Background
Diffuse large B-cell lymphoma (DLBCL) is the most common non-Hodgkin’s lymphoma, accounting for 30–40% of adult non-Hodgkin’s malignant lymphoma [
1]. Although patients diagnosed with DLBCL are potentially curable with chemotherapy, the disease proves to be fatal in approximately 50% of patients [
2]. Recently, given that there has been an increasing trend in the incidence of DLBCL, it is imperative to develop specific and effective treatments related to the pathogenesis of the disease.
Previous studies have reported that the phosphatidylinositol 3-kinase (PI3K) signaling pathway plays a critical role in regulating the growth and survival of DLBCL cells [
3], and that constitutive phosphorylation of PI3K resulted in the activation of signaling that represented frequent events both for main pathway components and their downstream substrates [
4]. Activated PI3K/AKT signaling pathways have been reported to be associated with decreased disease-free survival (DFS) and a poor response to treatment in patients with DLBCL [
5]. This suggests that the PI3K/AKT pathway is potentially an important tumorigenic signaling route and an unfavorable prognostic factor in DLBCL.
PI3Ks consist of a large and complex family that contains three classes, I, II, and III. Of them, Class I PI3K is the most studied and plays a key role in the development and progression of tumors [
6]. Class I contains the class IA catalytic subunits PIK3CA, PIK3CB, PIK3CD, and class IB catalytic subunit PIK3CG and the regulatory subunits PIK3R1, PIK3R2, and PIK3R3, while class II contains the catalytic subunits PIK3C2A, PIK3C2B, and PIK3C2G [
4]. However, how each subunit precisely contributes to the progression and maintenance of tumors is largely undetermined. The PI3K/AKT signaling pathway can be activated by two main mechanisms: activating mutations and amplifications [
7]. Amplification of genes encoding the catalytic subunits of PIK3CA, PIK3CB, PIK3CD, and PIK3CG has been reported in numerous solid tumors [
7]. In lymphomas, PIK3CA has been reported to be amplified in 15/22 (68%) cases of mantle cell lymphoma (MCL) [
8], 9/161 (5.6%) cases of chronic lymphocytic leukemia (CLL) [
9], and mutated in 1/76 (1.3%) cases of DLBCL [
10]; while PIK3CD has been reported to be mutated in 3/73 (4.1%) cases of DLBCL [
11]. However, there have been few reports available regarding CNVs or mutations of other PI3K/AKT subunits and their contribution to the activation of the PI3K/AKT pathway in DLBCL.
In the present study, we focused mainly on the various PI3K/AKT subunits and profiled their CNVs using the NanoString nCounter assay and investigated their protein expression by immunohistochemistry (IHC). Furthermore, we analyzed the association of CNVs and protein expression with clinicopathological parameters in DLBCL. We also studied various members of the PI3K/AKT pathway simultaneously in the same set of DLBCL clinical samples as well as in a panel of lymphoma cell lines to investigate their involvement in the pathogenesis of DLBCL.
Materials and methods
Tissue specimens, cell lines, and patient information
A total of 70 fresh frozen samples including 60 DLBCLs and 10 lymph node reactive hyperplasias (RHs) were collected for CNV detection on the NanoString nCounter platform (NanoString Technologies, Seattle, WA, USA). DLBCL cell lines (DOHH2, OCI-Ly1, OCI-Ly8, and Toledo) and Burkitt’s lymphoma (BL) cell lines (Raji, Namalwa) were included. Corresponding DLBCL formalin-fixed, paraffin-embedded (FFPE) tissues were collected for IHC detection of PI3K catalytic subunits. All cases used in the present study were retrieved during an 8-year period (2005–2012) from our tumor bank and the Department of Pathology in the Shanghai Cancer Center. Diagnoses were reviewed by two pathologists (Wenli Cui and Ying Cai) based on the 2010 World Health Organization classification. The other relevant clinical pathological information including primary site, B symptoms, bulky disease, performance status, lactate dehydrogenase (LDH) activity, stage, International Prognostic Index (IPI) were collected. The DLBCLs were classified into GCB and non-GCB subtype according to Hans algorithm [
12]. 13/60 (22%) patients received R-CHOP (Rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) or R-CHOP-like therapy, and 47/60 (78%) patients received CHOP or CHOP-like therapy. Patient clinical information was extracted from hospital records. Research protocols for this study were approved by the Ethics Committee at Fudan University Shanghai Cancer Center (Shanghai, China).
NanoString nCounter assay
Genomic DNA from fresh frozen tissue and cultured cell lines was extracted using a DNA extraction kit (Qiagen, Venlo, Netherlands) following the manufacturer’s protocol. Only DNA samples with an OD A260/280 ratio between 1.7 and 1.9, which indicates optimal purity for DNA, were used for further study. For detection of CNVs in the PI3K/AKT pathway, a panel of custom-compiled gene probes related to the pathway, including PI3K catalytic subunits PIK3CA, PIK3CB, PIK3CD, PIK3CG, PIK3C2A, PIK3C2B, and PIK3C2G, and regulated subunits PIK3R1 and PIK3R2, as well as AKT subunits AKT1, AKT2, and AKT3 were designed using NanoString nCounter technology and subsequently analyzed on the NanoString nCounter platform. NanoString probes were designed for the 12 genes according to different exons located in different regions (Table
1). Three probes were designed for each gene. Each assay contained six positive dsDNA control probes, eight negative control probes, and 10 invariant control probes (INVs) designed for autosomal genomic regions predicted not to contain common CNVs.
