A Phase I study of alpelisib in combination with trastuzumab and LJM716 in patients with PIK3CA-mutated HER2-positive metastatic breast cancer

Abstract

Background:

Activating mutations in PIK3CA promote resistance to HER2-targeted therapy in breast cancer, however inhibition of PI3K alone leads to escape via feedback upregulation of HER3. Combined inhibition of HER2, HER3 and PI3K overcomes this mechanism preclinically.

Methods:

This phase I study investigated the maximum tolerated dose (MTD) of alpelisib given in combination with trastuzumab and LJM716 (a HER3-targeted antibody) in patients with PIK3CA-mutant HER2-positive metastatic breast cancer (MBC) using the continual reassessment method. Secondary analyses included efficacy and exploratory correlative studies.

Results:

Ten patients were treated initially with daily alpelisib (Arm A). Grade ≥3 adverse events seen in ≥2 patients included diarrhea (n=6), hypokalemia (n=3), abnormal liver enzymes (n=3), hyperglycemia (n=2), mucositis (n=2), and elevated lipase (n=2). The MTD of alpelisib in Arm A was 250mg daily. This prompted the opening of Arm B in which 11 patients received intermittently dosed alpelisib. Grade ≥3 adverse events seen in ≥2 patients included diarrhea (n=5), hypokalemia (n=3), and hypomagnesemia (n=2). The MTD of alpelisib in Arm B was 350mg given 4 days on, 3 days off. Among 17 patients assessed, 1 had a partial response, 14 had stable disease, and 2 had disease progression at best response. Five patients had stable disease for >30 weeks. mRNA profiling of pre-and on-treatment tissue demonstrated PIK3CA target engagement by alpelisib via induction of downstream signaling and feedback pathways.

Conclusions:

Combination treatment with alpelisib, trastuzumab and LJM716 was limited by gastrointestinal toxicity. Further efforts are warranted to target the PI3K pathway in HER2+ MBC.

Introduction:

Human Epidermal Growth Factor Receptor 2 (HER2) is overexpressed in 15–20% of breast cancers and is associated with an aggressive phenotype and poor prognosis if untreated [1]. While the HER2-targeted antibody trastuzumab has contributed to clinically meaningful survival gains in HER2-positive breast cancer, resistance ultimately develops in most patients with advanced disease [2]. Studies suggest that resistance to HER2 targeted therapy can emerge from mutational activation of the PI3K pathway, and that pre-existing mutations in PIK3CA portend poor response to therapy in HER2-positive breast cancer [3–5]. For example, in a correlative analysis of the CLEOPATRA study, 32% of patients had PIK3CA-mutated HER2-positive metastatic breast cancer (MBC), and those patients exhibited significantly worse progression-free survival on HER2-targeted treatment than patients with wild-type PIK3CA [6, 7]. These and other data provide an initial rationale for targeting PI3K in the clinical setting of resistance to HER2-targeted therapy.

While targeting PI3K alone in HER2+ breast cancer cells results in inhibition of AKT activity, it also appears to stimulate increased HER3 expression and activation of the parallel MAPK pathway via relief of feedback inhibition, which leads to resistance to PI3K inhibitors when used as monotherapy [8, 9]. HER3 is a closely related member of the ErbB receptor family, and co-expression of HER2 and HER3 is notably correlated with worse overall survival in metastatic breast cancer [10–12]. Even in the absence of PIK3CA mutations, HER3 plays an important role in HER2-driven tumors via the formation of HER2:HER3 heterodimers, which trigger potent activation of downstream pathways involved in cell proliferation and survival [13–15]. Interfering with HER3 expression or its heterodimerization with HER2 leads to decreased proliferation and suppression of downstream signaling cascades, which may overcome this resistance mechanism to PI3K inhibition [14, 16]. Combination therapy inhibiting HER2, HER3, and PI3K is highly effective in xenograft models of breast cancer [17].

