A Novel MLL5 Isoform That Is Essential to Activate E6 and E7 Transcription in HPV16/18-Associated Cervical Cancers Free

Cervical cancer is a major cause of death among women worldwide. More than 90% of cervical cancers are related to the human papillomavirus (HPV) infections, implicating its role as the causative agent in cervical cancer (1, 2). A common feature among cervical cancers is the selective upregulation of 2 HPV oncogenes, namely E6 and E7, which encode for multifunctional proteins that are known for their abilities to inactivate cellular tumor repressors, p53 and pRb, respectively (3, 4). It has been shown that E6 and E7 expressions in cells may lead to genomic instability, a hallmark of cancer development (5). In addition, cell-cycle arrest and checkpoints are deregulated in these cells, due to the loss of tumor suppressors p53 and pRb (4). These 2 oncoproteins are translated from a bicistronic mRNA under the control of the HPV long control region (LCR), located upstream of the E6 open reading frame (6–8). Despite extensive studies conducted to elucidate the regulatory property of LCR which involves a complex system of both viral and human transcription factors, a complete understanding of the mechanism is yet to be achieved (9–11).

Mixed lineage leukemia 5 (MLL5) gene was first discovered during the search for candidate myeloid leukemia tumor suppressor genes from a commonly deleted 2.5-Mb segment within chromosome band 7q22 (12, 13). MLL5 lacks DNA-binding motifs of A-T hooks as well as the methyltransferase homology motifs that are commonly found in other MLL protein members. These may suggest that MLL5 does not bind directly to DNA but instead modulates transcription indirectly via protein–protein interactions through its PHD and SET domains (13). Even though the other MLL family members are known to be involved in histone 3 lysine 4 (H3K4) methylation activity (14–16), several reports suggested that MLL5 lacks such intrinsic methyltransferase activity (15, 17). Nonetheless, Fujiki and colleagues suggested that a short N-terminal isoform of MLL5 (609 amino acids) with both the PHD and SET domain possesses GlcNAcylation-dependent histone lysine methyltransferase (HKMT) activity (18). Furthermore, studies reported role of MLL5 in histone modification through indirect mechanism or forming complexes (19, 20). Three independent studies reported the genetic analysis of Mll5 deficiency in mice that suffer from mild growth retardation but do not develop spontaneous leukemia (17, 21, 22). However, a recent clinical study reported that higher MLL5 expression levels were associated with better prognosis in acute myeloid leukemia (23). We have previously shown that overexpression or knockdown of MLL5 impeded cell-cycle progression and proposed that MLL5 may participate in the cell-cycle regulatory network at multiple stages of the cell cycle (24, 25). We showed that MLL5 is a substrate of Cdc2 kinase, and phosphorylation of MLL5 is required for mitosis progression (26).

During the course of studying the restoration of p53 protein and reduction of retinoblastoma (Rb) protein phosphorylation upon knockdown of MLL5, we found an intriguing link between the downregulation of E6/E7 oncoproteins and MLL5 levels in HPV16/18-positive cervical cancer cell lines. We further characterized a novel MLL5 isoform that plays a role in activating E6/E7 through the association with the AP-1 transcription factor in HPV-LCR. The novel role of MLL5 isoform in cervical cancer makes it a potential therapeutic target and a biomarker for human cervical cancers.

Human cervical carcinoma SiHa (HPV16⁺), HeLa (HPV18⁺), and C33A (HPV⁻), embryonic kidney cells HEK293T were cultured as monolayer in Dulbecco's Modified Eagle's Medium (Gibco) whereas human cervical carcinoma CaSki (HPV16⁺) in RPMI (Gibco). The respective media were supplemented with 10% FBS (Hyclone), 2 mmol/L glutamine (Gibco), and 100 units/mL penicillin/streptomycin (Gibco) at 37°C with 5% CO₂. Plasmid transfection was carried out with TransIT-LT1 transfection reagent (Mirus) according to the product manual.

Putative MLL5 isoform was cloned by SMARTer RACE cDNA Amplification kit (Clontech) according to the product manual. PolyA⁺ mRNA was extracted by Oligotex Direct mRNA Midi kit (Qiagen) and used as templates for both 5′ and 3′ rapid amplification of cDNA ends (RACE). Gene-specific primers were used for both 5′ and 3′ RACE recognized exon 7 of MLL5 and are as follows: Reverse primer for 5′RACE: 5′-TTTCCCTTTTCCGGCGTTGT-3′ and forward primer for 3′RACE: 5′-CAACGCCGGAAAAGGGAAAAT-3′. RACE products were cloned into pCR2.1-TOPO (Invitrogen) for sequencing.

