PIK3CA Mutations: Are They a Relevant Target in Adult Diffuse Gliomas?

Abstract

Gliomas are the most common and lethal malignant primary brain tumors in adults, associated with the highest number of years of potential life lost. The latest WHO classification for central nervous system tumors highlighted the need for new biomarkers of diagnosis, prognosis, and response to therapy. The PI3K/Akt signaling pathway is clearly implicated in tumorigenesis, being one of the most frequently altered pathways in cancer. Activating PI3KCA mutations are oncogenic and can influence both prognosis and treatment response in various tumor types. In gliomas, however, studies have reported inconsistent PIK3CA mutational frequencies, ranging from 0% to 30%. Furthermore, the impact of these alterations on glioma diagnosis, prognosis, and therapy response remains unclear. Current evidence suggests that PIK3CA mutations may represent early and constitutive events in glioma development, associated with worse glioblastoma prognoses, earlier recurrences, and widespread disease. Among these, the hotspot mutation H1047R has been particularly associated with a more aggressive phenotype while also modulating the neuronal microenvironment. In this review, we examine the clinical relevance of PIK3CA mutations across different cancers, with a particular focus on their emerging role in glioma. Moreover, we also discuss the therapeutic potential and challenges of targeting PIK3CA mutations in the context of glioma.

1. Introduction

Gliomas are the most common malignant primary brain tumors in adults (80–85%), representing ~23% of all primary central nervous system tumors [1,2]. Arising from the malignant transformation of glial cells, such as astrocytes and oligodendrocytes or their precursors [3,4,5], these tumors develop primarily in the brain, but can also appear in the spinal cord, brain stem, or cerebellum [2]. Diffuse gliomas are especially known to be highly heterogeneous, aggressive and infiltrative [6], being associated with approximately 20 years of potential life lost, the highest number out of most cancers [7]. Particularly, only ~7% of patients with glioblastoma multiforme (GBM), the most common and aggressive glioma subtype, survive five years after diagnosis [2]. The gold standard glioma treatment consists of maximal surgical resection followed by radiotherapy and chemotherapy, which is not efficient [8]. Even though these neoplasms do not typically metastasize to other organs, gliomas frequently invade surrounding brain tissue, presenting a diffuse phenotype. Thus, complete surgical resection becomes impossible, consequently leading to tumor recurrence and dismal prognosis [6,9]. Furthermore, due to the highly heterogeneous nature of these tumors, chemotherapy and radiotherapy resistance also occurs, preventing tumor cell elimination [10,11].

Until 2016, gliomas were classified into three main histological subtypes—oligodendroglial, astrocytic, and oligoastrocytic tumors [12,13]. This classification was based exclusively on histological criteria, by comparison of morphological characteristics between tumor cells and glial cells. Furthermore, a malignant grading system was applied, stratifying these entities on a scale of I, associated with lower anaplasia, low proliferation, and the most favorable prognosis, to IV, comprising often incurable, rapidly progressing, mitotically active neoplasms, with the highest degree of anaplasia [12]. Both histological classification and tumor grading were achieved using techniques such as light microscopy of hematoxylin and eosin-stained tumor sections and immunohistochemistry with different markers of lineage, proliferation, and differentiation, such as glial fibrillary acidic protein (GFAP), oligodendrocyte lineage transcription factor 2 (OLIG2), Ki67, and cytokeratin [12,14].

However, this approach to glioma stratification was flawed, and consequently, it was difficult to precisely evaluate tumors, diagnose patients, predict clinical outcomes, and find adequate therapies [15]. Molecular differences and the complexity among glioma subtypes became more and more evident as technological advancements were made, especially involving the genomic platforms for mRNA expression profiling and genome sequencing [16,17,18].

Hence, in 2016 the World Health Organization (WHO) incorporated isocitrate dehydrogenase (IDH) gene mutations and 1p/19q codeletion as essential biomarkers to adequately stratify gliomas, while also refining the previously established histological criteria [13]. The latest 2021 WHO classification of central nervous system tumors relies even further on these biomarkers to classify diffuse gliomas into three main types: IDH-wildtype GBM, IDH-mutant astrocytoma, and IDH-mutant and 1p/19q codeleted oligodendroglioma [19]. This update eliminates the previous diagnosis of IDH-mutant GBM, reclassifying it as astrocytoma, IDH-mutant, Grade 4. Consequently, all GBMs are now defined as IDH-wildtype. Other relevant molecular characteristics that aid in glioma diagnosis include epidermal growth factor receptor (EGFR) amplification, telomerase reverse transcriptase (TERT) promoter mutations or concurrent gain of chromosome 7 and loss of chromosome 10 in GBMs, and cyclin-dependent kinase inhibitor 2A/B (CDKN2A/B) homozygous deletion in IDH-mutant astrocytomas [19].

The introduction of these new biomarkers facilitated the prediction of patient outcome, as IDH mutations have been linked to favorable prognoses and sensitivity to chemotherapy [16,20,21,22]. Additionally, the 1p/19q codeletion has also been associated with better outcomes [23]. Overall, segregation among the established molecular glioma subgroups according to prognosis, diagnosis, and response to therapy became more accurate [16,22,24]. But even so, patient survival remains dismal, therapeutic strategies are still not efficient, and heterogeneity prevails in some molecular subgroups [6,16]. New molecular biomarkers are thus needed to improve glioma stratification and as therapeutic targets, to develop more efficient therapies.

Throughout the years, several molecular alterations have been proposed as biomarkers of prognosis or response to therapy, such as O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation, EGFR amplification, Phosphatase and Tensin Homologue (PTEN) deletion, and alpha thalassemia/mental retardation syndrome X-linked (ATRX) and TERT mutations [15,16,22,24]. Nonetheless, the clinical impact of these biomarkers has not been clear or coherent across multiple studies [16,25,26], and so research into new potential biomarkers in gliomas is still extremely relevant.

