A novel carcinogenic PI3Kα mutation suggesting the role of helical domain in transmitting nSH2 regulatory signals to kinase domain

Background Mutations in PIK3CA, which encodes p110 subunit of PI3K class IA enzyme, are highly frequent in breast cancer. Here, we aimed to probe mutations in exon 9 of PIK3CA and computationally simulate their function. Method PCR/HRM and PCR/sequencing were used for mutation detection in 40 breast cancer specimens. The identified mutations were queried via in silico algorithms to check the pathogenicity. The molecular dynamics (MD) simulations were utilized to assess the function of mutant proteins. Result Three samples were found to harbor at least one of the E542K, E545K and L551Q mutations of which L511Q has not been reported previously. All mutations were confirmed to be pathogenic and MD simulations revealed their impact on protein function and regulation. The novel L551Q mutant dynamics was similar to that of previously found carcinogenic mutants, E542K and E545K. A functional role for the helical domain was also suggested by which the inhibitory signal of p85α is conducted to kinase domain via helical domain. Helical domain mutations lead to impairment of kinase domain allosteric regulation. Interestingly, our results show that p110α substrate binding pocket of helical domain in mutants may have differential affinity for enzyme substrates, including anit-p110α drugs. Conclusion The novel p110α L551Q mutation could has carcinogenic feature similar to previously known mutations. Each system contains about 113,000 atoms. Amber force field ff14SB was used in the simulations. Simulations were run using GPU-accelerated version of Amber16. 4000 cycles of energy minimization were done, with the first 1500 cycles with restraints on PIK3CA . Systems were heated up to 310 K within 40 ps in 4 steps and simulations were continued to 100 nanoseconds (ns). SHAKE algorithm was used for hydrogen bonds and 2 femtoseconds time step was applied. Coordinates were captured every 10 ps.

Background Mutations in PIK3CA, which encodes p110 subunit of PI3K class IA enzyme, are highly frequent in breast cancer. Here, we aimed to probe mutations in exon 9 of PIK3CA and computationally simulate their function. Method PCR/HRM and PCR/sequencing were used for mutation detection in 40 breast cancer specimens. The identified mutations were queried via in silico algorithms to check the pathogenicity. The molecular dynamics (MD) simulations were utilized to assess the function of mutant proteins. Result Three samples were found to harbor at least one of the E542K, E545K and L551Q mutations of which L511Q has not been reported previously. All mutations were confirmed to be pathogenic and MD simulations revealed their impact on protein function and regulation. The novel L551Q mutant dynamics was similar to that of previously found carcinogenic mutants, E542K and E545K. A functional role for the helical domain was also suggested by which the inhibitory signal of p85α is conducted to kinase domain via helical domain. Helical domain mutations lead to impairment of kinase domain allosteric regulation. Interestingly, our results show that p110α substrate binding pocket of helical domain in mutants may have differential affinity for enzyme substrates, including anit-p110α drugs. Conclusion The novel p110α L551Q mutation could has carcinogenic feature similar to previously known mutations.

Background
Breast cancer is the second cause of death due to cancer in women worldwide (1). In 2018, breast cancer accounted for 30% of 1.7 million new cases of cancer that were reported in the US (2).
The Phosphatidylinositol 3 kinase (PI3K) pathway is one of the most important oncogenic signaling pathway in breast cancer tumorigenesis and the majority of breast tumors harbor at least one molecular defect in this pathway (3). PI3K pathway plays a key role in various cellular functions such as survival, proliferation, differentiation, and angiogenesis (4). Class IA PI3K enzymes are composed of one p110 catalytic subunit which binds to a p85 regulatory subunit, to form a functional enzyme.
They are activated by receptor tyrosine kinases (RTKs) that remove inhibitory effect of p85 on p110 and therefore, PI3K becomes active. Among the members of class IA, p110α and p85α are expressed in most cells, encoded by PIK3CA and PIK3R1 genes, respectively (5). Large genomic studies such as The Cancer Genome Atlas (TCGA) have revealed that PIK3CA is the most frequently mutated gene in breast cancer (6,7). This data shows that ~40% of all breast tumors have at least one PIK3CA mutation that could be a representation of PI3K oncogenic pathway. The majority of hotspot gain-offunction PIK3CA mutations are clustered in exon 9 and 20 of the gene that correspond to the helical and kinase domains of p110α, respectively (8). Pathogenicity of E542K and E545K mutations located in exon 9 have been characterized by several studies suggesting that these mutations alter the interaction of helical domain and nSH2 and consequently, diminish the inhibitory effect of p85α on p110α (9). Another study (10) has shown that the mechanism of action of p110α mutations is not as simple as alterations in helical domain interactions with nSH2. In addition, a recent study (11) has suggested that p85α allosterically regulates p110α kinase activity via helical domain. Generally this change makes hyper-activation of PI3Kα followed by activation of AKT leading to an uncontrolled cell division (9,12).
Since breast cancer is a genetically complex and highly heterogeneous disease with diversity in therapeutic responses, our aim was to investigate the hotspot region in exon 9 of PIK3CA for pathogenic mutations using High Resolution Melting (HRM) followed by PCR+Sanger sequencing as a fast and cost-effective method for screening. Finally, we employed molecular dynamics simulation technique to study the potential function of the detected novel variants.

