Proliferative, pro-inflammatory, and angiogenesis regulator gene expression profile defines prognosis in different histopathological subtypes of nodal peripheral T-cell lymphoma

Nodal peripheral T-cell lymphoma (PTCL) is an aggressive and heterogeneous malignancy with poor prognosis. We studied the prognostic impact of the expression profile of genes related to cell proliferation (CCNA2, TOP2A, and CHEK1), pro-inflammatory activity (NFkB1 and IKBkB), and angiogenesis (VEGF1) in nodal PTCL outcomes, as well as the ability of this genomic panel to discriminate different histological subtypes. We investigated the relative expression of regulator genes in 63 nodal PTCL patients. CCNA2, TOP2A, CHEK1, and NF-kB1 proteins were also assessed by immunohistochemistry. The median patient age was 47 years, 57.1% were male, 34.9% were diagnosed with PTCL-NOS, 28.6% with ALK-/ALCL, 22.2% with ALK+/ALCL, and 14.3% with AITL. The proliferative genes were associated with worse 3-year OS and PFS in PTCL-NOS and better 3-year PFS in ALK-/ALCL. Expression of CCNA2≥median and overexpression of CHEK1 protein (HR 3.793; p = 0.007) were associated with worse OS for all the cohort of nodal PTCL (HR 1.418; p = 0.001). The genomic expression profile tested in this study was not able to discriminate the different subtypes of nodal PTCL, although it showed a distinct prognostic significance between PTCL-NOS and ALCL-ALK. Overexpression of the CCNA2 gene and CHEK1 protein were associated with poor prognosis in the total nodal PTCL cohort.


INTRODUCTION
Peripheral T-cell lymphoma (PTCL) belongs to a heterogeneous group of rare neoplasms of mature T lymphocytes, representing 10% to 15% of lymphoid malignancies [1]. The incidence of these lymphomas varies according to the geographic region; with a higher frequency in Southeast Asia and in Central and South Research Paper www.oncotarget.com America, where they represent up to 20% to 25% of all non-Hodgkin lymphomas (NHLs) [2]. The most recent World Health Organization (WHO) classification of lymphoid and hematopoietic tissue neoplasms divides PTCL according to clinical aspects into predominantly nodal, extra nodal, primary cutaneous, and disseminated or leukemic subtypes [3].
The predominantly nodal PTCL represents 60% to 70% of all PTCL, and includes peripheral T-cell lymphoma, not otherwise specified (PTCL-NOS), angioimmunoblastic T-cell lymphoma (AITL), and anaplastic large cell lymphoma (ALCL), which is further subdivided intoanaplasticlymphoma kinase (ALK) positive and negative variants [4]. However, overlapping morphological and immunophenotypical aspects among the different subtypes of nodal PTCL are not uncommon, leading to inappropriate diagnosis [3].
In some cases, cytogenetic and molecular features may also help to individualize a subtype of PTCL, but this kind of analysis is not routinely available in most centers. Therefore, novel biological markers are needed to better differentiate PTCL subtypes. Moreover, few biological prognostic markers have been described for PTCL, such as ALK expression associated witht(2;5)(p23; q35) involving the gene ALK on chromosome 2, and ALCL [5]. Similarly, IRF4 and DUSP22 rearrangement are associated with ALK-negative ALCL and deletion of chromosome 6q in nasal type natural killer lymphoma [6,7].
Except for ALK-positive ALCL, nodal PTCL patients have short survival, with a 5-year overall survival (OS) of 30% to 40%. This is partially related to the presence of worse clinical and laboratory features at diagnosis in these patients. Nodal PTCL patients are usually older, with high levels of dehydrogenase lactic enzyme (LDH), high risk of International Prognostic Index (IPI), and bone marrow infiltration. Nodal PTCLs may exhibit high P-glycoprotein (gpP) expression, which is associated with multiple drug resistance (MDR) phenotype and p53 mutation [8]. The main prognostic indexes used for nodal PTCL are the IPI [9], and the prognostic index for PTCL(PIT) [10].
To better characterize the molecular heterogeneity and physiopathology of nodal PTCLs, and to provide more rationality in its treatment, we evaluated the prognostic impact of cyclin A2 (CCNA2), DNA topoisomerase Iiα(TOP2A), checkpoint kinase 1 (CHEK1), factor nuclear kappa B (NF-kB1), inhibitor of nuclear factor kappa B kinase subunit β (IKBkB), and vascular endothelial growth factor 1 (VEGF1)gene expression and CCNA2, TOP2A, CHEK1 and NF-kB1 proteins in nodal PTCL at a single cancer treatment center in Brazil. This is a hypothesisdriven study, and the genes chosen were selected based on the findings of Cuadros et al. [11] which implicated such genes in the biology and pathogenesis of these neoplasms.
We also evaluated the potential of these biomarkers to discriminate different histological subtypes of nodal PTCL.

