Endogenous pAKT activity is associated with response to AKT inhibition alone and in combination with immune checkpoint inhibition in murine models of TNBC

Abstract

Triple-negative breast cancer is a heterogeneous and challenging-to-treat breast cancer subtype. The clinical introduction of immune checkpoint inhibitors for TNBC has had mixed results, and very few patients achieved a durable response. The PI3K/AKT pathway is frequently mutated in breast cancer. Given the important roles of the PI3K pathway in immune and tumor cell signaling, there is an interest in using inhibitors of this pathway to increase the response to immune checkpoint inhibitors. This study sought to determine if AKT inhibition could enhance the response to immune checkpoint inhibition. We further sought to understand underlying mechanisms of response or non-response to AKT inhibition in combination with immune checkpoint inhibition. Using four murine TNBC-like cell lines and corresponding orthotopic mouse tumor models, we found that hyperactivity of the PI3K pathway, as evidenced by levels of phospo-AKT rather than PI3K pathway mutational status, was associated with response to AKT inhibition alone and in combination with immune checkpoint inhibition. Additional mutations in other growth regulatory pathways could override the response of PI3K pathway mutant tumors to AKT inhibition. Furthermore, we observed that AKT inhibition enhanced the response to immune checkpoint inhibition in an already sensitive model. However, AKT inhibition failed to convert checkpoint inhibitor-resistant tumors, to responsive tumors. These findings suggest that analysis of both the mutational status and phospho-AKT protein levels may be beneficial in predicting which TNBC tumors will respond to AKT inhibition in combination with immune checkpoint inhibition.

1. Introduction

Triple-negative breast cancer (TNBC) remains one of the most aggressive and difficult-to-treat subtypes of breast cancer for several reasons, including its heterogeneity and a lack of targeted therapies. TNBCs are broadly defined by a lack of HER2 amplification and a lack of expression of the estrogen receptor (ER) and the progesterone receptor (PR). This failure to express amplified levels of HER2 and the absence of ER and PR hormone receptor expression limits treatment options available to TNBC patients. As a result, surgery followed by radiation and chemotherapy, or neoadjuvant chemotherapy followed by surgery and radiation serve as the standard of care. While a subset of advanced TNBC patients respond to radiation and chemotherapy, the effects are often short-lived, with a median survival time until recurrence of approximately 9 months [1]. To identify subgroups of TNBC that respond differently to available therapies, work by Lehmann identified four subgroups of TNBC: basal-like 1 (BL1), basal-like 2 (BL2), mesenchymal (M), and luminal androgen receptor (LAR) [2]. Their data suggested that patients with BL1 tumors are more likely to achieve a pathological complete response after neoadjuvant chemotherapy [1], while the LAR group might benefit from androgen receptor antagonists combined with inhibition of PI3K [3]. Moreover, the M subtype has several properties, including high mutational load, low T cell infiltrate, low PD-L1 expression, and transcriptional repression of MHC-1. When tumor-bearing mice were treated with polycomb repressor complex 2 (PRC2) subunit enhancer of zeste homolog 2 (EZH2) or embryonic ectoderm development (EED) inhibitors, major histocompatibility complex class 1 (MHC-1) expression increased, and there was an enhanced response to chemotherapy [2].While there remains an overwhelming reliance on standard chemotherapeutic agents for treatment of TNBC, recently the immune checkpoint inhibitor (ICI), anti-PD-1, has been added to the treatment armament. The KEYNOTE-355 trial led to the approval of anti-PD-1 (pembrolizumab) in combination with chemotherapy in TNBC [3], while the results of the KEYNOTE-522 trial led to the approval of pembrolizumab plus chemotherapy as a neoadjuvant and adjuvant treatment for high-risk, early-stage TNBC regardless of PD-L1 expression [4,5]. While these studies revealed promise for ICI for the treatment of TNBC, the benefit of ICI therapy for breast cancer patients is somewhat disappointing compared to the durable therapeutic responses in a subset of cancer patients with other tumor types [6]. Therefore, there is a great need to determine how to increase the clinical benefit of ICI for TNBC patients. One strategy is to identify combinations of chemotherapy and/or targeted therapies directed to driver mutation or driver signaling pathways that induce immunogenic tumor cell death and render tumors more responsive to ICI.

Despite the heterogeneity within the TNBC designation, the phosphatidylinositol 3-kinase/protein kinase B (PI3K/AKT) pathway is one of the most frequently mutated pathways in this tumor type, making this pathway an attractive therapeutic target. Around 50% of breast cancers have one or multiple activating alterations in the PI3K/AKT pathway [7–9]. While TNBC has a lower PIK3CA mutation rate (~16%) than non-TNBC (34%–42%) [10], about 15% of TNBCs, compared to 8% in non-TNBCs, have a truncating mutation or deep deletion of the phosphatase PTEN (phosphatase and tensin homolog) that negatively regulates PI3K/AKT activity [11]. Loss of PTEN is a feature frequently associated with ICI resistance in tumors [11]. Furthermore, mutations in PIK3R1, the gene encoding the p85α regulatory subunit of PI3K, are more common in TNBC (9.41%) compared to non-TNBC (1.67%) [7–9]. Clinical trial studies suggest that metastatic breast cancer patients with a more robust, initial anti-tumor T cell response have a more extended therapeutic response to PI3K inhibition that may sensitize tumors to ICI [11]. Similarly, both PI3K and AKT inhibitors are reported to provide additive therapeutic effects with the chemotherapy drug paclitaxel (PTX) to reduce breast cancer proliferative and metastatic effects [12,13] Consistent with this notion, pan-PI3K inhibition with BKM120 (buparlisib) enhanced the response of anti-PD-1 therapy in the C57BL/6J PyMT orthotopic tumor model through increased recruitment of CD8+ T-cells [14]. Toxicity and efficacy concerns prevented some pan-PI3K inhibitors, particularly BKM120, from advancing clinically [15]. Our group recently showed that PI3K inhibition in preclinical mouse models either through a pan-PI3K inhibitor, copanlisib, or PI3K/mTOR inhibitor, gedatolisib, but not PI3Kα-specific inhibitor, alpelisib, offers promise for enhancing the response of TNBC to immune checkpoint inhibitor therapy by modulating the tumor microenvironment (TME), especially the T-cell response [16]. [16].

The toxicity associated with pan-PI3K inhibition in the clinic [15] led to the studies described herein using AKT inhibition (AKTi) as a less toxic strategy to inhibit the PI3K pathway. AKT is a serine/threonine kinase that acts downstream of PI3K and can directly phosphorylate more than 80 substrates upon its phosphorylation and activation at T308 by the 3-phosphoinositide-dependent kinase 1 (PDK1) and S473 by mammalian target of rapamycin complex 2 (mTORC2) [17]. AKT signaling plays an important role in immune and tumor cell signaling [18–20], suggesting that AKT inhibitors could be combined with immune checkpoint inhibitors. For example, an allosteric AKTi, MK-2206, increased the density of CD8+ T-cells in tumors from a clinical study in patients with hormone receptor (HR)-positive, HER2− breast cancer [21]. Furthermore, MK-2206 was included in the I-SPYII clinical trial that examined the response of locally advanced breast cancers to neoadjuvant treatment with multiple inhibitors, and the estimated pathological complete response rate (pCR) in the MK2206 treatment arm was 40% compared to 22% in the chemotherapy control group [22]. Importantly, early clinical trials of other pan-AKTi, ipatasertib (IPAT) and capivasertib (CAPI), showed acceptable safety profiles, with the most frequent adverse events being gastrointestinal-related and these events were resolvable with minimal intervention [23,24]. IPAT and CAPI are ATP-competitive inhibitors with high affinity for all three isoforms of AKT (IPAT IC₅₀ AKT1/2/3 = 5nM, 18nM, 8nM) [25] (CAPI IC₅₀ AKT1/2/3 = 3nM, 7nM, 7nM) [26]. In pre-clinical, mechanistic studies, ATP-competitive inhibitors of AKT exhibit anti-proliferative and pro-apoptotic effects[26,27]. While AKT inhibitors show acceptable safety profiles in clinical trials, measures of efficacy have been mixed. The Phase II LOTUS trial for first line IPAT plus PTX for inoperable locally advanced/metastatic TNBC showed a trend toward increased overall survival (OS) with the combination treatment for PTEN low and PIK3CA/AKT1/PTEN-altered subgroups (25.8 vs. 22.1 mo), although small sample size in these subgroups prevented detailed biomarker analysis [27]. In the subsequent phase III IPATunity130 trial of IPAT in combination with PTX for PIK3CA/AKT1/PTEN-altered breast cancer, the addition of IPAT failed to improve progression free survival in either the TNBC or HR+, HER2− cohort [28]. However, in the phase II PAKT trial of CAPI in combination with PTX as a first-line treatment for metastatic TNBC, the addition of CAPI significantly improved progression-free survival (PFS) (5.9 v 4.2 months) as well as OS (19.1 v 12.6 months) [24]. The ongoing phase III CAPItello-290 study is hoping to confirm these findings in a larger cohort which will allow for detailed biomarker analysis (NCT03997123). In the FAIRLANE phase II trial of neoadjuvant IPAT plus PTX for the treatment of early TNBC, the addition of IPAT to PTX therapy failed to statistically increase pCR over that of PTX alone, although there was a trend toward improving overall response rate as analyzed by MRI in the IPAT treatment group [29]. A more recent analysis of the potential biomarkers for response to neoadjuvant IPAT plus PTX in TNBC patients from the FAIRLANE trial revealed that patients with high baseline pAKT exhibited an enriched benefit to IPAT therapy regardless of PI3K mutational status [30].[30].

An alternative approach involves combining IPAT or CAPI, both ATP-competitor inhibitors of AKT, with ICI therapy for the treatment of TNBC. CAPI is currently being investigated in combination with durvalumab (anti-PDL1) in metastatic TNBC (NCT03742102), and there are ongoing trials for IPAT in combination with PTX and atezolizumab (anti-PDL1) in metastatic TNBC (NCT04177108). To gain insight into the parameters involved in tumor response to inhibition of AKT with IPAT in combination with ICI, we evaluated this combination in four murine models of TNBC. We observed that PI3K pathway activation, as indicated by levels of phospho-AKT but not genetic mutation of the PI3K pathway, was associated with response to AKT inhibition alone or in combination with ICI in the murine TNBC-like models tested in our studies. We further sought to understand the mechanisms underlying response or resistance to AKT inhibition with or without ICI therapy.

