Stat3-positive tumor cells contribute to vessels neoformation in primary central nervous system lymphoma

With the aim of elucidating the relationship between Stat3 expression and tumor vessels abnormalities in the PCNLs, in this study we evaluated Stat3 and pStat3 expression by Real-time PCR and by immunohistochemistry in biopsy sections from PCNSL patients. Correlations of the expression levels with the presence of aberrant vessels were analyzed by confocal laser microscopy analysis, using FVIII as endothelial cell marker, CD133 and nestin as cancer stem cell (CSC) marker, CD20 as tumor cell marker, and Stat3. In addition, we investigated Stat3 mutations in lymphoma cells to clarify the role of the constitutive expression of Stat3 and of its phosphorylated forms. Results showed that in PCNSL, putative endothelial cells lining the vessels are heterogeneous, expressing FVIII/ pStat3/CD133 (presumably originally they are vascular progenitor cells), as well as FVIII/CD20/CD133 (presumably originally they are tumor cells). Finally, we detected a fraction of the FVIII+ endothelial cell that co-expressed Stat3 bearing a tetraploid karyotype, while no amplification signal for the Stat3 gene was detected.


INTRODUCTION
Primary Central Nervous System Lymphoma (PCNSL) has an aggressive course and poor prognosis, with a median survival of less than 20-24 months [1,2]. PCNSL accounts for 1-3% of all primary brain tumors [3], but its incidence has increased in recent decades. The majority of PCNSL are diffuse large B-cell Lymphomas [4] and lymphomatous cells are typically localized in the perivascular areas [3].
Angiogenesis has a prognostic value in human NHLs [5], and PCNSL [6]. Expression of vascular endothelial growth factor (VEGF) in PCNSL cells is correlated with microvascular density, with longer survival and blood-brain barrier alterations [6]. We have previously demonstrated that different type of cells are involved in the formation of vascular wall in PCNSL [7]. Cancerlike stem cells (CSCs) may be involved in formation of tumor vessels, including brain tumors, through a cross-talk between endothelial cells and CSCs [8][9][10].
With the aim of elucidating the relationship between Stat3 expression and tumor vessels abnormalities in the PCNSL, we evaluated Stat3 and pStat3 expression by Real-time PCR and by immunohistochemistry in biopsy sections from PCNSL patients. Correlations of the expression levels with the presence of aberrant vessels were analyzed by confocal laser microscopy analysis, using FVIII as endothelial cell marker, CD133 and nestin as CSC marker, CD20 as tumor cell marker, and Stat3. In addition, we investigated Stat3 mutations in lymphoma cells to clarify the role of the constitutive expression of Stat3 and of its phosphorylated forms.

Endothelial and tumor cells express Stat3
Stat3 immunohistochemical expression showed high levels of total and phosphorylated Stat3 protein in tumor brain compared to normal brain. Stat3 showed an intense staining in tumor tissues ( Figure 1A) compared to normal brain lacking Stat3 labeling ( Figure 1C).Tumor vessels were lined by Stat3 + cells and tumor cells were identified near to the vessel wall ( Figure 1B). After Real-time PCR ( Figure 1D) a significant difference in Stat3 expression between tumor and control brain tissues was observed. To better identify changes in the expression of phosphorylated and total form of Stat3, immunohistochemical reactions were performed in the same areas of tumor tissue. Stat3 displayed a higher expression (Figure 2A) compared to pStat3, which strongly labeled the nuclei of the endothelial cells and clusters of tumor perivascular cells in the same area ( Figure 2B). pStat3 was not detected in the control brain tissue ( Figure 2C). Morphometric analysis showed an increased expression of Stat3 and pStat3 in the tumor brain tissues compared with control tissues ( Figure 2D). After dual confocal immunofluorescence reactions, tumor vessels appeared lined by endothelial cells expressing both FVIII and Stat3 signals ( Figure 3A-3C) while in the control brain tissue only FVIII + vessels were detectable ( Figure 3D).
Overall, these data show, for the first time, that tumoral FVIII + vessel in PCNSL are formed by CD20 + / CD133 + cells ( Figure 7) as well as that FVIII + cells express both CD133 and pStat3 ( Figure 6). Moreover, morphometric analysis shows a significant increased expression of Stat3, CD133, nestin and CD20 protein in PCNSL compared with control brain (Figure 8).