FLAG-MLL5-CT (amino acid 1,113 to 1,858) was previously constructed (26). Green fluorescent protein (GFP)-tagged MLL5 C-terminal vector (GFP-CT) was generated by cloning MLL5 C-terminal region into pEGFP-C1 (Clontech) in frame with SalI and BamHI. PCR amplicons were digested with BamHI and NotI and cloned into pEF6/V5-His vector (Invitrogen) for FLAG-MLL5β, whereas PCR amplicons were digested with SalI and BamHI and cloned into pEGFP-C1 vector for GFP-MLL5β. To generate constructs for luciferase assays for the LCR promoter activity, a 958-bp (nucleotides 7,091 to 119) fragment containing the LCR and p-105 promoter was cloned into pGL3-basic (Promega) with XhoI and HindIII sites. Deletion constructs were generated in a similar manner with appropriately designed primers. Mutant constructs were generated using the Quick Change Site-Directed Mutagenesis Kit (Stratagene).

The transfection of siRNAs was carried out with Lipofectamine RNAiMAX (Invitrogen). Four MLL5-specific siRNA duplexes (#1, #2, #3, and #4) targeting nucleotide positions at 1,063, 1,147, 3,027, and 6,807, respectively, from the transcription starting point [National Center for Biotechnology Information (NCBI) reference sequence: NM_182931.2], were used. Scrambled siRNA (sense: 5′-UUCUCCGAACGUGUCACGUdTdT; antisense: 5′-ACGUCACACGUUCGGAGAAdTdT) was used as a control. All the siRNAs were generated from 1st BASE.

A total of 0.4 million cells were loaded in each lane. The antibody against the central region of MLL5 (amino acid 1,157–1,170, designated as α-MLL5-1157; ref. 25) is used to probe for full-length MLL5. α-MLL5-227 antibody raised against the N-terminal region of MLL5 (amino acid 227–241) was used to probe for MLL5β (24). α-HPV16/18-E6 (SC-460), α-HPV18-E7 (SC-1590), α-p53 (SC-126), α-pRb (SC-50), and α-actin antibodies (SC-1616) were purchased from Santa Cruz Biotechnology and α-p21 antibody (#2946) was from Cell Signaling.

Total RNA was extracted with TRIzol (Invitrogen). The RNA was treated with DNase (Ambion; #2222) prior to cDNA synthesis using the First Strand cDNA Synthesis Kit with oligo(dT) primer (Invitrogen). Gene expression was measured with the iQ5 qPCR (BioRad) with the SYBR Green PCR Master Mix (BioRad) and in-house designed primers. Primers used include MLL5: forward 5′-CCACCACAAAAGAAAAAGGTTTCTC-3′, MLL5: reverse 5′-GTGTTGGTAAAGGTAGGCTAGC-3′; MLL5β: forward 5′-GAAAACCCAGAGTGCCCTGTTCTA-3′, MLL5β: reverse 5′-CAATATACGCGAGACTAGTCTT-3′; HPV18-E6: forward 5′-GTGCCAGAAACCGTTGAATCC-3′, HPV18-E6: reverse 5′-CGACGCCAGCTATGTTGTGAAATCGTCG-3′; HPV18-E7: forward 5′-CGTCGCAACATTTACCAGCCCGACG-3′, HPV18-E7: reverse 5′-GAATGCTCGAAGGTCGTCTGC-3′; HPV16-E6: forward 5′-CTGCAATGTTTCAGGACCCA-3′, HPV16-E6: reverse 5′-TCATGTATAGTTGTTTGCAGCTCTGT-3′; HPV16-E7: forward 5′-AAGTGTGACTCTACGCTTCGGTT-3′, HPV16-E7: reverse 5′-GCCCATTAACAGGTCTTCCAAA-3′; and GAPDH: forward 5′-GTGAAGGTCGGAGTCAACG-3′, GAPDH: reverse 5′ TGAGGTCAATGAAGGGGTC-3′.