2. Materials and Methods

2.1. Mouse Cell Line Details

All cell lines were maintained in a humidified 5% CO₂ incubator at 37°C. The polyoma middle T antigen (PyMT) breast cancer cells were provided by the Hal Moses laboratory. The MMTV-PyMT tumor cell line was established from a breast tumor growing spontaneously in a MMTV-PyMT C57BL/6J female mouse. The PyMT tumors that develop are of luminal origin and as tumor progression ensues, the tumors move from being ER+/PR+ and HER2+ to being HR−/HER2+, and finally are triple-negative-like, with no/minimal expression of ER, PR or HER2, but are androgen-receptor positive, such that the cell line is categorized as luminal androgen receptor subgroup of triple-negative-like breast cancer [31]. The PyMT cells were maintained in DMEM/F12 medium supplemented with 10% FBS. The 6DT1 cell line was provided by Kent Hunter (NIH/NCI) and was maintained in DMEM/F12 supplemented with 10% FBS. The 4T1 cell line was purchased from ATCC (Cat. CRL-2539) and maintained in RPMI medium supplemented with 10% FBS. The E0771 cell line was purchased from CH3 BioSystems (Cat. #940001) and cultured in RPMI medium with 10% FBS. The 6DT1, 4T1, and E0771 cell lines are also classified as models of TNBC. Cell cultures were tested monthly for mycoplasma (e-Myco TM Plus, LiliF Diagnostics), and only mycoplasma-free cells were used for experiments outlined in these studies.

2.2. In vitro viability and proliferation assays

IPAT IC50s for viability for PyMT, 6DT1, E0771, and 4T1 cells were determined by CellTiterBlue assay (Promega Cat. #G8080). Breast cancer cells were seeded in 96-well plates with 5,000 cells per well in 200uL complete media (10%FBS + 1%PenStrep DMEM-F12 for PyMT and 6DT1; 10%FBS + 1%PenStrep RPMI for 4T1 and E0771). Twenty-four-hours later, the cells were treated with IPAT diluted to working concentrations in DMSO, at a range of doses (0uM [DMSO control], 1uM, 5uM, 10uM, 15uM, 20uM, 50uM) in triplicate. After 72 hours, CellTiterBlue was added to wells according to the manufacturer’s instructions. Fluorescence was recorded at 560_Ex/590_Em nm. The dose values were log10 transformed and plotted on the x-axis. The IC50s for cell viability were calculated using a non-linear regression of the data plotted at log¹⁰ IPAT vs response (fluorescence) using Prism 9.2. Values are presented as mean +/− SEM of three independent experiments. In vitro cell viability in response to capivasertib and/or selumetinib (ARRY-142866) [32] was analyzed through crystal violet assay as previously described [33]. CellTrace Deep Red in vitro cell proliferation assays were performed according to manufacturer’s instructions (Invitrogen Cat. #C34553). CellTrace Deep Red fluorescent intensity was analyzed using a 4-Laser Fortessa and the data were analyzed using FLowJo 10.5.3 software.

2.3. Whole-exome DNA sequencing

TWIST mouse whole exome sequencing was performed by the VANTAGE genomics core at Vanderbilt University Medical Center on DNA samples from the four cell lines: PyMT, 6DT1, E0771, and 4T1. According to the manufacturer’s instructions, DNA was extracted from cell culture lysates (Qiagen, Cat. #69504). PI3K and MAPK pathway mutations were assessed from a list of 96 and 92 respectively commonly described genes as described in the publicly available GeneCardsSuite Pathway unification database [34]

2.4. Western blot analysis

For cultured cells, 1.5–2×10⁶ cells were lysed in RIPA buffer (Sigma-Aldrich, Cat. #R0278) with cOmplete^™, EDTA-free protease inhibitor cocktail and incubated on ice for 20 min. For tumor samples, 35–50 mg tumor tissue was homogenized with Precellys 24 tissue homogenizer in RIPA buffer (Sigma-Aldrich, Cat. #R0278) with cOmplete^™, Mini EDTA-free protease inhibitor cocktail (Roche, Cat. #04693159001) and PhosSTOP^™ phosphatase inhibitor cocktail (Roche, Cat. #0406837001). Lysates were centrifuged at 14,000g for 10 min and Pierce^™ BCA protein assay was performed on the supernatant to quantitate the amount of protein in each sample. 35–40ug of total protein per sample was resolved by SDS-PAGE and transferred to nitrocellulose membrane with the Trans-Blot Turbo Transfer System (Bio-Rad). When using fluorescent detection methods, blots were blocked with Intercept^® (TBS) Blocking Buffers (LICOR, Cat. #927–60001) at room temperature for 1 hour followed by overnight primary antibody (1:400–1:2000 dilution with 50% Intercept^® [TBS] Blocking Buffers with 505 0.1% Tween 20/TBS [TBST-T] buffer) incubation at 4°C. A list of primary antibodies is available in TableS5. After washing 4 times with TBS-T buffer, blots were incubated with IRDye^® 800CW goat anti-rabbit (LI-COR, Cat. #926–32211, 1:15,000 dilution with 2% polyvinyl pyrrolidone in TBST-T buffer) and IRDye^® 680RD goat anti-mouse (LI-COR, Cat. #926–68070, 1:20,000 dilution with 2% polyvinyl pyrrolidone in TBST-T buffer) secondary antibodies at room temperature for 1 hour, and then were washed with TBST-T buffer 5 times. Blots were scanned with OdysseyR CLx Imaging System (LI-COR). Signals were quantified with Image Studio Lite Ver 5.2(LI-COR). If stripping was needed to probe for an additional protein, the blot was stripped with NewBlotTM Nitro Stripping Buffer (LI-COR, Cat. #928–40030, 1:1 dilution with TBS buffer) for 1 hour. When using Enhanced chemiluminescence (ECL) detection method, blots were blocked with SuperBlock^™ blocking buffer (Thermo, Cat. #37515) for 1 hour at room temperature followed by overnight primary antibody (1:1,000 dilution with Superblock^™) incubation at 4°C. After washing four times with TBS-T buffer, blots were incubated with anti-rabbit HRP conjugate for 2hrs (Promega, Cat. #W401B). Blots were visualized on an Invitrogen iBrightFL1000.

2.5. Animal care

Experiments using mice were approved and executed under the guidelines of the Vanderbilt University IACUC. All animals were housed and cared for in the AAALAC-accredited Animal Research Center at Vanderbilt University, and routinely monitored by laboratory and veterinary staff. Animals were euthanized by CO₂ asphyxiation using isoflurane overdose followed by cervical dislocation in accordance with the Guide for the Care and Use of Laboratory Animals. Protocols were in place for early and humane endpoints should experimental animals display signs of illness. With oversight from the veterinary staff, mice were monitored daily, mouse weight was evaluated weekly, and tumor volume was measured twice a week to determine when and if the animals should be euthanized.

2.6. Mouse tumor models

Animal studies were approved by the Vanderbilt IACUC. Mice were purchased from Jackson Laboratories (C57BL/6J #000664, BALB/cJ #000651, FVB/NJ #001800, Nu/J #002019). The mice were all 6–10-week-old females weighing 20–26g. Mice were anesthetized before all procedures with 2.5% isoflurane administered with a precision vaporizer. The day before surgery, the mice were ear-tagged, and hair was removed by depilatory cream followed by 70% alcohol. On the day of the surgery, ketoprofen (5–10 mg/kg) was administered by subcutaneous (SQ) injection. Breast cancer cells (30,000–300,000 cells/mouse depending on cell line and study) were surgically implanted or directly injected into the fourth mammary fat pad as previously described [35]. For surgical implantation, the wound was closed by staples, which were removed 10–14 days post-surgery. Typically, tumors were approximately 50mm³ by 14 days post-implantation. Mouse weight and body condition scores were assessed weekly. Tumors were measured twice a week and tumor volume was estimated as [length × width × width × 0.5]. Treatments began when the tumor reached a volume of approximately 50mm³-100m³. IPAT, 100mg/kg), capivasertib (130mg/kg), and selumetinib (100mg/kg) were prepared in 5% DMSO and 95% corn oil and administered by oral gavage (PO) in a total volume of 100μL. Corn oil with 5% DMSO was used as vehicle control. IPAT was administered seven days a week, capivasertib BID for four days on, three days off, and selumetinib five days a week. PTX (PTX) was prepared in PBS and administered every 5^th day by retroorbital injection (RO). PBS was used as the vehicle control, and BNP ophthalmic ointment was applied to the eye after injection. ICI consisting of anti-PD-1 (200ug/mouse (BioXCell, Cat. #BP0146) or 250ug/mouse (BioXcell, Cat. and anti-CTLA-4 (100ug/mouse) antibodies, or the recommended isotype control antibodies, were administered in InVivo Pure pH 7.0 dilution buffer every 3^rd day by intraperitoneal (IP) injection. Details on all drugs used in in vivo studies are available in TableS4.

2.7. Immunohistochemisty analysis

Immunohistochemistry (IHC) analysis on 10% buffered formalin-fixed, paraffin-embedded tissue sections was performed by the Vanderbilt Translational Pathology Shared Resource using standard protocols. Slides were placed on the Leica Bond IHC stainer. All steps besides dehydration, clearing and cover slipping were performed on the Bond stainer. Slides are deparaffinized. Heat induced antigen retrieval was performed on the Bond using Epitope Retrieval 2 solution for 15 minutes. Slides were placed in a Protein Block (Ref# x0909, DAKO, Carpinteria, CA) for 10 minutes. The sections were incubated with primary antibody for one hour. The Bond Refine Polymer detection system was used for visualization. Slides were then dehydrated, cleared and cover slipped. Antibody information is available in Table S5. Whole slide imaging was performed in the Digital Histology Shared Resource at Vanderbilt University Medical Center (www.mc.vanderbilt.edu/dhsr). 20x Brightfield images were acquired on a Leica SCN400 slide scanner.

2.8. Flow cytometric analysis

To phenotype immune cells in the tumor, approximately 10⁶ cells were isolated from dissociated tumors, rinsed, and stained with markers of interest using our standard protocols. Live/Dead-Ghost Dye Violet 540 (Tonbo Biosciences, Cat. #13-0870-T100) was used to gate out dead cells. Antibodies used for cytometry are available in Table S5. Cells were washed and re-suspended in fixation buffer (PBS containing 1% formalin). A total of 200,000 single-cell events were collected using a 4-Laser Fortessa and the data were analyzed using FLowJo 10.5.3 software. For t-distributed stochastic neighbor embedding (t-SNE) analysis, live CD45+ leukocytes were concatenated after downsampling to ~25,000 events for subsequent high-dimensional data analysis to normalize the contribution among samples under different treatments. Then, samples were analyzed in parallel by t-SNE, and manually gated leukocytes populations were overlaid onto the total t-SNE map using FlowJo 10.5.3.

