Phosphorylation of pRb: mechanism for RB pathway inactivation in MYCN‐amplified retinoblastoma

Abstract A small, but unique subgroup of retinoblastoma has been identified with no detectable mutation in the retinoblastoma gene (RB1) and with high levels of MYCN gene amplification. This manuscript investigated alternate pathways of inactivating pRb, the encoded protein in these tumors. We analyzed the mutation status of the RB1 gene and MYCN copy number in a series of 245 unilateral retinoblastomas, and the phosphorylation status of pRb in a subset of five tumors using immunohistochemistry. There were 203 tumors with two mutations in RB1 (RB1 −/−, 83%), 29 with one (RB1 +/−, 12%) and 13 with no detectable mutations (RB1 +/+, 5%). Eighteen tumors carried MYCN amplification between 29 and 110 copies: 12 had two (RB1 −/−) or one RB1 (RB1 +/−) mutations, while six had no mutations (RB1 +/+). Immunohistochemical staining of tumor sections with antibodies against pRb and phosphorylated Rb (ppRb) displayed high levels of pRb and ppRb in both RB1 +/+ and RB1 +/− tumors with MYCN amplification compared to no expression of these proteins in a classic RB1 −/−, MYCN‐low tumor. These results establish that high MYCN amplification can be present in retinoblastoma with or without coding sequence mutations in the RB1 gene. The functional state of pRb is inferred to be inactive due to phosphorylation of pRb in the MYCN‐amplified retinoblastoma without coding sequence mutations. This makes inactivation of RB1 by gene mutation or its protein product, pRb, by protein phosphorylation, a necessary condition for initiating retinoblastoma tumorigenesis, independent of MYCN amplification.


Introduction
In 1971, Alfred Knudson studied a series of cases of retinoblastoma, a childhood-onset ocular tumor, and proposed the "two-hit-model" [1]. According to this model, retinoblastoma is caused by biallelic inactivation of a single gene, retinoblastoma1 (RB1). These two "hits" or mutations originally defined the concept of "tumor suppressor" genes. In heritable cases of retinoblastoma, the first hit is an inherited germline mutation, whereas the second hit is a somatic mutation; in sporadic cases, both hits are somatic in origin in a single retinal cell which initiates tumorigenesis [1].
Recently, a small subset of retinoblastoma with no detectable mutation in RB1, and amplification of the MYCN gene was identified [2]. It was hypothesized that MYCN amplification initiated retinoblastoma tumorigenesis in the presence of functional pRb protein [2].
It is known that pRb can be inactivated by various mechanisms including genetic mutations and phosphorylation [3]. To define which of the identified mechanisms is present in MYCN-amplified retinoblastoma, we first investigated MYCN copy number and the mutations present in RB1 in a series of 245 cases of unilateral retinoblastoma. We compared and contrasted the clinical features of the two groups of tumors classified by their MYCN ORIGINAL RESEARCH Phosphorylation of pRb: mechanism for RB pathway inactivation in MYCN-amplified retinoblastoma amplification status. To explore the possibility of pRb inactivation by phosphorylation as an alternate pathway, we used immunohistochemical staining to evaluate the expression of four proteins: SKP2, a target of MYCN amplification, p27, a substrate for SKP2 ubiquitination thereby inhibiting pRb phosphorylation, total pRb, and phosphorylated pRb (ppRb).

Retinoblastoma specimens
A total of 245 unilateral retinoblastomas that had undergone complete screening for coding sequence and promoter region mutations in the RB1 gene were studied [4][5][6]. Ninety-four tumors were collected following enucleation at Wills Eye Hospital (CLS), Thomas Jefferson University, Philadelphia, PA. These specimens were submitted to the histopathology laboratory at Wills Eye Hospital (RCE) for routine processing, diagnosis, and descriptive analysis. Frozen or formalin-fixed paraffin-embedded samples were sent to the Genetics Diagnostic Laboratory (GDL), Perelman School of Medicine, University of Pennsylvania (AG), Philadelphia, PA for genetic testing. An additional 151 cases of unilateral retinoblastoma were submitted to the GDL for genetic testing by 62 different sites in the US, Canada, Thailand, and Chile. Tumors used in this study were collected between 1982 and 2014. A small subset of 41 tumors was included in a previous study [7] and is indicated by an asterisk in Table S1. Pathology reports were available for 111 of the retinoblastoma samples (Table S2).
The Institutional Review Board of the University of Pennsylvania approved this research in accordance with an assurance filed with and approved by the U.S. Department of Health and Human Services. Since all patients were under the age of 18 years, written informed consent for use of tissues and data for research was obtained from a parent or legal guardian of all patients prior to genetic testing.
DNA isolation and screening of the RB1 gene DNA was isolated from frozen and formalin-fixed, paraffinembedded retinoblastoma specimens using Qiagen DNeasy Blood and Tissue kits following manufacturer's protocols (Valencia, CA).
Mutation analysis of all 27 coding exons of the RB1 gene, plus promoter and flanking intronic regions was performed by Sanger sequencing as previously described [4,8]. Methylation status of the promoter region, estimation of exonic copy number, and loss of heterozygosity (LOH) were carried out as previously described [4,8].
MYCN copy number determination and amplicon size MYCN copy number was determined using an Applied Biosystems Taqman copy number assay (Hs00824796_cn, Life Technologies) using qPCR of tumor DNA following manufacturer's protocols. To characterize genome-wide chromosomal changes and to determine the size of the MYCN amplicons, 38 retinoblastoma samples were genotyped using CytoScan HD SNP arrays following the manufacturer's instruction (Affymetrix, Santa Clara, CA). Array specific CEL files were generated in GeneChip Command Console Software. The CytoScan HD array data were imported into Affymetrix Chromosome Analysis Suite 3.0 (ChAS) software for analysis using filters for marker count = 50 and amplicon size = 100KB.

