Although cytomorphology was not significantly different between groups, Group 1 showed a predominance of large centrocytes/centroblasts (6/11, 54%), whereas Groups 2 and 3 were mainly composed of small centrocytes (2/10, 20%) (Figure 4A). Morphological patterns were also strongly associated with these groups, since all cases (10/10) in Groups 2 and 3 had a nodular/mixed pattern, while most Group 1 tumors had a diffuse pattern (9/11) (P = 0.005) (Figure 4D). Group 1 tumors also showed a trend toward higher T‐cell infiltration (Figure 4E), though not statistically significant.

Interestingly, Group 2 tumors displayed a significant association with CREBBP mutations (4/5, 80%), which were rare in Group 1 (2/11, 18%) (Figure 4F). Indeed, comparing Group 2 with the rest of the cohort confirmed a significant enrichment in CREBBP mutations, all of them affecting the KAT11 domain (4/5 vs. 5/20, P = 0.04). Notably, tw out of five CREBBP‐mutated cases not classified as Group 2 harbored mutations outside the KAT11 domain (p.Pro1948fs and p.Leu2409Ile).

Finally, tumor location in skin differed by group: Group 1 tumors were mainly on the back (5/11), while 9/11 in Groups 2 and 3 were on the head (Figure 4G). This was more evident when considered sun‐exposed areas: all Group 2 and 3 tumors were in photoexposed areas versus 2/11 in Group 1 (P = 0.005; Figure 4G,H). We did not find differences in the relapse patterns (Figure 4I), indicating that different pathological and genetic features converge into similar clinical behaviors. Strikingly, the only two patients that showed systemic spread in our cohort (2/24 = 8.3%) were classified in Group 3, displaying isolated TNFRSF14 mutations without any other common mutations associated with systemic spread.

Spatial and temporal clonal dynamics differ in PCFCL and FL_CI

To investigate the clonal and genetic dynamics driving the frequent relapses in PCFCL, we analyzed sequential tumor samples from four patients—two PCFCL and two FL_CI. Three patients were sampled at two time points, and one FL_CI patient at three (Figure 5A). Importantly, BCR sequencing demonstrated identical immunoglobulin gene rearrangements across all samples from each patient, supporting a shared clonal origin in both PCFCL and FL_CI cases (Supporting Information S3: Table 4).

Figure 5.

Spatial and temporal clonal dynamics in follicular center lymphoma. (A) Variant allele frequency (VAF) (%) of all mutations detected across multiple time points in two systemic follicular lymphoma with cutaneous involvement (FL_CI) patients. Shared mutations (red) are linked across samples. (B) VAF (%) of all mutations across two time points in two primary cutaneous follicle center lymphoma (PCFCL) patients. Shared mutations are highlighted in red. (C) Heatmap summarizing pathological changes in the corresponding longitudinal biopsies. (D) VAF (%) of all mutations detected in two synchronous tumors (nose and arm) from the same patient. Shared mutations are linked in red. (E) Histological and immunophenotypic comparison of synchronous biopsies. The images show two skin lesion biopsies from the same patient, taken at the same time but in different locations. The first biopsy (A–D), corresponding to a skin lesion on the nose, shows a predominantly nodular dense infiltration by lymphoid cells, composed of medium‐ to large‐sized cells with occasional clear cytoplasm (A HE). The IHC study shows that the infiltrate corresponds to a B‐cell population (A CD20), BCL2 negative (B BCL2) with CD23 positivity in isolated cells (C CD23) and a proliferative index below 30% (D Ki‐67). The second biopsy (E–H), corresponding to a skin lesion on the arm, shows a diffuse dermal infiltration by a monotonous population of medium‐sized lymphoid cells (E HE). The IHC study demonstrates a B‐cell population (inset E CD20), which is diffusely positive for BCL2 and CD23, with a proliferative index around 30% (H Ki‐67). Scale bars: (A) 50 µm, (B) 200 µm, (C) 50 µm, (D) 100 µm, (E) 100 µm, (F) 200 µm, (G) 50 µm, and (H) 100 µm.

