Part IV · Molecular biology and therapeutic vulnerabilities · Chapter 17

PI3K, AKT, mTOR, and related networks

A pathway whose every successful trial is a combination, and why that is a finding rather than a convenience.

1 · PIK3CA, AKT1, and PTEN alterations

Three genes account for most activation of this pathway in breast cancer, and they are altered in three different ways.

PIK3CA encodes p110alpha, the catalytic subunit. Activating mutations cluster in the helical and kinase domains, and they occur in roughly 40% of hormone receptor positive, HER2-negative breast cancer1. They are typically clonal and present in the primary tumour. That makes them unlike ESR1 mutations, which are acquired under aromatase inhibitor exposure and are the subject of Endocrine therapy and the emergence of ESR1 alterations. The testing implication follows directly. There is no pressure-dependent window for PIK3CA testing, because the alteration was there before treatment started.

AKT1 carries a single recurrent hotspot. The E17K substitution in the pleckstrin homology domain is transforming and drives constitutive membrane localisation2. It is uncommon, and its detection depends on whether the panel covers that codon.

PTEN is the negative regulator, and it is lost rather than activated. Loss occurs by deletion, by mutation, by promoter methylation and by post-translational routes. This is the measurement problem in the section. A sequencing panel reports PTEN mutations. Copy number analysis reports deletions. Immunohistochemistry reports absent protein however it was lost. The three assays identify overlapping but not identical populations, and a trial that used one of them has not characterised the others.

Grouping matters clinically because trials have grouped. In CAPItello-291, an AKT pathway alteration was defined as an alteration in PIK3CA, AKT1 or PTEN. On that definition 289 of 708 randomised patients, or 40.8%, were pathway-altered3. That number is a property of a composite definition and of the assays behind it, not a statement about how often the pathway is active.

2 · Isoform-specific biology and the therapeutic window

Class I PI3K has four catalytic isoforms, and only p110alpha is recurrently mutated in breast cancer. That same isoform transduces insulin receptor signalling in liver, muscle and fat. The therapeutic window of this class is set by that single fact. Hyperglycaemia on an alpha-selective inhibitor is not an off-target effect. It is the target, hit in the tissue where its signal is wanted.

The numbers make the point concrete. With alpelisib and fulvestrant, grade 3 or 4 hyperglycaemia occurred in 36.6% of patients against 0.7% on placebo with fulvestrant. Discontinuation for adverse events occurred in 25.0% against 4.2%1. One patient in four stopped the drug for toxicity.

Inavolisib is also alpha-selective, and it additionally promotes degradation of mutant p110alpha4. With inavolisib added to palbociclib and fulvestrant, grade 3 or 4 hyperglycaemia occurred in 5.6% of patients against 0% on placebo, and no grade 3 or 4 rash was observed4. The contrast with alpelisib is striking. It is also not a comparison. Different trials, different populations, different backbones and different dosing cannot be placed side by side, and no head-to-head study exists.

Capivasertib acts one step down, on all three AKT isoforms. Pan-isoform inhibition forecloses selectivity as a source of window, so tolerability has to come from dose and schedule instead. Rash was the most frequent grade 3 or higher adverse event in the capivasertib arm of CAPItello-2913.

Caution

Cross-trial toxicity comparisons within this class read as if one agent were cleaner than another. They are not evidence of that. Hyperglycaemia rates depend on the population's baseline glycaemic status, on the monitoring schedule, on the grading convention and on what the combination partner contributes. The only defensible statement is that each agent's rate is what that trial observed in that population.

3 · mTORC1 and mTORC2 signaling and feedback reactivation

The kinase mTOR operates in two complexes that do different jobs. The rapamycin-sensitive complex, mTORC1, controls translation and growth through S6 kinase 1 and 4E-BP1. The second complex, mTORC2, phosphorylates AKT at serine 473, one of the two phosphorylation events required for full AKT activity.

The feedback loop is the central fact. Active mTORC1, acting through S6 kinase 1, suppresses insulin receptor substrate 1. Inhibiting mTORC1 lifts that suppression. Upstream receptor tyrosine kinase signalling then rises and AKT is activated5. A drug aimed at the bottom of the pathway switches the top of it back on.