Eight primary carcinoma specimens (6 squamous cell carcinomas and 2 adenocarcinomas) were pretreated biopsies from women with carcinoma of cervix prepared by Dr. S.K. Tay, a gynecologist in Singapore General Hospital. Among the 8 samples, 7 of them are positive for high-risk HPV, namely, 6 with HPV16 strain and 1 with HPV18 strain. Patients were informed and agreed to the use of their biological sample for scientific research in accordance with Singapore regulations. Total RNA and genomic DNA were extracted from the samples by the TRIzol and Wizard Genomic DNA Purification Kit (Promega).

A Dual-Luciferase Reporter Assay (Promega) was used to measure the transcription activity of the promoter region of interest. HEK293T and HeLa cells were cotransfected with both pGL3 and pEGFP, along with pRL-TK as the internal control. Cells were then harvested 48 hours posttransfection, and each sample was read in triplicate by using a luminometer (Tecan). Normalization of luciferase reading by Renilla reading was carried out before comparisons were made.

FLAG-MLL5β or FLAG-CT was transfected into HeLa cells for 48 hours before cells were being harvested for chromatin immunoprecipitation (ChIP) as described previously (27) except for the modification in the sonication step. Each sample was sonicated at 40% amplitude for 15 minutes consisting of 15 cycles of 30 seconds sonication with 30 seconds cooldown interval to generate DNA fragments of around 300 to 500 bp. Pull down was done by α-FLAG (Sigma-Aldrich) and α-mouse immunoglobulin G (IgG) antibodies. PCR amplicons were run on a 2% agarose gel to check for enriched region compared with IgG pulled down samples.

We noted a marked accumulation of p53 protein in HPV18-positive HeLa cells upon MLL5 knockdown (25). Because E6 oncoprotein has been known to target p53 for degradation via formation of complex with E3 ubiquitin-protein ligase E6AP, this led us to speculate that MLL5 knockdown may have an effect on the expression of E6 protein. Indeed, E6 protein level in HeLa was found to decrease in a time-dependent manner (Fig. 1A). Decrease in E6 protein expression upon MLL5 knockdown was less likely to be a result of the disruption of E6–E6AP–p53 complex because the protein level of E6AP remained at a similar level regardless of changes in MLL5 level. Interestingly, a significant decrease in E7 protein and an increase in hypophosphorylated form of Rb were also detected in HeLa when MLL5 was knocked down (Fig. 1A). Similar results were observed in 2 other HPV16–positive human cervical cancer cell lines CaSki and SiHa (Fig. 1A).

To rule out the possibility of an off-target effect, besides the MLL5-siRNA#1 used in Fig. 1A, 3 other siRNAs targeting different sites spanning across MLL5 mRNA were used (Fig. 1B). Surprisingly, massive p53 accumulation occurred only in the presence of MLL5-siRNA#1 and #2, even though MLL5 was knocked down with similar efficacy with all 4 siRNAs (Fig. 1C). Consistent with previous data, massive p53 restoration only occurred in MLL5-siRNA#1 but not MLL5-siRNA#4 in all 3 HPV-positive cell lines (Fig. 1D). Nonetheless, a small but detectable amount of p53 protein was observed in MLL5-siRNA#4 knocked down samples, which could be attributed to the effect of full-length MLL5 knockdown as seen in our previous observations (25). Next, we attempted to discern the effect of MLL5 knockdown with MLL5-siRNA#1 and #4 on both E6 and E7 mRNA levels. A marked downregulation of both E6 and E7 mRNA was observed in all 3 HPV-positive cell lines when MLL5-siRNA#1 was used but not MLL5-siRNA#4 despite the comparable knockdown efficiency of both siRNAs on MLL5 mRNA (Fig. 1E). The role of E2 which is a known E6/E7 repressor was disregarded because E2 was shown to be absent in these HPV-positive cells due to the disruption of E2 gene upon the integration of HPV DNA (28). These observations led us to speculate that a putative MLL5 isoform comprising the N-terminal region, but lacking the central and C-terminal regions, may be present.

