Immunohistochemical Algorithms and Gene Expression Proling in Primary Cutaneous B-cell Lymphoma

Objective: to assess whether immunohistochemical (IHC) algorithms used to classify the cell of origin (COO) of nodal Diffuse Large B-cell lymphoma (nDLBCL) in Germinal Center type (GCB) and non-GCB subtypes may be applied to Primary Cutaneous B-cell lymphoma (PCBCL) too, and which of these algorithms performs better on PCBCL. Design: retrospective case control study. Participants : fourteen PCBCL, including Primary Cutaneous follicle centre lymphoma (PCFCL) and primary cutaneous diffuse large B-cell lymphoma, Leg type (PCDLBCL-LT) and 14 nDLBCL were evaluated for 7-year period (January 2011 to December 2017). Primary cutaneous marginal zone cell lymphoma (PCMZL) cases were not included in the present study. Intervention: evaluation of immunohistochemical CD10, BCL6, MUM1/IRF4, BCL2, MYC and Ki-67 expression and classication according to three different algorithms. Gene expression proling (GEP) was performed on the same series using Lymph2Cx assay (Nanostring). The data obtained were compared and analysed. Results: all the IHC algorithms showed 13 GCB and 15 non-GCB. GEP showed 12 GCB, 12 activated B cell–type and 4 unclassied. Conclusions: the PCBCL were classiable as GCB and non-GCB like the nDLBCL as IHC algorithms were concordant to GEP and produced the same results. our series support the reliability of both GEP and IHC algorithms in the prognostic and predictive evaluation of PCBCL with accuracy comparable to that reported for nDLCBL Qiagen) and digital GEP was performed on 200 ng RNA using NanoString technology (Seattle, WA) with 23-Gene Signature for COO Classication Lymph2Cx (Nanostring Technologies, Seattle, WA, USA). GEP was performed on Affymetrix U133 plus 2.0 microarrays at Polo Tecnologico Pharmadiagen Srl Laboratory (Pordenone, Italy). The data algorithm nCounter-based Lymph2Cx Assay Misclassication Rate was 2%. GEP analysed cases were blindly classied according to the COO assignment [6,7]. primers MBR breakpoint nodal Diffuse, large, Lymphoma; Activated-B-Cell; Gene Proling.

CD79A  The pathological history in all the patients was negative for previous lymphoma.
After the histological diagnosis clinical staging was assessed in each of them by ow cytometry of peripheral blood, immunoglobulin gene rearrangements in lesional skin or in peripheral blood, complete and differential blood cell count, routine serum biochemistry analysis with lactate dehydrogenase (LDH), serum protein electrophoresis, liver function tests, computed tomography-uorodeoxyglucose-positron emission tomography (CT-FDG-PET) scan and bone marrow biopsy. The staging and therapy were assessed according to the European Society for Medical Oncology (ESMO) recommendations for primary cutaneous lymphoma [13]. All the patients, after speci c therapies, underwent strict clinical follow-up.
Immunoreactivity was evaluated without any knowledge of the patient's survival or other clinical data. IHC results were merged with the histological features of PCBCL and nDLBCL.

Gene Expression Pro ling
To assign the PCBCL to the GCB or activated B cell (ABC) groups, samples were analyzed with the Lymphoma/Leukemia Molecular Pro ling Lymph2Cx assay, a digital gene expression (NanoString)-based test for COO assignment in FFPE tissue (FFPET).   Each sample meeting the QC threshold is reported as one of the three molecular subtypes GCB, or ABC or Unclassi ed within the equivocal zone. Subtypes are determined using a Linear Predictor Score (LPS), which is computed by summing the products of 15 weighted gene coe cients and the gene expression measurements and applying pre-de ned thresholds [6,7]. Subtypes were ABC when LPS was higher or equal to 2433.5 (≥ 2433.5); GCB when LPS was lower or equal to 1907.8 (≤ 1907.8) and unclassi ed when LPS value was in equivocal zone (tab. 3).