2.9. Cytokine analysis

Tumor lysates were prepared as previously described [36], and 2–4 tumors per treatment group were pooled together. Protein concentrations of tumor lysates were determined by and Pierce^™ BCA protein assay, and 150ug of protein was used for each sample. Lysates were subjected to analysis with the G-series 3 Mouse Cytokine Antibody Array (RayBio, Cat. #AAM-CYT-G3–8). Samples were run in duplicate and normalized to an internal control as per manufacturers’ protocols. The glass chip was scanned on the Cy3 channel of a GenoPix 4000B scanner (Genopix 6.1, Molecular Devices).

2.10. Statistical analysis of tumor growth

Tumor growth curve data were summarized in figures using the mean ±S.E. for the error bar chart. Student’s t-test with unequal variance and analysis of variance (ANOVA) with Tukey’s HSD test were used when comparing two groups or more than two groups, respectively. The mixed-effects model was used to account for the correlations between repeated measurements of tumor volumes over time within each mouse. With least-square mean, the difference in tumor growth between treatments was compared using Wald test. With pairwise comparison, the p-value was adjusted using the Holm-Bonferroni method to correct inflated Type I error. For the synergic analysis, a mixed effect model with individual and interactive effects was used to assess interaction effect of two drugs on tumor volume over days. The likelihood ratio test was performed to determine if the interaction effect were statistically significant. When an interaction effect existed, a synergistic effect (i.e., combo effect > drug A effect + drug B effect) was further examined using general linear hypothesis. To better meet the normality assumptions for these parametric methods, the square root or natural log transformation was implemented to ameliorate the heterogeneity evident in the data. Standard residual analysis was evaluated. All tests are two-tailed, and the statistically significant level is set at 0.05. The analyses were performed using R version 4.2.0.

2.11. RNA isolation and Nanostring gene expression analysis

Tumor samples were collected in RNAlater and stored according to manufacturer’s instructions (Thermo Fisher, Cat. #AM7020). RNA was extracted using Qiagen RNeasy kit (Cat. #74004). RNA concentration was determined using a NanoDrop. Gene expression analysis was performed using the Nanostring nCounter^® mouse PanCancer Immune Profiling Panel (Cat. #115000142). The assay was performed using the nCounter Sprint Profiler at the Vanderbilt VANTAGE facility. Background correction was applied to target gene counts and target gene counts were normalized to housekeeping genes. “Spiked in” positive and negative controls were included for data normalization. Probes below background threshold were discarded from the analysis. Data was analyzed using nSolver and R Software (version 3.3.2). Pathway analysis and Cell type profiling were performed using nCounter Advanced Analysis 2.0 software.

3. Results

3.1. IPAT alone and in combination with ICI inhibits PyMT tumor growth

Given our previous work showing the pan-PI3K inhibitor, BKM120, enhanced response to anti-PD-1, we were interested in determining whether inhibition of the PI3K downstream kinase, AKT would also enhance response to ICI therapy. PyMT cells (1×10⁵) were orthotopically implanted into the fourth mammary fat pad of 8–10 week old female C57BL/6J mice. Therapeutic treatments were initiated when tumors grew to a volume between 50mm³-100mm³, and were continued for approximately 2-weeks. Treatment groups were as follows: (1) Vehicle(V)+IgG, (2) V+anti-PD-1(250ug, IP, 1x/3 days), (3) V+anti-CTLA4 (100ug, IP, 1x/3 days) (4) V+ICI (anti-PD-1(250ug)+anti-CTLA4(100ug) (IP, 1x/3days) (5) IPAT (100mg/kg, PO, QD)+IgG, (6) IPAT+anti-PD-1, (7) IPAT+anti-CTLA4, and (8) IPAT+ICI (anti-PD-1+anti-CTLA4). Compared to V+IgG control-treated mice, statistically significant reductions in tumor growth were observed in mice treated with V+ICI (p=0.033), IPAT+IgG (p<0.001), IPAT +anti-PD-1 (p<0.001), IPAT+ anti-CTLA4 (p<0.001), and IPAT+ICI (p<0.001) (Figure 1a). Additionally, IPAT+ICI significantly reduced tumor growth when compared to both V+ICI (p<0.001) and IPAT+IgG (p<0.001), suggesting additive effects of the therapies (Figure 1a). While the combination of IPAT and ICI reduces tumor growth more than either as a single agent, further statistical analysis revealed that the effect of IPAT+ICI does not result from the synergy (Figure S1). Furthermore, IPAT+anti-PD1 reduced tumor growth compared to V+anti-PD1 (p<0.001) and IPAT+anti-CTLA4 significantly reduced tumor growth compared to V+anti-CTLA4 (p<0.001) (Figure 1a). IPAT+anti-PD1, IPAT+anti-CTLA4 and IPAT+ICI treatment also significantantly reduced tumor weight at end point (Figure 1b). Additional statistical comparisons of tumor growth rate were made between all combinations of treatment groups and are available in Table S1. Caspase-3 IHC staining of paraffin-embedded tumor tissue was performed to assess apoptosis in response to the various treatment combinations. Histological analysis revealed high amounts of necrosis in tumors from all treatment groups at the endpoint of the experiment and necrotic tissue was excluded from analysis (Figure S2). Caspase-3 staining was quantified as percent positive staining in non-necrotic tumor area (Figure 1c). While IPAT+ICI and IPAT+anti-PD1 trended toward increasing capase-3 positivity compared to V+IgG, the combination of IPAT+anti-CTLA4 significantly increased caspase-3 positivity compared to V+IgG (p<0.05) (Figure 1c). There was not a significant difference in caspase-3 positivity between IPAT+anti-PD1 and IPAT+anti-CTLA4 (Figure 1c). In summary, while the PyMT tumors responded to AKTi as well as ICI alone, the combination of AKTi+ICI treatment led to the most robust reduction in tumor growth.

Figure 1. IPAT alone and in combination with ICI inhibits PyMT tumor growth.

a. Growth of PyMT tumors in response to various treatment combinations: (1) Vehicle(V)+IgG, n=5 (2) V+anti-PD-1(200ug, IP, every 3 days), n=5 (3) V+anti-CTLA4 (100ug, IP every 3 days), n=5 (4) V+ICI(anti-PD-1(200ug)+anti-CTLA4(100ug) (IP, every 3 days), n=6 (5) Ipat (100mg/kg, PO, QD)+IgG, n=5 (6) Ipat+anti-PD-1, n =5 (7) Ipat+anti-CTLA4, n=5 and (8) Ipat+ICI (anti-PD1+anti-CTLA4) n=6. Mice were treated for 13 days. Tumor growth was analyzed with a mixed effects model and analyzed on the natural log scale to better meet normality assumptions. The mean and SE of tumor volume over time for each treatment group is shown. p-values adjusted using the Holm-Bonferonni method *=p<00.05, **=p<0.01, ***=p<0.001

b. Tumor weight measured at the experiment endpoint. Statistical analysis: One-way ANOVA with Tukey’s post-hoc analysis * = p < 0.05

c. Representative 40x images of IHC analysis of caspase3 (red) expression in formalin fixed tumor sections. Quantitated as %caspase3 + staining in viable tumor tissue. Statistical analysis: One-way ANOVA with Dunnett’s post hoc analysis * = p<0.05.

d. Growth of PyMT tumors in response to the follow treatments: (1) Vehicle(V)+IgG, (2)PTX (RO, 10mg/kg, qw) + IgG, (3) IPAT (PO, 100mg/kg, QD) PTX (RO, 10mg/kg, qw)+ IgG, (3 ) PTX+ ICI (anti-PD-1(200ug) and anti-CTLA-4 (100ug), IP, every 3 days), (4) IPAT+ PTX + ICI, (5) IPAT + PTX + IgG, or (6) IPAT + PTX + ICI. Tumor growth was analyzed with a mixed effects model and analyzed on the natural log scale to better meet normality assumptions. The mean and SE of tumor volume over time for each treatment group is shown. n=5 mice per group. p-values adjusted using the Holm-Bonferonni method *=p<0.05, **=p<0.01, ***=p<0.001

e. Tumor weight measured at endpoint of experiment shown in c. Statistical analysis: One-way ANOVA with Tukey’s post-hoc analysis

f. PyMT rechallenge study. Mice from the experiment shown in c that exhibited full tumor regression were re-injected with PyMT cells 75 days after the initial experiment endpoint, The mice remained tumor free for 150 days, at which point the experiment was terminated.

In clinical trials, IPAT is being investigated in combination with standard-of-care chemotherapy, PTX, as a therapy for TNBC. Therefore, we examined the response of AKTi+PTX+ICI treatment in the PyMT orthotopic injection mouse model as described above. Treatments were as follows: (1) V+IgG, (2) PTX (RO, 10mg/kg, 1x/5 days) + IgG, (3) IPAT (PO, 100mg/kg, QD) + PTX + IgG, (4) PTX+ ICI (anti-PD-1,200ug and anti-CTLA-4, 100ug, IP, 1x/3 days), (5) IPAT+ ICI, (6) IPAT + PTX + ICI. Treatments were initiated when the tumor measured approximately 50mm³ and continued for 25 days. We observed statistically significant reductions in tumor growth in mice treated with PTX+ICI (p=0.047), IPAT+ICI (p=0.047), IPAT+PTX+IgG (p=0.003), and IPAT+PTX+ICI (p<0.001) compared to V+IgG-treated mice (Figure 1d). IPAT+PTX+ICI also significantly reduced tumor growth compared to PTX+ICI (p<0.001) (Figure 1d). Tumor weight was assessed at the experiment endpoint and tumors from IPAT+ICI, IPAT+PTX, and IPAT+PTX+ICI treatment groups weighed significantly less than tumors from the vehicle+IgG control group (Figure 1e). In summary, the triple combination therapy, AKTi+PTX+ICI, resulted in a significant reduction in PyMT tumor volume compared to any two combined treatments,, including PTX+ICI, IPAT+ICI, or IPAT+PTX+IgG, suggesting an increased therapeutic benefit of IPAT+PTX+ICI (Figure 1d). Impressively, complete tumor regression (CR) was observed in 6 of the 7 mice treated with IPAT+PTX+ICI. After 75 days, 5 mice that exhibited complete tumor regression in response to IPAT+PTX+ICI were rechallenged with PyMT tumor cells (1×10⁵) and monitored for tumor regrowth. All five mice exhibited immune protection by fully rejecting the growth of the PyMT tumor cells, remaining tumor-free for >150 days (Figure1f).

