Molecular characterization of subcutaneous panniculitis-like T-cell lymphoma reveals upregulation of immunosuppression- and autoimmunity-associated genes

Altogether, the study included 20 SPTL patients from three different European nations (Table 1), of whom four pre-treatment skin samples were analyzed by gene expression microarray, five samples by quantitative RT-PCR, and 23 samples by immunohistochemistry (IHC). Of one of the patients included in the array analysis, also an affected inguinal lymph node was biopsied and available for immunohistochemistry. Additionally, three Finnish SPTL patients were followed during treatment with oral prednisolone and low-dose methotrexate (Table 1) and altogether, a set of eight follow-up skin samples were obtained and analysed with the microarray. The demographic details of the patients are given in Table 1 and clinical presentation of lesions in Figure 1.

For the microarray analysis, fresh subcutaneous tissue samples were successfully obtained from four patients with newly diagnosed SPTL (cases 1–4, Table 1). The first samples were taken at the time of diagnosis, before any treatment (pre-treatment), together with a sample for TCR (T-cell receptor) rearrangement analysis (http://​www.​hus.​fi/​sairaanhoito/​laboratoriot/​Sivut/​default.​aspx). The follow-up samples were first obtained three to six months after the start of systemic treatment (treatment1) and second follow-up samples six to 12 months after the initiation of the therapy (treatment2) when a complete clinical response was reached (biopsy from the region of pre-existing lesions). In addition, one treatment1 sample was obtained from a patient (case 5, Table 1) with no matching pre-treatment or treatment2 sample. The control RNA for microarray studies consisted of two samples from normal subcutaneous fat tissue (FAT1 and −2, derived from patients undergoing dermatologic surgery) and two cases of non-malignant panniculitis, erythema nodosum (EN1 and −2). The study was approved by the Medical Ethical Review Board of Helsinki University Central Hospital.

For the confirmatory immunohistological studies six cases of lupus erythematosus profundus (LEP, i.e. lupus panniculitis), and 13 cases of EN were included. All LEP controls were female (mean age 38.3 years) and two (=33%) of them had a pre-existing collagenosis with immunosuppressive therapy (steroids and hydroxychloroquine). Eleven of 13 (85%) EN cases were female (mean age 37.9 years). In seven (54%) cases the etiology of EN was an infection (Yersinia in 67% of them), in two (15%) EN was the first symptom of a systemic disease (colitis ulcerosa and sarcoidosis), and in four, the etiology remained uncertain regardless of thorough examinations. All of the samples were taken prior to immunosuppressive treatment (except for the two LEP cases with a pre-existing collagenosis).

The fresh, skin biopsies were immediately immersed in RNALater™ (RNA Stabilization Reagent, Qiagen, Valencia, CA) and stored at −20/70°C. The subcutis and the deeper dermis of the skin biopsy were dissected for the RNA isolation, which was performed with RNeasy Mini Kit/RNeasy Lipid Tissue Mini Kit (Qiagen) according to manufacturers’ instructions. In SPTL samples, the amount of malignant T-cells was over 50% of the mononuclear cell infiltrate, based on cytomorphology in histopathological analysis.

The RNA used was intact and of high quality (RIN 8.0-10), as confirmed with Agilent 2100 Bioanalyzer at the Biomedicum Helsinki Functional Genomics Unit (FuGU, http://​www.​helsinki.​fi/​fugu/​). The gene expression arrays (Human Exon 1.0ST, Affymetrix) were performed at FuGU according to manufacturer’s instructions. Microarray data are available in the ArrayExpress database [www.​ebi.​ac.​uk/​arrayexpress] under accession number E-MTAB-910. (Username: Reviewer_E-MTAB-910, Password: wiknooqq, Experiment E-MTAB-910).

Data from microarrays were pre-processed using background correction and quantile normalization [8]. For each gene, probe set intensities were summarized to obtain a single expression value. Differential expression analysis was done for the following comparisons: (1) pre-treatment SPTL (n = 4) vs. normal subcutaneous fat tissue (n = 2), (2) pre-treatment SPTL (n = 4) vs. EN (n = 2), and (3) pre-treatment SPTL (n = 4) vs. combined controls (aforementioned normal subcutaneous fat tissue and EN; n = 4). In each comparison, genes with median fold change (FC) >4 (<0.25) and p-value of t-test <0.05 were considered as differentially expressed. Data analysis was done using the Anduril bioinformatics framework [9]. Due to the small number of samples available for this rare disease, false discovery correction for p-values was not used. Rather, the key findings were confirmed with qRT-PCR and immunohistochemistry.