Genetic analysis of Stat3 in PCNSL
To better understand the overexpression of Stat 3 in the vessel wall and tissues of PCNSL specimens, genetic analysis of the Stat3 gene was performed. Interphase FISH analysis and the FICTION technique on the FFPE tumor samples, and Stat3 gene sequencing on genomic DNA, were carried out. The FICTION technique detected red and green aberrant fluorescent spots in the nuclei of CD20 + cells ( Figure 9A). In detail, the Interphase FISH analysis, showed four red (Chr:17q21.2, Stat3) and four green (Chr:9q34.11) fluorescent spots in the nuclei of PCNSL cells, indicative of a tetraploid karyotype ( Figure 9B). After FICTION with FVIII and Stat3 dual immunofluorescence reaction, and FISH with the Stat3 gene (red) and chromosome 9q34.11 (green) probes, tumor vessels showed a merge fluorescent colocalization signal in some vascular tracts, while in other tracts the signals were separate ( Figure 9C-9D). Moreover, Stat3 + cells with a tetraploid karyotype were detected in the vessel wall ( Figure 9E).
All Stat3 exons and adjacent intronic regions were sequenced on genomic DNA isolated from biopsy samples derived from PCNSL specimens, but no pathogenic variants were found. Only two Single Nucleotide Polymorfisms (SNPs) were detected, the Stat3 rs2293152 G-allele (minor allele count/MAF = 0.346) and rs3830585 dupT-allele (minor allele count/MAF = 0.402).