3.2. Immune cell profile of PyMT tumors treated with AKTi + ICI

To better understand how AKTi may be enhancing the effects of ICI in the PyMT model, the intratumoral immune cell infiltrate was analyzed by nanostring nCounter^® analysis, a multiplex gene expression platform. The mouse PanCancer Immune Profiling Panel includes 770 target genes which were analyzed using the nsolver program. Pathway enrichment analysis indicated significant upregulation of six pathways in the AKTi+ICI treated tumors compared to V+IgG tumors: antigen processing, complement pathway, interferon pathway, MHC, microglial function, and TLR pathway (Figure 2a), defined by the gene lists in Table 1. Conversely, the basic cell functioning score was significantly decreased in the AKTi+ ICI group compared to the V+IgG group (Figure 2a). Cell type profiling revealed few differences in the abundance of specific cell types although there was a significant enrichment of the neutrophil cell score in the AKTi+ICI group compared to V+IgG (Figure 2b). Notably, the CD8 probe was discarded due to low cell counts below background detection of the assay. T-cell infiltrate was further examined by CD3 IHC staining of paraffin-embedded tumor tissue sections, confirming the low levels of total T-cells detected in the mRNA analysis (Figure 2c). While few T cells were present, AKTi plus ICI treatment produced a trend toward enhancing the infiltration of T-cells into the TME in the PyMT model compared to V+IgG treated tumors both by gene expression and IHC (Figure 2c). Treatment with AKTi alone, however, did not increase CD3+ T-cell infiltrate, and the increase caused by AKTi+ICI was very modest (<1%) (Figure 2c). CD8 IHC staining of paraffin-embedded tumor tissue additionally confirmed the extremely low presence of this population indicated by the gene expression analysis (Figure S3). The tissue sections and samples used for RNA analysis are obtained at the end point of the experiment where it is apparent upon dissection that the tumors contain a considerable amount of necrotic tissue. Therefore, it is quite possible that there was a T-cell infiltrate at earlier timepoints in the study that was not captured by this end point analysis based upon the response to ICI.

Figure 2. AKTi+ICI alters immune cell signaling in the PyMT model.

a. Pathway signature scores summarizing expression levels of related groups of genes from nanostring nCounter gene expression platform. Statistical analysis: One-way ANOVA with Dunnett’s post hoc analysis. From left to right, statistical comparisons represent V+IgG compared to IPAT+IgG, V+IgG compared to V+ICI, and V+IgG compared to IPAT+ICI. * = p< 0.05, ** = p<0.01. n = 3 biological replicates

b. Cell type signature scores summarizing expression levels of related groups of genes from nanostring nCounter gene expression platform. Statistical analysis: One-way ANOVA with Dunnett’s post hoc analysis. From left to right, statistical comparisons represent V+IgG compared to IPAT+IgG, V+IgG compared to V+ICI, and V+IgG compared to IPAT+ICI. * = p< 0.05. n= 3 biological replicates

c. Paraffin embedded tumor tissue sections from PyMT mice treated with V+IgG, V+ICI, Ipat+IgG, or Ipat+ICI, as described above, were analyzed for CD3 infiltrate by IHC staining (brown). Representative 20x images are shown and the scale marker is 150um.

Quantification of CD3 IHC calculated as %CD3+ nuclei using QuPath digital analysis software. Statistical analysis: One-way ANOVA with Tukey’s post-hoc analysis * = p <0.05

Table 1.

PyMT pathway signature score gene list from nCounter analysis

Pathway	p-value1 V+IgG v. V+IPAT	p-value V+IgG v. V+ICI	p-value V+IgG v. IPAT+ICI	Gene names
Antigen Processing	0.3448	*0.0399	*0.0279	H2-Aa, Cd74, H2-Ob, H2-Ab1, H2-Eb1, Fcgr3, Fcer1g, Slc11a1, Fcgr2b, Fcgr1, Tapbp, Icam1, Relb, H2-T23, H2-DMb1, H2-DMa, Nod2, Tap2, Psmb8, Tap1, Mr1, Ccr7, H2-DMb2, H2-Q10, Psmb9, H2-M3, H2-D1, H2-K1, Cd1d1, Nod1, H2Q2
MHC Pathway	0.4508	0.0536	*0.0382	H2-Aa, Cd74, H2-Ob, H2-Ab1, H2-Eb1, Fcgr3, Fcer1g, Fcgr2b, Fcgr1, Tapbp, Ciita, Klrk1, H2-T23, Ctsh, H2-DMb1, H2-DMa, Tap2, Pml, Tap1, Mr1, H2-DMb2, H2-Q10, Lag3, H2-M3, H2-D1, H2-K1, Cd40lg, Cd1d1, Nlrc5, H2-Q2
Complement Pathway	0.1262	0.0777	*0.0203	C1qb, C1qb, Cfd, C1s1, C4b, C5ar1, C3ar1, C1ra, C1qbp, Cfp, Cd59b, Cd55, Cfh, C2, Serping1, C3, Cfi, C9, Cfb, A2m
Interferon Pathway	*0.0184	*0.0483	**0.0024	Ifit1, Ifi44, Ulbp1, Ifit2, Ifit3, Ddx58, Irf8, Irf7, Ifi35, Ifih1, H2-Aa, H2-Ab1, Irgm2, Tbk1, Gbp5, Nlrc5, Ccr7, Ciita, Eomes, H60a, Ifi27, Ifnar1, Ifnar2, Ifngr1, Ifitm2, Ifna4, Tmem173, Fadd, Nos2, Sh2d1b1, Ifitm1, Ifna1, Cxcl16, Mavs
TLR Pathway	0.176	0.2237	*0.0187	Tlr8, Ticam2, Tlr1, Traf6, Nfkbia, Myd88, Mapkapk2, Prkce, Map3k7, Gfi1, Tlr3, Cd86, Tlr7, Tlr9, Tbk1, Irak1, Ticam1, Traf3, Tirap, Tlr6, Tlr2, Tlr4, Tlr5, Irak2, Irf3
Microglial Function	**0.0075	*0.0331	**0.0027	Cx3cr1, Nod2, Tlr7, Tlr6, Casp1
Basic Cell Function	*0.0169	0.3743	*0.0195	Isg15, Oasl1, Cmpk2, Emr1, Rsad2, Stat2, Rrad, Arg2, Usp18, Tank, Ythdf2, Oas2, Herc6, Oas3, Ncf4, Trem2, Ubc, Pdgfc, Map2k2, Dusp4, Arg1, Isg20, Ddx60, Hsd11b1, F13a1, Psmb7, St6gal1, Reps1, S100a8, Zfp13, Ewsr1, Mapk11, Ctsl, Pla2g6, Notch1, Raet1c, Chil3, Dock9, Tab1, Hcst

¹. One-way ANOVA with Dunnett’s post hoc analysis.

^*= p< 0.05,

^**=p<0.01 n= 3 biological replicates

3.3. T-cell independent anti-tumor effects of AKTi

Our previous studies with the pan-PI3K inhibitor, BKM120, showed that pan-PI3K inhibition enhanced the PyMT response to anti-PD1 therapy by recruiting CD8+ T-cells into the TME [14]. Given this previous finding, along with the complete tumor regression observed in 6 of 7 mice treated with AKTi+PTX+ICI in the PyMT model and the suggestion of immunological memory posed by the rechallenge study, we were interested in the role that cytotoxic T-cells may be playing a role in this response even though at the end point of our mouse experiments, few CD3+ cells were present. To specifically explore the contribution of CD8+ T-cells to this triple-therapy response, we conducted a CD8+ cell depletion study in the C57BL/6J immunocompetent mice with PyMT orthotopic tumors. Tumors were allowed to grow to a volume of 50mm³-100mm³ and then anti-CD8α (200ug, IP, 1x/3 days) or rat IgG2b control ab (200ug, IP, 1x/3 days) was administered three days prior to the start of therapeutic treatments (day −3). Anti-CD8α or rat IgG2b control ab were administered every three days for the remainder of the study. Therapeutic treatments were as follows: 1) V+IgG+IgG2b, 2)V+IgG+anti-CD8α, 3)IPAT+PTX+ICI (Combo) + IgG2b 4) Combo + anti-CD8α. Among the seven tumor-bearing mice that received treatment with IPAT+PTX+ICI, four mice exhibited a complete response (CR) 15-days after treatment initiation while the remaining three mice exhibited a partial response (PR) with either a 40% (n=2) or 97% (n=1) inhibition of tumor growth compared to vehicle+IgG treated mice post 24-days of treatment (Figure 3a). Tumor weight was also assessed at the experimental endpoint, and both combo and combo+anti-CD8α treatment groups exhibited significant tumor weight reductions compared to V+IgG+anti-CD8α (Figure 3b). While no tumors regressed with CD8+ depletion, IPAT+PTX+ICI treatment still partially inhibitied PyMT tumor growth (Veh+IgG+anti-CD8+ versus IPAT+PTX+ICI+anti-CD8+, p=0.012), indicating AKTi may also directly affect tumor growth, independent of CD8+ T-cells.

Figure 3. T-cells are not required to achieve a partial response to IPAT +/− ICI in the PyMT model.

a. Growth of individual PyMT tumors. Female C57BL/6J mice were injected with 100,000 PyMT cells into the 4^th mammary fat pad on either flank (n=7–9 per group). Anti-CD8α (200ug, IP, every 3 days) or rat IgG2b control ab (200ug, IP, every 3 days) was administered 3 days prior to treatment initiation (day −3) and continued every 3 days. Treatment with either V+IgG or Combo [Ipat (100mg/kg, PO, QD)+PTX (RO, 10mg/kg, qw)+ICI (anti-PD-1 (200ug)+anti-CTLA4 (100ug))] was initiated on day zero and continued for 24 days.

b. Tumor weight at endpoint of experiment from (a). Statistical analysis: One-way ANOVA with Tukey’s post-hoc analysis * = p <0.05, ** = p< 0.01, *** = p<0.001.

c. Tumor samples were dissociated for immune profiling by flow cytometric analysis. Live CD45+ leukocytes were concatenated after downsampling to ~25,000 events per each treatment group for t-SNE analysis.

d. Confirmation of CD8 depletion. Tumor samples were analyzed for CD8+ (T_c) cells gated on live, CD45+ cells. Activation status of CD8+ t-cells was analyzed by CD69 flow cytometry analysis. The combo treatment did not alter CD69%+ T_c population.

e. t-SNE plot of representative samples showing an induction of NK and T-cell populations in partially responding tumors.

f. MFI of CD69 staining in Tc and NK cell populations. While there was not a change in the percentage of CD69+ cells, there was an increase in the CD69 MFI per cell.

g. Growth of PyMT tumors in the Nu/J mouse model which lacks T-cells. Once tumors reached a volume of 50mm³-100mm³, mice were treated with vehicle (n=5) or ipatasertib (n=5) (100mg/kg, PO, QD). Tumor growth was analyzed with a mixed effects model and analyzed on the natural log scale to better meet normality assumptions. The mean and SE of tumor volume over time for each treatment group is shown. Tumor weight was measured at the experiment endpoint.