Statistical analysis
Statistical tests were carried out using the Fisher's exact probability or chi-square test for contingency table analysis. Wilcoxon Mann-Whitney U-test or Kruskal-Wallis test was used to compare the ages of patients in the different mutation categories. All analyses were two-tailed and were carried out using IBM SPSS Statistics, version 23, vassarstats (http://vassarstats.net/) or SISA (http://www. quantitativeskills.com/sisa/).
The age at which retinoblastoma was diagnosed was available for 239 cases (range = 0.5-136 months). There was no significant difference in the age of diagnosis and the presence of zero, one, or two RB1 gene mutations Categories of mutations in RB1 +/− and RB1 −/− tumors Point mutations  Fig. 2). Genome architecture, including whole genome chromosome copy number gains and losses, were determined for 38 retinoblastomas: 14 MYCN-amplified and 24 MYCNlow tumors. Classic retinoblastoma-related chromosome copy number changes, including 1q gain, 6p gain, and 16q loss [2,[11][12][13][14][15][16], were observed in 14 (37%), 15 (40%), and seven (18%) tumors (Fig. 3A), respectively; however, there was no significant difference in the fraction of tumors with these copy number changes between MYCN-amplified and MYCN-low retinoblastomas ( Table 2). Figure 3B depicts copy number changes in the MYCN region on chromosome 2, showing the length of the MYCN amplicon in 14 tumors with MYCN amplification. The size ranged between 1 and 5 MB in 13 tumors, and was 11 MB in one tumor. The size of the minimum overlapping region that included the MYCN gene was 948KB and included the MYCNOS (noncoding anti-sense RNA) gene and FAM49A exons 2-11. Differences in copy number gains or losses for all other chromosomes were not significant.
Copy number gains of MDM4 and OTX2, and loss of BCOR in retinoblastoma have been reported previously [7,12,17]. In our collection of retinoblastomas, MDM4 and OTX2 copy number gains were found in 16 (42%) and five (13%) of 38 retinoblastoma, respectively. Loss of BCOR was seen in three of 34 tumors (7.9%) ( Table 2); all three carried two or one RB1 mutations. No significant association was found between the presence of these copy number gains or losses, and MYCN amplification. In the previous study of MYCN-amplified retinoblastoma, a lower frequency of copy number changes in four genes characteristic of retinoblastoma (KIF14,  Table S1 for UPEN-RB-IDs for samples labeled 1-245 in the Figure).   [2]. We found no significant difference in the number of copies of KIF14, DEK, and CDH11 between MYCN-amplified and MYCN-low tumors; however, there was a significant difference in copy number gains in E2F3 where a smaller fraction of MYCN-amplified tumors showed a gain (3/14, 21%) compared to MYCN-low (14/24, 58%, P = 0.04) tumors (Table 2).
Pathology reports were available for 111 retinoblastomas and are summarized in Table S2: 12 were MYCN amplified and 99 were MYCN low. There was no significant association between the tumors carrying zero, one, or two RB1 mutations and the presence of high-risk features defined as invasion of the optic nerve to the level of the retrolamina, massive uveal/choroidal invasion and/or invasion of the anterior chamber, iris or ciliary body (P = 0.73) [18,19]. In addition, there was no significant difference between the degree of tumor differentiation and MYCN amplification status (P = 0.16) or tumors carrying zero, one, or two RB1 mutations (P = 0.77).
Among the 12 MYCN-amplified tumors with pathology reports, four (33%) had high-risk features compared to 23 of 99 MYCN-low tumors (23%, P = 0.48). One of the four MYCN-amplified, high-risk tumors, (UPENN-RB-175) was RB1 +/+ and was diagnosed at 10 months, by which time the tumor had invaded both the retrolaminar portion of the optic nerve and showed massive choroidal invasion. However, overall, the age of diagnosis was not significantly different between the four patients whose tumors had high-risk features and MYCN amplification and the 23 high-risk, but MYCN-low tumors (P = 0.16).
In terms of histological differentiation, 11 (92%) MYCNamplified tumors were poorly differentiated consisting of neuroblastic cells and few, if any rosettes; one tumor was described as mixed with respect to differentiation (Table  S2 and S3). This distribution of the degree of differentiation was not significantly different from that found in MYCN-low retinoblastoma: 63 (68%) undifferentiated/ poorly differentiated, 21 (23%) well or moderately well differentiated with numerous Flexner-Wintersteiner rosettes, and seven (7.6%) with mixed status (P = 0.16, compiled data not shown).