2. The Role of PI3K in Cell Signaling

Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) activating mutations have recently shown great promise as potential therapeutic targets in glioma [27,28,29]. PIK3CA is a 21-exon gene located on chromosome 3q26 that codes for the p110α subunit of Phosphatidylinositol-4,5-bisphosphate 3-kinase (PI3K) class I [30]. PI3Ks are a family of lipid kinases, acting as signal transducers in various pathways that are divided into three different classes according to function, structure, and substrate specificity—Class I, Class II, and Class III [31]. Class II and Class III PI3Ks have been less explored than Class I. Class II PI3Ks are monomeric kinases mainly involved in the regulation of membrane trafficking and angiogenesis through the production of phosphatidylinositol-3-phosphate (PI3P) and phosphatidylinositol-4,5-bisphosphate (PIP₂) [32,33]. Class III PI3K consists of a single catalytic subunit, Vps34, and a regulatory subunit, Vps15, and can generate PI3P from phosphatidylinositol (PI) [34]. It plays an important role in autophagy regulation [34], but it is thought to also promote cell growth by interacting with the mechanistic target of rapamycin (mTOR) [35].

PI3K Class I, however, comprises some of the most well-studied kinases. These enzymes convert PIP₂ to phosphatidylinositol-3,4,5-trisphosphate (PIP₃) [30], and are divided into two subclasses, IA and IB, depending on the type of receptor to which PI3K binds. Class IB PI3Ks are activated by G-protein coupled receptors (GPCR) [31,36]. These are heterodimeric kinases, with a p110γ catalytic subunit and a p101 regulatory subunit [37]. On the other hand, Class IA PI3Ks are mostly activated by receptor tyrosine kinases (RTK). Specifically, Class IA has been the most studied so far in cancer research, since it promotes cell survival and proliferation [38]. These kinases are also heterodimeric, containing two distinct subunits: a catalytic subunit (p110α, p110β and p110δ, encoded by PIK3CA, PIK3CB, and PIK3CD, respectively) and a regulatory subunit (p85α, p85β and p55γ, encoded by PIK3R1, PIK3R2, and PIK3R3, respectively) [31]. Regulation and location can vary between all these kinases. For instance, p110δ is mainly expressed in leukocytes, being involved in immunity [39].

Class IA PI3K is a key player in the PI3K/Protein Kinase B (Akt) pathway, a signaling pathway responsible for regulating a multitude of important biological processes, such as cell survival, proliferation, apoptosis, angiogenesis, and metabolism [31] (Figure 1). PI3K is activated when an RTK (e.g., EGFR) binds to a specific ligand, usually a growth factor [30]. Additionally, active GTPases from the Ras superfamily can directly bind to PI3K and promote its activation. Studies have shown that Ras acts synergistically with phosphorylated RTKs to enhance PI3K signaling [40]. PI3K phosphorylation then leads to the detachment of regulatory subunit p85 from catalytic subunit p110 and consequently to the release of catalytic inhibition, making it possible to modulate downstream signaling [41]. Thus, PI3K can phosphorylate PIP₂ to PIP₃, which in turn recruits serine/threonine kinase Akt to the cell membrane, one of the first steps in its activation [42]. To be fully active, Akt needs to be phosphorylated at serine and threonine residues. Phosphatidylinositol-Dependent Kinase 1 (PDK1) is a kinase that is also recruited by the second messenger PIP₃, responsible for Akt phosphorylation at threonine residue Thr308 [43,44]. The mechanistic target of rapamycin complex 2 (mTORC2) is responsible for Akt phosphorylation at serine residue Ser473 [45]. Akt is, therefore, one of the leading players in this pathway, controlling multiple downstream targets. Glycogen synthase kinase 3 α and β (GSK3α/β) is one of those targets, becoming inactivated upon phosphorylation by Akt, ultimately leading to an increase in glycogen synthesis and cell proliferation [31,46]. Akt also inhibits Forkhead box O (FOXO) transcription factors, culminating in the inhibition of apoptosis, cell cycle arrest, and catabolism [31,47]. Additionally, Akt releases mTORC1 inhibition by the tuberous sclerosis complex 1 (TSC1)-TSC2 system when it phosphorylates TSC2. In turn, mTOR can regulate several downstream targets, including the activation of ribosomal protein S6 kinase (S6K) and inactivation of the repressor of mRNA translation eukaryotic translation initiation factor 4E-binding protein (4E-BP1), inducing biosynthesis and inhibiting autophagy (Figure 1) [48,49].

PTEN is considered the antagonist of this pathway, as well as a tumor suppressor, since it is a serine/threonine phosphatase that catalyzes the dephosphorylation of PIP₃ to PIP₂ [50].

Understanding the real complexities of this pathway is quite challenging, especially because it intertwines with several other signaling pathways, such as mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) [51,52], 5′ adenosine monophosphate-activated protein kinase (AMPK) [53], WNT/β-catenin [54], and transforming growth factor β (TGFβ) [55]. Regarding the role of this pathway in cancer, there are added degrees of complexity that we must consider. Innumerable genetic alterations that lead to PI3K/Akt constitutive activation, promoting proliferative self-sufficiency, might occur throughout these signaling cascades.

3. PIK3CA Gene Alterations

The PI3K/Akt pathway is one of the most frequently altered pathways in cancer, usually due to genomic alterations in EGFR, PIK3CA, AKT, and PTEN [56,57,58,59]. In a 2019 study that analyzed 60,991 solid tumor samples using next-generation sequencing (NGS), 44% of the samples analyzed harbored at least one molecular alteration (base-pair substitutions, insertions/deletions, copy-number alterations, and rearrangements) in one of 18 PI3K-related genes [56]. These alterations, considered pathogenic or likely pathogenic, potentially lead to the constitutive activation of the pathway, whether through overactivation of proteins that promote PI3K/Akt signaling or loss of its antagonist, inducing profound changes in intricate signaling networks, particularly involving cell survival, metabolism, and proliferation [58].

PIK3CA activating mutations are among the most common PI3K/Akt pathway alterations in cancer [56,57,60]. These are oncogenic alterations that, both in vitro and in vivo, promote cell invasion, proliferation, and angiogenesis while inhibiting apoptosis independently from upstream signaling [61,62,63,64].