Complete tumor regression in the IPAT+PTX+ICI treatment group prevented endpoint analysis of immune cell populations in the complete responders, however the three remaining partial responders from this treatment group were evaluated with detailed t-distributed stochastic neighbor embedding (t-SNE) immune profiling using flow cytometry. Analysis confirmed the CD8-depletion antibody successfully eliminated the CD8+ population both in the vehicle+IgG+anti-CD8+ group (0.064%, 95% XI, −0.026, 0.153) and the IPAT+PTX+ICI+anti-CD8+ tumors (combo) (%0.097, 95% CI, −0.004, 0.239) (Figure 3d). The t-SNE analysis showed the treatment of IPAT+PTX+ICI significantly increased the frequency of CD8+ T cells (Tc) from 1.148% in the vehicle+IgG-treated tumors to an average of 8.902% in the two partially responding tumors (p=0.028, Figure 3c.d). Interestingly, the combo-treated tumor with the greatest inhibition in tumor growth (~97% reduction in tumor size) had a high CD8+ T-cell infiltrate (~20%) compared to the two tumors with a ~40% partial response (~2% CD8+ T-cell infiltrate) (Figure 3d,e). Additionally, compared to the vehicle+IgG treated tumors, there was an induction of NK cell content in the TME of the two 40% PR tumors treated with IPAT+PTX+ICI (Figure 3e) [37]. While the total percentage of CD8+ T-cells increased in the IPAT+PTX+ICI group compared to V+IgG control (Figure 3d), there was no significant difference in the percentage of CD69+ T-cells relative to the total CD8+ T-cell population. However, the mean fluorescence intensity (MFI) of CD69 on tumor-infiltrating T-cells was increased from MFI=284 in the V+IgG treated tumors to MFI = 1,372 and MFI=1,596 in the combo-treated tumors with a 40% PR and in the tumors with a 97% PR, respectively (Figure 3d,f). An earlier report showed that the intensity of early activation marker CD69 was rapidly up-regulated in CD8+ T cells upon stimulation and returned to naïve levels within six cell divisions, suggesting CD69 could be used to evaluate the proliferative potential of T-cells upon activation [37]. In addition to serving as a T-cell activation marker, CD69 is also required for activated NK cell-mediated lysis and killing of tumor targets [38]. CD69 intensity in the tumor-infiltrating NK cells was increased from MFI=215 in the vehicle+IgG treated tumors to MFI=296 and MFI=358 in the combo-treated tumors with a 40% PR versus tumors with a 97% PR, respectively (Figure 3f). These data reinforce the notion of a causal relationship between treatment-induced CD8+ content in the TME and the therapeutic benefit induced by IPAT+PTX+ICI. However it is also notable that even in the absence of a substantial T-cell infiltrate, IPAT+PTX+ICI can still evoke a partial anti-tumor response (Figure 3d,e).

Given the partial response of PyMT tumors to IPAT+PTX+ICI even in the presence of a CD8 depleting antibody, we sought to further probe the immune-cell independent effects of AKTi through using the athymic nude mouse (Nu/J) model. This model lacks all T-cells due to a mutation that causes defective thymic development [39]. Eight-week-old female Nu/J mice were orthotopically implanted with PyMT tumors and when the tumors became palpable (~50mm³) mice were treated with IPAT daily (100mg/kg, QD) by oral gavage. After 11-days of treatment, IPAT treatment led to a statistically significant decrease in tumor growth compared to vehicle-treated mice (p=0.027) (Figure 3g), further confirming that anti-tumor effects of AKTi can occur in the absence of T-cells.

3.4. AKT inhibition is not effective in three additional TNBC-like mouse models

TNBC is a heterogenous disease and response to therapy varies with the subtype of TNBC and the genetic profile of the tumor [2]. Thus, it is important to examine the response to potential new therapeutic combinations in more than one preclinical model. To accomplish this, we evaluated the effects of IPAT alone and in combination with PTX and/or ICI in three additional TNBC-like mouse models, 6DT1(FVB mice), E0771 (C57BL/6J mice), and 4T1 (BALB/cJ mice). The 6DT1 and E0771 models were previously characterized as luminal, claudin-low, while the 4T1 models is characterized as luminal, high proliferation [40]. We first determined the optimal dosage of cells to deliver to mice to ensure sufficient time to evaluate therapeutic response. Based upon these results, 6DT1 (5×10⁴), E0771 (3×10⁴), or 4T1 (5×10⁴) cells were orthotopically implanted into the fourth mammary fat pad of 8–10 week-old female FVB, C57BL/6J, or BALB/cJ mice, respectively with a minimum of 5 mice per treatment group. Treatments were initiated when tumors measured between 50mm³ and 100mm³. IPAT was ineffective in suppressing tumor growth of these three additional murine tumor models tested, 6DT1, E0771, and 4T1 (Figure 4a–c). Despite not decreasing tumor growth, IPAT did increase levels of pAKT as measured by western blot analysis of tumor lysates, in accordance with IPAT’s known mechanism of action whereby phosphorylated AKT is not subjected to phosphatase activity (Figure 4d) [41]. Notably, vehicle treated PyMT tumors have higher levels of pAKT than vehicle treated 6DT1, 4T1, or E0771 tumors (Figure 4d). However, pPRAS40, a direct downstream target of AKT, was only downregulated upon IPAT treatment in the PyMT model, suggesting the activation of compensatory pathways in the 4T1, 6DT1, and E0771 models (Figure 4d). While IPAT alone was ineffective in the three additional models, ICI alone significantly reduced tumor growth in the 6DT1 model compared to Vehicle+IgG (p<0.001) (Figure 4a); however, the addition of IPAT did not further enhance the ICI response. In the 6DT1 model there is variability in the ICI response, and in a repeat of the experiment, ICI failed to decrease growth, and the addition of ipatasertib did not alter ICI sensitivity (Figure S4). IPAT treatment failed to alter the sensitivity to ICI in the 4T1 and E0771 models as well (Figure 4b,c). While the E0771 tumors were largely non-responsive to AKTi and/or ICI, one mouse in the AKTi+ICI group exhibited tumor regression, suggesting potential heterogeneity in the response among individual tumors. However, this response is likely driven by a response to ICI rather than AKTi, as there is also mouse-to- mouse variability in ICI response in the E0771 model [40]. Detailed statistical comparisons between all treatment groups for each study can be found in Table S1.

Figure 4. AKTi is not effective in three additional TNBC-like mouse models with low pAKT.

Tumor growth curves plotted on a log axis and corresponding tumor weight at experiment endpoint for each of the three additional models tested. For experiments a-c, tumor growth was analyzed with a mixed effects model and analyzed on the natural log scale to better meet normality assumptions. The mean and SE of tumor volume over time for each treatment group is shown. n= 5–6 mice per treatment group. p-values for comparisons of tumor growth over time adjusted using Holm-Bonferonni method. *=p<00.05, **=p<0.01, ***=p<0.001

a. Growth of 6DT1 tumors in response to various treatment combinations: (1) Vehicle(V)+IgG, (2) IPAT (PO, 100mg/kg, QD), (3 ) V + ICI (anti-PD-1(250ug) and anti-CTLA-4 (100ug), IP, every 3 days), (4) PTX (RO, 10mg/kg, qw) + ICI (5) IPAT+ICI, (6) IPAT + PTX + ICI

b. Growth of 4T1 tumors in response to various treatment combinations: (1) Vehicle(V)+IgG, (2)PTX (RO, 10mg/kg, qw) + IgG, (3) IPAT (PO, 100mg/kg, QD) PTX (RO, 10mg/kg, qw)+ IgG, (3 ) PTX+ ICI (anti-PD-1(250ug) and anti-CTLA-4 (100ug), IP, every 3 days), (4) IPAT+ PTX + ICI, (5) IPAT + PTX + IgG, or (6) IPAT + PTX + ICI

c. Growth of E0771 tumors in response to various treatment combinations: (1) Vehicle(V)+IgG, (2)PTX (RO, 10mg/kg, qw) + IgG, (3) IPAT (PO, 100mg/kg, QD) PTX (RO, 10mg/kg, qw)+ IgG, (3 ) PTX+ ICI (anti-PD-1(250ug) and anti-CTLA-4 (100ug), IP, every 3 days), (4) IPAT+ PTX + ICI, (5) IPAT + PTX + IgG, or (6) IPAT + PTX + ICI

d. Western blot analysis of tumor lysates from 4T1, PyMT, 6DT1, or E0771 vehicle or IPAT treated tumors probing for pAKT(S473)/pan-AKT and pPRAS40. For each model (PyMT, 4T1, 6DT1, E0771), IPAT was compared to Veh treatment using an unpaired t-test * = p < 0.05, ** = p <0.01, ***p=<0.001. Levels of pAKT(S473)/pan-AKT from vehicle treated 4T1, 6DT1, and E0771 tumors were compared to vehicle treated PyMT tumors using a Brown-Forsythe and Welch ANOVA test.