Table 1
PI3K/AKT subunits, their localization and exons for probes
PIK3CA | 3q26.3 | 9,10,20 |
PIK3CB | 3q22.3 | 8,11,22 |
PIK3CD | 1p36.2 | 1,4,10 |
PIK3CG | 7q22.3 | 2,9,10 |
PIK3R1 | 5q13.1 | 8,13,15 |
PIK3R2 | 19q13.2-q13.4 | 2,8,15 |
PIK3C2A | 11p15.5-p14 | 1,10,32 |
PIK3C2B | 1q32 | 3,14,35 |
PIK3C2G | 12p12 | 2,16,26 |
AKT1 | 14q32.32 | 2,4,13 |
AKT2 | 19q13.2 | 4,5,11 |
AKT3 | 1q44 | 4,7,12 |
The NanoString nCounter assay was performed according to NanoString’s standard protocol. Briefly, 600 ng of fragmented genomic DNA per assay was hybridized with the capture and reporter probes in 30 μL total volume and incubated overnight at 65°C for at least 16 h. The target and probe complexes were washed and immobilized in the cartridge. Genomic DNA was fragmented into small pieces (200–800 bp) and denatured to produce single strands. The custom CNV CodeSet was then hybridized to the fragmented denatured DNA sample in a single multiplexed reaction (up to 800 genomic loci per CodeSet). Hybridized DNA-CodeSet complexes were purified using the fully automated nCounter prep station, and reporters were counted using the nCounter digital analyzer. The data were normalized to the INVs and to positive and negative controls in each hybridization reaction. Finally, data analysis was performed using nSolver software.
Copy number was determined by averaging over three probes per region. If the average copy number was below 1.4, the gene was considered as one copy; if between 1.5 and 2.4, considered as two copies; and if between 2.5 and 3.4, considered as three copies, according to the manufacturer’s protocol.
Immunohistochemistry
IHC was performed to detect expression of the four PI3K catalytic isoforms and pAKT on routinely processed FFPE specimens from which tissue microarrays (TMAs) were constructed using double cores (core diameter was 1 mm) from selected areas of 60 cases of DLBCL and 10 cases of RH. Briefly, serial sections from FFPE samples were collected onto poly-L-lysine coated slides and processed with a standard manual streptavidin-peroxidase technique using a biotin-free detection system (Dako, Carpinteria,CA, USA) after a heat-induced antigen retrieval procedure with EDTA for pAKT and citric acid for other antigens for 3 min. Primary antibodies were incubated with TMAs overnight. Omission of the primary antibody and its replacement with an antibody diluent were used as the negative control. A ready-to-use kit (EnVision™; Dako) was used according to the manufacturer’s instructions.
The source and dilution of the primary antibodies used were as follows: PI3K p110α (#4249, clone C73F8, 1:400), PI3K p110β (#3011, clone C33D4, 1:400), and PI3K p110γ (#5405, clone D55D5, 1:200) were from Cell Signaling Technology (Beverly, MA, USA), PI3K p110δ (sc-7176, clone H-219, 1:200) was from Santa Cruz Biotechnology (Dallas, TX, USA), and pAKT (Ser 473) (#2118-1, clone EP2109Y, 1:150) was from Epitomics (Burlingame, CA, USA).
Scoring of immunostaining
Expression of the four PI3K catalytic isoforms was evaluated blindly and independently by two pathologists (Wenli Cui and Ying Cai). The staining intensity was scored from 0 to 3: 0, no appreciable staining in tumor cells (negative); 1, slight staining in tumor cells (weak positive); 2: moderate staining in tumor cells (moderate positive); 3: distinct staining in tumor cells (strong positive) [
5]. Cases were considered positive if ≥30% of the tumor cells were stained with an antibody, as described in a previous study involving IHC evaluation of TMAs [
12].
Statistical analysis
Overall survival (OS) was measured from date of diagnosis to date of death of any cause, or latest follow-up. Survival analysis was performed using Kaplan–Meier survival curves with a statistical software package (IBM SPSS Statistics 20.0, Armonk, Town of North Castle, NY, USA) and compared with the use of logistic statistics (log-rank). The Fisher’s exact or χ2 test was used for statistical analysis of categorical data. The Cox proportional hazard regression model was used for univariate analyses. A P-value <0.05 was considered statistically significant.
Discussion
Given the important involvement of the PI3K/AKT pathway in the pathogenesis of tumors, and given the paucity of datas regarding CNV in PI3K/AKT gene members in DLBCL, we first investigated CNVs using NanoString nCounter’s method [
13] in 12 members of the PI3K/AKT signaling pathway in human DLBCL using an nCounter CNV assay. It was found that all PI3K and AKT subunits besides PIK3R1 had CNVs to a different extent, generally, with the frequency ranging from 8.3% to 23%. Among which, CNVs of PIK3CA and PIK3CB were significantly associated with inferior overall survival. Additionally, protein expression of p110α, p110β, p110γ, and p110δ using IHC method were also found to be associated with decreased survival.