To validate our hypothesis, 5′ and 3′ RACE was carried out to identify the isoform. For 5′ RACE, a PCR product of approximately 1 kb was amplified from all 6 cell lines used (Fig. 2A). The amplicon was sequenced, and data showed no difference as compared with the full-length MLL5, indicating that the putative MLL5 isoform has the same 5′-end as the full-length MLL5. Subsequently, 3′ RACE was carried out with an extension time of 7 minutes, allowing the complete amplification of full-length MLL5. All cell lines tested showed a 6-kb band that corresponds to the full-length MLL5 (Fig. 2B). Interestingly, an additional amplicon of approximately 1.6 kb was observed exclusively in HPV16/18-positive cell lines, CaSki, SiHa, and HeLa. Upon sequencing and alignment with MLL5 gene, the 1.6-kb band has the same sequence as the full-length MLL5 from the start codon to part of exon 14 (GenBank accession number: NM_182931), where it was truncated by a 26-bp sequence that introduced a stop codon followed by poly-A tail (Fig. 2C and D). We denoted this novel MLL5 isoform (503 amino acids) as MLL5β.

Next, a wider panel of human cell lines was used to detect the presence of MLL5β. Consistent with our previous result, MLL5β was only detected in HPV16/18-positive human cervical cancer cell lines but not in other human cancer cell lines (Fig. 2E). It is worthy to mention that MLL5β was not seen in the normal diploid cell line WI-38 and HPV-negative cervical cancer cell line C33A. Besides that, 8 human primary cervical carcinoma specimens were tested for the presence of MLL5β. Using MLL5β gene–specific primer for cDNA synthesis, MLL5β was successfully detected in all 8 human primary cervical carcinoma samples (Fig. 2F). Furthermore, 3′ RACE was carried out on all 8 human primary cervical carcinoma samples, and the identical 26-bp sequences can be detected in all 8 samples at the same location (Supplementary Fig. S1).

We used anti–MLL5-227 antibody, raised against the N-terminal region of MLL5 (amino acid 227–241), to probe for the MLL5β (24). A protein band migrating at approximately 70 kDa was successfully knocked down by both MLL5-siRNA#1 and #2 but not MLL5-siRNA#3 and #4 (Fig. 3A). Importantly, knockdown of the 70-kDa MLL5β protein corresponded to the restoration of p53 level. Next, a new siRNA duplex that specifically targets the 26-bp sequence exclusively found in MLL5β mRNA, designated as MLL5β-siRNA, was synthesized. MLL5β was successfully knocked down and this corresponded to an increase in p53 protein level (Fig. 3B). Besides that, we conducted a similar knockdown experiment on HPV-negative cells HCT116 and U2OS and confirmed that p53 accumulation only occurs in HPV-positive cells (Supplementary Fig. S2). Consistent with our hypothesis, the E6/E7 levels were found to be significantly reduced in all 3 human cervical cancer cell lines when MLL5β-siRNA was used (Fig. 3C). It is noteworthy that exogenously expressed GFP-MLL5β in HeLa cells has no effects on the expressions of E6 and E7 transcript (Supplementary Fig. S3). Subsequently, we asked whether exogenous expression of MLL5β can abrogate the MLL5β-siRNA–mediated p53 restoration. Effect of p53 restoration when knocked down with MLL5β-siRNA was rescued when GFP-MLL5β was overexpressed but not GFP-tagged full-length MLL5 (GFP-MLL5-FL; Fig. 3D). GFP-MLL5β plasmid was mutated appropriately (indicated by asterisk) so that the expression cannot be inhibited by MLL5β-siRNA.

Previous studies have shown that proteins, such as epidermal growth factor and IFN regulator factor-2, are able to regulate E6/E7 gene transcription through regulatory region, termed LCR or upstream regulatory region (29, 30). The position of the LCR is shown in the genomic map of HPV18 along with other viral genes (Fig. 4A). The regulatory region, rich in cis-regulatory elements, often acts as binding sites for various transcription factors and has been established to play important roles in HPV gene expression and replication (31). Therefore, dual-luciferase assay was conducted to address whether MLL5β is involved in the transcription regulation of bicistronic E6/E7 expression through the LCR. The complete region of HPV18-LCR (full length) starting from downstream of L1 gene to the origin of replication of E6/E7 (nucleotide number 7,018 to 119; Fig. 4A) was cloned into pGL3 vector upstream of the luciferase gene.