Catalytic subunit p110α, encoded by PIK3CA, has five distinct domains: the adaptor-binding domain (ABD), Ras-binding domain (RBD), C2 domain, helical domain, and kinase domain (Figure 2) [38]. PIK3CA mutations are found mainly at two of these domains: the helical domain (hotspot missense mutations E542K and E545K, located in exon 10) and the kinase domain (hotspot missense mutation H1047R, located in exon 21) [60]. Exons 10 and 21 of PIK3CA, according to Ensembl Transcript ID: ENST00000263967.4, RefSeq: NM_006218.4, are usually designated as coding exons 9 and 20, respectively, since the entire exon 1 is an untranslated region. Mutations in both of these hotspots can lead to oncogenic transformation by enhancing PI3K activity via one of two very distinct mechanisms [65]. H1047R blocks auto-inhibition by the C-terminal tail of the protein while also enhancing plasma membrane attachment. Furthermore, the amino acid substitution affects the ATP-binding site, making it more exposed and facilitating ATP binding [66]. Binding to regulatory subunit p85α is indispensable for the gain of function of kinase domain mutations, which are Ras-independent. On the other hand, E542K and E545K mutations prevent the binding of the p110α helical domain to the regulatory subunit p85α, being dependent on the interaction with Ras [65,67].

Alterations in the other PI3K domains, though less common, can still impact the protein’s function. ABD, for instance, interacts with the kinase domain within the protein’s three-dimensional structure, and mutations in this region may induce conformational changes in the kinase domain, thereby altering its enzymatic activity [68]. Additionally, mutations in the ABD or C2 domain can destabilize binding with the regulatory subunit, potentially leading to its disengagement and subsequent constitutive kinase activation [69]. The C2 domain, also important for PI3K’s association with the membrane, where its substrate PIP₂ resides, may also undergo mutations that increase membrane affinity and enhance enzymatic activity [70,71]. Notably, RBD alterations are rarely reported, suggesting they are uncommon and that their impact on Ras binding to PI3K and subsequent oncogenicity potential remains unknown.

PIK3CA mutations in cancer were first reported by Samuels and colleagues in 2004 in a small cohort of 35 colorectal cancer patients [60]. This initial discovery was expanded to include multiple cancer types, such as GBMs, gastric, breast, and lung cancers, leading to the recognition of PIK3CA as a key oncogene. Since then, extensive research has been directed toward understanding the frequency of PIK3CA mutations across different cancer types and their clinical implications on prognosis and response to therapy.

The PI3K/Akt pathway is particularly relevant in breast cancer, where PIK3CA alterations occur in roughly 21% to 47% of cases [56,59,72,73,74,75,76,77,78,79,80,81,82], making it the most frequently mutated oncogene in this cancer type. These mutations are predominantly found in hormone receptor (HR)-positive and human epidermal growth factor receptor 2 (HER2)-positive breast cancers [79,81,83,84]. However, their prognostic impact in breast cancer remains unclear. Some studies report PIK3CA mutations as independent negative prognostic factors [76,85] while others associate them with improved patient outcomes [75] and some find no significant correlation [59,78,79,81]. These findings suggest a context-dependent impact, likely influenced by breast cancer subtypes, and with specific mutations, such as the hotspot H1047R, linked to more aggressive disease and poorer prognoses compared to E345K [74,86,87,88]. Additionally, PIK3CA hotspot mutations may also contribute to therapy resistance in breast cancer, with associations found between these mutations and resistance to chemotherapy (e.g., paclitaxel, anthracyclines) [64,88,89] and HER2-targeted therapies like trastuzumab [82,90,91,92].

These mutations have also shown to be quite relevant in colorectal cancer, present in about 10% to 32% of tumors [56,60,93,94,95,96,97,98,99,100,101], and are often associated with worse prognoses [100,101,102]. However, there are some contradicting studies that do indicate no significant impact on clinical outcomes or tumor progression [94,95,96,97,98,103]. PIK3CA mutations are also implicated in acquired resistance to anti-EGFR therapy in metastatic colorectal cancer patients [101,104], similar to the resistance to trastuzumab seen in breast cancer. This aligns with the role of EGFR and HER2 in PI3K/Akt signaling, where downstream PIK3CA mutations can override inhibition of the RTKs targeted by these therapies. Additionally, these mutations seem to also be associated with chemoresistance in colorectal cancer [99].

Alterations in PIK3CA are also present in gastric (2–25%) [56,60,105,106,107,108,109,110,111], endometrial (16–51%) [112,113,114,115], head and neck (6–21%) [56,116,117,118,119,120], and esophageal cancers (5–21%) [121,122,123,124,125,126]. In gastric and endometrial cancers, these mutations are generally not significantly associated with patient prognosis [105,106,107,108,109,110,113,114,127]. However, specific exon 10 mutations have been linked to poorer outcomes in endometrial cancer [115], and a meta-analysis reported that PIK3CA mutations negatively impact survival in these patients [128]. Notably, while these mutations are not considered prognostic markers in gastric cancer, they have been associated with increased tumor aggressiveness [109]. In contrast, in esophageal cancer, PIK3CA mutations are regarded as independent favorable prognostic factors [121,122,123].

Given the complexity of the PI3K pathway within the broader network of signaling cascades, it is expected that PIK3CA mutations may exert paradoxical effects in different oncogenic contexts, even within the same type of cancer. Not only that, but the observed variability in PIK3CA mutational frequencies and prognostic associations across many studies is likely influenced by differences in methodology (e.g., Sanger sequencing versus NGS), the specific exons analyzed (e.g., only hotspot exons 10 and 21 versus all coding exons), and the diversity of patient cohorts.