Based on these data, we hypothesized that concurrent inhibition of PI3K, HER2 and HER3 would be a safe and effective combination for the treatment of HER2-positive breast cancer with PIK3CA mutations. Alpelisib is an oral class I PI3K inhibitor which strongly inhibits the PI3Kα isoform (and much less the β, δ and γ isoforms) of PI3K, now approved in combination with fulvestrant for the treatment of estrogen receptor-positive, HER2-negative PIK3CA-mutated metastatic breast cancer (MBC) at a dose of 300mg daily [18, 19]. LJM716 is a fully human monoclonal antibody against HER3, which traps HER3 in an inactive conformation, and can inhibit both ligand-induced and HER2-mediated activation of HER3 and its downstream signaling pathways [20, 21]. To explore the role of PI3K inhibition in HER2+ MBC, we conducted a phase I clinical trial of alpelisib in combination with trastuzumab and LJM716 in patients with treatment refractory HER2+ MBC harboring PI3KCA mutations. The primary objective of this study was to define the safety of this regimen and the maximum tolerated dose (MTD) of alpelisib as part of this triplet therapy. Secondary endpoints included clinical activity and exploratory correlative studies from pre- and on-treatment biopsies to evaluate target engagement and feedback pathways.

Materials and Methods

Patient Selection

Eligible patients were women 18 years of age or older who had a histologically or cytologically confirmed diagnosis of metastatic HER2-positive breast cancer (defined in accordance with ASCO/CAP guidelines [22]) with a documented PIK3CA mutation by tumor tissue analysis with next-generation sequencing. Patients were also required to have Eastern Cooperative Oncology Group (ECOG) performance status of 0–1, measurable or non-measurable disease, adequate end organ function, and fasting plasma glucose < 140 mg/dl. All patients must have received prior pertuzumab and T-DM1 unless they were deemed ineligible for these treatments. Any number of prior lines of chemotherapy in the metastatic setting was allowed. Patients were excluded if they had prior grade 3 hypersensitivity to cremophor or trastuzumab, had active central nervous system metastases, clinically significant cardiac disease or impaired cardiac function (ejection fraction <50%), diabetes that was sub optimally controlled, or a history of prior treatment with a PI3K or AKT inhibitor. See protocol for full inclusion/exclusion criteria. The study was approved by the institutional research ethics board of Memorial Sloan Kettering Cancer Center. All participants signed an IRB-approved Informed Consent document before they entered the study. This trial was registered at clinicaltrials.gov (identifier NCT02167854). The study was conducted in accordance with the US Food and Drug Administration regulations, the International Conference on Harmonization Guidelines for Good Clinical Practice and the Declaration of Helinski.

Study Design and Treatment Plan

The primary objective of this study was to determine the MTD of alpelisib when administered in combination with LJM716 and trastuzumab. Cycles were 28 days in length. Dose levels for each individual drug were chosen based on interim data from early-phase trials of LJM716 monotherapy (NCT01598077), LJM716 combination with trastuzumab (NCT01602406), and LJM716 in combination with alpelisib (NCT01822613) which were ongoing at the time of trial design. The trastuzumab dose was fixed at 2mg/kg and administered intravenously weekly (on days 1, 8, 15, and 22 of each cycle); patients whose last dose of trastuzumab was >21 days before enrollment received a loading dose of trastuzumab at 4 mg/kg intravenously over 90 minutes on Day 1. The dose of LJM716 was fixed at 20mg/kg and administered intravenously weekly (on days 1, 8, 15, and 22 of each cycle), with the only exception in dose level 1, where LJM716 would be given at a dose of 10mg/kg weekly. Patients in Arm A received alpelisib orally daily, as part of a dose-escalation protocol, starting with 250mg daily (Table 1). Doses were assigned using the Continual Reassessment Method (CRM) and a total sample size of 18 patients per arm was planned [23]. Suggested dose modifications and treatment holds were pre-specified in the protocol for anticipated toxicities. Guidelines were also provided for the use of supportive medications. For example, antidiarrheal medication was recommended at the first sign of abdominal cramping, loose stools or overt diarrhea.

Table 1.