Two consequences follow. Rapalogues are cytostatic rather than cytotoxic, partly because they inhibit 4E-BP1 phosphorylation incompletely. And their benefit, while real, is bounded. BOLERO-2 randomised 724 patients with hormone receptor positive advanced breast cancer, all with recurrence or progression on a non-steroidal aromatase inhibitor, to everolimus with exemestane or placebo with exemestane. At the interim analysis, median progression-free survival was 6.9 months against 2.8 months by local investigator assessment, and 10.6 months against 4.1 months by central assessment6.

That discrepancy deserves attention rather than averaging. The same patients and the same events, read by two conventions, gave medians 3.7 months apart in the everolimus arm and 1.3 months apart in the control arm. Any comparison of this result with a trial that used only one assessment method is comparing measurement as much as biology.

ATP-competitive mTOR inhibitors hit both complexes and remove the serine 473 escape. They also remove whatever tolerability the partial mechanism of rapalogues provided, which is why this generation has not displaced everolimus in practice.

4 · RAS and MAPK involvement in breast cancer

RAS mutations are uncommon in breast cancer relative to pancreatic or colorectal disease. The pathway is nonetheless activated, by other routes, and disproportionately in tumours that have already failed endocrine therapy.

Sequencing of 1,918 breast cancers, of which 1,501 were hormone receptor positive, allows the comparison to be made directly. Among 692 tumours previously exposed to hormonal therapy, alterations in MAPK pathway genes and in the oestrogen receptor transcriptional machinery were enriched. Activating ERBB2 mutations and NF1 loss-of-function mutations were more than twice as common in endocrine-resistant tumours. Alterations of this kind were present in 22% of tumours overall7.

Two features of that finding matter more than the frequency. The alterations were mutually exclusive with ESR1 mutations. And they were associated with a shorter duration of response to subsequent hormonal therapy7.

Mutual exclusivity is the structural point. Escape from endocrine therapy through a reactivated receptor and escape through a parallel growth pathway are alternatives, not stages. A patient whose circulating tumour DNA shows no ESR1 mutation at progression has not necessarily escaped by a mechanism that is yet to appear. They may have escaped by the other route entirely, and that route is not on most endocrine-resistance panels.

5 · MYC, FGFR, and other recurrent amplification-driven dependencies

Amplification-driven dependency differs from mutation-driven dependency in a way that shapes what a drug can do. A mutation creates a protein that behaves differently. An amplification creates more of a protein that behaves normally. The first is a qualitative target. The second is a quantitative one, and partial inhibition of a quantitative target restores the pre-amplification state rather than removing the signal.

FGFR1 amplification occurs in approximately 10% of breast cancers, and in 16% to 27% of luminal B tumours. Amplified lines show enhanced ligand-dependent signalling through MAPK and PI3K-AKT, basal ligand-independent signalling, and resistance to 4-hydroxytamoxifen that is reversed by silencing FGFR18. FGFR1 signalling suppresses progesterone receptor expression, which is why amplified cancers are frequently progesterone receptor negative8. That makes a routine immunohistochemistry result a weak surrogate for an amplification nobody measured.

MYC amplification at 8q24 is recurrent and is not directly druggable. Its dependencies are indirect, and they reach into translation and metabolism rather than into kinase signalling. Those are developed in Metabolism, proteostasis, and stress responses.

The measurement caveat is the same for all of them. An amplification call depends on the copy number threshold, on tumour purity and on ploidy correction. It also depends on where the amplicon lives, because extrachromosomal amplification produces wider cell-to-cell variance than intrachromosomal amplification does. That argument is made in full in Amplicon topology, extrachromosomal DNA, and the mechanistic origin of instability and applies here unchanged.

6 · Hippo signaling and FAT1 loss

Hippo is a growth-control network that runs parallel to PI3K rather than through it. Its output is the transcriptional coactivators YAP and TAZ. When Hippo signalling is intact, YAP and TAZ are held in the cytoplasm. When it fails, they accumulate in the nucleus and drive transcription.

FAT1 is an atypical cadherin acting upstream. In a genomic analysis of 348 ER-positive breast cancers treated with CDK4/6 inhibitors, loss-of-function mutations in FAT1 and in RB1 were linked to resistance. FAT1 loss produced marked elevation of CDK6, and suppressing CDK6 restored sensitivity. The induction was mediated by the Hippo pathway, with YAP and TAZ accumulating on the CDK6 promoter. Alterations in other Hippo components also promoted resistance9.