RESULTS
Of the 94 patients initially enrolled in this study, 7 (7.5%) were excluded due to unavailability of medical records or after histology revision. Two cases previously diagnosed as nodal PTCL were changed to primary cutaneous ALCL and another to adult T-cell leukemia/ lymphoma. Therefore, clinical and epidemiological data were available for 87 (92.5%) patients, but 24 (27.5%) were excluded due to lack of paraffin blocks and failure at RNA extraction or cDNA amplification, remaining 63 (72.5%) patients available for gene expression study and 54 (62.6%) for IHC analysis.
None of the six genes exhibited a prognostic correlation with OS and PFS for ALK-positive ALCL or AITL. Overexpression of VEGF1 was associated with worse 3-year PFS for ALK-negative ALCL ( The analyzed gene panel was not able to discriminate any specific histological subtypes of nodal PTCL. However, all genes, except VEGF1, demonstrated a typical and linear pattern of distribution of median expression intensity in decreasing order: PTCL-NOS > ALK-ALCL > ALK+ ALCL ( Figure 4).

Multivariate analysis
In multivariate analysis, CCNA2 expression ≥ median (

DISCUSSION
In this study, we showed that overexpression of CCNA2 gene and CHEK1 protein is associated with worse OS for nodal PTCL independent of the IPI, and ECOG ≥ 2 is the main independent variable for 3-year PFS. Our results are partly in agreement with those obtained by Cuadros et al. [11], who analyzed the gene expression profiles of 35 nodal PTCL patients (23 PTCL-NOS and 12 AITL) using two different microarray platforms.
These authors identified five distinct clusters comprising genes related to cell proliferation, inflammatory processes, and B-cell and T-cell lineages. They observed that a predominance of genes related to cell proliferation and cell cycle control, such as CCNA, CCNB, TOP2A, and PCNA, were significantly associated  with poor prognosis [11]. However, in our study, univariate analysis showed that TOP2A protein ≥ 70% was associated with better OS for the whole cohort of nodal PTCL patients and worse prognosis for the PTCL-NOS subgroup. This contradictory result may be explained by the differences between the studies; Cuadros et al. studied only PTCL-NOS and AITL histological subgroups, whereas we analyzed more cases and included patients with an ALCL diagnosis.
The TOP2A gene encodes the enzyme topoisomerase 2 alpha, which is a common target of anticancer therapy (e.g., etoposide). This enzyme participates in the topological arrangement of DNA during RNA transcription. Consequently, TOP2A inhibition may disrupt the DNA double helix during the S phase of the cell cycle and subsequently induce apoptosis of malignant cells [12].
In T-cell lymphoma, the addition of etoposide to the CHOP backbone has had contradictory results. Schmitz et al. [13] demonstrated that this approach improves survival for ALK-positive ALCL patients, but not for other subtypes of PTCL. However, a recent study was not able to show similar results [14]. As many patients in our study were treated with CHOP, we suppose that the inclusion of etoposide has contributed to the overcome of worse prognosis associated with TOP2A overexpression, with exception to the PTCL-NOS group.
For PTCL-NOS, univariate analysis revealed that overexpression of CCNA2, IKBKB, and CHEK1 conferred shorter 3-year OS. These results were not surprising, as IKBKB encodes a protein that phosphorylates the inhibitor of the inhibitor/NF-kB complex. Consequently, NF-kB is activated, triggering strong proliferative and inflammatory responses in the cell [15]. The constitutive activation of NF-kB in various types of solid tumors and hematological cancers, such as Hodgkin lymphoma (HL) and multiple myeloma, has been previously described [16,17]. NF-kBrelated genes cooperate to increase tumor progression, cell migration, inflammation, and vascular proliferation, and have been used as targets in cancer therapy [18,19]. In contrast to our data, one study that analyzed the expression of 109 genes associated with NF-kB pathways in 62 PTCL patients identified two different subgroups of PTCL, and the subgroup presenting up-regulated NF-kB genes was associated with better OS [20].
Interestingly, as CHEK1 is known as a DNAdamage response (DDR) coordinator [21], individuals carrying inactivated mutations in this gene are more susceptible to developing many types of cancer, as seen in Li Fraumeni syndrome patients. In our study, the expression of CHEK1 was associated with worse prognosis for PTCL-NOS and better survival for ALKnegative ALCL patients. Physiologically, CHECK1 activation triggers the cell cycle checkpoint, cell cycle arrest, and DNA-repair mechanisms, as well as cell death to avoid formation of damaged cells [22].
Recently, CHEK1 was tested as a target for anticancer drugs. Ravi et al. showed that the small molecule proteasome inhibitor ixazomib inhibits tumor growth in xenograft T-cell lymphoma, and HL in mice by downregulating CHEK1 and Myc [23]. The authors' conclusion was that CHEK1, as well as Myc, work as targets for ixazomib in cancer, and chemotherapy protocols including anti-CHEK1 drugs may represent a new approach for T-cell lymphoma therapy. They also showed that inhibition of CHEK1 activates caspase-dependent apoptosis in mouse models of B-cell lymphoma [23].
Hoglund et al. described the ability of Myc to induce CHEK1 transcription via a mechanism independent of the DNA damage response, such as ATM and p53 [24]. Furthermore, as we found that CHEK1 was related to a shorter OS for nodal PTCL in the multivariate analysis, and for PTCL-NOS in the univariate analysis, we hypothesize that ixazomib should be tested in these diseases, as proposed by Boonstra et al. in his article [25]. As the treatment options available for T-cell lymphoma are scarce, we consider that clinical trials with new drugs are urgently needed for these highly aggressive groups of diseases with very poor prognosis.