Planned dose levels and observed DLTs per dose level

Dose Levels by Drug				Arm A (n=10) Daily alpelisib		Arm B (n=11) Intermittent alpelisib
Dose Level	Alpelisib	LJM716	Trastuzumab	Patients per dose level	DLTs	Patients per dose level	DLTs
1	200mg	10mg/kg	2mg/kg	-	-	-	-
2	200mg	20mg/kg	2mg/kg	-	-	-	-
3	250mg	20mg/kg	2mg/kg	9	1 - Increased ALT	-	-
4	300mg	20mg/kg	2mg/kg	1	1 - SVT	3	-
5	350mg	20mg/kg	2mg/kg	-	-	8	2 - Macular Edema, Diarrhea*; Hyperglycemia

Dose level 3 was the starting dose for arm A, dose level 4 was the starting dose in Arm B.

^*The DLTs macular edema and diarrhea were observed in the same patient. DLT= Dose-Limiting Toxicity; ALT = Alanine aminotransferase; SVT = supraventricular tachycardia

For determination of the MTD, evaluable patients must have been followed through at least one cycle or have a dose-limiting toxicity (DLT). Therapy was continued until disease progression or unacceptable toxicity. Due to the observed toxicity pattern of continuously dosed alpelisib in this combination, Arm A was closed to accrual, and the research protocol was amended to allow for the opening of Arm B, in which patients received intermittent dosing of alpelisib (on the same dose escalation protocol) starting at 300mg for 4 days followed by 3 days off, such that alpelisib was given on days 1–4, 8–11, 15–18, and 22–25 of each cycle. LJM716 and trastuzumab were given at the same dose and schedule in arm B as they were in arm A. A similar CRM design was used to evaluate intermittent dosing in Arm B, independent from Arm A. In both arms, therapy was continued until disease progression or unacceptable drug toxicity.

The MTD was defined as the dose that leads to DLT in approximately 25% (target/acceptable rate) of patients with refractory solid tumors. Approximately 11 patients were treated on Arm A during dose escalation. Approximately eighteen patients were planned on Arm B during the dose escalation phase to determine the toxicity profile, DLT, and the MTD for alpelisib. DLT is a binary outcome and is defined below. The recommended dose for expansion (RDE) was based on an assessment of the MTD, review of required dose modifications and consideration of toxicities occurring beyond cycle 2 of the study. An expansion cohort of 30 patients at the MTD was planned but not executed. See Figure 1 for study schema.

Figure 1. Study Schema.

Arm A (n=10) included patients who were treated with trastuzumab, LJM715, and daily alpelisib; alpelisib dose was determined by the continual reassessment method (CRM). Based on the observed toxicity profile of this regimen, arm B (n=11) was opened to investigate intermitted dosing of alpelisib. IV = Intravenous; HER2+= Human epidermal growth factor receptor 2-positive; MBC = Metastatic breast cancer; GI=Gastrointestinal.

Toxicity and Treatment Response Assessment

All patients who received at least one dose of alpelisib, LJM716 and/or trastuzumab were evaluable for toxicity assessment, which was evaluated using the NCI Common Terminology Criteria for Adverse Events version 4.0. A DLT was defined as an adverse event or abnormal laboratory value assessed to be at least possibly related to the study medication occurring during Cycle 1 (≤ 28 days following the first dose of alpelisib) according to criteria defined in the study protocol. As an additional safety measure to assess for late-appearing toxicity, a confirmatory dedicated safety assessment was performed at the Cycle 3, day 1 visit.

Treatment response was assessed with CT scans initially after 2 cycles (56 days) and then every 3 cycles thereafter using the RECIST version 1.1 criteria [24]. The best overall response was defined as the best response recorded from the start of treatment until disease progression or withdrawal from the study. All patients who received at least one week of alpelisib were evaluable for response. Overall response rate (ORR) was defined as complete response (CR) + partial response (PR).