Immunohistochemical staining pattern
We compared the relative levels of expression of SKP2, p27, pRb, and ppRb proteins in four MYCN-amplified and one MYCN-low retinoblastomas by immunohistochemistry (IHC) staining of tumor sections with antibodies specific for these proteins (Table 3 and Fig. 4, Figure S1 and Table S4).
The histopathological features, as well as pRb and ppRb (Ser608) IHC staining pattern in five retinoblastoma tumor sections and matched normal retina (defined as retinal tissue not in proximity to the tumor) are visualized in Figure 4 and described in Table 3. About 25-50% of cells in the inner (INL) and outer (ONL) nuclear layers of the normal retina were moderately to strongly stained with both pRb and ppRb antibodies in the MYCN-amplified tumors UPEN-RB-175 (Fig. 4, panels A,C), UPEN-RB-93 (Fig. 4E,G), and UPEN-RB-40 (Fig. 4I,K) carrying zero or one RB1 mutations. This would indicate that the pRb protein that is expressed is phosphorylated. Over 90% of cells in the INL and ONL of UPEN-RB-201 (MYCNamplified but RB1 −/− ) were strongly positive for pRb; however, only about 70% of cells in the INL (Fig. 4M), and less than 5% in the ONL were ppRb immunoreactive (Fig. 4P). The INL of UPEN-RB-112 (MYCN-low and RB1 −/− ) had about 25% pRb-positive cells, but only very few ppRb-positive cells (Fig. 4S), and the ONL was essentially negative for both proteins (Fig. 4U) (Table 3).
Tumor sections of UPEN-RB-175 (MYCN-amplified, RB1 +/+ ) stained strongly for pRb and ppRb (Fig. 4B,D), while the two MYCN-amplified RB1 +/− stained strongly with pRb antibody (Fig. 4F,J), but showed weaker staining of fewer cells with ppRb antibody (Fig. 4H,L). Interestingly, the well-differentiated region of UPEN-RB-201 with many Flexner-Wintersteiner (FW) rosettes was mostly negative for both pRb and ppRb (Fig. 4O,R); however, the poorly differentiated section had scattered positive cells (Fig. 4N,Q). UPEN-RB-112 tumor was well differentiated with many FW rosettes and essentially negative for pRb and ppRb (Table S3 and Fig. 4). In all tumors, the immunoreactive pRb and ppRb proteins were primarily concentrated in the nucleus with very little staining of the cytoplasm. Figure 5 shows staining of the retinal section that seeds the retinoblastoma in two tumors with MYCN amplification, UPEN-RB-93 and UPEN-RB-40. With respect to pRb expression, there are relatively few scattered positive cells in both the ONL and INL; however, numerous darkly stained tumor cells can be seen originating from the adjacent weakly stained INL. Both tumors carry only one RB1 mutation; so, some expression of pRb is to be expected. In contrast, there is very little expression of ppRb in the INL of either tumor but significant expression is seen in the adjacent tumor section for both tumors. These results appear to show that tumor cells expressing both pRb and ppRb proteins may originate in the retinal INL, and not from the ONL, However, it is also possible that these cells may originate elsewhere and migrate into the INL, or preferentially invade the INL at the edges of a large tumor.