We confirmed the expression of three relevant genes, CXCR3, IDO-1, and IFNG by quantitative RT-PCR. The SPTL RNA samples (n = 5) were extracted either from fresh RNA later stabilized skin tissues (cases 2–3 in Table 1 used also in the arrays) or from formalin-fixed-paraffin-embedded (FFPE) skin tissues (cases 1, 5–6 in Table 1, NucleoSpin FFPE RNA 740969.10 Macherey-Nagel GmbH, Germany) according to manufacturer’s instructions. Three EN RNA samples were used in qRT-PCR as reference tissue. Two samples were the same used in the arrays (EN1 and EN2) and the third was extracted from the FFPE sample. The reverse transcription into cDNA was performed using SuperScript® VILO cDNA Synthesis kit (11754–050, Invitrogen). We used the following Taqman Assays (IDO-1; Hs00984148_m1, 66 bp n = 5, IFNG; Hs00989291_m1, 73 bp n = 4, CXCR3; Hs00171041_m1, 111 bp n = 4) and iQ Supermix (170–8860, Bio-Rad) and LightCycler 1.5 System (Roche) for the amplifications. The size and the purity of the amplicons were checked with agarose gel electrophoresis (2,5% SeaKem® LE agarose, Rockland, ME USA 1xTBE). The relative expression levels were normalized to reference gene GAPDH (Taqman assay 4310884E, 118 bp) and further compared to expression levels in reference tissue, erythema nodosum according to 2^-∆∆CP-method [10].

As part of routine diagnostics at the Dermatopathology laboratory of Helsinki University Central Hospital, all tissue samples were immunostained for the following markers (manufacturer and dilutions given in parenthesis): CD3 (Novocastra, New Castle, UK; 1:100), CD4 (Novocastra; 1:150), CD5 (Novocastra; 1:25), CD7 (Novocastra; 1:100), CD8 (Novocastra; 1:25), CD30 (Dako, Glostrup, Denmark; 1:25), CD56 (Zymed, South San Fransisco, CA, USA; 1:50), GZMB (Monosan, Uden, The Netherlands; 1:100), TIA1 (Biocare, Birmingham, UK; 1:200), Ki-67 (MIB-I antibody, Dako, Glostrup, Denmark; 1:50), and TCR alpha/beta (GeneTex, TX, USA; 1:100) according to the manufacturers’ instructions, and visualized with DakoEnvision (Glostrup, Denmark).

Additionally, immunohistochemical (IHC) detection of the following proteins CXCL9 (Abcam, Cambridge, UK; 1:500), IL2RB (Abcam, 1:200), IDO-1 (Chemicon International Inc. USA; 1:100, clone MAB5412), FoxP3 (SpringBioscience; 1:50 clone SP97), and CXCR3 (Abcam, 1:500), was performed according to the manufacturers’ instructions and ImmPRESS Universal Antibody (anti-mouse Ig/anti-rabbit Ig, peroxidase) Polymer Detection Kit (Vector Laboratories, Burlingame, California) and NovaRED (Vector Laboratories, Burlingame, CA) or AEC (Abcam) chromogens. Furthermore, double IHC stainings were performed for CD8 (1:100)/IDO-1 (1:100), and also for CD68 (Spring Bioscience, Pleasanton, CA, USA, 1:200)/IDO-1 (1:100) according to the manufacturers’ instructions and using Vector Elite PK-6101 Rabbit IgG (Vector Laboraties)/Permanent HRP Green Kit KDB10049 (Nordic BioSite AB, Täby, Sweden) and VECTASTAIN AP Mouse IgG Kit(Vector Laboratories, AK-5002)/Permanent AP-Red Kit, BCB20041 (Biosite), respectively. Double IHC staining for CD8/CXCR3 was performed using MACH2 Double Stain 2 Mouse-HRP + Rabbit-AP Polymer Detection Kit (cat.nro 901-MRCT525-021709, Biocare Medical, Concord California) with BCIP/NBT and AEC as chromogens, respectively. Moreover, double immunofluorescence (IF) staining for CD11c (Bio SB, Santa Barbara, CA, 1:50)/IDO-1(1:100) was performed according to the manufacturers’ instructions and using AlexaFluor-594 anti-rabbit antibody (red, Abcam, 1:1000) and AlexaFluor-488 anti-mouse antibody (green, Abcam 1:1000), respectively. Immunofluorescence stainings were analyzed and photographed using Leica Confocal Microscopy (Leica Microsystems). The IHC detection was carried out on the total of 42 FFPE tissue samples, obtained from 20 SPTL patients (23 samples), six cases with LEP and 13 cases of EN. For each IHC, several technical controls were included for both positive and negative reactions. The positive staining result was graded as follows: − indicates <10%, +10-25%, ++ 25-50%, and +++ over 50% of the lymphocytes expressed the given marker.