DISCUSSION
Avascular phase followed by a vascular phase characterizes growth in both solid and haematological tumors [25,26]. Different mechanisms are involved in tumor angiogenesis: sprouting angiogenesis; non-sprouting angiogenesis (intussusceptive microvascular growth); postnatal vasculogenesis and vasculogenic mimicry.
Mounting evidence has shown that Stat3 is strongly linked to tumor angiogenesis and metastasis [27,28] and is related to poor prognosis in different tumors [15,29]. Anti-tumor effects of Stat3 knockdown has been demonstrated by means of small interfering RNA (siRNA), micro-RNA (mi-RNA), or small molecule inhibitors [30][31][32][33]. Inhibitory molecules, which downregulate Stat3 phosphorylation and induce apoptosis in tumor cells have been developed [34][35][36][37][38][39][40][41][42]. Moreover, previous studies have shown that metformin, an antidiabetic drug, could inhibit Stat3 phosphorylation and reduce risk of development in many type of cancer [43,44]. Therefore Stat3 is a prognostic indicator and a therapeutic target. In this study we observed pStat3 expression in tumor PCNSL vessels and cells, moreover we found high levels of total and phosphorylated Stat3 protein in tumor brain compared to normal brain. Elevated levels of the phosphorylated protein have been reported in human glioblastoma multiforme (GBM) samples [45] and in GBM cell lines [47]. IL-6 and IL-10 activate Stat proteins [47]. IL-6 contributes to tumor metastasis via the JAK/Stat3 signaling pathway [48,49], and IL-10 phosphorylates of Stat3 in PCNSL cells [50], while the role of IL-6 in PCNSL remains poorly defined.   We found that pStat3 was elevated in PCNSL brain samples compared to normal tissues. Overexpression of Stat3, as shown by Real-time PCR and immunohistochemical reactions, might mediate the cellular response to cytokines and growth factors [51,52].
In this study we also demonstrated Stat3 expression in FVIII + endothelial cells that lined PCNSL vessels and in FVIII + cluster of cells in PCNSL brain samples while in control samples endothelial cells were FVIII + . Moreover in tumor tissues, cells co-expressing pStat3 and CD133, and vessels co-expressing CD133, pStat3, CD20 and FVIII markers, were detected. Therefore, putative endothelial cells lining these vessels are heterogeneous, expressing FVIII/pStat3/CD133 (presumably originally they are vascular progenitor cells), as well as FVIII/CD20/CD133 (presumably originally they are tumor cells), as already suggested for CSCs [53]. Interestingly, Stat3 is required for the maintenance of pluripotency in murine stem cells , and is necessary for self-renewal of murine neural stem cells [58,59]. Neural stem/progenitor cells has been identified in bioptic specimens from different tumors [60]. Moreover, neural stem cells and progenitors cells have been labeled by nestin [61] and CD133 [62,63].
The data are in agreement with previously published studies showing that tumor cells can give rise to tumorassociated endothelial microvessels in human B-cell lymphomas, multiple myeloma, and malignant melanoma [64][65][66][67]. In this work we have demonstrated, for the first time, that FVIII positive tumor vessels are formed also by CD133, nestin and pStat3 positive cells, suggesting that in PCNSL CSCs might contribute to the tumor vasculature. CSCs reside in a vascular niche named the CSC niche [68],   and stimulate tumor angiogenesis. Tumor vasculature, in turn, supports CSC self-renewal and maintenance. CSCs produce high levels of VEGF [69], and recruit endothelial precursors for revascularization and tumor re-growth [70].
In addition, we further analyzed the tumor vasculature by combining FVIII and Stat3 dual immunofluorescence reactions and fluorescence in situ hybridization using the Stat3 locus specific probe (Chr:17q21.2) and the probe for chromosome 9q34.11.
We detected a fraction of FVIII + endothelial cell that coexpressed Stat3, bearing a tetraploid karyotype, while no amplification signal for Stat3 gene was detected. This polyploidy, a common feature of tumor cells, supports the notion that the tumor vascular endothelium in our PCNSL samples is of neoplastic origin. We also identified the same chromosomal aberration in CD20 + tumor cells. These findings confirm that in PCNSL, either the tumor cells can differentiate into endothelial cells.  Genomic analysis of tumors demonstrates significant genetic intra-and inter-tumor heterogeneity [71]. Colorectal cancer endothelial cells overexpress specific transcripts as a result of qualitative differences in gene profiling as compared with endothelial cells of the normal colorectal mucosa [72]. Further studies in glioma [73] and in invasive breast carcinoma [74] demonstrated a distinct gene expression pattern related to tumor endothelial cells, which cells acquire genotype alterations, leading to altered anti-angiogenic targets and resistance [75], and the contiguity of tumor cells and endothelial cells may be responsible for the genotype alterations [76].
Overall, these data suggest that in PCNSL, putative endothelial cells lining the vascular wall are heterogeneous, and are formed by two cellular populations, expressing FVIII/ pStat3/CD133 (presumably originally they are vascular progenitor cells), as well as FVIII/CD20/CD133 (presumably originally they are tumor cells).

Patients and tissue samples
The clinical and anatomical features of the patients investigated in this study are reported in Table 1. All patients were affected by histologically proven primary human diffuse large B-cell PCNSL. Surgical specimens were fixed in 4% formaldehyde, routinely processed and paraffin-embedded. Three samples of histologically normal brain, removed in the course of surgical exposure, were used as control. The study was approved by the local Ethics Committee at the University of Bari Medical School, and all the patients gave their informed consent in accordance with the declaration of Helsinki.

