Patient inclusion and follow-up
The retrospective SARC TCGA sarcoma cohort contained 265 patient samples in total (see transcriptomic data below). Subtypes included 105 leiomyosarcoma, 58 dedifferentiated liposarcoma, 49 undifferentiated pleomorphic sarcoma/malignant fibrous histiocytoma/high-grade spindle cell sarcoma, 25 myxofibrosarcoma, 10 synovial sarcoma, 9 malignant peripheral nerve sheath tumor and 9 other types with frequencies less than 2%. Neoadjuvant radiotherapy and chemotherapy were exclusion criteria for the cohort [
7]. The publicly archived dataset on sarcoma (TCGA, provisional) can be found at
http://www.cbioportal.org/.
For the Karolinska STS cohort, 33 patients, 18 male and 15 female, with STS of the trunk or the extremities were prospectively included between 2013 and 2015. Patients were diagnosed with a high-grade STS through a standardized multidisciplinary approach at the Sarcoma Center Karolinska, Karolinska University Hospital [
23]. Median age was 69 years (24–90). The most common histological diagnosis was undifferentiated pleomorphic sarcoma, present in 17 patients, whereas 7 patients were diagnosed with myxofibrosarcomas, 2 with angiosarcomas, 2 with malignant peripheral nerve sheath tumors, 2 with synovial sarcomas, 2 with malignant solitary fibrous tumors and 1 with leiomyosarcoma. Twenty-two tumors were deep seated, whereas 9 were subcutaneous. Twenty-two tumors were located in the lower extremities, 7 in the trunk or pelvis, and 4 in the upper extremities. Metastasis at diagnosis was present in one patient of the cohort. No patients received neoadjuvant radiotherapy or chemotherapy before tumor resection and collection of samples. Patient surveillance followed existing guidelines for high-grade STS [
23]. Clinical examination and chest X-ray were done every 3 months for the first 2 years of follow-up, then bi-annually. Local recurrence and lung metastases were documented. Median follow-up was 48 months. Of the 33 patients, 23 were still alive at the last follow-up. Metastases were detected in seven patients during follow-up and local recurrence in two patients. Overall survival was 85% at 12 months and 66% at 36 months, whereas metastasis-free survival was 70% at 12 months and 57% at 36 months.
Immunostaining
Formalin-fixed paraffin-embedded tumor sections with an average size of 0.8 cm2, and a thickness of 4 µm, were deparaffinized and rehydrated before heat-induced epitope retrieval at 110 °C for 5 min in a Decloaking NxGen Chamber TM (BioCare Medical). The unmasking buffer was selected according to the antibody product sheet recommendations with a preference for pH 6 (S2369, DAKO) when more than one buffer was listed. Sections were allowed to cool down for 30 min, equilibrated in TBS–Tween 20 (0.1%), and endogenous peroxidase activity was quenched by 3% H2O2 for 10 min if horseradish peroxidase-linked reagents were to be used. A 20-min incubation step with serum-free ready-to-use block (X0909, DAKO) was performed before applying the primary antibody. The antibodies used were directed against CD163 (NCL-L-CD163, Novocastra, 1:200), CD80 (MAB140, RnD, 20 µg/ml), CD8 (M7103, DAKO, 1:75), FOXP3 (ab20034, Abcam, 10 µg/ml, IHC), FOXP3 (12653, Cell Signaling, 1:100, immunofluorescence with liquid permanent red as below), CD68 (M0876, DAKO, 1:50), CD20 (M0755, DAKO, 1:200), CD19 (M7296, DAKO, 1:50) and PAX5 (12709, Cell Signaling, 1:100). Secondary detection reagents were chosen considering the species origin of the primary antibody and the type of enzyme label/visualization method preferred. For immunofluorescence, if not otherwise stated, Alexa Fluor 488 (A11001, Life Technologies, 1:300) and Alexa Fluor 594 (A11037, Life Technologies, 1:300) were used before mounting with Prolong diamond antifade mountant with DAPI (P36962, Life Technologies). For IHC, ImmPress reagent anti-mouse IgG, peroxidase (MP-7402, Vector laboratories), ImmPress reagent anti-rabbit IgG, peroxidase (MP-7401, Vector laboratories), ImmPress reagent anti-mouse IgG, alkaline phosphatase (MP-5402, Vector laboratories) or ImmPress reagent anti-rabbit IgG alkaline phosphatase (MP-5401, Vector laboratories) was used. Chromogenic substrates were DAB peroxidase substrate (SK-4100, Vector laboratories), liquid permanent red (K0640, DAKO), Vector blue alkaline phosphatase (SK-5300, Vector laboratories) or a combination of these according to colors shown in each image. Protocols for double labeling were optimized for sequential IHC with a second heat-induced epitope retrieval step at 80 °C for 5 min, followed by 20 min cooldown. If indicated in images, cell nuclei were counterstained with Mayer’s hematoxylin (01820, Histolab) before dehydration and mounting in permanent VectaMount mounting medium (H5000, Vector laboratories).