The genetic analysis revealed striking differences between the two subtypes. FL_CI cases showed high genetic stability over time. In patient FL_CI_C012, eight mutations were shared across Time 1, Time 2, and Time 3 samples, with a few mutations unique to each sample. Patient FL_CI_C004 had five shared mutations between both time points, with a few private mutations per sample (Figure 5A). In contrast, PCFCL cases showed marked genetic divergence. PCFCL_C054 shared only two mutations between Time 1 and Time 2, with four and nine private mutations, respectively. PCFCL_C048 shared six mutations but also had six unique mutations to Time 1 and five to Time 2 (Figure 5B). In agreement, FL_CI samples were consistent over time in morphology and expression of main IHC markers, such as CD10, BCL2, and CD23 (Figure 5C). PCFCL cases, however, showed variable BCL2 and CD23 expression, and one case (C054) showed a cytomorphological shift from medium‐large cells to small centrocytes (Figure 5C).

Next, we aimed to study the nodal and cutaneous involvement in FL_CI, comparing the mutational profiles of paired lymph node and skin biopsies when available (7/10 cases) (Supporting Information S2: Figure 6). Although the total number of mutations was similar in most cases, the mutational landscapes differed substantially between compartments, with many tissue‐specific mutations. Analysis of recurrent FL‐associated genes showed partial concordance between tissues. CREBBP status was concordant in most cases (5/7 mutated in both tissues and 1/7 non‐mutated), consistent with its role as an early driver event. KMT2D mutations were present in both compartments in four cases but restricted to either skin or LN in three cases, while EP300 mutations were shared in two cases and detected only in LN in two others. Further studies are needed to validate these findings, as technical differences between samples may have influenced the analysis, including higher DNA quality in OCT‐frozen cutaneous samples compared with archival FFPE lymph node samples.

The analysis of synchronous tumors in patient C050 (Figure 1A) revealed that both lesions showed the same BCR clone (Supporting Information S3: Table 4) but different genetic and pathological profiles. Strikingly, this includes different mutations in STAT6, three out of four mutations affecting the STAT6 domain (Figure 5D). Interestingly, we detected differences in BCL2, CD23, and Ki‐67 expression, higher in the arm lesion (Figure 5E, bottom). All these results together suggest that FL_CI tumors tend to evolve from a stable clonal population, whereas PCFCL follows more variable genetic and phenotypic trajectories in which relapses and even synchronous lesions may represent diverging evolutionary paths from the same original clone.

DISCUSSION

Our study integrating genetic, molecular, and pathological features of PCFCL and FL_CI confirms the differences between both types of tumors and shows higher heterogeneity in PCFCL than previously described. ²¹ Our cohort has similar age and sex distributions than previous reports. FL_CI patients developed skin involvement following a prior diagnosis of nodal lymphoma, though rare cases have been reported where cutaneous lesions are the inaugural disease manifestation. ²² None of the patients with PCFCL deceased of the disease, whereas 20% of patients with FL_CI deceased, with both subgroups having a similar follow‐up time. Although extranodal involvement is not generally associated with poor prognosis in FL, other high‐risk features—such as early relapse may have contributed to the unfavorable outcomes observed in these cases. ²³ We also confirmed higher BCL2 rearrangement, BCL2 and CD10 expression in FL_CI versus PCFCL, but found no significant differences in Ki‐67 expression, highlighting the limitations of diagnosis based on staining alone.7, 14, 24, 25, 26, 27, 28, 29, 30, 31, 32 The importance of our results should be understood in the context of Zhou et al. ⁷ genetic and pathological criteria to predict the likelihood of systemic involvement by cutaneous follicular lesions. This criterion was based on three parameters: the presence of BCL2 rearrangement, a low proliferative index (Ki‐67 < 30%), and mutations in two or more chromatin‐modifying genes (CREBBP, KMT2D, EZH2, and EP300). The fulfillment of two or more criteria was associated with systemic spread in their series. In our cohort, this criterion is not fully confirmed, since the two PCFCL cases with systemic dissemination lacked mutations in two or more epigenetic regulators and showed no evidence of BCL2 translocation, which would classify it as skin‐restricted. Interestingly, these two cases shared CD10 positivity and TNFRSF14 mutations, being part of the Group 3 PCFCL cases. Further studies are needed to clarify features linked to systemic progression.