Tumor samples and DNA extraction
Tumor tissue was collected from 40 patients diagnosed with primary breast cancer ranging from 35 to 65 years of age. The study was approved by the ethics committee of Isfahan University of Medical Sciences and written informed consent was obtained from all the patients.
The tumor tissue was flash frozen in liquid nitrogen and DNA was extracted using Qiagen DNeasy Blood & Tissue kit (Qiagen, Germany).

PCR/HRM
PCR/HRM was used to identify samples with potential mutations on the hotspot region of exon 9 (from aa513 to aa554). The primer set that was used for amplifying the region of interest contained forward 5′TGACAAAGAACAGCTCAAAGCA-3′ and reverse 5′-AGCACTTACCTGTGACTCCA-3 that produced a 96 bp amplicon. HRM analysis was performed using Corbett Research Rotor Gene-6000 (Qiagen-Germany) and the HRM Master Mix (Type-it HRM PCR Kit Qiagen-Germany). PCR was performed by incubating the reaction at 95°C for 5 minutes followed by 45 cycles of 95°C for 10 seconds, 64°C for 30 seconds and 72°C for 10 seconds. In the final step, HRM analysis was carried out from 70-85°C with ramping at 0.2°C/s. The PCR/HRM curve was analyzed via Q 5plex HRM software to compare unknown samples with wild type (WT) control.

Sanger sequencing
As a second approach to verify the detected hotspot mutations via PCR/HRM, DNA from all the HRM positive samples was amplified by 5′-CATCTGTGAATCCAGAGGGGA-3′ as the forward and 5′-AGCACTTACCTGTGACTCCA-3′ as the reverse primers producing an amplicon of 201bp. PCR products were purified using QIAquick PCR purification kit (Qiagen-Germany) and sequenced by ABI prism 3730 sequencer (Applied Biosystems, Waltham, MA, USA). The sequencing data was analyzed via Chromas version 2.33 software.2In Silico Prediction

Algorithms
The pathogenicity of novel mutations in exon 9 of PIK3CA was investigated by several in silico tools that predict the deleterious effect of coding single nucleotide variants (SNVs). These include PolyPhen2 (13), SIFT (14), MutationTaster (15), PROVEAN (16), PhD-SNP (17), I-MutantDDG-Seq Suite (18) and Hope software (19). In addition several tumor mutation databases were searched for identified mutations including Catalog of Somatic Mutation in Cancer (20), My Cancer Genome (21) and TCGA (22). Simulation boxes were made using Amber16/LEaP (24). Protein complexes were solvated into an octahedral TIP3P water model box with at least 15Å distance from box boundaries. Na + and Clions were added to neutralize and maintain 150 mM physiologic saline concentration. Each system contains about 113,000 atoms. Amber force field ff14SB was used in the simulations.

Molecular Dynamics Simulations
Simulations were run using GPU-accelerated version of Amber16. 4000 cycles of energy minimization were done, with the first 1500 cycles with restraints on PIK3CA. Systems were heated up to 310 K within 40 ps in 4 steps and simulations were continued to 100 nanoseconds (ns). SHAKE algorithm was used for hydrogen bonds and 2 femtoseconds time step was applied. Coordinates were captured every 10 ps.

Discovery of novel mutation in exon 9 of PIK3CA
Initial screening via PCR/HRM was performed on DNA extracted from 40 tumor samples and the HRM positive samples were subsequently investigated by Sanger sequencing for mutations on exon 9. Four tumors displayed an altered melting pattern in PCR/HRM of which three were confirmed to harbor at least one of the E542K, E545K or L551Q mutations. E542K and E545K variants were previously reported as pathogenic mutations in My Cancer Genome (21), cosmic (20) and TCGA (22), however, L551Q mutation did not exist in any of the cancer databases. Interestingly, the sample with L551Q variant had a pathogenic mutation, E542K, in hotspot region of exon 9 (Table 1). This prompted us to investigate the pathological effect of E542K, E545K, and the newly detected L551Q by further in silico studies.