CNVs have been found in some genes that are components of the PI3K/AKT pathway in various tumors [
14‐
16]. However, little has been reported about CNVs of the PI3K/AKT in DLBCL. Here we have for the first time detected CNVs in almost all members of the PI3K/AKT signaling pathway in DLBCL. We have revealed CNVs profiles and genetic abnormality as common event in the PI3K/AKT signaling pathway in DLBCL, providing evidence and possible explanation for the pathogenesis of DLBCL on genetic level.
There have been reports showed that PIK3CA gene was happened to CNVs in ovarian cancer [
17], cervical cancer [
18], and gastric cancer [
19]; and that CNVs in PIK3CA were reported to be an independent factor for predicting poor OS of patients with ovarian cancers [
17] and gastric cancers [
19]. Our result showed that each CNVs of PIK3CA and PIK3CB has significant shorter survival. Either CNVs of PIK3CA or PIK3CB had significant shorter survival too, indicating either PIK3CA or PIK3CB has significant effects on overall survival. Both CNVs of PIK3CA and PIK3CB had the effect trend on survival effect, there was no statistically significance.
In an earlier study, the authors discovered that amplification of PIK3CD was significantly higher in patients who were sensitive to rituximab than that in patients who were resistant to rituximab [
20], suggesting that CNVs of PIK3CD may be a useful marker used for chemotherapy sensitivity when selecting the appropriate chemotherapeutic treatment for patients with DLBCL. Alizadeh et al. found that PIK3CG was highly expressed in GCB (germinal center B-cell like), and patients with GCB DLBCL had a significantly better OS than those with ABC (activated B-cell like) DLBCL [
21], which was consistent with our detection that PIK3CG expression was significantly higher in GCB cell lines (n = 4).
Among all of the members that we analyzed, the CNVs of PIK3CA and PIK3CB were positively and significantly associated with prognosis compared with other members. Depending on previous research, non-GCB represented poor overall survival than GCB. Although CNVs of PIK3CA and PIK3CB were apparently higher in non-GCB group (80% (8/10) and 67% (8/12), respectively) than that in GCB group, there was no statistically significant. Copy number amplification of either PIK3CA or PIK3CB located in the same region (3q) showed a correlation with poor survival in DLBCL patients, indicating that this region of amplification has significant correlation with survival in DLBCL. There was no significant correlation between the CNV and protein level other than PIK3CA, indicating that CNV of these genes might partially contribute to the aberrant expression of PI3K isoforms supposedly. The inconsistency we discovered here is highly similar to an earlier report by Ye ZQ and colleagues whose finding is that there was inconsistency between CNV and protein differential expression for the most genes [
22].
In our present study, not only were CNVs detected in clinical sample of DLBCL, but also it were detected in a panel of six cell lines, including DOHH2, OCI-LY1, OCI-LY8, Toledo, Raji, and Namalwa. From the results obtained at the cell-line level as well as in clinical tissues, it can be seen that CNV was a common event in almost all components of the PI3K/AKT signaling pathway.
Using TMA, we found that of all the members subjected to IHC, p110α, p110β, p110γ, and p110δ protein expression rates were 80%, 81.6%, 81.6%, and 81.6%, respectively. This was similar to the findings of Meadows’s findings in Hodgkin Lymphoma (HL) where they showed that the positive expression of p110α, p110β, p110γ, and p110δ was 97.2%, 29.2%, 54.4% and 80.6%, respectively [
23]. The expression of p110α has been reported to be an independent factor for predicting decreased OS for patients with ovarian cancers [
17] and gastric cancers [
19], although there are no related reports with respect to lymphoma. Protein expressions of p110α, p110β, p110γ, and p110δ showed significant correlation with poor survival.
With regard to AKT, whose active form is phosphorylated AKT (pAKT). It has been reported that p110α amplification was closely associated with pAKT expression. In the study, CNV of PIK3CA was highly associated with aberrant p110α protein expression, which subsequently associated with pAKT, indicating p110α was main isoform for activation of the downstream core protein AKT in DLBCL. PAKT has been extensively reported to be associated with poor prognosis in different types of cancer [
15,
24‐
26]. Expression of pAKT has shown a trend towards decreased 5-year survival for patients with DLBCL (
P = 0.05) [
5], while another study showed that high pAKT expression had decreased OS (
P = 0.036) [
27]. In our research, high pAKT expression was associated with poor survival, although statistical significance was not reached. Due to limited number of cases in the study, large cohort study is required to further investigate their relationship and validate our findings.
Competing interests
The authors declared that they have no competing interests.
Authors’ contributions
WC and YC performed experiments and were responsible for data collection, analysis, interpretation of the results, and writing the manuscript. RB and WW were responsible for conducting the data analysis in cooperation with ZL and PW. MS provided clinical samples for performance of experiments. XZ were responsible for experimental design, analysis and interpretation. All authors have read and approved the final manuscript.