t(14,18)(q32;q21) Polymerase Chain Reaction
To evaluate the t(14,18)(q32;q21) causing the BCL2-IGH rearrangements, genomic DNA was extracted by 8 sections (5-10 μm thick) using the Qiagen QIAamp DNA FFPE Tissue Kit (Qiagen, cat. n. 56404), following the manufacturer's instructions, after treatment with Qiagen Depara nization Solution (Qiagen, cat. n. 19093). The kit combines the selective binding properties of a silica-based membrane by allowing the genomic DNA puri cation from FFPET sections without overnight incubation. Once para n was dissolved in xylene, samples were lysed under denaturing conditions with proteinase K and incubated at 90°C to reverse formalin cross-linking. The DNA was bound to the membrane and contaminants owed through by centrifugation at 8000 rpm; DNA was then washed to remove the residual contaminants by centrifugation at 8000 rpm and then eluted from the membrane in 20-100 μl of Buffer ATE by centrifugation at 13200 rpm. In less than 30 minutes, the pure and concentrated DNA was ready for use in ampli cation reactions or for storage at -20°C.
The PCR analysis of t(14,18)(q32;q21) was carried out to evaluate BCL2-IGH rearrangement, according to Rambaldi A. et al. [20]. Brie y, the rst round of ampli cation was in 50 μl of nal volume of 1X GoTaq(R) Green Master Mix (2X; Promega, Madison, WI, USA, cat. n. M7123) using 100 ng of DNA with primers MBR for major breakpoint region, or mcr for minor cluster region and JH consensus region.
Samples were ampli ed on Veriti® Thermocycler (Applied Biosystems, Carlsbad, California, USA) with an initial denaturation at 95°C for 7 minutes, followed by either 27 cycles of denaturation for MBR or 30 cycles for the mcr. Each cycle was performed with 1 minute of denaturation at 94°C and 1 minute of annealing at 55°C for the MBR and 55°C for mcr ampli cations and 1 minute of extension at 72°C. A nested PCR reaction was performed using 5 μl of 1:10 dilution of the rst-round ampli cation product using oligonucleotide primers internal to the original primers, such as MBR nested, mcr nested and JH nested of which nucleotide sequences were described by Rambaldi A. et al. [20]. PCR reactions were performed according the following PCR temperature pro le: denaturation for 7 minutes at 94°C, 30 cycles (1 minute at 94°C, 1 minute at 58°C and 1 minute at 72°C for each cycle) and nal extension at 72°C for 10 minutes. A 25 μl aliquot of PCR product was veri ed on a 2% agarose gel containing ethidium bromide in Tris-borate electrophoresis buffer and visualized under UV light. Speci cally, positive and negative PCR controls were processed, and samples were run in duplicate to ensure PCR reproducibility.
Data comparison IHC data were compared to the GEP data to validate results and to one or more IHC algorithms to be used as GEP surrogate. Concordance percentage was calculated to determine the reproducibility between the GEP and IHC in PCBCL and nDLBCL. To evaluate the in uence of biological, histological, immunohistochemical, or molecular data on the distribution of cases with GEP and IHC, ANOVA analysis was carried out with a signi cance of 5% and with two tails for independent data. Statistical analysis was performed using Graphpad Prism 8.0®. Different clinical and histological features were evaluated in discordant cases as possible cause of disagreement.
The ow cytometry of peripheral blood, immunoglobulin gene rearrangements in lesional skin or in peripheral blood, complete and differential blood cell count, routine serum biochemistry analysis with LDH, serum protein electrophoresis, liver function tests were negative and CT-FDG-PET scan and bone marrow biopsy did not show systemic diffusion of corresponding PCBCL. As far the staging is concerned, all cases were T3-N0-M0-B0 according to ESMO recommendations [13]. All the patients underwent speci c therapies, namely PCFCL with solitary or localized skin lesion received radiotherapy at the dose of 24-30 Gy, PCFCL with multifocal lesions received rituximab and/or steroids, the PCDLBCL-LT received rituximab, cyclophosphamide, deoxyrubicin, vincristine, prednisone (R-CHOP).
All the PCBCL patients were alive at the moment of the study except one (case n. 1c) who died due to disease progression and eventually central nervous system involvement. All the above-reported data are summarized in Tables 1 and 2.

Histological features and classi cation
Four PCBCL (29%) showed monomorphic, diffuse in ltration of the dermis by large, roundish or irregular, nucleolated cells (monotonous proliferation of centrobasts and immunoblasts), with numerous mitoses. The cells did not show epidermotropism and small lymphocytes were scant or absent ( Fig. 1A and Tab. 1); these cases were diagnosed as PCDLBCL-LT. Among these cases, 2 PCDLBCL-LT (14% of PCBCL) showed reactive T cells, whereas in the remaining 2 cases T lymphocytes were few or absent (Tab. 1). Only one (25%) PCDLBCL-LT case showed skin ulceration and 2 (14% of PCBCL) cases showed necrosis and nuclear debris (Tab. 1).
Clinical and histological data of the 14 PCBCL and 14 nDLBCL are summarized in Tables 1 and 2.
The 4 out of 28 IHC vs GEP discordant cases were: 1 PCDLBCL-LT (case 4c) that resulted non-GCB at IHC algorithms and unclassi ed at GEP (Tabs. 1, 2 and 3); 2 nDLBCL (cases 3n and 4n) were non-GCB at IHC and unclassi ed at GEP, while the other case (13n) was GCB at IHC algorithms and unclassi ed at GEP (Tabs. 1, 2 and 3). Among discordant cases, 4c PCDLBCL-LT case was a 54-year-old female with a 15 mm lesion on the leg and local relapse, with reactive T cells, diffuse growth pattern, necrosis and nuclear debris, Ki-67 of 60% and IHC BCL2 positivity (Tab. 1). The 2 nDLBCL (cases 3n and 4n) that were non-GCB at IHC algorithms and unclassi ed at GEP were a female and a male aged 64 and 73 respectively, with 17 mm and 60 mm lesions on para-aortic lumbar and inguinal nodules, with reactive T cells (Tab. 2). The female nDLBCL (case 3n) showed diffuse growth pattern, Ki-67 of 70% and IHC BCL2 positivity, while the male nDLBCL (case 4n) showed diffuse growth pattern, nuclear and necrotic debris and starry sky pattern, Ki-67 of 90%, IHC BCL2 and t(14;18) positivity (Tab. 2). The 13n nDLBCL case GCB at IHC algorithms and unclassi ed at GEP was a male aged 59 with a 8 mm lesion on inguinal nodules, diffuse growth pattern, nuclear and necrotic debris and starry sky pattern, Ki-67 of 80% and t(14;18) positivity (Tab. 2).