Histoscoring of images
Each tissue section (approximately 0.8 cm2) was assigned an IHC score, where 0, 1, 2 and 3 indicated negative, low, moderate or high abundance, respectively, of each cell type. Stratification into two groups (low versus high) was done before performing survival analysis. Histoscoring was performed blinded to patient outcome (Monika Ehnman, PhD) and an independent observer (Yifan Zhang, MD, pathology resident) assisted with scoring and estimating the number of positive cells/section for B cell markers. The visual assessment was based on whole tissue section analysis and not hotspot analysis. Due to low cellular abundance and a strong trend toward prognostic significance in the first analysis, two additional sections from all tumor samples were immunostained for CD20. This resulted in an average tissue area of 2.4 cm2/tumor analyzed under the microscope in total. If no CD20 signal was detected in two out of three sections, the tumor obtained the lowest IHC score. Cohen’s kappa coefficient (κ) was calculated to measure inter-observer agreement (low/negative versus high/positive) for CD20.
RNA extraction and quantitative real-time PCR
Tumor pieces were collected in RNAlater™ and RNA was subsequently isolated by Trizol followed by the GeneElute™ Mammalian Total RNA Miniprep Kit (Sigma-Aldrich) protocol including an on-column DNase digestion step. For cDNA synthesis, SuperScript II First-Strand Synthesis System for RT-PCR (Invitrogen) was used. SYBRgreen Universal PCR Master Mix (Applied Biosystems) was used in the PCR reaction with primers (Sigma-Aldrich) as follows: TTT GTC AAC TTG AGT CCC TTC AC (CD163, fwd), TCC CGC TAC ACT TGT TTT CAC (CD163, rev), ACG GCG CTG TCA TCG ATT (IL10, fwd), GGC ATT CTT CAC CTG CTC CA (IL10, rev), GCC CAG CAC TTC ACG CAT CAG (PTGS2, fwd), AGA CCA GGC ACC AGA CCA AAG ACC (PTGS2, rev), GCA GGT CGA GGA CTA TTT CTT TCA (NOS2, fwd), CGT AAG GAA ATA CAG CAC CAA AGA TA (NOS2, rev), CTC ATG GGC ACG GTG ATG (NOS3, fwd), ACC ACG TCA TAC TCA TCC ATA CAC (NOS3, rev), TGACACTGGCAAAACAATGCA (HPRT, fwd) and GGTCCTTTTCACCAGCAAGCT (HPRT, rev) with the latter primer pair used for normalization.
Statistical analysis
Statistical analyses were done on pseudonymized data in the Karolinska STS cohort using SPSS software version 20. Overall survival (OS) was computed from the date of diagnosis to the date of last follow-up or death, and metastasis-free survival (MFS) from the date of diagnosis to the date of last follow-up or first distant metastasis. Survival analysis was per Kaplan–Meier, the parameters tested were dichotomized around the median, and the log-rank was used for comparison between groups. Hazard ratios between groups were calculated using a multivariate Cox regression analysis (proportional hazards model), where prognostic factors identified in the univariate survival analysis were balanced for sex and age. A Spearman rank test was used for analysis of immune cell marker correlations and correlations with standard prognostic markers. All tests were double sided, no P value correction was applied, and a P value of < 0.05 was considered significant.
For the SARC STS cohort transcriptomics, the
Z score normalized gene expression values for MS4A1, CD19, IL10, PTGS2 and CD163 along with clinicopathological annotations were downloaded from cBioportal [
24,
25]. Complete data were available for 258 patients in the SARC sarcoma dataset (TCGA, provisional). Gene expression values were split across all tumors into equal sized tertile groupings. All gene expression analyses were performed in R version 3.4.2 using the dplyr (version 0.4.1), survival (version 2.41-3) and survplot (version 0.0.7) packages.
P values were adjusted for multiple testing using the Benjamini and Hochberg method with the
P.adjust function of the R stats package version 3.5.0. The cBioportal software analysis tool identified mutually exclusive mRNA upregulations (none significant), and gene pairs with co-occurrent mRNA upregulations (2 significant) by Fisher’s exact test, and the results are presented with, and without, Bonferroni adjusted
P values. All tests were double sided, and a
P value of < 0.05 was considered significant.