We first compared SPTL skin samples to normal subcutaneous fat tissue to exclude the effect of normal fat gene expression. We identified altogether 968 genes differentially expressed in untreated/pre-treatment SPTL skin samples, out of which 589 were up regulated and 379 were down regulated. The three most strikingly overexpressed genes in the SPTL lesions were chemokine (C-X-C motif) ligand 10 (CXCL10; fold change: 171), guanylate binding protein 5 (GBP5; FC: 78), and indoleamine 2,3-dioxygenase (IDO-1; FC: 71). Likewise, the expression levels of CXCL11 (FC: 41) and CXCR3 (FC: 10), were elevated. Moreover, interleukin 2 receptor β (IL2RB; FC: 46), chemokine (C-C motif) ligand 5 (CCL5 = RANTES; FC 35), and interferon gamma (IFNG; FC 17) were highly expressed. The data also shows the overexpression of perforin 1 (PRF1; FC 33), different granzymes (e.g. GZMA; FC 25, GZMB; FC 23), and members of the SLAM family (SLAMF1; FC 6, SLAMF6; FC 36, SLAMF7; FC 29, and SLAMF8; FC 23). Interestingly, of the TRIM family genes (TRIM59; FC 10 and TRIM14; FC 6) were upregulated in this comparison (Table 2).

We then compared the pre-treatment SPTL samples with in silico “combined controls” of normal subcutaneous fat tissue and non-malignant panniculitis, erythema nodosum (EN) samples, i.e. the median values yielded by the latter two groups. A visualization of 290 genes that are overexpressed between SPTL samples and combined controls is presented as a heatmap (Figure 2a). In this comparison, the aforementioned genes stayed overexpressed, but to a slightly weaker extent (Table 2). Three most overexpressed genes were IGJ, IDO-1, and CXCL10 (Table 2). To note, the overexpression of CXCL9 (FC 20) was seen only in this combined comparison (Table 2). Furthermore, the 99 genes that are annotated with the Gene Ontology term “Defence response” are visualized in a heatmap as well (Figure 2b).

When comparing SPTL samples against inflammatory EN samples only, six genes (PRF1, KLRD1, IGJ, KIR2DS4, GZMB, and IDO-1) showed the highest expression and appeared equally overexpressed by 20-fold. Also, IFNG, IL2RB, and CXCR3 stayed overexpressed but to a slightly weaker tendency as observed earlier (Table 2). Of the CXCR3 ligands, only CXCL11 reached significant overexpression in this comparison (Table 2). This would suggest some similarity between SPTL and EN regarding the CXCR3 pathway involved in the development of autoimmune diseases (reviewed by [11]).

Consistent with previous findings, the expression of GZMB gene encoding the cytotoxic protein GZMB, used in the routine diagnostics of SPTL, was constantly approximately 20-fold overexpressed in all comparisons. Likewise, Fas ligand (FASLG, TNF superfamily, member 6) was equally overexpressed by 5-fold in all comparisons. In addition, up regulation of RASGRP1, a nucleotide exchange factor specifically activating Ras, was observed in SPTL samples (FC 21, FC 8, and FC 4). From the NK gene family, only NKG7 (natural killer cell group 7 sequence) was overexpressed by 23-, 18-, and 14-fold, respectively. Other up regulated genes in this comparison remained in line with the other comparisons. Of the down regulated genes, several T-box transcription factors (TBX18, TBX15) were represented. Among the miRNA family (hsa-miR-199a-2, hsa-miR-410, hsa-miR-487-b, and hsa-miR-3665) were down regulated by 5-10-fold, respectively. On the other hand, miR-219-1 was overexpressed (FC 4). The list of discussed, deregulated genes is summarized in Table 2.