Next, HEK293T and HeLa cells were cotransfected with pGL3-HPV18-LCR and one of the following: pEGFP-empty, pEGFP-MLL5β, or pEGFP-MLL5-CT (GFP-tagged MLL5 C-terminal region) for 48 hours, along with pRL-TK as an internal control. A HPV-negative cell line HEK293T was included for the luciferase experiment because it does not express endogenous MLL5β and does not contain any copy of integrated HPV18-LCR in its genome. Therefore, any changes in the luciferase expression can be attributed directly to the effect of exogenously expressed MLL5β. In HEK293T cells, relative luciferase activity was significantly increased when GFP-tagged MLL5β was overexpressed, providing evidence that MLL5β is able to activate the transcription of the LCR (Fig. 4B). However, in HeLa cells, relative luciferase activity is high even when GFP protein was overexpressed and exogenously expressed MLL5β does not further increase the relative luciferase activity (Supplementary Fig. S4). It is likely that the endogenous MLL5β present in HeLa cells has saturated the binding sites of the pGL3-HPV18-LCR luciferase reporter. To test this hypothesis and illustrate the specific role of MLL5β in HPV18-LCR activation in HeLa, we knocked down MLL5β before introducing pGL3-HPV18-LCR. A significant decrease in luciferase activity was observed when the endogenous MLL5β was diminished, suggesting that MLL5β plays an essential role in transcription activation of HPV18-LCR (Supplementary Fig. S5). Besides this, we have shown that the amount of HPV18-LCR is the limiting factor for MLL5β-mediated E6/E7 activation as an increase in relative luciferase activity can be observed when an increasing concentration of pGL3-HPV18-LCR was used (Supplementary Fig. S6).

Next, we constructed various fragments of HPV18-LCR with pGL3 vector to dissect the region responsible for LCR activation through association with MLL5β. As shown in Fig. 4B and Supplementary Fig. S4, relative luciferase activity of fragment A in both HEK293T and HeLa cell lines exhibited the largest decrease, suggesting that the distal region from nucleotides 7,018 to 7,305 of LCR contains the essential elements required for the activation of HPV18-LCR by MLL5β.

To show that MLL5β would interact with HPV18-LCR to regulate its transcriptional activity, ChIP experiment was set up in HeLa cells. Significant enrichment of the amplicon with primer pair 2 was observed in the α-FLAG-MLL5β pulled-down eluates but not in the FLAG-MLL5-CT (Fig. 4C), indicating an interaction between MLL5β and HPV18-LCR.

Because there is no conserved DNA-binding motif found on MLL5 and recent study has shown that MLL5 or its isoform (MLL5α, 609 amino acids) can form complex with other proteins to exert its H3K4 activity (13, 18, 19), we hence hypothesized that one or more transcription factors might be involved in the MLL5β-mediated E6/E7 gene activation. An in silico study using PROMO (32) virtual laboratory was first conducted to screen for putative transcription factor–binding sites on the HPV18-LCR fragment amplified by primer pair 2. Four putative transcription factor–binding sites with high matching scores were identified (Fig. 5A). Further fragments of HPV18-LCR were designed according to the distribution of the 4 transcription factor binding sites and cloned into pGL3-basic vector for dual-luciferase assays in HEK293T cells. Results showed no significant reduction in the relative luciferase activity except for fragment A3 (Fig. 5A), suggesting that the region (nucleotides 7,310 to 7,350) may comprise candidate transcription factors associating with MLL5β for activation of LCR. Similar reduction in relative luciferase activity for fragment A3 was also observed in HeLa cells (Supplementary Fig. S7).

Two highly matched transcription factor–binding sites can be found in fragment A3, namely, AP-1 and SP-1. To verify the involvement of these transcription factors in MLL5β-mediated activation, we generated pGL3 vectors carrying HPV18-LCR with mutated AP-1- and SP-1–binding sites and transfected them into HeLa cells. AP-1 mutant showed a significant reduction in relative luciferase activity but not SP-1 mutant, suggesting that AP-1 may participate in MLL5β-mediated E6/E7 activation (Fig. 5B). We have also tested whether HPV16-LCR has similar interactions with MLL5β through the distal AP-1 site. HPV16-LCR was cloned into pGL3 vector, and similar luciferase experiment was set up in HPV16-positive SiHa cells. As shown in Supplementary Figs. S8 and S9, the region from nucleotides 7,018 to 7,362 that contains a conserved AP-1 site is also important for the MLL5β-mediated E6/E7 transcription activation.