3.5. Genetic differences and pAKT levels in PYMT, 6DT1, EO771 and 4T1 models influence response to AKT inhibition

Given the in vivo variability of response to IPAT across our four models, we performed whole-exome sequencing on the PyMT, 6DT1, E0771, and 4T1 cell lines to understand the genetic diversity. The PI3K pathway mutational status of each cell line is of particular interest since results from the LOTUS trial suggested that patients with PI3K pathway mutations may benefit from treatment with the AKTi, IPAT [42]. Analysis of 98 PI3K pathway genes revealed a distinct difference in the number of PI3K pathway-related genes. The PyMT cell line has eight mutations in the PI3K pathway whereas the 6DT1, E0771, and 4T1 have 31, 21, and 31 mutations in this pathway, respectively (Figure 5a). Of note, the 6DT1 cell line has a mutation in Pik3ca, the catalytic subunit of PI3K (Figure 5a), and none of the cell lines have a mutation in Pten. A complete list of PI3K pathway mutations can be found in Table S2. Interestingly, the frequency of PI3K pathway mutations does not correlate with the in vivo or in vitro sensitivity to IPAT the models tested in the present study. Despite having few PI3K pathway mutations, the PyMT cell line is the most sensitive to AKTi both in vitro, with an IC50 of 10.04uM (Figure 5b), and in vivo where reductions in tumor growth were observed with single agent IPAT (Figure 1a). Conversely, the 4T1 cell line has the most PI3K pathway mutations, yet has an IC50 more than double that of the PyMT cell line (23.36μM) (Figure 5b), and did not respond to IPAT in vivo (Figure 4b). Toxicity of IPAT to normal breast epithelia has previously been characterized in the MCF10a cell line which is described as IPAT resistant with a 39% maximum reduction of cell viability at 20μM [27]. Furthermore, IPAT significantly reduced proliferation of the PyMT cell line by >3-fold after 48hr and >4-fold after 72hrs of treatment (Figure 5c). Modest decreases in proliferation were observed in the 6DT1, E0771, and 4T1 cell lines after 48hrs of IPAT treatment, but this effect was lost by 72hrs (Figure 5c). However, the lack of genetic mutation does not exclude the possibility of aberrant pathway activation. Therefore, we analyzed baseline levels of PI3K pathway activation in the cell lines by based upon quantitation of pAKT by western blot. Despite having no genetic mutations in the PI3K pathway, the PyMT cell line has the highest basal levels of pAKT (Figure 5d). However, in accordance with the mechanism of action whereby pAKT levels are increased upon treatment with the ATP-competitive AKT inhibitors, IPAT and CAPI, in all four cell lines (Figure 5d). The mechanism of action of IPAT involves inhibiting AKT kinase activity while stabilizing and enriching the pAKT in a conformation in which the phosphorylated AKT residues are inaccessible to phosphatases [41].

Figure 5. Genetic differences and pAKT levels in PyMT, 6DT1, E0771, and 4T1 cell lines influence response to AKTi.

a. PI3K pathway mutational status of four TNBC-like cell lines. Venn diagram depicting number of nonsynonymous mutations in PI3K pathway genes per cell line. PI3K pathway genes identified from a list 96 well described genes as described in the methods.

b. IC50 of the four cell lines to 72hr treatment with ipatasertib. n=3 biological represents and error bars represent SD.

c. Cells were loaded with CellTrace Deep Red (CTDR) proliferation dye and treated with DMSO or 10 μM IPAT for 48hr or 72hrs. Mean fluorescent intensity (MFI) was measured by flow cytometry to track cell proliferation. Data shown as fold-change relative to DMSO treated control for each cell line at each time point. Statistical analysis: Student’s t-test *=p<0.05, **=p<0.01, ****=p<0.0001

d. Western blot analysis of endogenous levels of pAKT(S473) and levels of pAKT in response to 24hr treatment with either ipatasertib (1μM) or capivasertib (1μM). Statistical analysis: Student’s t-test * = p<0.05, ** = p<0.01, *** = p< 0.0001

e. Western blot analysis of changes in expression of pAKT over time in response to ipatasertib treatment (1uM) in the 4T1 and PyMT cell lines. n=3 biological replicates. Statistical analysis: One-way ANOVA with Tukey’s post hoc analysis *=p<0.05, **=p<0.01, ***=p<0.001

f. Western blot analysis of changes in expression AKT phosphorylation targets in response to ipatasertib treatment (1uM). N=3 biological replicates. Statistical analysis: One-way ANOVA with Tukey’s post hoc analysis *=p<0.05, **=p<0.01, ***=p<0.001

The time course of AKT inhibition was further explored in vitro using the PyMT and 4T1 cells lines as models of high and low basal pAKT expression, respectively. Upon IPAT treatment, PyMT cells significantly increased the levels of pAKT-S473 and pAKT-T308 as soon as 5 min post-treatment and the elevated level of pAKT was maintained for at least 24hrs (Figure 5e). In the 4T1 cells, an increase in the level of pAKT was observed 60 min after IPAT treatment (Figure 5e). Multiple substrates downstream of the AKT/mTOR pathway, including p70S6K, S6, 4EBP1, FoxO3a, GSK3β, and PRAS40 have been shown to play a role in the activation of cell growth, migration, metabolism, survival, and proliferation [44–46]. By probing the level of each of these proteins in the cultured PyMT and 4T1 cells by western blot, we found that the PyMT cells exhibited a significantly higher level of intrinsic phospho-p70S6K, phospho-S6, phospho-FoxO3a, GSK3β and pPRAS40 and PRAS40 compared to 4T1 cells (Figure 5f). Importantly, IPAT treatment resulted in a more rapid (5–30min) downregulation of phospho-4EBP1, phospho-GSK3β and phospho-FoxO3a in PyMT cells, compared to 4T1 cells (60min to 24hrs). The more rapid action of IPAT in the PyMT cells compared to the 4T1 cells is likely explained by the fact that IPAT binds to pAKT, and since basal pAKT is high in the PyMT cell line, there is more substrate available for IPAT to bind. The kinetics of IPAT-induced downregulation of phosho-p70S6K, phospho-S6, and phospo-PRAS40 was comparable between the PyMT and 4T1 cells. In addition, there was a significant reduction on the total PRAS40 protein over the 24hr timecourse and an increase in total 4E-BP1 protein (Figure 5f). When 4E-BP1 is unphosphorylated, it remains bound to eIF4E, preventing eIF4E-initiated cap-dependent translation [47]. This decrease in translation through this mechanism could be responsible for the observed decrease in total PRAS40, however the data are inconclusive as to whether or not PRAS40 translation is eIF4E dependent. 4E-BP1 in its unphosphorylated state is also subject to ubiquitin mediated degradation [48], and it is possible that we are observing a compensatory upregulation in total 4E-BP1 while observing a decrease in phosphorylated 4E-BP1.

3.6. Immunological profile of AKTi+ICI non-responsive models

In the PyMT model, there was a trend toward increased T-cell infiltrate associated with a greater tumor volume reduction (Figure 2c), the strongest partial responders to AKTi+PTX+ICI exhibited increased T-cell infiltrates (Figure 3e), and CD8⁺ T-cell responses are required for a CR in AKTi+PTX+ICI treated tumors. Therefore, we wanted to determine how AKTi+/− ICI affected T-cell infiltration in these three AKTi, non-responsive TNBC models. Immunohistochemistry staining for CD3+ T cells in paraffin-embedded tumor tissue sections revealed that AKTi plus ICI failed to significantly increase T-cell frequency in 4T1, 6TD1, and E0771 tumors (Figure 6a). Furthermore, cytokine analysis was performed on tumor lysates to characterize the baseline cytokine profile of PyMT, 6TD1, E0771, and 4T1 tumors. Compared to vehicle-treated PyMT tumors, vehicle treated 4T1, 6DT1, and E0771 tumors have significantly higher levels (defined as greater than a 1.5-fold increase) of IL-6, IL-10, IL-13, and IL-17 (Figure 6b), suggesting that the three IPAT non-responsive models may have a more immunosuppressive barrier to overcome than does the PyMT model. While IL-6, IL-10, IL-13, and IL-17 may play varied roles in tumor progression, they are broadly considered immunosuppressive cytokines [49–51] [52]

Figure 6. Immunological profile of AKTi+ICI non-responsive models.

a. Paraffin embedded tumor tissue sections from 4T1, 6DT1, and E0771 mice treated with V+IgG, Ipat+IgG, or Ipat+ICI, as described in Figure 4, were analyzed for CD3 infiltrate by IHC staining (brown). Representative images are shown, and the scale marker is 150μm. Quantification of CD3 IHC calculated as %CD3+ nuclei using QuPath digital analysis software.

b. Lysates from vehicle-treated tumors were analyzed for an array of cytokines. When possible, up to 5 tumors were pooled per group, and the assay was run in duplicate and normalized to an internal control as described in the Methods. Data are shown as fold-change relative to PyMT samples. Fold-change greater than 1.5 as indicated by the horizontal dotted line are considered significant. IL-6, IL-10, IL-13, and IL-17 are significantly upregulated in 4T1, 6DT1, and E0771 tumors compared to PyMT tumors.

c. Pathway signature scores summarizing expression levels of related groups of genes from nanostring nCounter gene expression platform. Statistical analysis: One-way ANOVA with Dunnett’s post hoc analysis. From left to right, statistical comparisons represent V+IgG compared to IPAT+IgG, V+IgG compared to V+ICI, and V+IgG compared to IPAT+ICI. * = p< 0.05, ** = p<0.01. n= 3 biological replicates

d. Cell type signature scores summarizing expression levels of related groups of genes from nanostring nCounter gene expression platform. Statistical analysis: One-way ANOVA with Dunnett’s post hoc analysis. From left to right, statistical comparisons represent V+IgG compared to IPAT+IgG, V+IgG compared to V+ICI, and V+IgG compared to IPAT+ICI. * = p< 0.05. n = 3 biological replicates

To further understand the immunological profile of a non-responsive tumor, nanostring nCounter^® gene expression analysis was performed on RNA isolated from 4T1 tumors from the study shown in Figure 4b. The 4T1 samples were chosen for further analysis as an example of a tumor that consistently did not respond to either AKTi or ICI. Pathway enrichment analysis was performed on the panel of 770 genes in the mouse Pancancer Immune Profiling panel, and of the six pathways significanlty upregulated in AKTi+ICI treated PyMT tumors (Figure 2a), all but two are unchanged by AKTi, ICI, or AKTi+ICI in the 4T1 model. Two pathways – Interferon and TLR – are downregulated in the 4T1 model in response to treatment (Figure 6c). While unaffected in the responsive PyMT model, the NK-cell function score is downregulated in AKTi+IgG treated 4T1 tumors compared to V+IgG treated tumors (Figure 6c). Genes used to define the pathways can be found in Table 2. Interestingly, the cytotoxic cell score, defined by Gzmb, Ctsw, Klrd1, Gzma, and Prf1 transcript abundance was significantly downregulated in the IPAT+IgG and IPAT+ICI groups compared to V+IgG in the 4T1 model (Figure 6d). Our results suggest that while IPAT may have a modest beneficial effect on the recruitment of T-cells, it is not enough to turn a tumor that is not responsive to ICI, into an ICI-responsive tumor. However, AKTi may further extend the response to ICI therapy if the tumor exhibits significant constitutive AKT activity, as evidenced by the PyMT response.