Discussion
PCBCL are a rare and heterogeneous sub-group of primary cutaneous lymphoma; the latest WHO classi cation of lymphoid tumours has classi ed PCBCL in PCMZL, PCFCL and PCDLBCL-LT [1][2][3]. PCMZL is a de ned entity with speci c prognosis and therapeutic indications. The distinction between PCFCL and PCDLBCL-LT may be less de ned whereas prognosis and treatment may be different [13]. Therefore, an accurate sub-classi cation of these entities, like that of nDLBCL, might be useful. nDLBCL is the largest and most heterogeneous group of non-Hodgkin lymphomas (NHLs); they may arise de novo or from a pre-existing differentiated NHL. Evidence shows that their prognosis and response to therapies are in uenced by the corresponding COO; namely, GCB has a better response to therapy and longer disease-free survival than ABC and unclassi ed cases.
Numerous studies have investigated the COO classi cation of nDLBCL and its clinical relevance either by IHC or GEP, but little corresponding data on PCBCL are available so far [8,9,12,[21][22][23][24][25][26][27][28][29][30][31][32][33][34][35]. This is probably due to the heterogeneity of PCBCL, their relatively recent classi cation and their low incidence when compared to nDLBCL. Despite that, a COO sub-classi cation of PCFCL and PCDLBCL-LT would be desirable for prognostic and predictive evaluations as PCFCL are generally treated with radiotherapy only where PCDLBCL-LT require chemotherapy. A large and comprehensive study on GEP of PCBCL has been performed by Hoefnagel et al. [8]. In this study differences in GEP of PCFCL and PCDLBCL-LT have been detected and considered to be the cutaneous counterpart of nDLBCL GCB and ABC types, respectively. For this purpose, Hoefnagel et al. [8] investigated the whole genome of PCBCL and concluded that the different expression of few genes determines the distinction of PCBCL in 2 subtypes. This distinction was validated by quantitative PCR of seven genes and by the MUM1/IRF4 expression only by IHC. In our study we analysed 14 PCBCL and 14 nDLBCL cases by GEP, focusing on a small number of genes (23 Lymphoma/Leukaemia-related and 5 housekeeping genes) of the Lymph2Cx assay of Nanostring Technologies (Seattle, WA, USA) GEP, that are involved in cell cycle apoptosis, B-cell differentiation, B-cell receptor signalling, antigen presentation, chromatin regulation/DNA methylation [6,7]. IHC and GEP data of two parallel series of PCBCL and nDLBCL were compared to validate IHC data and to analyse the performance of three IHC algorithms frequently used in clinical practice. As previously reported, most PCFCL were CD10+ (70%, 7 out of 10) and BCL6+ (90%, 9 out of 10), but only 6 out of 10 (60%) showed BCL2 positivity. Only 3 PCFCL were MUM1/IRF4+ (80%, 4 out 5) while the all 4 PCDLBCL-LT were MUM1/IRF4+ (100%) and BCL2+ (100%). Therefore, BCL2 expression seems to differentiate PCDLBCL-LT from PCFCL with diffuse growth pattern and predominance of centroblasts, since BCL2 was strongly expressed in PCDLBCL-LT, as already observed elsewhere [39][40][41]. Moreover, the molecular analysis of t(14;18) con rmed that the BCL2-IGH translocation was negative in PCBCL [23,42]. Therefore, according to the literature, BCL2 over-expression in PCBCL seems to be associated with BCL2 ampli cation rather than with a translocation; conversely BCL2 over-expression is rare in PCFCL lacking either ampli cation or translocation [39].
In addition to the GEP, some studies have been conducted to identify PCBCL genetic differences by using whole genome sequencing or comparative genomic hybridization array analysis [29]. Different alterations were found including gains such as 18/18q; 7/7p; 3/3q; 1p; 12/12q; 13/13q or loss such as 6q of different chromosomes. Oncogene gains (SAS/CDK4, MYCL1, MYC, FGFR2, BCL2, CSE1L, PDGFB) or losses (AKT1, IGFR1, JUNB, FGR, ESR, ABL1, TOP2A, ERBB2, CCNE1, BCR) were also detected as demonstrated by Mao et al. by comparative genomic hybridization and microarray-based genomic analysis [29]. Multiple genomic mutations in the NF-kB and B-cell signalling pathways have been characterized in PCDLBCL-LT [26], similarly to MYD88 somatic mutation in PCDLBCL-LT [31]. MicroRNA studies on PCBCL have demonstrated the overexpression of microRNA 17-92 cluster and