4. PIK3CA Mutations as Biomarkers in Glioma

4.1. Mutation Frequency

The frequency and impact of oncogenic PIK3CA mutations has also been explored in glioma, yet questions remain about their true relevance in this context. Studies report a wide range of mutation frequencies in GBM, from 0% to 30% [18,27,28,60,129,130,131,132,133,134,135,136] (

Table 1. Previous research on PIK3CA mutation frequency and clinical impact in glioma subgroups.

Classification	Glioma Subgroup	Sample Size	Method	Region Evaluated	Mutation Frequency	Prognostic Effect (U/M)	Reference
Histological	GBM	30 (Multifocal)	Illumina HiSeq	All coding exons	30%	Unfavorable (U)	[27]
		130 (Solitary)			10%
		15	Sanger sequencing	All coding exons	27%	---	[60]
		60	Illumina HiSeq	All coding exons	21.7%	---	[136]
		38	Sanger sequencing	Exons 2, 3, 5, 6, 8, 10, 13, 14, 19 and 21 *	18%	---	[130]
		291	Illumina HiSeq	All coding exons	11%	---	[18]
		157	SNaPshot® multiplex system	6 known hotspots	8.3%	Unfavorable (M)	[28]
		70	PCR-SSCP direct sequencing	All coding exons	7%	---	[132]
		116	Affymetrix microarray	All coding exons	6%	---	[131]
		105	Sanger sequencing	Exons 10 and 21 *	5%	---	[129]
		97	PCR-SSCP direct sequencing	Exons 2, 3, 5, 6, 8, 10, 13, 14, 19 and 21 *	5%	---	[133]
		40	Ion semiconductor sequencing	All coding exons	5%	---	[134]
		30	PCR-SSCP	All coding exons	0%	---	[135]
	Astrocytoma	52	Ion semiconductor sequencing	All coding exons	9.6%	Unfavorable (U)	[134]
		31	Sanger sequencing	Exons 10 and 21 *	3%	---	[129]
	Oligodendroglioma	21	Sanger sequencing	Exons 10 and 21 *	14%	---	[129]
		66	PCR-SSCP direct sequencing	Exons 2, 10 and 21 *	5%	---	[137]
		17	Ion semiconductor sequencing	All coding exons	0%	---	[134]
Molecular	GBM, IDH-wildtype	567 (TCGA)	Illumina HiSeq	Exons 10 and 21*	2%	Not significant (M)	[29]
		239	Sanger sequencing		3%
	GBM, IDH-mutant	25 (TCGA)	Illumina HiSeq	Exons 10 and 21 *	8%	---	[29]
		11	Sanger sequencing		9%
	Astrocytoma, IDH-wildtype	39	Sanger sequencing	Exons 10 and 21 *	10%	---	[29]
	Astrocytoma, IDH-mutant	56	Sanger sequencing	Exons 10 and 21*	5%	---	[29]
	1p/19q codeleted + IDH-mutant	84	Illumina HiSeq	All coding exons	20%	---	[138]
		107	Illumina HiSeq	All coding exons	14%	Unfavorable (M)	[139]
		49	Sanger sequencing	Exons 10 and 21 *	10%	---	[29]

* PIK3CA exon designation is done according to Ensembl Transcript ID: ENST00000263967.4, RefSeq: NM_006218.4. Exons 2, 3, 5, 6, 8, 10, 13, 14, 19 and 21 are often designated in the literature as coding exons 1, 2, 4, 5, 7, 9, 12, 13, 18, and 20, respectively. U/M, univariable or multivariable analysis; GBM, glioblastoma multiforme; PCR-SSCP, Polymerase Chain Reaction—Single Stranded Conformation Polymorphism; TCGA, The Cancer Genome Atlas.

). In their pioneer study, Samuels et al. reported a 27% PIK3CA mutation frequency in a small GBM cohort (n = 15) using Sanger sequencing. However, they may have overestimated mutation prevalence due to the limited sample size [60]. Gallia et al. also analyzed a small cohort of 38 primary GBM cases, reporting a 17% mutation rate, though they excluded some PIK3CA exons, potentially underrepresenting true mutation rates [130]. In contrast, larger cohorts often yield lower frequencies. Broderick and colleagues observed only a 5% mutation frequency in a GBM cohort of 105 samples, albeit analyzing only exons 10 and 21 [129]. However, similar frequencies (5–7%) were obtained by other authors, using different detection methods and sample sizes (from 70 to 116 samples). These studies managed to analyze more PIK3CA exons, with some even being able to sequence all coding exons [131,132,133,134]. Of note, even studies covering all coding exons have yielded inconsistent findings, with one study identifying no PIK3CA mutations in 30 GBM samples [135].

Recent advances, such as NGS, have enabled a more comprehensive coverage, revealing a PIK3CA mutation frequency of 10–11% in larger cohorts (n = 130 [27], and 291 [18], respectively), though rates as high as 30% have been reported in certain multifocal GBM cases [27]. Tanaka et al., using a multiplex detection system focused on six known PIK3CA hotspots (R88, E542, E545, Q546, H1047, G1049), found an 8.3% PIK3CA mutation rate in 157 GBM samples [28]. On the other hand, Saadeh et al. reported a higher frequency of 21.7% using whole exome sequencing on a smaller cohort of 60 GBM [136].

Research regarding PIK3CA mutation frequencies in lower-grade glioma has been similarly inconclusive, with mutation rates reported from 0% to 14% depending on the histological subtype and methodology used [129,134,137] (Table 1). Broderick et al. analyzed the hotspot PIK3CA exons 10 and 21 and identified a 3% mutation rate in 31 astrocytomas and a higher 14% frequency in 21 oligodendrogliomas [129]. Hartmann et al. reported a lower PIK3CA mutation rate of 5% in oligodendrogliomas in a slightly bigger cohort (n = 66) by analyzing exons 2, 10, and 21 [137]. On the other hand, more recently Wang et al. analyzed all coding exons and reported a 9.6% mutation rate in 52 astrocytomas but found no PIK3CA mutations in a small subset of 17 oligodendrogliomas [134].