p110α L551Q dynamics displays a carcinogenic activity
The impact of L551Q was first predicted to be disease causing using various variant annotation tools that is described in Table 2 in details. We then used molecular dynamics simulation to study the structural effects of all mutations we found in p110α helical domain. Although the p110α-p85α is a large complex, the p110α kinase and p85α nSH2 domains are the main functional domains of the complex, with the p110α helical domain mediating their interactions (Figure 1). We restricted our simulation systems to these three domains that have the central role in PI3Kα activity to investigate how the kinase activity of p110α is regulated by p85α. domains. Indeed, dynamic behavior of L551Q is compatible with that of E542K and E545K mutants.
Thus, these results strongly suggest that the novel p110α L551Q mutation could harbor carcinogenic features similar to previously known mutations.

nSH2 allosterically regulates p110α catalytic activity through helical domain
We then asked how a small change in helical domain can affect dynamics of kinase domain. To answer that, we utilized community network analysis to investigate the pathways that connect motions of kinase to helical and nSH2 domains. Edge betweenness measurements revealed that E542 mediates the communication between nSH2 and helical domains in WT. Then D538 in the helical domain, in vicinity of E542, interacts with N996 in the kinase domain, and directly spreads the nSH2 regulatory signal throughout the kinase domain ( Figure 3A). So, these three residues (E542, D538, and N996) have a central role in conducting the regulatory signal from nSH2 domain to kinase domain.
As D538 is spatially oriented in the middle of the E542 and N996, changing the repulsive force of E542-D538 to an attractive one by E542K mutation influences the orientation of D538 relative to N996, and consequently disrupts the interaction between helical and kinase domains. Therefore, E542K mutant cannot conduct the nSH2 regulatory signal to kinase domain, but instead, diverts the signal to the helical domain, leading to the detachment of kinase domain from helical and nSH2 domains. On the other hand, E545K and L551Q mutations change helical-nSH2 and helical domain internal interaction respectively, both resulting in an impaired interaction of D538-N996.
For further confirmation of these results, we tried to find suboptimal pathways from nSH2 domain to three kinase domain residues including: K776 in the p-loop which identifies p110α substrate specificity, H917 in the catalytic loop which accommodate ATP and catalysis the enzymatic reaction, and K941 in the activation loop, which makes the initial interactions with substrates. In WT, signals from nSH2 domain allosterically regulate catalytic activity by two main pathways: one passes through kinase domain H12, and the other through helical domain (data not shown). In all mutants, these allosteric pathways become significantly longer ( Figure 3B), because the nSH2 regulatory signal has been trapped in the helical domain.
Collectively, communication network analysis shows that the helical domain mutations disrupt the connection between helical and kinase domains, resulting in detachment of kinase domain from rest of the complex. Consequently, nSH2 signal is trapped in the helical domain since it cannot find its way towards kinase domain which makes it inaccessible to nSH2 domain in all three mutants.

P110α substrate affinity is different in each mutant
Some of the known mutants of p110α has been proved to result in resistance to anti-p110α drugs (26,27). Thus, it is clinically important to find out whether breast cancer patients with the novel carcinogenic p110α L551Q mutation may be drug resistant. To this end, we investigated the flexibility and shape of the p110α substrate binding pocket (SBP) using pairwise root-mean-square deviation (RMSD) and radius of gyration (R g ) calculations, respectively. RMSD distribution plot (Figure 4) shows that the flexibility of the SBP has been slightly reduced in all mutants, specifically in E545K, compared to WT. For an enzyme with various substrates, a more rigid SBP could result in higher affinity for a specific substrate. To further inspect this hypothesis about p110α catalytic site, we compared distributions of R g of p110α SBP in all simulated systems ( Figure 4B). R g variation shows the major difference between E542K and E545K mutants, suggesting that the shape of apo-enzyme SBP differs among mutants (p < 2.2e-16).
Accordingly, geometric analysis suggests that all helical domain mutations result in a slightly rigid p110α SBP with different shapes. This proposes that a specific therapeutic compound may not have the same affinity for all p110α mutants. Further biophysical experiments are warranted to test this hypothesis.

