The tuberous sclerosis complex (TSC) is a multisystem disease genetically characterized by a loss of function in either of the two tumor suppressors, TSC1 (hamartin) or TSC2 (tuberin) (1, 2). We recently used comprehensive genomic profiling of TSC patient tumors and found that ~85% carried mutant TSC2, ~12% had mutant TSC1, and ~3% had no identified mutation in either TSC gene (3). Loss of TSC1 or TSC2 results in benign tumor formation in various organs (4), including subependymal nodules (SEN), subependymal giant cell astrocytomas (SEGA) and cortical tubers of the brain, smooth muscle tissue and cystic changes in the lungs (lymphangioleiomyomatosis, LAM), skin fibromas and angiofibromas, and cardiac rhabdomyomas in infants (5, 6). The kidney is the most common location for lesions in TSC patients, with up to 80% of patients developing renal cysts and angiomyolipoma (RA). RA tumors are associated with spontaneous hemorrhage and require lifelong surveillance (7). Further, the multisystem and numerous organ manifestations can be associated with severe morbidity and potentially death with kidney disease as the highest cause of mortality in TSC patients (8).

Loss of TSC1 or TSC2 leads to constitutive activation of the mammalian target of rapamycin (mTOR) (9), a master regulator of nutrient and energy status in cells. This permits aberrant cell division and growth. Accordingly, TSC tumors show dramatic therapeutic sensitivity to rapamycin (sirolimus) or other rapamycin analogs (rapalogs). However, while sirolimus is effective against various TSC-based tumor lesions, this allosteric mTOR inhibitor is primarily cytostatic, and tumors regrow upon cessation of therapy (10, 11). Furthermore, rapamycin treatment in patients is associated with a decrease in angiomyolipoma volume of ~50%, and unfortunately, these benefits are reversed after treatment is withdrawn (10, 12, 13). However, rapamycin side effects are mild to moderate, and after continuous treatment for 3 years, efficacy was maintained without new or additional significant side effects (14). Due to this cytostatic effect and favorable safety profile, most patients may require a lifelong treatment regimen of rapamycin (15). Therefore, identifying additional therapeutic options for TSC patients that would sensitize TSC2-deficient angiomyolipoma cells might be preferable over lifelong therapy. Here, we sought to identify new TSC2-dependent vulnerabilities using chemical biology approaches and validation in mouse models.

To identify therapeutic vulnerabilities in TSC2-deficient tumors, we used a pair of isogenic cell lines derived from a renal angiomyolipoma (RA) cell line 621-101 that were either TSC2-deficient (621-102, control) or TSC2-rescued (621-103, TSC2 expression) ( Figure 1A ) (16). First, we performed chemical compound screens in each cell line to identify compounds that compromised cell viability greater in the TSC2-deficient setting when compared to the TSC2-rescued cells ( Figure 1 and Supplemental Table 1 ). We used a molecularly targeted library consisting of 480 compounds from a collection of diverse, active, cell-permeable small-molecule inhibitors from preclinical research and clinical trials, including kinase inhibitors, natural products, and chemotherapeutic agents, screened at six concentrations (0.1 nM to 10 µM) for 72-h treatments ( Supplemental Table 1 ). Next, we generated six-point dose–response curves for each molecularly targeted compound to generate cellular half maximal effective concentration (EC₅₀) values and prioritized compounds that were effective (reduced cell viability > 60% at any concentration) and particularly potent in TSC2-deficient cells (EC₅₀ < 1 µM). By plotting EC₅₀ values in TSC2-deficient versus TSC2-rescued cells ( Figure 1B ), we identified two compounds targeting the serine/threonine-protein checkpoint kinase 1 and 2 (CHK1 and CHK2), LY2603618 and AZD7762, as selective for TSC2-deficient cells, which were selected for additional interrogation. AZD7762 is equally potent against CHK1 and CHK2, and generally with good selectivity (>10-fold) against 164 kinases. Kinases with <10-fold selectivity were in the same family of kinases as CHK1/2, the CAM kinases, and some non-receptor tyrosine kinases (22). For LY2603618, CHK1 maintained >100-fold selectivity over the next target (PDK1) tested, and >1,500-fold selectivity over CHK2 (23). Dose–response curves for these inhibitors displayed >17-fold and >10-fold reductions in EC₅₀ dependent on TSC2 status for LY2603618 and AZD7762, respectively ( Figures 1C, D ). Notably, TSC2-deficient cells treated with the dual CHK1/2 inhibitor, AZD7762 ( Figure 1D ), have a more potent and complete cellular EC₅₀ when compared to the more selective CHK1 inhibitor, LY2693618 ( Figure 1C ), suggesting that CHK1/2 inhibition may be more effective at reducing the viability of TSC2-deficient cells. We note that AZD7762, while selective for CHK1/2, may be eliciting increased toxicity through off-target kinase engagement (22). Specifically, AZD7762 (EC₅₀ = 37 nM) was 6-fold more potent for TSC2-deficient cells than LY2603618 (EC₅₀ = 220 nM), and thus, we decided to pursue dual CHK1/2 inhibition in a TSC2-deficient setting.