These notable inconsistencies across studies stem from differences in sample sizes, mutation detection methods, exon coverage, and outdated histological classification systems. Furthermore, the presence of mutations in a pseudogene with over 95% sequence homology to PIK3CA exon 10 may also confound these results [140,141], which none of the studies above mentioned.

Although several studies have examined the frequency of PIK3CA mutations in glioma, most were mainly based on outdated histological classifications, with few considering the updated molecularly defined subgroups. To date, only one study has investigated PIK3CA mutation rates across all glioma molecular subtypes defined by the 2016 WHO classification, using a large cohort of 394 adult diffuse gliomas, albeit only analyzing exons 10 and 21 [29]. Mutation frequencies ranged from 3% to 10%, with higher rates in 1p/19q codeleted and IDH-mutant oligodendrogliomas and IDH-wildtype astrocytomas (10%), and lower frequencies in GBM IDH-wildtype (3%), the most aggressive glioma subtype. Notably, similar frequencies were found in The Cancer Genome Atlas (TCGA) dataset—3% versus 2% in GBM IDH-wildtype, and 9% versus 8% in the now obsolete subgroup of GBM IDH-mutant. Compared with the findings reported by TCGA in 2015 [138] and Dono and colleagues [139], the observed PIK3CA mutation frequency in 1p/19q codeleted and IDH-mutant gliomas was lower in this cohort (10% versus 20% and 14%, respectively) [29], likely due to the more limited sequencing strategy. Thus far, no studies have addressed PIK3CA mutation frequency across the glioma subgroups defined by the 2021 WHO classification. As such, much remains to be explored regarding the potential of PIK3CA mutations as biomarkers in glioma.

4.2. Clinical and Biological Impact

While most studies have focused primarily on PIK3CA mutational frequencies, there has been limited research into their impact on prognosis and therapy response in glioma, which remains poorly understood. Moreover, research in this field appears significantly outdated—studies exploring the role of PIK3CA mutations considering the most recent glioma molecular stratification are severely lacking.

Regarding the impact of PIK3CA mutations on GBM patient outcomes, an association with poor prognoses has been reported [27,28]. Lee J. and colleagues demonstrated that patients with GBM harboring PIK3CA mutations had worse prognoses than wildtype patients, although no multivariable analysis was conducted [27]. Another study linked PIK3CA mutations to reduced progression-free survival of patients with GBM IDH1-wildtype, independent of other variables [28]. Additionally, by studying the imaging characteristics of PIK3CA mutant GBM, this same study found a significant association between these mutations and widespread disease at diagnosis, compatible with gliomatosis, multicentric lesions, or distant leptomeningeal lesions. However, no association with overall survival was found, consistent with the findings of Brito et al. [29], who showed that PIK3CA alterations were not independent prognostic factors in GBM IDH-wildtype, despite a trend toward shorter survival.

Beyond their potential prognostic relevance in GBM, PIK3CA mutations have also been described as early events in GBM development, found in all sectors of the tumor [27]. Moreover, analyses of matched primary and recurrent tumor samples have shown that these mutations constitute early events that are maintained through glioma progression, regardless of the therapy administered [29]. These findings suggest that PIK3CA mutations may represent a stable and therapeutically actionable target in these highly heterogenous tumors, while also hinting at a possible role in therapy resistance.

In lower-grade gliomas, the prognostic value of PIK3CA mutations has been explored to a limited extent. Only two studies to date have assessed the impact of these mutations on patient survival in this context. Wang and colleagues [134] reported that patients with PIK3CA-mutated lower-grade gliomas had significantly lower overall and progression-free survival than patients with wildtype tumors. However, the impact on patient prognosis was not independent from other variables, and the number of PIK3CA-mutated samples was low (5/69). Regarding IDH-mutant and 1p/19q-codeleted oligodendrogliomas, Dono et al. showed that PIK3CA mutations are associated with worse overall survival in a multivariable analysis [139].

While clinical data on the prognostic relevance of PIK3CA mutations in glioma continues to grow, research into their underlying pathogenic mechanisms remains scarce. McNeill and colleagues demonstrated that PIK3CA mutations promote astrocyte growth by potentiating PI3K signaling [142]. Interestingly, only the H1047R mutation induced anchorage-independent colony formation. Nevertheless, both helical and kinase domain mutations potentiated RAS-mutant astrocyte tumorigenesis in vivo. Similarly, Tateishi et al. showed that PIK3CA mutations seem to drive oligodendroglioma progression in an orthotopic mouse model [143]. Furthermore, another in vivo study identified H1047R and C420R as key players in neuronal microenvironment modulation [144], suggesting they are main drivers of gliomagenesis, acting through different mechanisms. However, most clinical studies overlook the differential impact of specific mutations like H1047R or E545K on patient outcomes, representing a gap in the current literature.

Overall, studies point to a potential important role of PIK3CA mutations in glioma, but the current body of evidence remains inconsistent. Robust multicentric studies, using the most recent 2021 WHO classification, are needed to clarify the role of these mutations in gliomagenesis, aggressiveness, prognosis, and therapy response. In this case, the acquisition of enough samples of each subgroup for a robust analysis might be a challenge, especially as all evidence points to PIK3CA mutations not being too abundant in glioma. Given the heterogeneous genetic landscape of gliomas, these mutations may ultimately be one of many within the signaling cascade, potentially lacking unique prognostic value without stratified, subtype-specific analysis. Additionally, examining how PIK3CA mutations may modulate treatment response, particularly to temozolomide, is critical, given the current evidence of these mutations conferring resistance to multiple therapeutic strategies in other cancers [64,82,88,89,90,91,92,99,101,104].

With glioma remaining an aggressive and virtually incurable type of cancer, understanding its biology is vital to identify effective therapies. Thus, the use of PIK3CA mutations as actionable targets in glioma should be explored further.

5. Could PI3Kα Inhibitors Offer a New Approach to Target Glioma?

Knowing that alterations in the PI3K/Akt pathway are so common and important in cancer, multiple antagonists have been developed throughout the years in the hopes of providing better and more efficient therapeutic strategies to cancer patients, especially as so many of them harbor upregulation of this pathway and often develop resistance to standard treatment [38].