Correlative and Exploratory Analyses

Pre-treatment and on-treatment biopsies (obtained prior to cycle 1 day 1, and on cycle 1 day 15 respectively), were obtained in a subset of patients on arm A. All samples were independently assessed for tumor content. Those with adequate tumor cellularity underwent RNA extraction and quantitative mRNA profiling using the NanoString platform [25]. Patients additionally underwent tumor sequencing with MSK-IMPACT, a high throughput, targeted-DNA-sequencing panel for somatic mutations using methods as previously described [26]. ER, PR and HER2 staining on pre-treatment tissues were performed as part of routine care using standard clinical grade assays and tabulated retrospectively in Figure 2. PFS Ratio was calculated for each individual patient as the PFS (in weeks) on trial divided by the PFS (in weeks) on the line of systemic therapy that immediately preceded clinical trial enrollment.

Figure 2. Swimmers Plot.

Swimmers plot describing the number of weeks patients remained on study in each arm, along with the reason for study discontinuation denoted by a symbol. One patient left study due to a coincident complicating medial issue, but all others left study due to disease progression or toxicity. Patient-level data regarding breast cancer receptor status in the most proximal pre-treatment biopsy, as well as PIK3CA mutation status. PFS ratio is defined PFS on the current trial/PFS on the line of systemic therapy immediately prior to enrollment. ER= Estrogen Receptor; PR= Progesterone receptor; PFS = Progression Free Survival; Mut. = mutation. PD = Progression of Disease; PR = Partial Response

Statistical Analyses

CRM assumes a simple model for the probability of a DLT as a function of dose, and uses the occurrence of dose limiting toxicities within cycle 1 in the patients enrolled in the trial to sequentially determine which dose to administer to a new patient [23]. At the end of the trial, the posterior probability of a DLT occurrence was calculated. All toxicities were summarized using frequencies. Exploratory and correlative analyses on mutations and efficacy are reported as descriptive statistics. Clinical outcomes are reported graphically in a Swimmer’s plot.

Results

Patient Characteristics and Demographics

Ten patients were enrolled onto Arm A and 11 patients were enrolled onto Arm B. Baseline patient characteristics for patients on Arms A and B are shown in Table 2. Median age was 57 years (range 47–68) in Arm A and 53 years (range 36–67) in Arm B. Median number of lines of prior therapy for MBC were 6.5 in Arm A and 5 in Arm B (range 3–10). Five patients (50%) on Arm A and 8 patients (72.7%) on Arm B had hormone-receptor positive breast cancer. All patients had received trastuzumab, pertuzumab, and T-DM1 prior to enrollment. Overall 6 patients (28.6%) had an ECOG performance status of 0, and 15 (71.4%) patients had an ECOG performance status of 1. Ten patients (47.6%) had visceral disease upon enrollment.

Table 2.

Demographics and Baseline Patient Characteristics

Demographics	Arm A n=10	Arm B n=11
Median age, years (range)	57 (47–68)	53 (36–67)
ECOG PS, N (%)
0	4 (40)	2 (18)
1	6 (60)	9 (82)
Median # of prior therapies for MBC (range)	6.5 (3–10)	5 (3–10)
Prior HER2-targeted therapy, N (%)
Trastuzumab	10 (100)	10 (100)
Pertuzumab	10 (100)	10 (100)
TDM-1	10(100)	10(100)
Lapatinib	4 (40)	6 (55)
Breast cancer pathology, N (%)
ER+/HER2+	5 (50)	8 (73)
ER−/HER2+	5 (50)	3 (27)
Sites of metastasis, N (%)
No visceral disease	4 (40)	7 (64)
Visceral involvement	6 (60)	4 (36)

ECOG PS = Easter Cooperative Oncology Group Performance Score; MBC = Metastatic breast cancer. ER+ = Estrogen receptor-positive; HER2+ = human epidermal growth factor receptor 2-positive.

Safety and Dose-finding

All 21 enrolled patients were evaluable for toxicity. In arm A (n=10 patients), grade ≥3 adverse events included diarrhea (n=6), hypokalemia (n=3), abnormal liver enzymes (n=3), hyperglycemia (n=2), mucositis (n=2), elevated lipase (n=2), nausea (n=1), anorexia (n=1), and supraventricular tachycardia (n=1). The MTD of alpelisib in this arm was 250mg daily. Based on this toxicity profile, daily alpelisib was not thought to be feasible, which prompted the investigation of an alternative schedule of intermittently dosed alpelisib (Arm B) based on preclinical models as discussed below. Eleven patients were treated in Arm B, in whom grade ≥3 adverse events included diarrhea (n=5), hypokalemia (n=3), hypomagnesemia (n=2), and anorexia (n=1). The MTD of intermittently dosed alpelisib was 350mg 4 days on, three days off. Overall, 6 patients withdrew study participation due to toxicity, 3 in arm A and 3 in arm B.