The contradictory results for CCNA2 and CHECK1 in the whole cohort/PTCL-NOS and ALK-negative ALCL subgroups are in agreement with theories that gene effects are dependent of the interaction of intracellular networks of molecules, cell connections, the microenvironment, and epigenetic factors [26]. Furthermore, CHEK1 is related to different prognosis in distinct types of cancer, and even in particular subgroups of the same tumors [27].
The reproducibility of gene expression profiling studies is strongly dependent on the amount of hypotheses and variables tested. Considering a large number of variables being analyzed, and a small population size, higher are the possibilities of finding false positive correlations and false positive results [28]. In addition, few studies have focused on demonstrating the role of genetic and epigenetic issues in PTCL; most of them have shown broad heterogeneity of results with low reproducibility.
PTCL is a rare disease of difficult diagnosis and hard to determine different histological subgroups. Consequently, the detection of recurrent genetic abnormalities is challenging. Furthermore, tumor sample quality must be high for PTCL diagnosis, since a broad IHC panel is often needed to establish an accurate diagnosis [5] Currently, there is not a gold standard microarray platform for gene expression analysis that is universally recommended in the literature. Researchers worldwide have been applying different sets of methods for this goal.
Surprisingly, ALK-negative ALCL patients presented higher survival rate than ALK-positive ALCL patients. At the time of study enrollment, our protocol did not incorporate consolidation with ASCT for highrisk ALK-positive ALCL in first CR, but only for ALKnegative patients in first CR/PR independent of the risk index. As half of our population of ALK-positive ALCL patients had a high risk IPI, we infer that this different approach for treatment could partially explain the contradictory outcome. Recently, d'Amore et al. showed that, inside the nodal PTCL, the ALK-negative ALCL subgroup benefits greatly from ASCT, with a 5-year OS of 70% versus 47% and 52% for PTCL-NOS and AITL, respectively [29].
Although ALK protein has been associated with a good prognosis, it is not always seen as an independent prognostic factor [30,31]. Other authors have demonstrated that the prognosis related to ALK is dependent on other variables, such as age, IPI, CD56 expression, and B2 microglobulin [32][33][34]. Recently, Sibon et al. [34] demonstrated in 138 cases of ALCL that patients younger than 40 years of age presented similar OS regardless of ALK status. Thus, the high proportion of young patients in our ALK-negative ALCL subgroup may also explain the better outcomes observed in this group.
De Leval et al. and Piccaluga et al. demonstrated that many genes involved in angiogenesis are overexpressed in AITL, especially VEGF1 [35,36]. In AITL, high levels of VEGF1 are associated with bone marrow infiltration by tumor cells, cutaneous involvement, and high risk IPI. These authors suggested that VEGF1 could be used as a biomarker of PD in AITL [37]. We did not obtain similar results, as VEGF1 expression was not associated with prognosis in our AITL population. However, in univariate analysis, we found that overexpression of VEGF1 was associated with worse PFS in ALK-negative ALCL. Moreover, we question whether VEGF1 should be explored as a molecular target for therapy in ALKnegative ALCL.
In agreement with the literature, the most common subtype of PTCL in our study was PTCL-NOS. However, we also found a higher prevalence of ALCL than AITL, which is the second most prevalent PTCL in Europe [5]. Interestingly, the higher prevalence of ALCL was in accordance with North American epidemiology, where ALK-positive ALCL predominates in comparison to AITL [38]. However, the lower prevalence of AITL in our population could be partially explained by the difficulty of diagnosis. AITL presents peculiar morphological features that require adequate specimens, preferably obtained by excision biopsy to allow identification of the arboriform aspect of the vases, of the uneven arrangement of FDCs, and of the paracortical expansion of EBV-positive large B immunoblasts [3].
We observed a high rate of discrepancy between previous diagnosis and after the centralized review in 10.3% (9/87) of cases. This was similar to what was previously described by the International T-cell Lymphoma Project, which reported a 10.4% discrepancy [5]. These data suggest that more specific diagnostic markers are necessary to improve the accuracy and reproducibility of the subtypes of PTCL classification. Moreover, in the histology review, we added more IHC markers to identify follicular helper T-cells, the neoplastic cells of AITL, as well as CXCL-13, PD-1, and ICOS. This approach allowed us to identify more cases of AITL, and more stringent morphological criteria allowed us to identify hallmark cells with strong expression of Ki-1 (CD30), which was essential for differentiating ALKnegative ALCL from CD30+ PTCL-NOS.
In conclusion, we found that overexpression of CCNA2 and CHECK1 protein are associated with worse OS in nodal PTCL. However, high expression of CCNA2 and CHEK1 had an opposite effect on the prognosis of PTCL-NOS and ALK-negative ALCL. No gene or protein studied was able to discriminate subgroups of nodal PTCL. Future studies with more patients are necessary to confirm these findings and improve our knowledge of PTCL.