To investigate whether MLL5β interacts with AP-1, coimmunoprecipitation was carried out with exogenously expressed GFP-MLL5β and FLAG-c-Jun where c-Jun is part of the heterodimer that forms AP-1. GFP-MLL5β was observed to interact with FLAG-c-Jun in HeLa cells (Supplementary Fig. S10) but such interaction in HEK293T cells requires HPV18-LCR to be cotransfected (Fig. 5C). Although studies have shown that full-length MLL5 lacks intrinsic H3K4 methyltransferase activity (15, 17), nonetheless, it contains a SET domain which possesses HKMT activity (33). Moreover, 4 other MLL protein family members were found to exert H3K4 methyltransferase activity through their SET domain (15, 16, 34). Therefore, to investigate whether MLL5β activates E6/E7 transcription through the SET domain, SET-inactivated (Y358A) MLL5β mutant was constructed and used in the dual-luciferase assay (18). Relative luciferase activity was found to decrease up to 2-fold when SET mutant was used (Fig. 5D). Nonetheless, only a marginal decrease in the luciferase activity was observed when the concentration of exogenously introduced SET mutant vector was increased. It is noteworthy that SET-inactivated MLL5β mutant can still interact with AP-1 in HeLa cells (Supplementary Fig. S10).

In this study, we identified a novel MLL5 isoform, MLL5β, resulting from an additional 26 nucleotides that introduces a stop codon in exon 14 of MLL5 mRNA. The truncated MLL5β isoform encoded a 503–amino acid polypeptide that is present in HPV16/18-positive cervical cancer cells and can be detected in human primary cervical carcinoma. MLL5β isoform associates with AP-1–binding site located at nucleotide 7,326 of the distal region of HPV18 LCR, upstream of the E6 and E7 promoter site. The association with AP-1 transcriptional factor is essential for the MLL5β-mediated bicistronic E6/E7 gene expression. Knockdown of MLL5β isoform led to downregulation of both E6 and E7 oncoproteins, resulting in the restoration of p53 protein levels and a reduction in Rb phosphorylation. Our study is the first to describe the presence of the novel MLL5β isoform and to illustrate the involvement of MLL5β in HPV16/18-related cervical cancers.

Although full-length MLL5 (1,858 amino acids) has been shown to lack intrinsic histone methyltransferase H3K4 activity (15, 17), an isoform of MLL5 comprising a total of 609 amino acids with PHD and SET domains exhibited GlcNAcylation-dependent HKMT activity in retinoic acid–induced granulopoiesis (18). Given the fact that viruses often cause diseases when they enter host cells and “hijack” the important cellular apparatus to redirect the cellular functions for their benefit, it is possible that the induction of MLL5β isoform may be an invasive strategy used by HPV to regulate E6/E7 promoter activity (Fig. 6). MLL5β interacts with AP-1 transcription factor and recognizes the AP-1–binding site on nucleotide 7,326 of the distal region of LCR (Fig. 6, dashed area). The SET domain in MLL5β was found to play a role in the activation. However, when SET domain was inactivated, the effect of E6/E7 gene activation was not depleted completely (Fig. 5D), suggesting that besides interaction with AP-1, other proteins may be required to play a cooperative role for the E6/E7 activation. It is possible that there might be an unidentified posttranslational regulation on MLL5β isoform that is required for full activation activity. When MLL5β was depleted through transient siRNA knockdown, no MLL5β–AP-1 complex can be formed, leading to the absence of the interaction with LCR. This resulted in downregulation of both E6/E7 transcripts as revealed by quantitative real-time PCR and Western blot analysis.

Depletion of MLL5β in HPV16/18-positive cervical cancer cell lines caused a dramatic reduction in the transcription of 2 important oncogenes, E6 and E7 (Fig. 1). Such reduction in E6/E7 expression allows for the restoration of p53 and pRb pathways, leading to the reactivation of cell-cycle checkpoints and apoptotic pathway. Because of the prominent roles of oncogenic proteins E6 and E7 in cervical cancer tumorigenesis, much efforts have been made to explore the therapeutic potentials of direct suppression of E6 and E7 expression through the use of siRNAs or overexpression of E2 proteins in vitro and in vivo (35–37). Given that MLL5β-siRNA was able to simultaneously reduce the E6/E7 transcripts, there is a greater potential for MLL5β to serve as a novel specific therapeutic target for human cervical cancers than siRNA targeting E6 or E7 alone. Besides, the fact that MLL5β can be detected only in HPV-positive cancer cells but not in normal diploid cells suggests that the potential side effects of MLL5β-siRNA therapy can be minimized. It will be interesting to explore the association of MLL5β expression with various stages of cervical carcinoma, and the result may lead to a novel biomarker for cervical cancer diagnosis. Moreover, HPV infections are known to associate with other cancers such as head and neck cancer. It will be intriguing to explore whether MLL5β plays a similar role in activating E6/E7 transcription in these HPV-related cancers.