Table 2.

4T1 pathway signature score gene list from nCounter analysis

Pathway	p-value1 V+IgG v. V+IPAT	p-value V+IgG v. V+ICI	p-value V+IgG v. IPAT+ICI	Gene names
Antigen Processing	0.3059	0.9944	0.0777	H2-Aa, Cd74, H2-Ob, H2-Ab1, H2-Eb1, Fcgr3, Fcer1g, Slc11a1, Fcgr2b, Fcgr1, Tapbp, Icam1, Relb, H2-T23, H2-DMb1, H2-DMa, Nod2, Tap2, Psmb8, Tap1, Mr1, Ccr7, H2-DMb2, H2-Q10, Psmb9, H2-M3, H2-D1, H2-K1, Cd1d1, Nod1, H2Q2
MHC Pathway	0.2839	0.9792	0.0876	H2-Aa, Cd74, H2-Ob, H2-Ab1, H2-Eb1, Fcgr3, Fcer1g, Fcgr2b, Fcgr1, Tapbp, Ciita, Klrk1, H2-T23, Ctsh, H2-DMb1, H2-DMa, Tap2, Pml, Tap1, Mr1, H2-DMb2, H2-Q10, Lag3, H2-M3, H2-D1, H2-K1, Cd40lg, Cd1d1, Nlrc5, H2-Q2
Complement Pathway	0.6162	0.1995	0.8685	C1qb, C1qb, Cfd, C1s1, C4b, C5ar1, C3ar1, C1ra, C1qbp, Cfp, Cd59b, Cd55, Cfh, C2, Serping1, C3, Cfi, C9, Cfb, A2m
Interferon Pathway	**0.0042	0.2717	*0.0114	Ifit1, Ifi44, Ulbp1, Ifit2, Ifit3, Ddx58, Irf8, Irf7, Ifi35, Ifih1, H2-Aa, H2-Ab1, Irgm2, Tbk1, Gbp5, Nlrc5, Ccr7, Ciita, Eomes, H60a, Ifi27, Ifnar1, Ifnar2, Ifngr1, Ifitm2, Ifna4, Tmem173, Fadd, Nos2, Sh2d1b1, Ifitm1, Ifna1, Cxcl16, Mavs
TLR Pathway	0.414	>0.9999	*0.0331	Tlr8, Ticam2, Tlr1, Traf6, Nfkbia, Myd88, Mapkapk2, Prkce, Map3k7, Gfi1, Tlr3, Cd86, Tlr7, Tlr9, Tbk1, Irak1, Ticam1, Traf3, Tirap, Tlr6, Tlr2, Tlr4, Tlr5, Irak2, Irf3
Microglial Function	0.718	0.531	0.718	Cx3cr1, Nod2, Tlr7, Tlr6, Casp1
Basic Cell Function	**0.0025	0.253	0.0738	Isg15, Oasl1, Cmpk2, Emr1, Rsad2, Stat2, Rrad, Arg2, Arg2, Usp18, Tank, Ythdf2, Oas2, Herc6, Oas3, Ncf4, Trem2, Ubc, Pdgfc, Map2k2, Dusp4, Arg1, Isg20, Ddx60, Hsd11b1, F13a1, Psmb7, St6gal1, Reps1, S100a8, Zfp13, Ewsr1, Mapk11, Ctsl, Pla2g6, Notch1, Raet1c, Chil3, Dock9, Tab1, Hcst
NK Cell Function	* 0.0463	0.9988	0.0891	Ccl3, Ccl4, Il12a, Il12b, Mertk, Ccl7, Il23a, Il21, Stat5b, Cd7, Ccl2, Flt3l, Itgb2, Ikzf1, Ncam1, Il2rb, Klrd1, Lag3, Slamf7, Klrc1, Klra1, Klra17, Klra27, Klra5, Klrb1c, Klra3, Klra20, Klra7, Klra15, Axl, Cd247, Ccl5, Cd244, Il11ra1, H2-M3, Ubp1, Sh2d1a, Klrk1, Klra6, Klra21, Il15, Il15ra, Il12rb1, Cd96, H60a, Klra4, Pvrl2, Klra2, Pvr, Mil2, Cd2, Sh2d1b1, Ifi27

¹. One-way ANOVA with Dunnett’s post hoc analysis.

^*= p< 0.05,

^**=p<0.01 n= 3 biological replicates

3.5. MAPK pathway activation effect on response to AKT inhibition

In the initial phase I clinical trial of IPAT, it was suggested that aberrant activation of the RAS/MAPK may be associated with non-responsive tumors. KRAS or NRAS mutations were associated with decreased time on the study, and pERK was upregulated in response to IPAT treatment [23]. Therefore, we sought to characterize the RAS/MAPK mutational profile of our non-responding models. Analysis of whole-exome sequencing data from each of the four TNBC-like cell lines studied here revealed differences in the number of MAPK pathway mutations among cell lines (Figure 7a). Notably, the 6DT1 cell line has a Kras(G12C) mutation and the E0771 cell line has both an Nras (Q61H) and a Kras (G12C) mutation (Figure 7a). The list of all MAPK pathway mutations can be found in Table S3. Compared to the PyMT cell line, the two ras mutant cell lines, the 6DT1 and E0771, have elevated basal pERK as measured by western blot analysis (Figure 7b). Consequenctly, both 6DT1 and E0771 cells are sensitive to MEK inhibition alone and in combination with AKT inhibition in vitro (Figure 7c). Therefore, the MEK inhibitor (MEKi), selumetinib was tested in combination with the AKT inhibitor, capivasertib (CAPI), and ICI in the 6DT1 orthotopic tumor model. The 6DT1 model was chosen because despite having a Pik3ca mutation, it has low basal pAKT as measure by western blot and does not respond to AKTi in vivo. 6DT1 cells (1×10⁵) were injected into the fourth mammary fat pad of 8–10 week old female FVB mice. Treatments were as follows: (1) V+IgG, (2) CAPI(130mg/kg, PO, BID 4 of 7 days)+IgG, (3) MEKi (100mg/kg, PO, 5x/week) +IgG, (4) CAPI+MEKi +IgG, (5) V+ICI (anti-PD-1, 250ug, and anti-CTLA4, 100ug, IP, once every3 days), (6) MEKi +ICI, (7) CAPI+ICI, and (8) CAPI+MEKi+ICI. Selumetinib (MEKi) significantly decreased tumor growth and tumor weight compared to both V+IgG Control (p<0.001) and V+ICI (p<0.001), but there was no further benefit by the addition of either AKTi or ICI to the MEKi (Figure 7d&e). Adding the MEKi to the AKTi also added substantial toxicity to the treatment regimen. Mice treated with the combination of capivasertib and selumetinib had a lower survival rate with or without ICI (Figure S5). Blood serum was collected from surviving mice at the experimental endpoint and kidney and liver enzymes were analyzed as another measure of toxicity (Figure S5). While adding a MEKi did not alter the sensitivity to the AKTi in the 6DT1 model, treatment with single-agent MEKi did significantly decrease tumor growth, suggesting that the MEK pathway is the dominant driver of tumorigenicity in the 6DT1 tumors.

Figure 7. MEK inhibition in combination with AKT inhibition in Ras mutant models.

a. MAPK pathway mutational status of four TNBC-like cell lines. Number of nonsynonymous mutations in MAPK pathway genes per cell line. MAPK pathway genes identified from a list 92 well described genes as described in the methods. Select genes with functional relevance to MAPK signaling in cancer are represented here as examples.

b. In vitro effect of AKTi (1μM ipatasertib) on protein expression of pERK1/2 and total ERK1/2 as measured by western blot. Quantitation of western blot analyzed with one-way ANOVA and Tukey’s post hoc analysis ****=p<0.0001.

c. In vitro effect of AKTi (1μM capivasertib), MEKi (1μM selumetinib), or AKTi+MEKi) on cell viability as measured by crystal violet assay after 24hrs of treatment in the E0771 (Kras Q12C & Nras Q61H) and 6DT1 (Kras G12C) cells lines ( n=3 biological replicates). Statistical analysis: One-way ANOVA with Tukey’s post hoc analysis * = p<0.05, *** = p <0.0001. Representative 10x image of crystal violet staining prior to solubilization and quantitation.

d. in vivo efficacy of combining AKT inhibition (capivasertib) with MEK inhibition (selumetinib) and/or ICI. Female FVB mice were implanted with 6DT1 tumor cells and treatments were initiated when tumors reached ~50mm^3., (n =7 mice per group). Treatments were as followed: (1) Vehicle + IgG, (2) capivasertib (130mg/kg, PO, BID 4 of 7 days )+ IgG, (3) selumetinib (100mg/kg, PO, 5 days of 7 ) + IgG, (4) capivasertib + selumetinib (Combo) + IgG, (5) Vehicle + ICI (anti-PD-1 (200ug) and anti-CTLA4(100ug)), IP, every 3 days, (6) selumetinib + ICI, (7) capivasertib + ICI, (8) Combo + ICI. Statistical analysis: Tumor growth was analyzed with a mixed effects model and analyzed on the natural log scale to better meet normality assumptions. The mean and SE of tumor volume over time for each treatment group is shown. p=values adjusted with Holm-Bonferonni method *** = p < 0.001

e. Tumor weight at the endpoint of experiment from (b). Statistical analysis: Student’s t-test compared to corresponding control. * = p <0.05, ** = p <0.01.