Currently, a plethora of inhibitors targeting this pathway is being researched and undergoing clinical trials [145]. So far, nine inhibitors targeting the PI3K/Akt/mTOR pathway have already been clinically approved by the U.S. Food and Drug Administration (FDA) for use in cancer treatment in specific contexts (Figure 3): temsirolimus and everolimus (mTOR inhibitors) [146,147], capivasertib (AKT inhibitor) [148], idelalisib and umbralisib (both PI3K inhibitors, selective for catalytic subunit p110δ) [149,150], copanlisib (pan-PI3K inhibitor targeting all four class I PI3Ks, with a preference for PI3Kα and PI3Kδ) [151], duvelisib (dual-PI3Kδ/PI3Kγ inhibitor) [152], and alpelisib and inavosilib (selective for catalytic subunit p110α) [153,154].

However, safety concerns have emerged regarding four out of the six PI3K inhibitors, which were approved for use in hematological malignancies given the enrichment of p110δ and p110γ in leukocytes. Umbralisib and copanlisib were both withdrawn after recent clinical trials failed to confirm clinical benefit and raised safety issues. Similarly, idelalisib had its approval for use in small lymphocytic lymphoma (SLL) and follicular lymphoma (FL) withdrawn and now faces restricted use in chronic lymphocytic leukemia (CLL) due to toxicity concerns. Duvelisib is approved for use in CLL and SLL but had its FL approval withdrawn. Many clinical trials based on these four inhibitors have been terminated. Consequently, new-generation PI3K inhibitors are under investigation for hematological malignancies with an emphasis on enhancing tolerability [155,156,157]. Still, in other oncological contexts, PI3K inhibition has shown great promise (summary in

Table 2. Current landscape of the main PI3K inhibitors in solid malignancies, including evidence from preclinical and clinical studies.

Inhibitor	Cancer Type	Cancer Subtype	Study Type	Treatment	Outcome	PIK3CA Status Dependent?	Reference
Alpelisib PI3Kα (orthosteric)	Breast	HR+ HER2-	Clinical-PhIII	+Fulvestrant	■ Favorable	♦ Yes	[158,159]
		HR+ HER2- PI3K altered	Clinical-PhII	Monotherapy	■ Favorable	N.E. (All-mut)	[160]
		Triple negative	Clinical-PhII	Monotherapy	■ No benefit	N.E.
		---	Preclinical-in vitro	Monotherapy	■ Sensitive	♦ Yes	[161]
	Brain	GBM	Preclinical-in vivo	±PI3Kβ inhibitor (AZD6482)	■ Sensitive	N.E.	[162]
			Preclinical-in vitro	±mTOR inhibitor (OSI-027)	■ Sensitive	N.E. (WT-only)	[163]
	Gynecological	PIK3CA-mut	Clinical-obs	Monotherapy	■ Favorable	N.E. (All-mut)	[164]
		PIK3CA-mut cervical	Clinical	Monotherapy	■ Favorable	N.E. (All-mut)	[165]
		Cervical	Preclinical-in vitro	Monotherapy	■ Sensitive	♦ Yes	[166]
	Liver	HCC	Preclinical in vivo + in vitro	±mTOR inhibitor (MLN0128)	■ Sensitive	♦ Yes	[167]
	Head and neck	PI3K altered SCC	Clinical-PhII	Monotherapy	■ Favorable	♦ Yes	[168]
		SCC	Clinical-PhIb/II	+Cetuximab	■ No benefit	N.E.	[169]
			Preclinical-in vivo	Monotherapy	■ Sensitive	♦ Yes	[170]
			Preclinical-in vivo	+Cisplatin	■ Favorable	N.E. (All-mut)	[171]
	Lung	SCC	Preclinical-in vivo	±CDK4/6 inhibitors	■ Favorable	♦ Yes	[172]
Buparlisib Pan-PI3K (mainly PI3Kα)	Breast	HR+ HER2-	Clinical-PhIII	+Fulvestrant	■ Favorable	♦ Yes	[173]
			Clinical-PhII	+Tamoxifen	■ Favorable	♦ Yes	[174]
		Triple negative	Clinical-PhII	Monotherapy	■ Minimal benefit	N.E.	[175]
	Esophagus	SCC	Clinical-PhII	Monotherapy	■ Favorable	N.E.	[176]
	Brain	GBM	Clinical-PhII	Monotherapy	■ Minimal benefit	♦ No	[177]
			Clinical-PhIb/II	+Carboplatin or lomustine	■ Minimal benefit	N.E.	[178]
			Preclinical in vivo + in vitro	±PARP inhibitor (rucaparib)	■ Favorable	N.E. (WT-only)	[179]
			Preclinical-in vivo	Monotherapy	■ Minimal benefit	N.E.	[162]
			Preclinical-in vitro	±MEK inhibitor (selumetinib)	■ Sensitive	♦ No	[142]
	Head and neck	SCC	Clinical-PhII	±Cetuximab	■ Favorable	N.E. (WT-only)	[180]
			Clinical-PhII	+Paclitaxel	■ Favorable	♦ No	[181,182]
	Lung	SCC	Preclinical-in vivo	±CDK4/6 inhibitors	■ Favorable	♦ Yes	[172]
Inavolisib PI3Kα (orthosteric)	Breast	HR+ HER2- PIK3CA-mut	Clinical-PhIII	+Palbociclib-Fulvestrant	■ Favorable	N.E. (All-mut)	[183]
			Clinical-PhIb/II	+Letrozole or Fulvestrant	■ Favorable	N.E. (All-mut)	[184]
		---	Preclinical in vivo + in vitro	±Palbociclib and/or fulvestrant	■ Favorable	♦ Yes	[185]
			Preclinical-in vitro	Monotherapy	■ Sensitive	♦ Yes	[161]
Paxalisib Pan-PI3K/ mTOR	Brain	GBM	Clinical-PhII	Monotherapy	■ Favorable	N.E.	[186]
			Preclinical in vivo + in vitro	±EGFR inhibitor (AZD-9291)	■ Favorable	N.E. (WT-only)	[187]
			Preclinical-in vivo	Monotherapy	■ Sensitive	N.E.	[188,189]
Pictilisib PI3Kα/δ	Colon	---	Preclinical-in vitro	Monotherapy	■ Sensitive	♦ No	[185]
	Brain	GBM	Preclinical-in vitro	±Temozolomide	■ Sensitive	N.E.	[190]
	Head and neck	SCC	Preclinical-in vitro	Monotherapy	■ Sensitive	♦ No	[191,192]
RLY-2608 PI3Kα (allosteric)	Breast	HR+ HER2- PIK3CA-mut	Clinical-case	+Fulvestrant	■ Favorable	N.E. (All-mut)	[193]
		---	Preclinical in vivo + in vitro	±Fulvestrant	■ Favorable	♦ Yes
Taselisib PI3Kα/δ/γ	Breast	HR+ HER2-	Clinical-PhIII	+Fulvestrant	■ Minimal benefit	♦ No	[194]
			Clinical-PhII	+Fulvestrant	■ Favorable	♦ Yes	[195]
			Clinical-PhIb	+HER2 inhibitors	■ Favorable	♦ No	[196]
		---	Preclinical-in vivo	Monotherapy	■ Sensitive	N.E. (All-mut)	[185]
	Colon	---	Preclinical-in vitro	Monotherapy	■ Sensitive	♦ Yes
	Head and neck	SCC	Preclinical in vivo + in vitro	Monotherapy	■ Favorable	♦ Yes	[191]
	Lung	SCC	Clinical-PhII	Monotherapy	■ No benefit	N.E. (All-mut)	[197]
Voxtalisib Pan-PI3K/ mTOR	Breast	HR+ HER2-	Clinical-PhI/II	+Letrozole	■ Minimal benefit	♦ No	[198]
	Brain	GBM	Clinical-PhI	+Temozolomide	■ Favorable	N.E.	[199]
			Preclinical in vivo + in vitro	±Temozolomide	■ Favorable	N.E. (WT-only)	[200]
		Low-grade glioma	Preclinical in vivo + in vitro	Monotherapy	■ Favorable	N.E. (WT-only)	[201]
	Gynecological	Ovarian	Clinical-PhII	+MEK inhibitor (pimasertib)	■ Minimal benefit	N.E.	[202]
		Endometrial	Preclinical-in vitro	±MEK inhibitor (pimasertib)	■ Sensitive	♦ No	[203]
	Prostate	---	Preclinical-in vitro	Monotherapy	■ Sensitive	N.E.	[204]