Study population
After local ethics committee approval, 94 patients between 16 and 86 years old with a diagnosis of predominantly nodal PTCL, treated at a reference center for cancer in Brazil, were retrospectively evaluated from January 2000 to December 2015. Patients with primary cutaneous CD30+ T-cell lymphoproliferative disease, heart failure grade III/IV, unavailable paraffin blocks or medical records were excluded.
Baseline clinical and disease features, including age, gender, Ann Arbor stage, number of extranodal sites involved with lymphoma, LDH dosage, Eastern Cooperative Oncology Group (ECOG)performance status, B symptoms, and bulky disease (tumor size ≥ 10 cm or over one-third of the cardiothoracic index) were taken from medical records by a specific researcher. HBV, HCV, HIV, and HTLV-1 serology, a comprehensive biochemical panel, including kidney and liver function, echocardiogram, bone marrow biopsy, and neck, chest, abdomen, and pelvis computed tomography (CT) were performed at diagnosis. The patients were stratified in clinical stages I, II, III, and IV according to Ann Arbor criteria [39].
The interim response, after the fourth cycle of chemotherapy and the final response at the end of treatment, was evaluated using the Cheson 2014 criteria [40]. The IPI and PIT scores were calculated for the cohort as described previously [9,10]. ALK-positive ALCL patients were treated with six to eight cycles of CHOP-21 (cyclophosphamide 750 mg/sqm IV on day 1, vincristine 1.4 mg/sqm [max 2 mg] IVon day 1, doxorubicin 50 mg/ sqm IV on day 1 and prednisone 100 mg/day PO from day 1 to 5). The majority of ALK-negative ALCL, PTCL-NOS, and AITL patients were treated with six to eight cycles of CHOEP-21 (CHOP plus etoposide 100 mg/sqm IV days 1 to 3).
Patients with bulky disease, sinus, bone, testis, breast, and Waldeyer's ring involvement underwent field radiotherapy with 3600 cGy at the end of chemotherapy. Autologous hematopoietic stem cell transplantation (ASCT) was indicated for consolidation in first line therapy for patients in complete remission (CR) or partial remission (PR), except for ALK-positive ALCL and patients >65 years of age or with clinical contraindication. ASCT was indicated for ALK-positive ALCL only in second CR or PR. After the end of treatment, patients were monitored by clinical and laboratory exams every 2 months in the first year, every 3 months in the second year, every 4 months in the third year, every 6 months in the fourth year, and annually after the fifth year.
If HE was suggestive of ALCL, with atypical lymphoid infiltrate of anaplastic large cells or CD30positive "hallmark cells">70% [41], IHC for PAX-5, CD15, and ALK was added to exclude Hodgkin lymphoma. PTCL-NOS was diagnosed by exclusion in cases exhibiting atypical and diffuse small to medium lymphoid infiltrate, lacking criteria for characterization of any other category of nodal PTCL [3], and after an extensive IHC panel for T-lymphoid markers with CD2, CD4, CD5, CD7, CD8, TCR α/β, and TCR γ/δ ( Table 4). All cases categorized as PTCL-NOS consistently presented negative serology for HTLV-1.