Our current model raised an interesting question of the possible mechanism behind MLL5β derivation. We have tried to map the exact chromosomal location of this 26-bp fragment through bioinformatics search. However, no significant and meaningful sequence match of this 26-bp fragment with HPV genome or human genome can be found. We hypothesized that MLL5β is generated through a more complicated process than simply a result of HPV infection and integration. For instance, DNA double-strand break repair enzymes are involved in viral integration sites, and studies have shown that deficiencies in some double-strand break repair enzymes caused aberrant junction sequences at the integration sites (38, 39). This possible aberration adds on the difficulty in relying on genomic database to identify the alternative splicing process or design appropriate primers for PCR reactions. In addition, current evidences show that such integration is a random event, but common DNA fragile sites are more susceptible to the integration event (40). If the generation of MLL5β is indeed a result of HPV integration, MLL5 locus is likely to be located near the common DNA fragile site. However, it is still difficult to explain how the HPV integration can lead to the same 26-bp site at the 3′ end of the MLL5β transcript in all 8 primary patient samples (Supplementary Fig. S1). We hence suspect that there might be other alternative mechanisms leading to the generation of MLL5β. On the basis of the fact that the 26-bp RNA sequence does not match exactly with any published genomic DNA sequences and this 26-bp site lacks the canonical pattern of 3′ untranslated region sequence, RNA editing may be the alternative explanation for such short region modifications in the 3′ end of MLL5β RNA (41, 42). These changes may arise from substitution or insertion/deletion RNA editing; however, more work has to be done to explore the possibility.

Ubiquitous transcription factors, such as AP-1 and SP-1, have well-established roles in HPV transcription regulation (43–45). As shown in Fig. 6, the distal region (dashed area) of LCR contains transcription termination signal, whereas the central region (gray area) contains majority of the transcription factor–binding sites termed specific enhancer region (46, 47). The proximal region (chequered area) contains early promoter and origin of replication (48). In our study, we identified the distal AP-1–binding site at nucleotide 7,326 as the important partner for MLL5β to interact with LCR. Only this AP-1 site plays an essential role in the transcription activation by MLL5β, and this activation is at least 5-fold stronger than the other 2 previously reported AP-1 sites which are located at the central region (nucleotide 7,609 and 7,793; refs. 46, 47; Fig. 4B). This could be attributed to the role of the chromatin structure in protecting the AP-1–binding sites at the central region from exposure to the AP-1 transcription factor but not the AP-1 at the distal region (49). Moreover, we have shown that this distal AP-1 site is also important in HPV16 (Supplementary Figs. S7–S9). These findings illustrate the identification of a novel AP-1–binding site which activates E6/E7 through formation of MLL5β–AP-1 complex.

In conclusion, we have successfully identified a novel MLL5 isoform (MLL5β) which plays a role in activating E6/E7 through the formation of a complex with AP-1 and binds to the distal region of the HPV18-LCR. Upon depletion of MLL5β with siRNA, E6 and E7 oncoproteins are downregulated, leading to the restoration of p53 and hypophosphorylated Rb, thereby reestablishing the cell-cycle checkpoint controls. Because of the simultaneous downregulation of E6 and E7 transcripts by this isoform, MLL5β may be a more effective therapeutic target than E6 or E7 individually. Moreover, our study provides a platform for further investigations on the novel mechanism in which isoforms may be derived upon viral infection and the possibilities of MLL5β as a biomarker for HPV-induced cervical cancers.

C.W. Yew and P. Lee are the recipients of research scholarships from Yong Loo Lin School of Medicine, National University Health System, National University of Singapore. No potential conflicts of interest were disclosed by other authors.

FLAG-c-Jun plasmid was a kind gift from Dr. Kanaga Sabapathy of National Cancer Centre, Singapore.

This work was supported, in part, by BMRC-A*STAR, Singapore, grant R-183-000-164-305; NMRC-A*STAR, Singapore, grant R-183-000-293-213; Academic Research Fund Tier 1 grant R-183-000-286-112; and Ministry of Education Academic Research Fund Tier 2 grant R-183-000-195-112 to L.-W. Deng.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.