Your Drug Mileage May Vary: The PIK3CA Double Mutation Story

“People have studied the PIK3CA gene for 20 years, yet no one has looked into how double mutations of the gene affects drug response in cancer treatment.”

This was Neil’s (@VasanNeil) answer when I asked him how his recent Science publication, “Double PIK3CA mutations in cis increase oncogenicity and sensitivity to PI3Ka inhibitors”, came about.

PIK3CA (Phosphatidylinositol-4,5-Bisphosphate 3-Kinase Catalytic Subunit Alpha) is a protein-coding gene that encodes for a subunit of the PI3K protein, which is involved in the growth, survival, movement, and structure of cells. PIK3CA is the most mutated oncogene in human cancer and is associated with a wide spectrum of cancer types including breast, colon, uterine, etc.

Mutated alleles of PIK3CA are potential drug targets. Two challenges exist in implementing PI3K inhibitors. The first is that patient tumors are not routinely sequenced to identify targetable mutations. The second is the lack of biomarkers that predict drug sensitivity or resistance. The drug alpelisib was developed to target PIK3CA in estrogen receptor-positive (ER+) breast cancer patients.

This is where the story begins.

Liz: How and when did your lab begin to look for double mutations in the PIK3CA gene?

Neil: We began the project around July 2016. We had two pieces of data that hinted at a double mutation being responsible for drug hypersensitivity. One was that in a phase 1 clinical trial testing alpelisib in metastatic breast cancer, nineof the 51 patients had double mutations and did better. The other was a breast cancer patient who was an exceptional responder to alpelisib monotherapy. We had detected double PIK3CA mutations with equal variant allele frequencies, hinting that cis-mutations (double mutations on the same allele) may be responsible for the better outcome.

Supplementary Figure S2B, C from Vasan et al. showing the frequency of multiple PIK3CA mutations in either breast cancer or pan-cancer cohorts.

Liz: Double mutations are not uncommon — your paper estimated it to be around 10–19% of all PIK3CA mutation cancers. How reliable is this estimate?

Neil: The 19% came from the SANDPIPER study, which uses circulating tumor DNA, which could lead to an overestimation by picking up clonal mutations. It is hard to know. Certain metastases have heterogeneity, so tumor sampling is limiting. Still, an estimate of 10–20% double mutations across all cancers would be a good guess.

Liz: And for PI3KCA, it would be either single or double mutations, but not three or more?

Neil: ~96% of the patients in the dataset we looked at that had multiple mutations had double mutations. It’s possible that too many mutations are deleterious.

Supplementary Figure 5A from Vasan et al. showing the most common PIK3CA double mutations, their cDNA distance, and whether each can be resolved using Sanger or SMRT Sequencing.

Liz: How did the use of long reads, specifically PacBio’s SMRT Sequencing, come about?

Neil: We did Sanger sequencing first, but the transcript is ~3 kb long and some of the mutations are far apart, and can only be phased by long reads. I had read about PacBio and just went looking for service providers on the website and discovered that one of the providers is Bobby Sebra who is also at Mt Sinai. I sent him an email and that was the beginning of it!

Supplementary Figure 5E from Vasan et al. showing the phasing report from SMRT Sequencing of PIK3CA amplicon in breast cancer cell lines. Three of the four cell lines showed major cis- double mutations. For the six patient samples in the clinical study, refer to Figure 1F.

Liz: You sequenced six tumor samples using PacBio and were able to phase all of them, showing that the double mutations are cis- (on the same allele). I noticed the phasing was analyzed using the PacBio Minor Variant software, which doesn’t natively handle alternative splice forms. Is there only one dominant transcript for PIK3CA?

Neil: There might be fusions (see Matissek et al 2017), but yes, alternative splicing seems rare.

Liz: You then went on to propose a potential mechanism for why double mutations affect the PI3K pathway. You showed that, compared to single mutations, double mutations have a lower melting temperature and have increased anionic and lipid binding. You hypothesize that double mutations result in cells being more dependent on the PI3K pathway for proliferation and survival, making them more susceptible to PI3Ka inhibitors.

Help me out with understanding Fig 5. in the paper, then. You show that taselisib is more effective than placebo, if you consider both single and double mutation patients. Now, if you separate the single and double mutation patients, the double mutation patients benefit from taselisib, but the single mutation patients appear to not respond to the drug significantly better than the placebo. Doesn’t this suggest that single mutation patients are not benefiting from taselisib?

Neil: Our assessment of patients with multiple mutations was a retrospective exploratory assessment of the SANDPIPER clinical trial and this was not an endpoint of the study. However, these data do corroborate our hypothesis that multiple mutant patients have increased tumor shrinkage and an increased response rate. We know from multiple randomized phase 3 clinical trials that patients with PIK3CA mutant breast cancers benefit from PI3K inhibitors, but this work raises the possibility that the multiple mutant patients derive a greater benefit. I would add that in this study, we showed that double mutations hyperactivate the PI3 kinase, but we also need to look at other genes that are also in the PI3K pathway as this may also have a role in the response of patients to PI3K inhibitors.

We would need to see the findings in this paper reproduced in other clinical trials. PI3K inhibitors are currently second-line treatments in estrogen receptor positive metastatic breast cancer and have significant on target toxicities due to inhibition of the insulin signaling pathway. One hypothesis based on our work is that multiple mutant patients may derive a greater benefit from PI3K inhibitor in earlier lines of treatment.

Liz: How has your experience with using long-read sequencing technology affected you as a clinician?

Neil: From SMRT Sequencing, we got the data we needed to phase the allelic configurations of the two mutations. I believe that we have been so focused on single mutations that we have missed important phenomena like double PIK3CA mutations and likely there are other higher order genetic and genomic patterns with clinical implications waiting to be discovered. Long-read sequencing will enable us to get past that hurdle and find different patterns. We would potentially use long reads to look at cancer at a higher order level such as fusions, mutational signatures, and epigenetics.

Liz: SMRT Sequencing is now being used in many clinically relevant applications. How do you see it moving into clinical practice?

Neil: It is important to use the right technology to ask the right questions. Long-read sequencing has taken off in areas such as microbiology and HLA typing. In cancer, it’s about studying high order or second-order genomic changes. The more we use long reads to query these questions the more it will become a standardized tool.

Liz: What you have done in this study could be applied to other genes as well, right?

Neil: Yes, this is just the tip of the iceberg for other ongoing work. We must think outside the box.