Molecular biology
Total RNA was extracted from FFPE tissues (4-μmthick) using the Qiagen FFPE RneasyKit (Valencia, CA, USA) with a modified deparaffinization step [42]. Briefly, sections were deparaffinized by two repeated incubations in xylene for 10 minutes, followed by two repeated incubations in 100% ethanol for 5 minutes, and then washed with distilled water for 30 seconds. After deparaffinization, the remaining steps of RNA extraction were followed according to the Qiagen FFPE RNeasy Kit manual. All RNA samples were analyzed spectrophotometrically (NanoDrop 1000 Spectrophotometer V3.7, Thermo Fisher Scientific, Wilmington, DE, USA). cDNA was synthesized from 1.0 µg of total RNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA) according

Statistical analysis
The univariate analysis to assess the association among categorical variables was performed using the Mantel-Haenszel chi-square test (mhodds). A Cox univariate analysis was performed to estimate the association between categorical variables and survival curves. The log-rank test was used to compare survival curves and to verify the association between categorical variables and survival curves. Variables that were significant in the univariate analysis were tested in a multivariate analysis using proportional hazard model stepwise backward selection to establish independent variables.
OS and progression-free survival (PFS) curves were estimated by the Kaplan-Meier method. OS was measured from the date of diagnosis to the date of death from any cause and was censored at the date of the last follow-up. PFS was assessed from the date of the diagnosis to the date of progression, death from any cause, or the last follow-up. The cutoff points for protein expression were determined by the association between their median expression, and sensitivity and specificity were estimated using receiver operator characteristic (ROC) curves. Statistical analyses were performed using STATA 12.0 software. P-values ≤ 0.05 were considered statistically significant.

Ethics approval
This study was performed in accordance with the Declaration of Helsinki and approved by the Faculdade de Medicina da Universidade de São Paulo (FMUSP) Ethics Committee.