Discussion
Our findings suggested that newly detected L551Q mutation may harbor carcinogenic features similarly to known mutations such as E545K and E542K located in p110α helical domain. Multiple in silico variant annotation tools showed that L551Q is likely to be disease causing. In addition, molecular dynamic simulation was used to study the functional impact of the three mutations E545K, E542K and L551Q on the protein structure. Our MD simulation studies revealed that the helical domain mutations prevent the allosteric regulation of p110α catalytic activity.
PIK3CA is one of the significantly mutated genes in breast cancer with ~40% frequency. The role of PIK3CA mutations in pathophysiology and drug resistance in breast cancer is not well known. Several studies (28,29) have suggested that helical domain mutations are associated with resistance of HER2-positive breast cancer to trastuzumab. Here, we investigated the exon 9 of PIK3CA as a hotspot region to helical domain mutations. For the first time, we found a novel mutation in p110α, L551Q, which could be functionally carcinogenic.
The interaction between p110α helical domain with nSH2 domain of p85α is not yet fully uncovered. Therefore, we performed an MD analysis comparing DCCM and PC1 between WT and mutants. Our results showed that helical domain mutations extremely affect dynamics of p110α-p85α complex. In addition, communication network analysis displayed that the helical domain mutations disrupt the connection between helical and kinase domains. The defective allostery in each mutant affects p110α substrate binding pocket (SBP) differently, resulting in various SBP shapes in each mutant.
Previous MD simulation studies of p110α helical domain mutants have shown that E542K and E545K mutations disrupt nSH2 interactions with helical domain, which results in helical domain detachment (30). We also observed a similar partial detachment. It has been shown that p85α stabilizes p110α, and the enzyme needs its regulatory partner for better functionality (31). However, helical domain mutations result in a p110α enzymatic activity independent of p85α (10). Thus, the helical domain mutations result in a p110α which benefits from the stabilizing effects of p85α, but does not receive its inhibitory signals.
Gkeka and colleagues (32) have found that SBP is more compact in WT than in H1047R, suggesting WT has an intrinsic resistance to substrates binding. Here we found that SBP shape in the kinase domain is different in WT compared to that of mutants, and even between mutants. The major difference in SBP shape is seen between E542K and E545K mutants. This may suggest that mutants of helical domain may have different substrate affinities. Notably, p110α inhibitors bind to this pocket and any structural changes in SBP could affect drug affinity for p110α (33). Thus, we suggest that genotyping PIK3CA in breast tumors prior to the treatment could have clinical implications, as we speculate that a specific p110α inhibitor could have different outcomes in each p110α mutant.
Indeed, p110α has role in drug resistance in cancer patients (34). However, our simulations do not include anti-p110α effects on helical domain mutants, but it would be of great interest if anti-p110α drugs affinity be measured for all of the helical domain mutants.
Various functional studies on PIK3CA mutations have been carried out of which, Meyer et al have investigated the overexpression of E545K in a transgenic mouse model. Notably, E545K mutation induces heterogeneous mammary carcinomas but with longer latency than H1047R (35). Another study on E545K and E542K in chicken embryonic fibroblasts suggested that the gain of function induced by helical domain mutations is independent of binding to p85 but requires interaction with RAS-GPT (10). Another in vivo study on induced tumors in the chorioallantoic membrane of the chicken embryo showed that the three prevalent mutants of p110α, E542K, E545K, and H1047R are able to promote angiogenesis and increase the activation of the Akt pathway (36,37).
A recent study by Vasan N et al. (38) showed that double PIK3CA mutations can result in increased activity of PI3K activity and downstream signaling pathways. Although we were not able to test whether E542K and L551Q are on the same allele, it is plausible that these two mutations may have synergistic effect on tumor growth. Future studies are warranted to investigate the function of double mutations in tumor cells proliferation and more importantly, response to PI3K inhibitors.

Conclusions
Our MD simulations show that L551Q mutation has a functional dynamics network similar to E542K and E545K mutations and thus, may have similar carcinogenic activity. However, we suggest that future in vitro/vivo studies are warranted to characterize the potential carcinogenic activity of L551Q mutation in breast cancer.

Ethics approval and consent to participate:
This manuscript does not contain any study with animal performed and all of studies involving All the patients were consented according to the ethical standards of the Isfahan University of Medical Sciences with reference number: ir.mui.rec.1396.3.131. Written consent was obtained from all the patients.

Consent for publication:
Written informed consent was signed by all the authors included in the study.

Availability of data and material:
The datasets used and/or analyzed during this study is available by the corresponding authors upon request.

Competing interests:
The authors declare that there is no conflict of interests.

Funding
This study was funded by medical university of Isfahan scholarship which provides analysis, collection, and interpretation of data.

Authors' Contributions
Designing the study and collection of data by SG, designing and analysis of the MD simulations by SA, writing the manuscript by SG, HM, SA, and AB, providing material and tumor samples by FK and HM, interpretation of the data by AB, final approval of study by MS and AB Tables     p110α substrate binding pocket flexibility and shape. A: SBP flexibility was measured by calculating distribution of pairwise RMSDs of each system. B: Distribution of SBP Rg among different mutants that could be an indication of SBP shape. Kruskal-Wallis test was used to test if the difference between the medians is significant.