4. Discussion

Our studies sought to distinguish the properties of TNBC tumors that may benefit from AKTi and whether combining an AKTi with ICI might result in enhanced inhibition of tumor growth. We analyzed the response to AKTi+/− ICI in four TNBC-like murine models: PyMT, 6DT1, E0771, 4T1 and determined that the endogenous pAKT levels are associated with response to AKTi in our models (Figure 8). Though PI3K pathway mutational status of a tumor may result in enhanced pAKT levels, confounding mutations in other pathways appear to prevent endogenous elevations in pAKT and result in insensitivity of the tumor to AKT inhibitors. This point is further illustrated in the results from the FAIRLANE trial of neoadjuvant IPAT plus PTX, where RPPA analysis of patient tumors revealed that pAKT levels were associated with clinical benefit regardless of PI3K mutational status, even though the overall patient population did not show pCR in response to IPAT plus PTX [29]. Baseline AKT activity as measured by pAKT levels was also found to correlate with sensitivity to the AKT inhibitor, capivasertib, in a study of HER2− breast cancer patient-derived xenografts [53]. The LOTUS clinical trial with IPAT specifically examined the relationship between PI3K pathway mutation and response and showed that mutational status was not a clear-cut indicator of pathway activity [42]. For example, it was previously shown that despite the frequency of PI3K pathway alterations in breast cancer, PI3K mutation does not always correlate with enhanced signaling through AKT [54]. In fact, PI3Kα mutant tumor cells often exhibit diminished AKT signaling because of preferential signaling of PDK1 through SGK3 rather than through AKT [55]. Nevertheless, clinical interest remains for stratifying patients by PI3K/AKT mutational status. In an ongoing clinical trial of IPAT plus atezolizumab in solid tumors, PI3K/AKT hyperactivity as determined by the presence of activating mutations in PIK3CA, AKT1, AKT2 or PTEN loss as determined by IHC is being used as inclusion criteria (NCT03673787). An additional study is examining IPAT in combination with PTX for treating non-breast solid tumors with genetic alterations in AKT (NCT05554380). However, our study described herein suggests that pAKT levels may be more predictive of a PI3K pathway addiction than PI3K pathway mutational status alone, and TNBC tumors with PI3K hyperactivity, as measured by endogenous pAKT activity, regardless of PI3K mutation, may benefit from treatment with an AKT inhibitor. However, we do not rule out the possibility that other pathway mutations might interrupt tumor response to AKT inhibition.

Figure 8. Proposed signaling schematic of AKTi+ICI responsive and AKTi+ICI non-responsive models.

Proposed characteristics of the AKTi responsive, PyMT model, and the AKTi non-responsive, E0771, 4T1, and 6DT1 models. Competitive inhibitors of ATP such as ipatasertib and capivasertib bind the ATP-binding pocket and stabilize the phosphorylated conformation of AKT while inhibiting the kinase activity of AKT. pAKT levels associate with response to AKTi and AKTi+ICI. Dotted lines represent translocation to the nucleus. Arrows with blunted ends represent inhibitory interactions while pointed arrows represent activating interactions. Thicker lines represent greater pathway involvement. MAPK pathway mutations may contribute to resistance to AKTi and AKTi+ICI in the AKTi non-responsive E0771, 6DT1, and 4T1 models

RAS mutations and upregulation of pERK in response to AKTi have been suggested as factors that mediate AKTi resistance [23], and there has been interest in combining AKTi with MEKi [56]. For example, the AKT and ERK signaling pathways convergently regulate the translation initiation factor 4E-BP1, and pathway activation in one is associated with resistance to inhibition of the other [47]. In its unphosphorylated state, 4E-BP1 remains bound to eIF4E, preventing the initiation of eIF4E cap-dependent translation [47]. However, the MAPK pathway can activate mTORC1, bypassing the effects of AKT inhibition to phosphorylate 4E-BP1 and allow the downstream initiation of eIF4E cap-dependent translation [57] (Figure 8). There is also evidence suggesting that activating mutations in KRAS may induce proviral integration site for Moloney murine leukemia virus-1 (PIM-1) expression [58], a Ser/Thr kinase that phosphorylates PRAS40 independent of AKT that has been implicated in resistance to AKT and PI3K inhibitors [59,60]. The 6DT1 model is not responsive to AKTi in vivo and even though it carries a PIK3CA mutation, it has low endogenous levels of pAKT. However, the 6TD1 cell line does have a KRAS mutation and responded well to single-agent MEKi in our study. Nonetheless, MEK inhibition failed to sensitize the 6DT1 tumors to AKTi, and there was no further benefit of adding ICI therapy to MEK inhibition. Despite evidence that MAPK/ERK signaling may contribute to AKTi resistance, combining MEKi + AKTi is not sufficiently effective to justify the considerable clinical toxicities that have been observed with the combination [61,62]

Given the ongoing clinical trials testing AKTi in combination with ICI (NCT03742102 and NCT04177108), as well as previous studies suggesting AKTi can increase CD8+ T-cells [21], we evaluated the combination of AKTi plus ICI in PyMT, 6DT1, 4T1 and E0771 models. AKTi in combination with ICI in the ICI-sensitive, pAKT high, PyMT model resulted in greater tumor inhibition than with either single treatment group. Cell death through apoptosis can be immunogenic [63,64] and this may explain how IPAT enhances the ICI response in the PyMT model. However, AKTi was not effective either as a single agent and did not enhance response to ICI or PTX in the pAKT low, 4T1, 6DT1, and E0771 tumors. There is considerable variability as to whether 6DT1, 4T1, and E0771 tumors respond to single-agent ICI. Thus, these tumors are not categorized as ICI-resistant models [65–68]. While all four models exhibited low levels of T-cell infiltrate, it is notable that vehicle-treated tumors from the 4T1, 6DT1, and E0771 models exhibited a greater CD3 infiltrate than vehicle-treated PyMT tumors (Figure 6a), indicating variability in the immune microenvironments of these different models. It is well documented that different mouse strains have different immunological characteristics[69,70], and in these studies we used mice on the FVB background (6DT1), the BALB/cJ background (4T1), and the C57BL/6J background (E0771 and PyMT). It also has been shown previously that different murine models of TNBC exhibit distinct intratumoral immune cell profiles [71]. In the present studies, AKTi did not significantly inhibit tumor growth or alter the tumor’s sensitivity to ICI.

It should be noted that while the strongest effect was seen in the PyMT group treated with AKTi+ICI, there was also significant inhibition of tumor growth with IPAT plus anti-PD-1 or with IPAT + anti-CTLA4 (Figure 2a). Either of these checkpoint inhibitors combined with IPAT shows inhibition of tumor growth equivalent to that of IPAT combined with both anti-PD-1 and anti-CTLA4, so the toxicity associated with combination anti-PD-1 and anti-CTLA4 may be avoided. Anti-PD-1 is currently the only FDA-approved checkpoint inhibitor for treating breast cancer. While anti-CTLA4 is under clinical investigation, the additional toxicity of including anti-CTLA4 with anti-PD-1 therapy is a major concern [72–74].

The effects of AKTi are partially mediated by T-cells, but there is a strong tumor-cell intrinsic effect at play, as evidenced by the anti-tumor effects of IPAT in the PyMT athymic nude mouse model and the partial response seen in the CD8+ cell depletion study. Many TNBCs, particularly those of the mesenchymal (M) and luminal androgen receptor (LAR) subtypes, exhibit low populations of tumor-infiltrating lymphocytes (TILs) [11,47][75]. Our data support the idea that low-TIL tumors may still benefit from AKTi provided that the tumor exhibits moderately high levels of endogenous pAKT. However, we cannot rule out the possibility of AKT inhibition effecting either early T-cell or NK cell recruitment. Loss of PI3K 110 delta in cytotoxic CD8+ T-cells was previously shown to increase expression of genes related to cell migration, particularly lymphocyte movement to and from lymphoid tissue [76]. Inhibiting AKT, an effector of PI3K signaling, may similarly affect expression of chemotactic factors that increase CD8+ T-cell migration into the tumor at early timepoints. Furthermore, t-SNE analysis of leukocytes in PyMT tumors growing in C57Bl/6 mice revealed that tumors with a partial response to AKTi+ICI exhibited an increase in the small population of NK cells (Figure 3e). These NK cells may be involved in the anti-tumor immune response in immune competent as well as in the T-cell deficient nude mouse model (Figure 3g).

We show here that high baseline pAKT levels may predict response to single-agent AKTi in TNBC tumors, and AKTi combined with ICI may have greater benefit than either agent alone for some TNBCs that are ICI-sensitive and express high pAKT. Overall, these studies support the idea that proteomic analysis of pAKT may be an indicator of which TNBC tumors are likely to benefit from AKTi when combined with genomic analysis. This work also supports a larger hypothesis that the addition of proteomic analyses to existing genetic analyses may better predict which targeted therapies will be most effective for a particular TNBC.

Supplementary Material

Supplementary Material

Figure S1. No statistical interaction between IPAT and ICI in the PyMT model

The experiment shown in figure 1a was examined for synergy between ipatasertib and ICI. The interaction plot for tumor growth rates with 95% confidence interval by treatment is shown. There is not sufficient evidence to show an interaction effect between the two treatments. That is, neither synergy nor antagonism occurs between ipatasertib and ICI.

Figure S2. Necrotic tumor tissue area in PyMT tumors

Representative whole-tumor image of caspase 3 IHC staining (red) from figure 1c. This tumor was from the IPAT+ICI treatment group. Tumor area inside of the yellow line is considered viable tissue. Tissue not within the yellow lines is considered necrotic and is excluded from analysis.

Figure S3. Low infiltration of CD8+ T-cells in PyMT tumors

Paraffin embedded tumor sections from the study shown in figure 1a were assessed by IHC for infiltration of CD8+ cells (red). Arrows indicate examples of positively staining cells.

Figure S4. IPAT does not alter the sensitivity of the 6DT1 model to ICI

No significant differences in tumor growth were observed between any of the treatment groups evaluated in this experiment.

Tumor growth curves plotted on a log axis and corresponding tumor weight at experiment endpoint for each of the three additional models tested. For experiments a-c, tumor growth was analyzed with a mixed effects model and analyzed on the natural log scale to better meet normality assumptions. The predicted mean and SE of tumor volume over time for each treatment group is shown. N = 5–6 mice per treatment group. Treatment groups include: (1) V+IgG, (2)PTX (RO, 10mg/kg, 1x/5 days) + IgG, (3) IPAT (PO, 100mg/kg, QD) PTX (RO, 10mg/kg, 1x/5 days)+ IgG, (3 ) PTX+ ICI (anti-PD-1(250ug) and anti-CTLA-4 (100ug), IP, 1x/3 days), (4) IPAT+ PTX + ICI, (5) IPAT + PTX + IgG, or (6) IPAT + PTX + ICI.

Figure S5. Toxicity of combining AKTi and MEKi in the 6DT1 model

a. The survival rate of each treatment group is shown for mice implanted with 6DT1 murine breast cancer cells and treated with vehicle controls (V), capivasertib (Capi or C), Selumetinib (Selum. Or S) and the combination of capivasertib and selumetinib (Combo) as described in figure 6c. The groups treated with control IgG or ICI were combined since there was no difference in survival between those two treatment groups. Groups receiving the combination of capivasertib and selumetinib exhibited worse survival.

b. Blood serum was collected from those mice that survived until the experiment endpoint and the enzymes aspartate transaminase (AST), alanine transaminase (ALT) and blood urea nitrogen (BUN) were analyzed as markers of toxicity. The dotted line shows accepted values.

Data Availability

The data that support the findings of this study are available upon request to the corresponding authors, AR & CY.