Legend: ■, Evidence supports added therapeutic benefit; ■, No evidence of added therapeutic benefit; ♦ Yes, response is dependent on PIK3CA mutational status; ♦ No, response is not dependent on PIK3CA mutational status; N.E., not evaluated (All-mut, all patients/models are PIK3CA mutated; WT-only, all patients/models are PIK3CA wildtype); GBM, glioblastoma multiforme; HCC, hepatocellular carcinoma; HER2, human epidermal growth factor receptor 2; HR, hormone receptor; obs, observational; SCC, squamous cell carcinoma.

).

Notably, several studies have reported PIK3CA mutations as predictive biomarkers of response to PI3K/Akt pathway inhibitors [120,158,172,173,174]. However, concerns have emerged regarding the use of PIK3CA status alone to guide therapeutic decisions. Indeed, it seems that increased sensitivity to PI3K/Akt pathway inhibitors is not universal to all types of cancer or to all classes of inhibitors, and the presence of different PIK3CA mutations might lead to distinct inhibitor sensitivities [161,168,170,177,182,185,192,193,205,206,207]. Moreover, there is still a scarcity of clinical trials explicitly designed to evaluate how PIK3CA mutations influence response to PI3K/Akt pathway inhibitor in different types of cancer.

In general, pan-PI3K inhibitors have shown limited efficacy in cancer therapy due to increased toxicity when compared with more targeted approaches [58,198,202]. Furthermore, efficiency is usually very limited since patients often acquire resistance to these inhibitors, either due to acquired mutations in regulatory genes or the activation of compensatory mechanisms and pathways. As such, the field has shifted toward isoform-selective PI3K inhibitors, which tend to exhibit better efficacy with fewer off-target effects [208].

Alpelisib (BYL719) is an orthosteric p110α selective isoform inhibitor from the 2-aminothiazole family, which blocks catalytic activity by binding to the ATP-binding site [209]. Researchers have shown particular interest in studying the response to this inhibitor in the presence of PIK3CA mutations, seeing as it is highly specific toward PI3Kα and so it is expected to reduce off-target toxicity [208]. Indeed, alpelisib has a more pronounced antitumor effect in the presence of PIK3CA mutations, which is amplified in the presence of double mutations [161]. In 2019, it was approved for clinical use in patients with advanced HR-positive and PIK3CA-mutant breast cancer, in combination with fulvestrant, a selective estrogen receptor degrader [153,158,159]. More recently, in 2024, inavolisib (GDC-0077)—another selective orthosteric PI3Kα inhibitor that additionally promotes mutant PI3Kα degradation [185]—emerged as an additional treatment option for HR-positive, HER2-negative, PIK3CA-mutant breast cancer, in combination with palbociclib, a CDK4/6 inhibitor, and fulvestrant [154,183].

Nevertheless, as with other targeted therapies, p110α selective isoform inhibitors do not always reach their full potential. One major limitation is the activation of compensatory signaling pathways, which can undermine treatment efficacy and lead to drug resistance [210,211]. Resistance mechanisms include upregulation of insulin, PIM kinases, IGF1R, PI3Kβ, and mTOR signaling, as well as PTEN loss and secondary PIK3CA mutations [206,212,213,214,215,216]. In HER2-positive, PIK3CA-mutant breast cancer, for instance, HER3 upregulation via HER2-induced phosphorylation is a key compensating mechanism [210,217]. Accordingly, combination strategies that co-target these alternative pathways or with already existing standard-of-care treatments can enhance the efficacy of PI3Kα inhibitors [145,171,199,218,219].