Discussion
Rushlow et al. reported a class of novel RB1 +/+ retinoblastoma with no genetic mutations in RB1, but having a high level of MYCN amplification (>29 copies) [2]. This led to the hypothesis that high level MYCN amplification is sufficient for initiation of tumorigenesis of RB1 +/+ retinoblastoma. In this manuscript, we investigated the characteristics of MYCN-amplified sporadic unilateral retinoblastoma and the mechanisms of inactivation of the RB1 gene in these tumors.
Newly diagnosed, sporadic unilateral retinoblastoma is caused by somatic mutations in over 80% of cases, and by low penetrance germline mutations in the RB1 gene in 15% of cases [4]. The sensitivity of mutation detection in coding sequences of RB1 is high, around 94% but not 100% [4]. The current methods of mutation screening of RB1 are limited to the promoter region, coding exons, and immediate flanking intronic regions, but not the entire introns which are known to harbor mutations [20,21] and/or rearrangements including chromothripsis [7] that can also inactivate the RB1 locus. Thus, based on current sensitivity of DNA-based genetic testing, the presence of a subset of retinoblastoma without a detectable mutation in the RB1 gene is not unexpected. Assuming a sensitivity for finding a coding sequence RB1 mutation to be 94% [4], one would expect to observe only one wild-type RB1 +/+ retinoblastoma among the 245 tested. However, we identified 13 (5.3%, P < 0.001) tumors without any RB1 mutation.
We identified 18 of 245 retinoblastomas (7%) with a high MYCN copy number between 29 and 110 copies (Fig. 1). Of these, eight tumors carried two mutations in RB1, four were RB1 +/− , and six were RB1 +/+ . This finding that 44% of MYCN-amplified tumors carried two RB1 mutations directly contrasts the observation by Rushlow et al. [2] who found no retinoblastomas with MYCN amplification among 93 tumors with two mutations in RB1 (RB1 −/− ). Further investigations into alternate mechanisms of inactivation of the RB1 gene product demonstrated phosphorylation of pRb at S608 in tumors without coding sequence mutations in one or both alleles of RB1 which is consistent with the observation that phosphorylation is the most common mechanism of pRb inactivation in most known cancers [22,23].
MYCN amplification in retinoblastoma was first described by Lee et al. [24] showing that the MYCN gene was amplified 10-200-fold in two primary retinoblastomas and in the Y79 cell line, but not in fibroblasts. Subsequently, Mairal et al. [14] assessed MYCN amplification karyotypically as double-minute chromosomes and by PCR in four of 20 primary retinoblastomas. Further studies in a genetically engineered mouse model of human retinoblastoma demonstrated a 6-400-fold amplification of MYCN in three of 16 aggressive retinoblastomas [25].
Amplification of MYCN has been described in about 20% of neuroblastoma where it identifies a highly aggressive subtype of the tumor [26,27]. In contrast, Lillington [28] found in a study of 25 primary retinoblastomas, that three tumors with MYCN amplification (>30 fold) were not associated with high-risk histological features. In our series, retinoblastoma cases with high MYCN copy number were associated with an earlier age of onset that was significantly different from that of retinoblastoma without MYCN amplification (Fig. 2). However, there was no difference in the presence of high-risk features between the two groups of tumors with and without the amplification.
In many cancers, the regulation of pRb becomes unbalanced, and elevated cyclin-dependent kinase (CDK) activity prevents pRb from being an effective brake on cell cycle control. One of the potent CDK inhibitors is p27, a protein which is degraded by the protein, SKP2. In addition, SKP2 is a known target of MYCN amplification in neuroblastoma [29]. The IHC staining results in our tumors indicate the absence of SKP2 protein in any significant amount and could signal high expression of p27 protein. However, the weak p27 immunoreactivity suggests that it is expressed in low levels in all MYCN-amplified tumors, except the RB1 −/− tumor, UPEN-RB-201, where it is highly expressed in both well and poorly differentiated areas of the tumor (Fig. S1, I, and J). Further comparison of p27 expression in UPEN-RB-201 and UPEN-RB-112 which is also RB1 −/− , but MYCN-low suggests that p27 is upregulated in the presence of MYCN amplification, but not in MYCN-low tumors in the absence of pRb. In essence, these observations rule out the activation of the MYCN-SKP2-p27 axis in three out of four MYCN-amplified retinoblastomas. Since it is known that p27 is degraded upon progression into S-phase when both pRb and p27 are phosphorylated [29], we investigated the phosphorylation status of pRb.
Alterations in pRb phosphorylation during late G1 phase allow the activation of E2F-dependent transcription that drives the cell cycle and is needed for DNA replication. It has been shown that the alteration in the structure of pRb by phosphorylation at multiple residues, including S608/S612 [30,31] causes specific conformational changes and are sufficient for inhibition of interaction with the transactivating domain of E2F, while phosphorylation at S795 inhibits binding to the marked box (MB) domain of E2F1 [29].
The observation that pRb is phosphorylated at S608 and/or S795 residue in the MYCN-amplified retinoblastoma provides an important insight into how retinoblastoma develops without mutational knock-out of the RB1 gene

Supporting Information
Additional supporting information may be found in the online version of this article: Figure S1. Expression of SKP2 and p27 in four MYCNamplified and one MYCN-low retinoblastomas detected by immunohistological staining with anti-SKP2 and anti-p27 Kip1 antibodies. Table S1. MYCN copy number and RB1 mutations found in 245 retinoblastomas. Table S2. Histopathological features of 111 retinoblastomas. Table S3. RB1 gene mutations and high-risk histological features in 18 tumors with MYCN amplification. Table S4. IHC staining of five retinoblastoma tumors with antibodies specific for SKP2 and p27.