To validate the primary chemical screen from 621-102 cells, we extended to a second patient-derived TSC cell line, UMB1949. We performed ten-point dose–response curves with AZD7762 in both TSC2-deficient cell lines, UMB1949 and 621-102. AZD7762 inhibited the growth of UMB1949 cells (EC₅₀ = 33 nM) with a similar potency as 621-102 cells ( Figure 2A ). To validate that the inhibitory effects on viability were, in fact, due to CHK1 or CHK2 engagement, we used siRNA gene knockdown of CHK1 or CHK2 in the aforementioned cell lines ( Figure 2B ). While CHK1 knockdown resulted in a ~35% reduction in viability after 72 h, knockdown of CHK2 reduced viability up to 65%, consistent with increased potency of the dual CHK1/2 inhibitor AZD7762 relative to the CHK1-selective LY2603618. Given the role of TSC2 as a negative regulator of mTORC1 in response to various cellular stresses (24), including growth factor deprivation, we wanted to explore media conditions that allowed for cell growth and proliferation in a TSC2-deficient setting ( Supplemental Figure 1A ). In response to serum-restricted culture media (0.1% FBS), TSC2-deficient cells proliferated and responded to rapamycin treatment, while the TSC2-rescued cells were arrested and were unresponsive to rapamycin treatment ( Supplemental Figure 1B ) (25). Under these optimized growth conditions, we performed a secondary chemical screen in the TSC2-deficient 621-102 cells with a ten-point dose–response to acquire EC₅₀ measurements ( Supplemental Table 2 ). Notably, the rapalog everolimus was a potent inhibitor ( Supplemental Figure 1C ) as were both CHK inhibitors, LY2603618 and AZD7762 ( Supplemental Figure 1D ).

To expand the repertoire of potent compounds against TSC2-deficient cells, the top eighty-eight most potent (EC₅₀) compounds with >60% reduction in viability were selected from the secondary chemical screen in 621-102 cells, as described above ( Supplemental Table 2 ). Next, we performed a follow-up compound screen in both the 621-102 and UMB1949 cell lines with ten-point dose responses to report EC₅₀ measurements ( Supplemental Table 3 ). Under the limited serum and growth factor conditions, both CHK inhibitors were among the most potent compounds tested, maintaining EC₅₀ < 7 nM for UMB1949 cells and EC₅₀ < 3 nM for 621-102 ( Supplemental Figures 1D–F ). The general overview of the chemical screens performed is shown in Supplemental Figure 1G .

To test the effects of the dual CHK1/2 inhibitor, AZD7762, on tumor growth, we turned to in vivo mouse models. We first established a cell line tumorgraft model of the patient-derived UMB1949 in NSG (NOD scid gamma) mice. Next, we treated mice with either vehicle or AZD7762 (12.5 mg/kg, 5× weekly, for 28 days) and measured tumor volume ( Figure 2C ). AZD7762 treatment caused a significant reduction in UMB1949 TSC2-deficient tumor size compared with the tumor volume of the vehicle control. Strikingly, the growth delay, calculated at the days required to reach tumor volume of 500 mm³, was not achieved in the AZD7762 treatment group over the 28 days of treatment, while the control group reached 500 mm³ on day five. This indicates complete tumor stasis with CHK1/2 inhibition, as mice were enrolled to begin drug treatment when tumors reached 400 mm^3, and final tumor measurements did not reach starting tumor volumes with 0% tumor growth in the AZD7762-treated group. As expected, tumor lysates from AZD7762-treated mice showed decreased CHK1 serine 296 auto-phosphorylation (pS296), confirming in vivo CHK1 target inhibition ( Figure 2D ). In addition, increased CHK1 serine 345 phosphorylation (pS345) by ATM/ATR indicates strong checkpoint activation, replication stress, or DNA damage response by CHK1 inhibition ( Figure 2D ).