There were 5 DLTs across both arms, all grade 3, observed in four patients (Table 1). DLTs included increased ALT (on alpelisib 250mg QD), supraventricular tachycardia (on alpelisib 300mg QD), diarrhea (on alpelisib 350mg intermittently dosed), macular edema and hyperglycemia (on alpelisib 350mg intermittently dosed, observed in the same patient). All DLTs were attributed to study treatments. Dose reductions of alpelisib were required in 11 patients, and repeated dose reductions were required in 3 (Tables 3 and 4). Overall, diarrhea presented the greatest challenge in terms of the toxicity of this combination regimen. This symptom was slightly improved but not addressed entirely by intermittent dosing of alpelisib. Dose expansion was not pursued based on this observed toxicity profile, and thus the RDE was not formally declared. The recommended phase 2 dose is the same as the MTD.

Table 3.

Toxicity by Grade

Toxicity	Arm A (n=10)				Arm B (n=11)
	G1 (%)	G2 (%)	G3 (%)	Total (%)	G1 (%)	G2 (%)	G3 (%)	Total (%)
Diarrhea	2 (20)	1 (10)	6 (60)	9 (90)	4 (36)	1 (9)	5 (45)	10 (90)
Hyperglycemia	6 (60)	1 (10)	2 (20)	9 (90)	9 (82)	-	1 (9)	10 (90)
Nausea	4 (40)	3 (30)	1 (10)	8 (80)	3 (27)	4 (36)	-	7 (64)
Hypokalemia	4 (40)	-	3 (30)	7 (70)	5 (45)	-	3 (27)	8 (72)
Anorexia	1 (10)	5 (50)	1 (10)	7 (70)	3 (27)	5 (45)	1 (9)	9 (82)
Mucositis	3 (30)	2 (20)	2 (20)	7 (70)	4 (36)	3 (27)	-	7 (64)
Hypomagnesemia	5 (50)	2 (20)	-	7 (70)	5 (45)	-	2 (18)	7 (64)
AST increased	1 (10)	-	2 (20)	3 (30)	5 (45)	-	-	5 (45)
Dysgeusia	5 (50)	-	-	5 (50)	4 (36)	1 (9)	-	5 (45)
Rash	2 (20)	2 (20)	-	4 (40)	1 (9)	1 (9)	-	2 (18)
Lipase increase	1 (10)	2 (20)	2 (20)	5 (50)	2 (18)	-	-	2 (18)
Dry mouth	3 (30)	1 (10)	-	4 (40)	3 (27)	-	-	3 (27)
Vomiting	3 (30)	1 (10)	-	4 (40)	2 (18)	4 (36)	-	6 (54)
ALT increased	3 (30)	1 (10)	1 (10)	5 (50)	4 (36)	-	-	4 (36)
Abdominal pain	2 (20)	1 (10)	-	3 (30)	-	1 (9)	-	1 (9)
Fatigue	1 (10)	2 (20)	-	3 (30)	3 (27)	4 (36)	-	7 (64)
SVT	-	-	1 (10)	1 (10)	-	-	-	-

G1, G2, G3 = Grade 1–3 toxicities using Common Terminology Criteria for Adverse Events version 4.0. AST= Aspartate transaminase; ALT = Alanine aminotransferase; SVT = supraventricular tachycardia

Table 4.