Emerging evidence suggests that secondary PIK3CA mutations may confer resistance to orthosteric inhibitors, such as alpelisib, which might be overcome by AKT inhibition or novel allosteric selective PI3Kα inhibitors [206]. Notably, a new mutant-selective allosteric PI3Kα inhibitor, RLY-2608, has shown great promise in circumventing some of the downfalls associated with orthosteric inhibitors, including toxicities related with glucose homeostasis and the skin [193].

In gliomas, however, the role of PIK3CA mutations in modulating sensitivity to PI3K pathway inhibition remains poorly understood. Buparlisib, a brain-penetrant pan-PI3K inhibitor with preferential activity against PI3Kα, is the most extensively studied PI3K inhibitor in GBM (Table 2). Nevertheless, it has shown limited efficacy in clinical trials, whether used as monotherapy [177] or in combination with other agents [178]. Of note, PIK3CA mutational status was not assessed in the latter study, and preclinical data suggests that specific PIK3CA mutations may modulate response to buparlisib when in combination with MEK inhibitors [142]. Therefore, it remains unknown whether combination therapy would be more efficient in the presence of PIK3CA mutations. Other brain-penetrant inhibitors such as paxalisib (GDC-0084) and voxtalisib (XL765), both dual PI3K/mTOR inhibitors, have shown more favorable outcomes in preclinical and clinical studies [186,187,188,189,199,200]. However, the predictive role of PIK3CA mutations was not assessed.

Regarding alpelisib, the p110α selective inhibitor that has shown such promise in PIK3CA-mutant breast cancer, less research has been conducted in the context of gliomas. Preclinical studies show that alpelisib combined with other PI3K/Akt pathway inhibitors more effectively reduces Akt phosphorylation in PTEN-null GBM models than alpelisib alone, suggesting that PTEN-null GBM cells can rely on both PI3Kα and PI3Kβ isoforms to maintain cell survival [162,163]. These findings are extremely relevant in the context of glioma, since PTEN loss is quite frequent in GBM [220,221]. Similar compensatory mechanisms have been described in other types of cancer, such as breast, colon, head and neck, and prostate, where PTEN-deficient tumors sustain proliferation through PI3Kβ and are thus most likely not to respond to PI3Kα inhibition [191,214,216,222].

Still, none of these studies have analyzed the effect of PI3Kα selective isoform inhibitors specifically in PIK3CA-mutated glioma. While alpelisib shows a clear clinical benefit in PIK3CA-mutant breast cancer, the complexity of signaling networks in glioma may require broader or combination-based strategies. Paradoxically, pan-PI3K inhibitors, though less favored in other types of cancer, may offer better coverage in glioma if used alongside the targeting of other signaling pathways, as shown in Table 2. Furthermore, the relationship between PTEN deficiency and PIK3CA mutations and how it affects alpelisib treatment response in glioma must be further explored. Curiously, in an ovarian cancer cell line harboring both PTEN loss-of-function and PIK3CA mutations, only the inhibition of the p110α isoform, and not p110β, resulted in cell growth suppression [222]. This finding seems to indicate that, in the presence of both alterations, PIK3CA mutations might triumph over PTEN loss-of-function, allowing the efficient targeting of the p110α isoform. Hence, understanding the connection between these alterations might be the key to the efficient targeting of the PI3K/Akt pathway in glioma.

6. Future Perspectives

Gliomas remain highly challenging and aggressive tumors, often recurring and causing significant patient morbidity and mortality. As such, identifying reliable biomarkers is imperative to better stratify these patients and develop new efficient therapies. Given the prominent role of the PI3K/Akt pathway in other cancers, where PIK3CA mutations sometimes serve as prognostic or predictive biomarkers, it is crucial to further investigate the real impact of these mutations on glioma pathogenesis, prognosis, and therapy response. Although recent studies have shed some light on the role of PI3KCA in glioma, our current understanding remains limited. Considering the updated 2021 WHO classification of central nervous system tumors, future research must aim to clarify the role of PIK3CA mutations across glioma molecular subtypes. Large, rigorous, multicentric studies involving extensive patient cohorts will be essential to determine the true impact of PIK3CA mutations within each subgroup. Advanced sequencing methods, including the generalized use of NGS, can facilitate a more comprehensive analysis of PIK3CA mutations, although assessment of the pseudogene with exon 10 homology must be prioritized for accurate results. These efforts could ultimately reveal whether PIK3CA mutations are genuine drivers of glioma pathogenesis or simply represent another layer of the extensive glioma molecular complexity.

The identification of PIK3CA mutations as early and constitutive events in glioma progression strongly hints that these alterations might be valuable therapeutic targets, even if they are relatively infrequent. Selective PI3Kα inhibition in PIK3CA-mutant low-grade gliomas could potentially suppress disease progression at an earlier stage. In GBM, the most prevalent and lethal glioma subtype, targeting PI3Kα could also serve as a promising therapeutic approach, particularly in multifocal cases where PIK3CA mutations are consistently present across all lesions. However, PI3Kα inhibitors, when used as monotherapies, often exhibit limited efficacy due to the presence of redundant resistance mechanisms, which should not be overlooked. Unlike in breast cancer, selective targeting of p110α in glioma may encounter challenges due to reliance on other PI3K isoforms, PTEN loss-of-function, and interactions with additional survival pathways. Therefore, investigating the interplay between PIK3CA mutations, p110β expression, PTEN loss, and other glioma-associated alterations is essential to refine PI3K-targeted approaches in glioma. Tailoring PI3K inhibition based on the tumor’s molecular profile could yield more precise and individualized treatments in glioma, rather than the more generalized approach that has been used in recent clinical trials. Furthermore, assessing whether PI3Kα inhibition could enhance glioma sensitivity to standard therapies like temozolomide and radiotherapy may also open new avenues for more effective combination treatment strategies.