The patient-derived UMB1949 cell line represents an additional cellular model of TSC with loss of TSC2 ( Figure 3A ), originally isolated from a male TSC patient with renal angiolipoma (18). Interestingly, we uncovered that UMB1949 cells reported by Lim et al. (18) were derived from a TSC patient that was part of our Martin et al. genomic profiling study (08-RA1) (3), which contained the same pathogenic frameshift deletion in TSC2 ( Figure 3A ). UMB1949 cell line identity matched to patient 08-RA1 was confirmed by rare variant analysis with minor allele frequency (MAF) < 0.01 in our exome data (3). To further establish a molecular understanding for the in vitro and in vivo models tested above, RNA-sequencing ( Supplemental Table 4 ) was performed on UMB1949 and 621-102 cell lines and compared to our existing transcriptional landscape data for both normal kidney (NK) and RA tumors (3). When examining the top 500 differentially expressed genes (DEG) between normal kidneys and RA tumors, the transcriptional profiles of the two cell lines resembled the patient tumors ( Figure 3B ). In addition, cell-type deconvolution software (26) found adipose tissue and smooth muscle signatures shared between the cell models and patient angiomyolipomas, while expanded fetal kidney signatures were prominent in the cell lines ( Figure 3C ), perhaps due to the immortalization and effect, excess growth factors, or nutrients in cell culture media. Importantly, each cell line contained signatures of blood vessels, smooth muscle, and adipose tissue, the defining features of renal angiomyolipomas.

We further evaluated CHK1/2 inhibition in a second mouse TSC model (27), in which 105K cells, a Tsc2-deficient cystadenoma cell line, derived from a renal tumor from a Tsc2 ^+/- mouse (19). Tsc2-deficient 105K cells were engrafted subcutaneously into BALB/c-nu immunodeficient mice, and again, mice treated with AZD7762 had a significant reduction in tumor volume relative to vehicle-treated mice ( Figure 4A ). In addition, complete tumor stasis with CHK1/2 inhibition was again observed, as mice were enrolled to begin drug treatment when tumors reached 100 mm³ and final tumor measurements at 30 days did not reach starting tumor volumes. The AZD7762-treated mice had a >20% reduction in tumor volume as compared to the vehicle treated group with a >300% increase in tumor volume.

Next, we treated an A/J mouse model with heterozygous germline deletion of Tsc2 (A/J Tsc2^+/- mice) that spontaneously develops kidney pathology with age (20). In this model, tumors develop and progress from cysts to cystadenomas, and the severity of kidney cystadenomas increases with age. At 8 months of age, the age at which these mice have high tumor burden, A/J Tsc2^+/- mice received AZD7762 or vehicle for 28 days ( Figure 4B ). After treatment, we harvested both left and right kidneys and performed histological analysis of cysts and cystadenomas to quantify overall cyst and tumor burden. Representative H&E-stained kidney sections used to quantify cysts and cystadenomas after treatment are shown ( Figure 4E ). Lesions were classified as cysts (0% < cell content < 25%) or cystadenomas (25% < cell content < 90%). AZD7762 treatment resulted in a 32% decrease in cystadenomas ( Figure 4D ) and a reduction in the progression from cyst to cystadenomas, as evidenced by the modest increase in the number of cysts ( Figure 4C ).

Here, we presented multiple chemical compound screens in TSC2-deficient cells as compared to TSC2-rescued cells. CHK inhibitors were potent inhibitors under all screening conditions tested, in particular in the restricted growth conditions that reveal the tumor-promoting properties of TSC2 loss. The antitumor potential of the dual CHK1/2 inhibitor, AZD7762, was demonstrated in three different animal models of TSC2-deficient tumors. The second-generation CHK inhibitor (prexasertib) is currently in clinical trials for various human cancers and appears to have mitigated toxicity concerns associated with the first-generation CHK inhibitor (28). Our data support further exploration of CHK inhibition in mTOR-driven pathologies and suggest a mechanistic connection between DNA damage response and mTOR signaling. Indeed, a growing body of evidence links mTORC1 activation to DNA damage repair (29–31). Consistent with these data, we previously reported a low mutational burden in TSC2-deficient patient tumors (3), suggesting efficient DNA repair mechanisms and perhaps a therapeutic vulnerability in tumors lacking functional TSC2.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material .

The animal studies were reviewed and approved by Porsolt’s Ethical Committee and Van Andel Institute Review Committee.

Designing research studies: KM., JM, DA. Conducting experiments: KM, JK, KS, C-YC, NS, MB, MK. Acquiring data: KM, RV, JM. Analyzing data: KM, RV, NS, JM, DA. Providing reagents: DA. Writing the manuscript: RV, KM, JM. Acquisition of funding: RV, JM, DA. All authors contributed to the article and approved the submitted version.

RV (K00CA245821) and JM (R21CA263133) have research support from the National Cancer Institute. This work was supported by grants and funding from the Michigan Strategic Fund, Tuberous Sclerosis Complex Alliance, Blue Cross Blue Shield of Michigan Foundation, and individual donors.

JM has consulting agreements with Merck, research support from Erasca, and scholarly activity and support from Translational Genomics Research Institute (non-profit organization). JK’s current affiliation is with AbbVie.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.