Toxicities requiring dose reduction

Arm	Starting Dose	1st Dose Reduction	Cycle#	Reason	2nd Dose Reduction	Cycle #	Reason
A	250mg	200mg	3	G2 mucositis
A	300mg	250mg	2	G2 palpitations	200mg	5	G3 diarrhea
A	250mg	200mg	2	G3 diarrhea
B	300mg	250mg	4	G1 diarrhea
B	300mg	250mg	3	G1 mucositis, G1 diarrhea
B	350mg	300mg	3	G3 diarrhea
B	350mg	300mg	4	G3 diarrhea	250mg	6	G2 diarrhea
B	350mg	300mg	1	G3 diarrhea, G3 macular edema	250mg	2	G3 macular edema
B	350mg	300mg	3	G1 hypomagnesemia
B	350mg	300mg	3	G2 diarrhea
B	350mg	300mg	1	G2 fatigue, G1 nausea

G1, G2, G3 = Grade 1–3 toxicities using Common Terminology Criteria for Adverse Events version 4.0

Efficacy

All enrolled patients were evaluable for efficacy. However, four patients discontinued study participation prior to the first on-study response assessment: three for toxicity and one for other medical reasons. Among patients who underwent response assessment, 1 exhibited a partial response, 14 exhibited stable disease and 2 exhibited disease progression at best response. The ORR among all evaluable patients was thus 1 of 21 (4.8%). The ORR among patients with measurable disease who underwent response assessment was 1 of 13 (7.7%). Five patients, 4 of whom were in Arm B, had SD for >30 weeks (Figure 2).

Exploratory and Correlative Analyses

MSK-IMPACT was performed on tumor tissue from 19 of 21 patients (14 of these samples were obtained pre-treatment); the remaining two patients had outside tissue-based testing performed using commercial assays. This testing confirmed somatic PIK3CA mutations in all enrolled patients. The most common PIK3CA mutation was the H1047 exon 20 mutation, discovered in 11 patients (52.4%). Two patients had 2 concurrent mutations in PIK3CA, one with H1047R and K111N mutations who remained on study for 36 weeks before disease progression, and another with E545K and T1025A mutations who remained on study for 20 weeks before disease progression. One patient with a concurrent PIK3CA amplification and a H1047R mutation remained on study for 32 weeks before disease progression. Among the three patients who remained on study the longest, one had notable concurrent mutations in NRAS and NF1 with FGFR3 amplification, another had an NF1 rearrangement, and the last had several copy number alterations including CCND1, FGF19, and FGFR3. The fold change in FGFR3 copy number was 2.1 for the patient with PFS of 57 weeks, and 2.0 for the patient with PFS of 75 weeks. No other patients in the cohort exhibited copy number changes in FGFR3. The cutoff for gene amplification using the MSK-IMPACT assay is 2.0, thus neither of these cases is considered highly amplified.

Pre- and on-treatment tissue samples were available for 8 patients, all of whom were enrolled on arm A of the study. Two patient sample sets were discarded: one because no tumor content was discovered in independent review of the biopsy sample and another because of technical issues, leaving 6 paired samples for analysis with quantitative mRNA profiling. Comparing with pre-treatment biopsies, on-treatment biopsies showed induction in the intrinsic pathway genes PIK3CA, AKT2 and AKT2, as well as in the feedback genes IGF1R and IRS2 in 6 of 6 patients. FOXO4 and IRS1 transcripts were further upregulated in 5 of 6 patients while ERBB3 transcripts were induced in 3 of 6 patients. ER-related genes ESR1, PGR1, and XBP1 were induced in ER-positive tumors, whereas MYC was induced in all 6 tumors. Responses in parallel pathway genes were mixed (Figure 3).

Figure 3. Correlative Analyses.

Patient level data are presented for the 6 study participants for whom adequate tissue for mRNA-profiling was available, including weeks on treatment, study stop reason, hormone-receptor and PIK3CA mutational status, in addition to the fold-change in selected mRNA transcripts in pre-vs. on-treatment study samples, taken prior to cycle-1 day 1, and on cycle-1 day-15 respectively. Values of 1 represent no change in transcript levels over time; values >1 represent increases and <1 represent decreases. Transcripts are grouped according to their functional classes. ID = Identification; POD = progression of disease; Tox = Toxicity; ER = Estrogen receptor.