Morphometric analysis of Stat3 and pStat3 expression
For each case, three slides stained for Stat3 and pStat3 expression were scanned using the wholeslide scanning platform Aperio Scanscope CS (Leica Biosystems, Nussloch, Germany). All the slides were scanned at the maximum available magnification (40 ×) and stored as digital high resolution images on the workstation associated with the instrument. Digital slides were inspected with Aperio ImageScope v.11 software (Leica Biosystems, Nussloch, Germany) at 20 × magnification and ten fields with an equal area were selected for the analysis at 40 × magnification. Stat3 and pStat3 expression was assessed with the Positive Pixel Count algorithm embedded in the Aperio ImageScope software and reported as positivity percentage, defined as the number of positively stained pixels on the total pixels in the image. The statistical significance of differences between the mean values of the percent labeled areas between patients and control brain tissues was determined by the 2way Anova test in GraphPad Prism 5.0 software (GraphPad software, La Jolla, CA, USA). Findings were considered significant at P values < 0.05.

Real-time PCR
Total RNA was extracted from FFPE blocks tissues of 20 patients and 3 normal control brain, using RecoverAll TM Total Nucleic Acid isolation kit (Ambion, Life Technologies, Inc., Austin, TX,USA) and then used to synthesize the first-strand c-DNA with the IScriptcDNA Synthesis kit (Bio-Rad Laboratories, Hercules, CA, USA), according to the manufacturer's instructions. For the detection of Stat3 expression, cDNA was amplified with the iTaq SYBR Green supermix using a ROX kit (Bio-Rad Laboratories). PCR amplification was performed using the Chromo4 real-time PCR Detection System (Bio-Rad Laboratories). Samples were normalized to human RPLPO (large ribosomal protein PO). Table 2 show the sequences of primers (Sigma-Aldrich) used for Stat3 amplification.

Dual and triple immunofluorescence-confocal laser scanning microscopy
Twelve micrometer thick deparaffinized brain sections were incubated for 30 minutes in a blocking buffer [BB; phosphate-buffered saline (PBS), pH 7.4, 1% bovine serum albumin, 2% fetal calf serum] and exposed to primary antibodies: (i) goat anti-Stat3 (ab5073, Abcam) and rabbit anti-FVIII (A0082, Dako), diluted 1:100 and 1:50 in BB, respectively, overnight at 4°C; (ii) mouse anti-pStat3 (sc56747, Santa Cruz Biotechnology) and rabbit anti-CD133 (ab19898, Abcam), diluted 1:50 and   A sequential scan procedure was applied during image acquisition of the two fluorophores. Confocal images were taken at 200 nm intervals through the z-axis of the section covering a total depth of 10 µm. Images from individual optical planes and multiple serial optical sections were analyzed, digitally recorded and stored as TIFF files using Adobe Photoshop software (Adobe Systems Inc., San Jose, CA, USA). Morphometric analysis was performed by two independent observers on ten randomly selected fields observed at 63 × magnification by using Cell^F as image analysis software (Olympus Italia,Rozzano,Italy).

Morphometric analysis of nestin + /FVIII − , nestin + / FVIII + , CD133 + /FVIIIand CD133 + /FVIII + vessels in PCNSL sections
nestin + /FVIII − , nestin + /FVIII + , CD133 + /FVIII − and CD133 + /FVIII + vessels were evaluated from five optical fields randomly chosen for each PCNSL samples at 200 × magnification. The images were acquired using the confocal fluorescence microscope (Leica) with an integrated camera. The software Cell^F as image analysis (Olympus Italia,Rozzano,Italy) was applied for the vessel counting. The values were presented as percentage of positive vessels for each staining over the total vessels number. The data were expressed as mean value ± SD.

Morphometric analysis of pStat3 and CD133 cell expression
The diameter of the pStat3\CD133 labeled and unlabeled cells were evaluated using Leica Confocal Multicolor Package (Leica Microsystems) on single optical planes of randomly chosen fields from ten sections per tumor and control samples with 63 × oil lenses. The measure was made by optimized contrast and brightness enhancement functions and digital filters. The Graph Pad Prim 5.0 statistical package (GraphPad Software, San Diego, CA, USA) was used for the analysis and P < 0.05 was considered as the limit for statistical significance. The data were expressed as mean value ± SD.