This is the clearest available example of pathway convergence in the chapter. A cadherin, a cytoskeletal signalling network and a transcriptional coactivator pair end at the promoter of a cell cycle kinase. Nothing about the route would be predicted from the target of the drug.

Two constraints belong with it. FAT1 loss is uncommon, and a mechanism found in a minority of resistant tumours does not explain resistance in general. And the alteration was identified in tumours already treated, so its frequency in that series is a frequency among the resistant rather than a prevalence. FAT1 also appears in the PALOMA-3 on-treatment alteration set described in Emergence of ESR1, PIK3CA, RB1, CCNE1, and FAT1 alterations on treatment.

7 · Cross-talk with endocrine and HER2 signaling

The relationship between this pathway and the oestrogen receptor is reciprocal. Blocking either arm raises the other, which is why single-agent activity in this class is modest and why every positive trial in it is a combination.

Suppressing PI3K induces oestrogen receptor dependent transcription. Occupancy of the receptor at promoters of upregulated genes rises, and ESR1 messenger RNA and receptor protein increase. That was shown in cell lines, in xenografts, in patient-derived models and in tumours from patients receiving the PI3Kalpha inhibitor alpelisib. Fulvestrant markedly sensitised ER-positive tumours to PI3Kalpha inhibition and produced tumour regression in vivo10. The epigenetic mechanism runs in part through the histone methyltransferase KMT2D11.

The traffic runs the other way as well. HER2 signalling destabilises p2712 and represses FOXO3a, which lowers oestrogen receptor expression13. Oestrogen receptor signalling is itself upregulated as an adaptive survival response in HER2-positive tumours under HER2-directed therapy14. Hyperactivation of EGFR and HER2 drives resistance to endocrine therapy and to CDK4/6 inhibition together15.

Read as a system, these are not three cross-talk observations. They are one property. The oestrogen receptor, HER2 and PI3K form a network with compensating inputs, and inhibiting any single node raises flux through the others. The clinical corollary is that the escape route is predictable from what was blocked, which is the premise of Therapy as selective pressure, the central premise.

8 · Pathway convergence and combination logic

Three randomised trials have converted the argument above into results, and each is worth stating with its denominator.

SOLAR-1 compared alpelisib and fulvestrant with placebo and fulvestrant. Among 341 patients with confirmed tumour-tissue PIK3CA mutations, median progression-free survival was 11.0 months against 5.7 months, hazard ratio for progression or death 0.651.

CAPItello-291 compared capivasertib and fulvestrant with placebo and fulvestrant in 708 patients. Median progression-free survival was 7.2 months against 3.6 months overall, hazard ratio 0.60. In the 289 patients with AKT pathway alterations it was 7.3 months against 3.1 months, hazard ratio 0.503.

In INAVO120, inavolisib was added to palbociclib and fulvestrant in 161 patients, against 164 receiving placebo with the same backbone. All had PIK3CA-mutated disease and had relapsed during or within 12 months of completing adjuvant endocrine therapy. Median progression-free survival was 15.0 months against 7.3 months, hazard ratio 0.434. At the final overall survival analysis, median overall survival was 34.0 months against 27.0 months, with a hazard ratio for death of 0.67 and a 95% confidence interval of 0.48 to 0.9416.

That last result is the first overall survival benefit for any agent directed at this pathway in breast cancer. Three things changed at once, so the attribution is not clean. The combination blocked PI3K, CDK4/6 and the oestrogen receptor together. The population was selected for early relapse on adjuvant endocrine therapy, which the 7.3-month control-arm median reflects. And the agent degrades mutant p110alpha rather than only inhibiting it.

Caution

The 15.0-month median in INAVO120 and the 11.0-month median in SOLAR-1 are not comparable. The populations differ in prior therapy, in relapse timing and in backbone. Comparing medians across trials in this class tends to credit the drug with what belongs to the entry criteria.

Whether the class is worth its burden outside the biomarker-selected group is the live controversy. SOLAR-1 answers part of it. In the cohort without a PIK3CA mutation the hazard ratio was 0.85, with a posterior probability of a hazard ratio below 1.00 of 79.4%1. Benefit was not established there, and the hyperglycaemia and discontinuation rates in Isoform-specific biology and the therapeutic window were paid regardless. The biomarker is doing as much work on the harm side as on the benefit side.