Our mutational analysis supports that PCFCL and FL_CI represent biologically distinct entities, with FL_CI aligning more closely with the spectrum of systemic FL. Although the differences did not reach statistical significance—likely due to the limited number of FL_CI cases—we observed a trend toward higher mutation frequencies in RRAGC, ATM, POU2F2, and SPEN, suggesting potential contributors to extranodal lymphomagenesis. Notably, SPEN is recurrently mutated in MALT lymphomas across anatomical sites, ³³ reinforcing its association with extranodal biology. Activating RRAGC mutations affecting the nucleotide‑binding domain were also more common in FL_CI (30% vs. 7% in systemic FL), implicating dysregulated mTORC1 signaling in its pathogenesis. ³⁴

Consistent with these genomic differences, immunoglobulin repertoire analysis further distinguished the two entities. FL_CI displayed an IGHV usage pattern similar to systemic FL, ¹⁹ , ²⁰ whereas PCFCL showed a pronounced bias toward IGHV1 family genes, supporting divergent immunogenetic backgrounds.

On the contrary, PCFCL, as a BCL2‐negative rearranged follicle center entity, represents a fundamentally different biological process. ³⁵ Our integrated analysis of genomic and clinicopathological of PCFCL has identified three subgroups. Most samples were Group 1, with mutations in immune escape genes, that were recurrent in 54% of these tumors. In this regard, the diversity of immune evasion mechanisms has been well documented. ³⁶ Genetic alterations in FAS, ³⁷ B2M, ³⁸ and CIITA ³⁹ , ⁴⁰ are common in B‐cell lymphomas, although they have been less extensively studied in FL. IRF8 is a more recently implicated gene in immune evasion. Its mutations impair MHC class II antigen presentation by downregulating CD74 and HLA‐DM, leading to selective suppression of CD4⁺ T‐cell activation. ⁴¹ Regarding the previously published PCFCL series to date, Zhou et al. ⁷ identified IRF8, FAS, and B2M mutations in only a minority of skin‐restricted cases (3/16, 1/16, and 2/16, respectively) in their exome‐sequenced FFPE cohort. However, the lack of detailed case‐level clinicopathologic annotation in that study precluded direct assessment of phenotypic correlations. In the Wobser cohort, ⁸ a 40‐gene targeted lymphoma panel performed on fresh‐frozen tissue did not include IRF8, B2M, or CIITA, thereby preventing evaluation of these alterations. Similarly, Barasch et al. ¹⁰ reported no mutations in these genes, although interpretation is limited by the high rate of technical failures (“not obtained/failed”) in their FFPE‐based NGS panel.

Interestingly, Group 1 tumors showed a trend toward higher T‐cell infiltrates, suggesting a dysfunctional but “hot” tumor microenvironment. To note, these differences align with tumor localization, contrasting Group 1 (non‐photoexposed areas) with Groups 2 and 3 (photoexposed areas). This spatial distribution suggests that immune surveillance in PCFCL may be compromised through distinct oncogenic contexts: intrinsic genetic alterations in Group 1, and extrinsic factors—such as chronic sun exposure—in Groups 2 and 3. Regarding the latter, while UV‐related mutational signatures are well‐documented in cutaneous T‐cell lymphomas, ⁴² evidence in B‐cell lymphomas remains limited. ⁷ Because our targeted sequencing approach and the observed low mutational burden prevent a direct assessment of UV‐induced damage, future whole‐exome or whole‐genome studies are necessary to confirm whether sun exposure drives these specific PCFCL subtypes.

