IVLBCL is a rare entity with poor representation of neoplastic cells in biopsy specimens. Histological analysis of pathological samples is mandatory for diagnosis. Several recent studies showed that random skin biopsy is the approach with the highest diagnostic yield and the lowest side effect and complication rate [6,7,8, 26]. Molecular studies are needed to better understand pathogenesis of this disease, to evaluate the prognostic impact of specific genomic alterations, to guide treatment decisions and new treatment development, and to accurately detect early relapse during follow-up.

In consideration of all these elements, we investigated whether skin samples collected during the diagnostic work-up could be used for whole-exome next generation sequencing. Several limitations had to be addressed to get appropriate material for comprehensive genomic studies. First, DNA quality is of paramount importance for subsequent analyses. To this issue, we used fresh-frozen material which is known to allow excellent tissue preservation and to avoid artifacts arising from formalin-fixation. Furthermore, high quality DNA allows to work with limiting DNA quantity. Second, the neoplastic cell fraction is extremely low in tissue biopsies, since tumor cells are located exclusively in small vessels which can’t be extensively and selectively sampled. To solve this problem, we were able to dissect subcutaneous tissue which is enriched in small vessels, and has otherwise low cellularity. This could be performed manually with a scalpel, without the need for laser capture microdissection equipment.

So far, most molecular studies have been limited to targeted-sequencing on DNA derived from either micro-dissected tumor cells or cfDNA [7, 15]. Tumor cell microdissection is an extremely time-consuming and operator-dependent procedure. On the other hand, cfDNA can be problematic too. For example, since no pathognomonic mutations have been found so far, the chance of nonspecific finding is to be considered. As such, MYD88 L265P mutation may be the fortuitous finding of other lymphoproliferative disorders -such as IgM monoclonal gammopathy of unknown significance (MGUS), lymphoplasmocytic lymphoma/Waldenstrom macroglobulinemia, or DLBCL-, rather than the rare IVLCBL. Similarly, CD79B Y196 may be found frequently (around 30%) in activated-B cell (ABC)-DLBCL, especially in those aggressive cases with extranodal clinical presentation, the so called “MCD” or “cluster 5” subgroups, as defined by Schmitz and Chapuy, respectively [13, 14]. Conversely, in our approach, these mutations can be directly assigned to the tumor clone present in the biopsy.

Targeted studies focused on a panel of genes, selected on the assumption of a non-GCB origin of IVLBCL, have been important in confirming aberrations in pathways such as the B-cell receptor (BCR)-signaling pathway, the Toll-like receptor/Interleukin-1 receptor (TLR/IL-1R) pathway, and the nuclear factor kappa B (NFκB) pathway. However, more appropriate and specific gene panels could be established only after comprehensive genome-wide studies. In fact, to date there are no genomic data able to explain the typical histological pattern of this disease nor to confirm mutations in genes whose products were previously described as potential players of intravascular distribution (such as CD29 or CD54) [5].

Until recently, the only attempt of comprehensive genomic studies was performed on DNA derived from both cfDNA and BMMNC, although the cohort was entirely formed of Japanese patients, and therefore may not be representative of the worldwide IVLBCL cohort. Yet no genomic features leading to the typical intravascular pattern have been found. Furthermore, analysis of bone marrow samples is limited to the cases with bone marrow involvement. More recently, a WES study was performed by Kodgule and colleagues on post-mortem tissues involved by IVLBCL [16]. In their study, they found recurrent mutations in the RAC2 gene, and postulated that this mutation may play a role in IVLBCL cells adhesive features. However, in the aforementioned WES study by Shimada and colleagues, RAC2 mutations were identified in only a minority of cases, and this particular finding was not emphasized [9]. We were not able to detect any RAC2 mutations despite thorough evaluation. Given that our analysis was confined to a single patient, and the two other studies reported RAC2 mutations in only a subset of patients, this finding underlines the possibility that, for some IVLBCL cases, exclusive intravascular location may be attributed to different mediators [9]. Further comprehensive studies are needed to evaluate the recurrence and significance of RAC2 mutations in IVLBCL.

In this study, we found that fresh-frozen subcutaneous sample allows comprehensive genomic studies. In fact, DNA quality and quantity were high, and the cancer cell fraction could be artificially enhanced by selection of the subcutaneous selection. The resulting tumor DNA fraction, albeit low, was amenable to sequencing analysis at medium depth. However, it was likely still too low to reliably identify CNAs. Confirming the feasibility and robustness of our approach, our results align with those obtained from workflows using different sequencing techniques, sample types and manipulations, and geographical areas. In fact, we confirmed the two most frequently described mutations in IVLBCL, CD79B and MYD88 [7, 9, 15, 16]. Moreover, we found a similar number of mutations as compared to the two other WES analyses [9, 16]. In line with other reports, we confirmed that our sample bears multiple mutations in genes targeted by SHM (PIM1, TMSB4X, MPEG1 and OSBPL10) [9, 16]. Since we could only analyze one sample, we cannot exclude that genes with multiple mutations in our dataset could also be artefactual or arise from other localized hypermutational processes such as those mediated by aberrant activity of apolipoprotein B mRNA editing enzyme catalytic polypeptide (APOBEC) [27]. Although limited by the small sample size which is partly due to the extreme rarity of this entity, we believe our study may represent an important proof-of-concept which could be further explored. Our methodology, which is time-efficient and cheap, can be adopted by any pathology laboratory, and could be easily integrated in the standard diagnostic workup of the patient, which is based on random skin biopsy even in the absence of skin lesions. The main limitations of our approach include its operator-dependent nature during the cutting of the biopsy specimen and the necessity to freeze the sample which can be resource-intensive. On the other hand, identifying subcutaneous tissue before cutting is easily feasible, and freezing IVLBCL biopsies should not strain resources, given its rarity.

Therefore, this approach, if validated, may be employed in future studies to better define the genomic landscape of this rare entity, through whole genome sequencing studies which can be carried out at increasing depths using modern technologies and can detect subclonal populations [28]. Extensive and widespread adoption of such pipeline may lead to recognize the specific aberrations associated with this entity, its different clinical presentations, response to treatment, and CNS relapse risk. In case pathognomonic mutations should be found, they could also be used as an important tool to confirm diagnosis and guide measurable residual disease assessments in follow-up. Unravelling the characteristic pathogenetic features may in turn lead to personalized treatments through the use of targeted panels [29,30,31]. Interestingly, MYD88 L265P has already been described as a target of ibrutinib, whereas little is known about CD79 mutations and polatuzumab vedotin efficacy. Moreover, since the microenvironment contribution on disease pathogenesis is likely limited, IVLBCL may be a good model to identify tumor-intrinsic key drivers, as opposed to DLBCL which is much more influenced by interaction with non-neoplastic cells most times.