In the set of the additional follow-up samples, obtained during the systemic therapy with prednisolone and low-dose methotrexate, a clear transition towards normalization of the most relevant genes, like IDO-1, was observed as early signs of response at the time when the malignant T cell population was still clearly detectable histologically (Additional file 1: Figure S1).

The expression levels of three selected deregulated genes, CXCR3, IDO-1, and IFNG were confirmed by quantitative RT-PCR from five cases and normalized to reference gene, GAPDH. The relative mRNA expression levels were then compared with the levels in reference tissue (erythema nodosum, EN) and results were presented as fold changes. IDO-1 mRNA showed overexpression by 30-350-fold in all SPTL samples compared with EN (Figure 3). The cytokine receptor CXCR3 showed overexpression by 10-50-fold and IFNG by 50-150-fold compared to EN. We did not detect any CXCR3 or IFNG expression in case 5, although the GAPDH levels were similar to other samples. Overall, the quantitative analysis not only confirmed the expression of IDO-1 and CXCR3 but also revealed even higher fold changes than the microarray analyses.

To further confirm the cellular origin of the deregulated gene products, we performed IHC of selected gene products in an extended series of patient and control tissue samples. IDO-1 was seen intensively expressed both in some morphologically malignant T-cells rimming the fat cells (Table 3, Figures 4b and 5c) as well as in nearby CD8⁻CD68⁻CD11c⁻ cells (Figures 5a,c,f) [12],[13], when using double immunofluorescence. Surprisingly, IDO-1 expression was not detected in CD11c + dendritic cells with confocal microscopy performed on double-IF stained sections (Figures 5f-h). We also searched for FoxP3-positive regulatory T-cells (Tregs), since IDO-1 has been reported to increase the proportion of Tregs in the tumor infiltrate [14]. Approximately 25-50% of the inflammatory lymphocytes were FoxP3-positive in SPTL samples (Figure 4f) with abundant expression of IDO-1 (Table 3). FoxP3⁺ lymphocytes were found also within the inflammatory infiltrates of EN and LEP, both with no IDO-1 expression, but at a considerably lower frequency (Table 3).

CXCL9 and CXCR3 were chosen as markers for CXCR3 pathway. CXCR3 protein was expressed almost exclusively in the malignant lymphocyte infiltrates of the SPTL samples (Table 3, Figures 4c,e, 5b). The non-malignant inflammatory infiltrates of LEP and EN control samples expressed CXCR3 protein, too, but to a varying extent. (Table 3, Figure 4d). CXCL9 was abundantly expressed in the malignant cells of 15 SPTL samples (Figure 4a, Table 3) while notably less so in all LEP and EN samples (Table 3). By double immunostaining of SPTL samples for CD8 and CXCR3, we confirmed that the malignant lymphocytes, typically rimming the adipocytes, mostly co-expressed both markers (Figures 4c and 5b). In LEP, the CXCR3 and CD8 were not co-expressed by the same cells (Figure 4d).

Interestingly, we could also examine the biopsies of affected lymph nodes of one of the SPTL patients (case 4, Table 1). An intense malignant T-cell infiltration, rimming the fat cells in the adipose tissue surrounding the nodes, was found. The pattern of IDO-1 and CXCR3 expression was similar to that seen in the cutaneous SPTL lesions (Figure 5d-e), as well as FoxP3 expression (data not shown). To conclude, markers of CXCR3 pathway, typically involved in autoimmune diseases, were expressed both in SPTL and in control (LEP and EN) cases, but in SPTL, the main source of CXCL9 and CXCR3-positive cells was the malignant, CD8+ lymphocyte infiltrate (Table 3).

Microarray data are available in the ArrayExpress database [www.​ebi.​ac.​uk/​arrayexpress] under accession number E-MTAB-910 (see Methods for more details).