A differential feature between Group 1 and Group 2/3 is the expression of CD10. Group 2 samples shared STAT6‐activated mutations in known functional domains (DNA binding or src homology domains). ⁴³ , ⁴⁴ Notably, these tumors closely resemble the provisional entity BCL2‐R–negative CD23⁺ follicle center lymphoma described in the 2022 International Consensus Classification. ¹⁵ Group 2 cases were positive for both CD10 and CD23, in agreement with the literature supporting a germinal center origin. ⁴⁵ , ⁴⁶ Moreover, Group 2 showed a higher frequency of CREBBP mutations, consistent with molecular data indicating the functional co‐occurrence of CREBBP and STAT6 mutations.47, 48, 49 These events were also detected by Wobser (2/10) ⁸ and Barasch (2/18), ¹⁰ the latter including one CD23⁺/CD10⁺ case located in sun‐exposed skin, paralleling our observations. However, there are some differences between Group 2 PCFCL cases and BCL2‐R–negative CD23⁺ follicle center lymphoma. First, Group 2 showed a nodular/mixed pattern, unlike the diffuse growth described for the systemic cases. Second, the subset B2 of sFL BCL2‐R–negative cases characterized by the presence of STAT6 and CREBBP mutations and the absence of TNFRSF14 and EZH2 mutations frequently harbor mutations for KMT2D, which were not detected in Group 2 of PCFCL. ³⁵ , ⁴⁵ The functional relevance of these findings in the skin niche remains unclear, but STAT6 constitutive activation can activate the expression of IL‐4–responsive genes, leading to an IL‐4–rich microenvironment of the skin. ⁵⁰ The lack of significant clinical differences of this group supports the idea of convergent oncogenic mechanisms leading to an overlapping clinical behavior. ⁴⁶ Group 3 shared a similar clinicopathological profile with Group 2 but was defined by recurrent TNFRSF14 mutations. This alteration was more consistently observed across prior studies (Zhou et al. ⁷ 5/18; Wobser et al. ⁸ 3/10; Barasch et al. ¹⁰ 6/15), including at least one case demonstrating CD23/CD10 positivity and sun‑exposed localization—mirroring our observations. Finally, the oncogenic mechanisms of Group 3 PCFCL remain unclear, but it notably shares features with a recently described lower genital tract variant of PCFCL. ⁵¹ These cases harbored TNFRSF14 mutations without CREBBP, STAT6, or KMT2D mutations, or BCL2 rearrangements, showing predominant nodular/mixed IHC patterns.

The analysis of clinical relapses and synchronous tumors in five patients revealed identical immunoglobulin rearrangements, confirming a common clonal origin. However, FL_CI demonstrated genetic stability over time, suggesting limited subclonal evolution and a possible external tumor reservoir. Although the number of cases is limited, PCFCL showed marked genetic and phenotypic divergence, consistent with prior studies in systemic FL subclonal complexity. ⁵² , ⁵³ Notably, despite the presence of different subclones, we observed convergent evolution in PCFCL, with distinct subclones independently acquiring mutations in the same gene (such as TMSB4X and STAT6) at different nucleotide positions, a phenomenon also described in other B‐cell malignancies ⁵⁴ , ⁵⁵ under immune selective pressure and likely influenced by the complex cutaneous microenvironment. These findings support a model in which PCFCL relapses often represent the expansion of genetically distinct subclones from a shared precursor, rather than bona fide relapses of a dominant clone, in line with recent reports in systemic FL. ⁵⁶ , ⁵⁷ The analysis of a patient with multifocal PCFCL also shows that synchronous tumors may share a common clonal origin, supported by identical immunoglobulin rearrangements and CREBBP common mutations in both lesions. Interestingly, both tumors shared pathological and genetical characteristics of the same PCFCL group, Group 2 (CD10⁺, CD23⁺, STAT6 mutations, absence of mutations in immune escape genes). However, these two cases showed striking differences in other relevant pathological features such as BCL2 and Ki‐67 staining. In summary, this synchronic case, together with the genomic behavior of the longitudinal PCFCL samples, suggests that PCFCL may originate from a progenitor outside the skin capable of seeding multiple locations. Clinically, our findings support obtaining new skin biopsies in relapsed or multifocal disease to detect distinct subclonal populations and gain insight into the biology of this neoplasm. However, given the very small number of sequential and synchronous tumors analyzed, these observations should be interpreted with caution, and larger cohorts are needed to confirm these patterns.