What escapes a blocked pathway is the question Part X inherits. Reactivated oestrogen receptor transcription is one route10, MAPK alteration a second7, and CDK6 elevation through Hippo a third9. A fourth is loss of the negative regulator itself. In a patient treated with alpelisib and autopsied after progression, all 14 metastatic sites had copy loss of PTEN, and the refractory lesions carried additional and different PTEN alterations that abolished protein expression17. Separate deposits reached the same null phenotype by separate mutations, which is convergent evolution observed inside one patient. Treatment selection on this pathway in metastatic disease is developed in Metastatic hormone receptor-positive, HER2-negative disease and Targeted therapy.

References

  1. Andre F, Ciruelos E, Rubovszky G, et al. Alpelisib for PIK3CA-mutated, hormone receptor-positive advanced breast cancer. N Engl J Med 2019 380:1929-1940. PMID 31091374
  2. Carpten JD, Faber AL, Horn C, et al. A transforming mutation in the pleckstrin homology domain of AKT1 in cancer. Nature 2007 448:439-444. PMID 17611497
  3. Turner NC, Oliveira M, Howell SJ, et al. Capivasertib in hormone receptor-positive advanced breast cancer. N Engl J Med 2023 388:2058-2070. PMID 37256976
  4. Turner NC, Im SA, Saura C, et al. Inavolisib-based therapy in PIK3CA-mutated advanced breast cancer. N Engl J Med 2024 391:1584-1596. PMID 39476340
  5. O'Reilly KE, Rojo F, She QB, et al. mTOR inhibition induces upstream receptor tyrosine kinase signaling and activates Akt. Cancer Res 2006 66:1500-1508. PMID 16452206
  6. Baselga J, Campone M, Piccart M, et al. Everolimus in postmenopausal hormone-receptor-positive advanced breast cancer. N Engl J Med 2012 366:520-529. PMID 22149876
  7. Razavi P, Chang MT, Xu G, et al. The genomic landscape of endocrine-resistant advanced breast cancers. Cancer Cell 2018 34:427-438. PMID 30205045
  8. Turner N, Pearson A, Sharpe R, et al. FGFR1 amplification drives endocrine therapy resistance and is a therapeutic target in breast cancer. Cancer Res 2010 70:2085-2094. PMID 20179196
  9. Li Z, Razavi P, Li Q, et al. Loss of the FAT1 tumor suppressor promotes resistance to CDK4/6 inhibitors via the Hippo pathway. Cancer Cell 2018 34:893-905. PMID 30537512
  10. Bosch A, Li Z, Bergamaschi A, et al. PI3K inhibition results in enhanced estrogen receptor function and dependence in hormone receptor-positive breast cancer. Sci Transl Med 2015 7:283ra51. PMID 25877889
  11. Toska E, et al. PI3K pathway regulates ER-dependent transcription in breast cancer through the epigenetic regulator KMT2D. Science 2017. PMID 28336670
  12. Yang HY, et al. Oncogenic signals of HER-2/neu in regulating the stability of the cyclin-dependent kinase inhibitor p27. J Biol Chem 2000. PMID 10859299
  13. Guo S, Sonenshein GE. FOXO3a regulates estrogen receptor alpha expression and is repressed by the Her-2/neu/PI3K/Akt signaling pathway. Mol Cell Biol 2004. PMID 15367686
  14. Giuliano M, et al. Upregulation of ER signaling as an adaptive mechanism of cell survival in HER2-positive breast tumors treated with anti-HER2 therapy. Clin Cancer Res 2015. PMID 26015514
  15. Belli S, et al. EGFR and HER2 hyper-activation mediates resistance to endocrine therapy and CDK4/6 inhibitors in ER+ breast cancer. Cancer Lett 2024
  16. Jhaveri KL, Im SA, Saura C, et al. Overall survival with inavolisib in PIK3CA-mutated advanced breast cancer. N Engl J Med 2025 393:151-161. PMID 40454641
  17. Juric D, Castel P, Griffith M, et al. Convergent loss of PTEN leads to clinical resistance to a PI(3)Kalpha inhibitor. Nature 2015 518:240-244. PMID 25409150