Discussion

We present the results from a phase 1 clinical trial testing a rational therapy combination in treatment-refractory PIK3CA-mutated HER2-positive MBC. These efforts were based on evidence that PIK3CA mutations confer resistance to HER2-targeted therapy in this population, but that single-agent PI3K blockade might be compromised by relief of feedback inhibition which could be overcome by concurrent HER2 and HER3 blockade [6, 8, 9, 27]. Specifically, we aimed to determine the safety and maximum tolerated dose of the α-specific PI3K inhibitor alpelisib when used in combination with fixed doses of LJM716 and trastuzumab. This study demonstrates the feasibility of targeting PI3K in PIK3CA-mutated HER2-positive MBC using an oral agent. While this general approach is worthy of further investigation, attention should be paid to the issue of overlapping drug toxicity when generating future combination therapies for this high-risk patient population.

The patients who participated in this study were heavily pre-treated, such that all had received trastuzumab, pertuzumab, and T-DM1, and approximately half had received lapatinib before enrollment. In Arm A of this trial, we found that the MTD of alpelisib was 250mg taken orally daily. However, daily dosing of alpelisib in this therapy combination was associated with significant gastrointestinal toxicity, including grade 3 diarrhea which was observed in 6 of 10 patients (60%) and was not sufficiently ameliorated with supportive interventions. Concurrent grade 3 hypokalemia was also noted in 3 patients, and grade 3 abnormal liver enzymes, lipase elevation, nausea, anorexia, hyperglycemia and mucositis were also observed. As discussed above, this toxicity prompted a protocol revision which allowed for the creation of Arm B, in which patients were treated with intermittently dosed alpelisib, 3 days on followed by 4 days off. This decision was based on preclinical data suggesting that the activity of PI3K inhibitors is retained with intermittent dosing, and that this strategy may enhance therapeutic index to allow for combination treatment with other drug partners [27–29]. The MTD of intermittently dosed alpelisib in Arm B was 350mg. In this arm, grade 3 diarrhea was observed in 5 of 11 patients (45%), and no grade 3 nausea, mucositis, LFT or lipase elevations occurred. While intermittently dosed alpelisib appeared less toxic than daily dosing, both schedules did generate notable safety signals which compromised the tolerability of this regimen and precluded dose expansion and further exploration of this triplet. We also note that baseline differences between study arms in factors such as the presence of visceral disease and number of prior therapies received may be partially explanatory of any observed differences in toxicity patterns.

Of note, other efforts to target the PI3K/AKT pathway in HER2-positive MBC have also been limited by toxicity [30–32]. HER2-targeted therapies such as trastuzumab are known to cause diarrhea, which appears to be worsened with dual blockade with other members of the ErbB receptor family [33–35]. The mechanism of diarrhea with these treatments appears to be distinct from that caused by chemotherapy, and involves the upregulation of ion channels in enterocytes without disruption of gut barrier function [36]. Gastrointestinal toxicity of any grade was further reported in 75.4% of patients who received alpelisib in SOLAR-1 at a dose of 300mg daily, the largest published trial of this drug in combination with fulvestrant in ER+ HER2-negative MBC [19]. It is possible that feedback upregulation of HER3 in normal tissues, such as the gut, can minimize the toxicity of dual PI3K and HER2 targeted therapy, and thus the addition of a HER3 inhibitor in this study exacerbated on-target, off-tumor toxicity. Future attempts to target PI3K in HER2-positive breast cancer might acknowledge this limitation, and utilize targeted agents with greater tumor specificity and less potential for overlapping drug-related toxicity.