Immunofluorescence and fluorescence in situ hybridization analysis (FICTION)
Fluorescence immunophenotyping and interphase cytogenetics (FICTION), a technique combining immunofluorescence and FISH, was carried out on 4-µmthick paraffin sections of the PCNSL samples. For the identification of the chromosomal Stat3 signal in tumoral cells FISH assay was performed and then the sections were stained with either anti-CD20 antibody, a cytoplasmic tumoral cells marker, or with anti-FVIII, an endothelial cells marker, and anti-Stat3. In brief, deparaffinized and rehydrated slides were next incubated at 96°C in Tris/ EDTA acid buffer solution for 15 minutes, washed in sterile water and treated in 0.01 N HCL solution at 37°C for 2 minutes. Enzymatic digestion was then performed by adding 200 µl 0.4% pepsin (Sigma-Aldrich) solution and incubating at 37°C for 7 minutes. Thereafter, tissue samples were washed with sterile water, dehydrated in ethanol, and air dried. Hybridization was done with the Stat3 probe and the centromeric probe of chromosome 9 as control. Probe and target were co-denatured at 75°C for 5 minutes, followed by overnight hybridization at 37°C on StatSpinThermoBrite (Abbott Molecular, Abbott Park, Illinois, USA). Post-hybridization washing was carried out at 72 ± 1°C in 0.4XSSC for 2 minutes and in 2XSSC/0.1% Tween at room temperature for 1 minute. Then the slides were counterstained and mounted with DAPI-Antifade (Cytocell, Tarrytown NY). After posthybridization washing, immunostaining was performed. The tissues were treated with normal goat serum at room temperature for 30 minutes and incubated overnight with mouse monoclonal anti-CD20, rabbit polyclonal anti-FVIII (M0755 and A0082, Dako) and goat polyclonal anti-Stat3 (ab5073,Abcam) primary antibodies at 22°C then washed in PBS buffer and incubated at room temperature for 2 hours, with Alexa Fluor 488 goat anti-mouse, Alexa Fluor 555 donkey anti-rabbit and Alexa Fluor 488 donkey anti-goat (Invitrogen) antibodies, respectively, used as secondary reagent. Afterwards, the slides were washed, counterstained with DAPI (Invitrogen), and mounted in Vectashield (Vector Laboratories Inc.). Images were captured using an Olympus BX51 microscope fitted with an Olympus DP70 camera, equipped with filter sets for DAPI (nuclei counterstaining), FITC (cytoplasmic immunofluorescence signal and green nuclear FISH signal) and TRITC (cytoplasmic immunofluorescence signal and red nuclear FISH signal). Hybridization signals were counted in 200 morphologically intact nuclei for each sample.

DNA extraction and sanger sequencing
Genomic DNA was extracted from biopsy sections derived from PCNSL specimens using the QIAamp Mini Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions, and quantified on a BioSpectrometer Plus (Eppendorf, Hamburg, Germany). The entire coding regions of Stat3 (RefSeq NM_139276.2), including all splice junctions and adjacent intronic sequences, were amplified by standard PCR protocols using Taq DNA Polymerase (Thermo Scientific, USA) and the primer pairs listed in Table 3. Direct sequencing was performed using the BigDye Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems) according to the manufacturer's instructions on an ABI 310 Genetic Analyzer (Applied Biosystems). The sequence analysis software Alamut® (Interactive Biosoftware) was used to interpret variants. Online databases including dbSNP (Database the Single Nucleotide Polymorphism Database), 1000 Genomes, ClinVar, EXAC (Exome Aggregation Consortium), COSMIC (Catalogue of Somatic Mutations in Cancer), ESP (Exome Sequencing Project), as well as on line search engines (e.g. PubMed, LOVD), were used to search for previously described variants.

Statistical analysis
Data are reported as means ± SEM. Student's t-test was used for two-group comparisons and Newman-Keuls multiple comparison post-test was used to compare all treatment groups following one-way ANOVA. The Graph Pad Prim 5.0 statistical package (GraphPad Software, San Diego, CA, USA) was used for the analysis and P ˂ 0.05 values were considered statistically significant.