In conclusion, PCFCL and FL_CI are distinct entities with different mutational profiles and prognoses. FL_CI genetically resembles systemic FL, while PCFCL may include different molecular subtypes related to specific mutations and tumor features. Despite shared clonal origins, FL_CI shows stable evolution, whereas PCFCL displays branched, subclonal, and convergent evolution. These differences support the need for detailed molecular analysis and re‐biopsy at relapse to guide management.

AUTHOR CONTRIBUTIONS

Andrea Combalia: Formal analysis; data curation; writing—original draft. Nuria Vidal‐Robau: Data curation; formal analysis; writing—original draft. Ferran Nadeu: Software; formal analysis. Cristina López: Data curation; writing—review and editing. Gerard Frigola: Formal analysis. Irene Lopez‐Oreja: Formal analysis. Noelia Garcia: Formal analysis. Melika Bashiri: Formal analysis. Pablo Mozas: Data curation. Armando Lopez‐Guillermo: Conceptualization. Itziar Salaverria: Writing—review and editing. Teresa Estrach: Conceptualization. Elias Campo: Conceptualization; writing—review and editing. Adriana Garcia‐Herrera: Conceptualization; supervision; writing—review and editing. Robert Albero: Conceptualization; writing—original draft; supervision.

CONFLICT OF INTEREST STATEMENT

A.C. has been a consultant for Takeda and Kyowa and received honoraria for speaking at educational activities for Kyowa. F.N. has received honoraria from AbbVie, AstraZeneca, Janssen, and SOPHiA GENETICS for speaking at educational activities and research funding from Gilead. A.L.‐G. served on the advisory board of Roche, Cel‐gene, Novartis, and Gilead/Kite and received grants from Celgene, Gilead/Kite. E.C. has been a consultant for GenMab; has received research support from AstraZeneca, and honoraria from ThermosFisher for speaking at educational activities; and is an inventor on a Lymphoma and Leukemia Molecular Profiling Project patent “Method for sub‐typing lymphoma subtypes by means of expression profiling” (PCT/US2014/64 161). F.N. and E.C. licensed the use of the protected IgCaller algorithm for SOPHiA GENETICS and Diagnóstica Longwood. The rest of the authors stated no conflicts of interest.

FUNDING

This work was supported by Ajut Josep Font to A.C. N.V.‐R. is a recipient of Contracte de Recerca “Clínic ‐ La Pedrera” grant from Hospital Clínic de Barcelona. R.A. is supported by the Spanish Ministry of Science through the Ramon y Cajal Program (RYC2021‐031 306‐I). R.A. is supported by the Spanish Ministry of Science through the Ramon y Cajal Program (RYC2021‐031 306‐I) and E.C. by the Spanish National Plan Project PID2021‐123054OB‐I00, funded by MCIN/AEI/10.130 39/501100011033 and by FEDER, “A way of making Europe”. This work was developed at the Centro Esther Koplowitz, Barcelona, Spain, and at Hospital Clinic de Barcelona, Barcelona, Spain.

Supporting information

Supporting Information.

Supporting Information.

Supporting Information.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.