With respect to the efficacy of the triplet, 1 partial response was observed among 21 evaluable patients. This response was seen at the first 8-week efficacy assessment, but unfortunately this patient exhibited disease progression 8 weeks later. Of the remaining 16 patients who underwent response assessment, best response was stable disease in 14 patients, and disease progression in 2 patients. Notably, 5 patients remained on therapy for >30 weeks, including two patients who remained free from disease progression for over a year. Four of the 5 patients with prolonged disease stability were in arm B of the trial. Of special note, the two patients were found to have dual PIK3CA mutations, and remained on study for 20 and 36 weeks respectively, reflecting the finding that double PIK3CA mutations sensitize breast cancers to PI3K inhibition [37]. Concurrent PIK3CA amplifications did not appear to correlate with therapy response in this small sample, and the presence of alterations in other receptor tyrosine kinases or the MAPK pathway did not preclude benefit from this triplet in our small sampleThese early data suggest that while this triplet does not appear uniformly cytotoxic, a subset of patients might obtain clinical benefit despite the dose-limiting toxicities that we observed. Conversely, seven patients exhibited early progression by the time of the 16-week disease assessment, suggesting de novo therapy resistance. However, these findings are highly limited by the small sample size and lack of randomization in this dose-finding study, as well as the number of patients who came off study due to excess toxicity.

Exploratory quantitative mRNA profiling demonstrated signs of PI3K target engagement by alpelisib in on-treatment biopsies, including the induction of intrinsic pathway as well as feedback genes. These findings support existing models of targeted therapy of PIK3CA-mutant breast cancer [9, 27]. The mixed changes we observed in ERBB2 and ERBB3 transcripts on treatment may be compensatory, though their significance is difficult to interpret in this small subset. Lastly, we noted ESR1 mRNA upregulation upon inhibition of PI3K specifically in ER-positive tumors. This well-described finding has been posited by our group and others as an escape mechanism to single-agent alpelisib in ER-positive breast cancer, and helps to explain the benefit of adding antiestrogen therapy to PI3K blockade in ER-positive HER2-negative MBC [19, 38–40]. If confirmed elsewhere, our findings may suggest that similar principles might apply in HER2-positive breast cancer, such that concurrent antiestrogen therapy may be beneficial for patients with PIK3CA-mutated “triple positive” MBC. As such, the safety and anti-tumor activity of anti-estrogen therapy in combination with CDK4/6 inhibitors [41] and HER2-targeted therapies such as trastuzumab, pertuzumab, and lapatinib [42, 43] has been recently established and is currently being evaluated in the first-line metastatic setting after induction therapy in a phase 3 trial for triple positive metastatic breast cancer (NCT02947685).

In conclusion, further study is warranted to develop strategies to target PI3K in HER2-positive MBC. This study provides evidence regarding the safety and preliminary efficacy of alpelisib when used with HER2 and HER3 blockade in this population. Preclinical studies suggest that patients with HER2-positive MBC with PIK3CA mutations are most likely to benefit from similar combination therapy regimens that inhibit parallel pathway activation. However, for this strategy to become a valid therapeutic approach, the significant hurdle of overlapping drug toxicity must be overcome. Carefully crafted combinatorial approaches with HER2-targeted oral tyrosine kinase inhibitors, alternative inhibitors of PIK3CA/AKT including at intermittent dosing schedules, or antibody-drug conjugates should be considered. Some such efforts are underway (NCT04208178, NCT04208178, NCT04108858, NCT03767335, NCT04253561), which hold promise to improve outcomes for this high-risk patient population.

Statement of Translational Relevance.

PIK3CA mutations confer resistance to HER2-targeted therapies in breast cancer. In HER2-positive breast cancer models, targeting PI3K inhibits AKT activity but stimulates HER3 expression and MAPK parallel pathway activation, leading to therapy resistance. Concurrently targeting HER2, HER3, and PI3K is effective preclinically but has not been investigated clinically. We conducted an investigator-initiated, single center dose finding trial of this rational chemotherapy-sparing triplet targeting HER2, HER3 and PI3K in patients with refractory HER2-positive PIK3CA-mutant metastatic breast cancer. The study was halted early due to toxicity (primarily gastrointestinal in nature) but found that intermittent dosing of alpelisib appeared better tolerated than daily dosing, and that PI3K target engagement could be confirmed using mRNA profiling. To overcome overlapping toxicities in future trials employing this therapeutic strategy along with carefully crafted combinatorial approaches with HER2-targeted oral tyrosine kinase inhibitors, alternative inhibitors of PI3K/AKT (including with intermittent dosing schedules), or antibody-drug conjugates should be considered.