its paralogs miR-106a-363 in PCDLBCL-LT, that have a causative role in lymphomagenesis, when compared to PCFCL [22]. Conversely miR-9-5p, miR-31-5p, miR-129-2-3p and miR214-3p are overexpressed in PCFCL when compared to PCDLBCL-LT [28]. Nonetheless the same authors assess that microRNAs -reported as overexpressed in the ABC type when compared to the GCB type -in nDLBCL are not differentially expressed between PCFCL and PCDLBCL-LT [28]. Another study has highlighted different pro ling of genes of apoptosis and cytotoxic immune response being the rst enhanced in PCDLBCL-LT and the second in PCFCL [41]. PCDLBCL-LT and PCFCL may be differentiated for speci c chromosomal abnormalities too [24]. According to their study, performed by array-based comparative genomic hybridization, ampli cation of 2p16.1 and deletion of 14q32.33 chromosomes frequently occur in PCFCL [24]. Conversely ampli cation of 18q21.31-q21.33 including the BCL-2 and MALT1 genes and deletions of 9p21.3 were speci c of PCDLBCL-LT [24]. All these studies produced signi cant information on PCBCL but do not have clinical implications in terms of prognostic and predictive evaluation.
With reference to the IHC differentiation of the PCFCL and PCDLBCL-LT, Hallerman et al. [25] observed that BCL2, OCT2 and MUM1 positivity was associated with poor prognosis and BCL6 positivity with a better prognosis, whereas the differentiation between PCBCL and PCDLBCL-LT was less straightforward. Similar data were reported by Sundram U et al. [33]: analysing 14 PCBCL, they noted that the patients were divided into two groups: one with BCL6 positivity, MUM1 negativity and overall survival of 176 months; the second with BCL6 negativity, MUM1 positivity and shortest overall survival (26 months). Another study showed that PCFCL and PCDLBCL-LT have a germinal centre B-cell differentiation stage signature when applying gene pro ling on 17 PCBCL [9].
Consequently, most PCFCL were GCB when applying the three IHC algorithms, while most PCFCL with diffuse growth pattern and all the PCDLBCL-LT were ABC by GEP. Therefore, PCFCL with the follicular growth pattern might be equivalent to the nDLBCL GCB type and the PCFCL with diffuse growth pattern and PCDLBCL-LT might be referable to the nDLBCL ABC type.
Data obtained in our series support the reliability of IHC algorithms in the prognostic and predictive evaluation of PCBCL with accuracy comparable to that reported for nDLCBL. With reference to the different applied algorithms, Hans' and Muris' algorithms seem to be the most suitable for PCBCL.
In conclusion, our study con rms that GEP classi cation of PCBCL reproduces the classi cation of nDLBCL. A good correlation has been assessed between the GEP classi cation and the IHC expression of CD10, MUM1/IRF4, BCL6, BCL2 and MYC. These data may be utilized for the prognostic evaluation of PCBCL, but further validation on larger series is required.

Declarations
Contributor and guarantor information: P.Z, I.C., A.P, S.P., M.T., C.S., A. I. and A.L.P. conceived and designed the study. A.L.P., P.C, C.P, A.C. and P.Z. equally contributed in writing the text of the article. P.C, A.C, and C.B. collected patient clinical and histological data. P.C and A.C provided photographic material and performed statistical analysis. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Copyright/license for publication: the Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of all authors, a worldwide licence to the Publishers and its licensees in perpetuity, in all forms, formats and media (whether known now or created in the future), to i) publish, reproduce, distribute, display and store the Contribution, ii) translate the Contribution into other languages, create adaptations, reprints, include within collections and create summaries, extracts and/or, abstracts of the Contribution, iii) create any other derivative work(s) based on the Contribution, iv) to exploit all subsidiary rights in the Contribution, v) the inclusion of electronic links from the Contribution to third party material where-ever it may be located; and, vi) licence any third party to do any or all of the above.
Patient consent: written informed consent to perform this study was obtained from all the participants.
Competing interests declaration: