Disruption of the Scaffolding Function of mLST8 Selectively Inhibits mTORC2 Assembly and Function and Suppresses mTORC2-Dependent Tumor Growth In Vivo

mTOR is a serine/threonine kinase controlling multiple cellular processes including cell growth, survival, metabolism, and cytoskeletal reorganization (1, 2). mTOR functions in two distinct complexes, mTORC1 and mTORC2, both of which share the mTOR kinase and mLST8, but also contain components unique to each complex (Raptor for mTORC1; Rictor and SIN1 for mTORC2). Although tremendous progress has been made in understanding the regulation and function of mTORC1/2, the specific role of mLST8 in each complex is less well defined. Although originally identified as an mTORC1 subunit (3), increasing evidence suggests that mLST8 is dispensable for mTORC1, but indispensable for mTORC2 activity (4, 5). However, the molecular mechanisms underlying the requirement for mLST8 in mTORC2 activation remain unclear.

In this report, we identify interaction points between mLST8 and mTOR, which when disrupted, specifically impair mTORC2 assembly and activity without affecting mTORC1. Furthermore, we found a direct interaction between mLST8 and SIN1 and further identified the region of the SIN1 NH₂-terminal domain that interacts with mLST8. We found that prostate cancer xenografts expressing the mLST8 mutant incapable of mTOR interaction displayed impaired mTORC2 signaling in vivo and reduced tumor growth, suggesting that targeted disruption of mLST8–mTOR interactions could be employed as a therapeutic strategy to selectively target mTORC2, while sparing the activity of mTORC1.

Human embryonic kidney 293T (HEK293T) cells, BT549, DU159, and PC3 cells were obtained from ATCC and maintained in DMEM containing 10% FBS. HEK293FT were obtained from Thermo Fisher Scientific and maintained in DMEM containing 10% FBS. Primary and immortalized human bronchial epithelial cells BEAS 2B cells were maintained in RPMI media containing 10% FBS. Normal human dermal fibroblasts (NHDF) were obtained from Lonza. NHDF cells were cultured according to the manufacturer's protocol using the FGM-2 bullet kit. All cells were cultured in a humidified incubator with 5% CO₂ at 37°C. Cell lines were used between passages 1 and 50 after thaw. Cell lines from ATCC were authenticated using short tandem repeat profiling. Mycoplasma testing was performed every 6 months, most recently in January 2019, using the PlasmoTest Kit from InvivoGen.

The mTOR E2285A knock-in under MLST8 knockout background cell line was generated using the CRISPR/Cas9 technology. The sgRNA targeting the genomic sequence close to the codon of the 2285 glutamate residue on mTOR was designed using the CRISPR design tool (http://crispr.mit.edu) and cloned into pSpCas9(BB)-2A-GFP vector. Single-stranded oligodeoxynucleotides (ssODN) were used as the template for the E2285A mutation and a synonymous change to the PAM site. For ease of genotyping, a SphI site near the target was mutated synonymously. The Cas9/sgRNA construct and the ssODNs were cotransfected into MLST8 knockout HEK293T cells. Forty-eight hours posttransfection, transfected cells were seeded into a 96-well plate by limited dilution. The genomic DNA of individual clones was extracted to amplify the DNA fragment containing the E2285 site. PCR products were digested by SphI-HF to screen for correct clones. Knock-in mutations were verified by the Sanger sequencing of cloned genomic PCR. sgRNA oligos, primers for genomic PCR, and ssODN are listed in Supplementary Table S1.

A total of 5 × 10⁶ cells suspended in 100 μL of Matrigel and PBS (1:1) were injected into the hind flanks of 6-week-old athymic nude mice (Foxn1^nu; Envigo). Tumor measurements were started 10 days after injection and were monitored for 3 weeks. Tumors were measured every 2–3 days with digital calipers and tumor volume calculated according to the formula [V = 4/3π(l/2)(h/2)(w/2)]. Data are presented as mean ± SEM (n = 10). Two-way ANOVA with Bonferroni Correction was used for statistical analysis. Experiments with mice were preapproved by the Vanderbilt Institutional Animal Care and Use Committee and followed all state and federal rules and regulations.

Additional Materials and Methods are available in the Supplementary Information.

We generated 293T cells carrying loss-of-function indel mutations in the MLST8 gene by CRISPR/Cas9-mediated genome editing, using multiple sgRNAs targeting human MLST8 exon 2, 4, or 7 (Supplementary Fig. S1A) and confirmed loss of mLST8 protein expression (Fig. 1A). mLST8 loss impaired coprecipitation of mTOR with the mTORC2 cofactors RICTOR and SIN1, without affecting the mTORC1 cofactor, RAPTOR. Consistent with the idea that mLST8 loss impairs mTORC2, mLST8-deficient cells displayed decreased pAKT^Ser473, the mTORC2 phosphorylation site, under steady state growth conditions (Fig. 1A), or in response to amino acid or serum induction (Fig. 1B). Phosphorylation of other mTORC2 downstream effectors such as PKC and NDRG1 were also reduced (Fig. 1A) with no impact on phosphorylation of the mTORC1 substrate S6K1^Thr389 or mTORC1 effector S6RP. In vitro kinase assays performed using mTOR immunoprecipitates (IP) from 293T cells lacking mLST8 showed strongly diminished phosphorylation of 6His-Akt (Fig. 1C). In contrast, RAPTOR IPs from cells lacking mLST8-phosphorylated GST-S6K1 in vitro at similar levels to those from control cells expressing mLST8 (Supplementary Fig. S1B). These findings agree with recent data demonstrating that mLST8 is important for mTORC2 activity, but is dispensable for mTORC1 signaling, despite being a component of the mTORC1 holoenzyme. Importantly, restoration of mLST8 expression in several individually selected mLST8-deficient 293T cell clones (293T-sgMLST8) rescued coprecipitation of RICTOR and SIN1 with mTOR, as well as pAKT^Ser473 (Fig. 1D). We confirmed our findings in a panel of human and murine cell lines (Supplementary Fig. S1C and S1D).

Previous analyses of the mLST8-mTOR kinase crystal structure identified structural areas of mTOR and mLST8 that might interact through formation of hydrogen bonds (6). In mLST8, the interaction points may occur across six of seven WD40 repeat domains that cooperate to form a beta-propeller structure (6). We introduced mLST8 point mutations at evolutionarily conserved residues predicted to be involved in hydrogen bonds, where important interactions with mTOR would most likely occur (Supplementary Fig. S2A). These mutations included mLST8-Q44E/N46L (mLST8^Mut1) and mLST8-W272F/W274F (mLST8^Mut2; Fig. 2A; Supplementary Fig. S2A). mLST8^WT, mLST8^Mut1, and mLST8^Mut2 were expressed in an mLST8-deficient 293T-sgMLST8 cell clone, as was mLST8^Mut1/2, harboring Q44E, N46L, W272F, and W274F mutations. Expression of each mLST8 species was confirmed by Western blot analysis (Fig. 2B). Coprecipitation of mTOR with RICTOR and SIN1 was seen in cells expressing mLST8^WT, mLST8^Mut1, and mLST8^Mut2, but not in cells expressing mLST8^Mut1/2. RAPTOR coprecipitation with mTOR was not affected by any mutation in mLST8 (Fig. 2B), consistent with the idea that mLST8 is not necessary for mTORC1 assembly. Notably, cells expressing mLST8^Mut1/2 exhibited far less pAKT^Ser473 as compared with cells expressing mLST8^WT, mLST8^Mut1, or mLST8^Mut2 (Fig. 2B). This was confirmed by in vitro kinase assays (Supplementary Fig. S2B).

A reciprocal approach was used to confirm these results, by introducing mutations into mTOR at the predicted site of its interaction with mLST8 amino acids 272 and 274 (Fig. 2A), and thus was expected to mimic the effects of mLST8^Mut2. Transfection of 293T cells with mTOR expression constructs confirmed that mTOR^E2285A produced a functional mTOR protein capable of coprecipitation with mTORC2 components, including wild-type mLST8 (Fig. 2C). The mTOR^E2285A mutation was knocked-in to the endogenous mTOR gene in a 293T-sgMLST8 cell clone, which was confirmed by sequencing of genomic DNA (Supplementary Fig. S2C). Expression of mTOR^E2285A with wild-type mLST8 was permissive for mTORC2 assembly, as assessed by coprecipitation of RICTOR with mTOR (Fig. 2D). However, a combination of mTOR^E2285A knock-in with mLST8^Mut1 expression completely abolished mTORC2 assembly, and blocked pAKT^Ser473. These data indicate that compromised mTORC2 integrity with mLST8^Mut1/2 were not due to disruption of overall conformation of mLST8 by introducing the mutations but caused by elimination of the interaction with mTOR.

Recent Cryo-EM structural analyses of yeast TORC2 revealed that LST8 is sandwiched between Avo1 (the yeast SIN1 homologue) and TOR kinase (7, 8). In contrast, cryo-EM studies of human mTORC1 show that mLST8 is not sandwiched or embedded within mTORC1, but rather protrudes from the structure core, leaving mLST8 highly exposed (6). This does not rule out a potential role of mLST8 in mTORC1 regulation but may explain why mLST8 is critical for assembly and function distinctly of mTORC2. We directly tested this hypothesis by mutational analysis of SIN1 and coimmunoprecipitation of mTORC2 components.

To determine whether mLST8 is a molecular bridge between mTOR and SIN1 in human mTORC2, we first assessed mLST8–SIN1 interactions by co-IP. 293T cell clones deficient for SIN1 (293T-sgSIN1) were generated by CRISPR/Cas9 gene editing of the SIN1-encoding gene, MAPKAP1 (Supplementary Fig. S3A–S3C). Phosphorylation of AKT^Ser473 was completely abolished in SIN1-deficient cells (Supplementary Fig. S3B). Next, SIN1 expression was reconstituted in 293T-sgSIN1 cells using expression of a SIN1-GFP fusion construct bearing wild-type SIN1 (SIN1^WT), or a SIN1 mutant bearing deletions of the NH₂-terminal 136 amino acids (SIN1^Δ1-136), conserved CRIM region (SIN1^ΔCRIM) or the PH domain (SIN1^ΔPH; Supplementary Fig. S3C). mTOR and RICTOR each coprecipitated with GFP in cells expressing SIN1^WT, SIN1^ΔCRIM, and SIN1^ΔPH, but not in cells expressing SIN1^Δ1-136 (Fig. 3A), suggesting that the SIN1 NH₂-terminal domain is key for interaction with mTORC2 cofactors. Furthermore, cells expressing SIN1^ΔCRIM or SIN1^Δ1-136 showed decreased pAKT^Ser473, consistent with decreased mTORC2 signaling. Notably, the CRIM domain of SIN1 has a known interaction site with AKT and other AGC kinases (9), explaining the reduced pAKT in these cells (Fig. 3A).

Examination of the SIN1 NH₂-terminus revealed that expression of a SIN1 fragment comprised solely of aa 1–136 was sufficient for coprecipitation of mTOR with SIN1 in cells otherwise deficient for SIN1, presumably through a direct interaction between SIN1 and mLST8 (Fig. 3B). Further characterization of the NH₂ terminal domain by expression of sequential deletion mutants (Supplementary Fig. S3C) in 293T cell clones lacking SIN1 expression identified aa 97–127, as the required region for mLST8 binding (Fig. 3C). Importantly, cells expressing mLST8^Mut1/2 retain a level of coprecipitation with SIN1 (Fig. 3D), despite the fact that mLST8^Mut1/2 does not support complex formation between mTOR and SIN1. Interestingly, SIN1 coprecipitation of RICTOR was not affected by disruption of mTOR-mLST8 binding (Fig. 3D), suggesting mLST8 brings SIN1 and RICTOR together with mTOR to build the mTORC2 complex. These data are in agreement with previous studies in yeast describing potential sites of interaction between Avo1 and LST8 within the aa 98–136 region of LST8 (Supplementary Fig. S3D; ref. 7). Collectively, these data support the hypothesis that mLST8 is the key molecular bridge between mTOR and the mTORC2-specific component, SIN1.

Given that mLST8 is required for mTORC2 signaling yet remains dispensable for mTORC1 signaling, it is possible that therapeutic approaches aimed at targeted disruption of mLST8–mTOR interactions might be used for selective targeting of mTORC2 in the setting of cancer. Notably, PTEN-null PC3 human prostate tumor cells require PI3K-mTORC2 signaling for cell survival (10), making these an ideal model in which to test this potential strategy for selective mTORC2 inhibition.

We generated mLST8-deficient PC3 cell clones (PC3-sgMLST8), which were confirmed by Western blot analysis to lack mLST8 expression, mTORC2 assembly, and pAKT^Ser473 (Fig. 4A). A PC3-sgMLST8 cell clone was reconstituted with mLST8^WT or with mLST8^Mut1/2, the mLST8 mutant lacking specific mTOR-interaction points (Fig. 2). Ectopic mLST8^WT, but not mLST8^Mut1/2, coprecipitated with mTOR in PC3-sgMLST8 cells, and rescued pAKT^Ser473 (Fig. 4A), confirming that this approach successfully blocked mTORC2 signaling within the setting of increased PI3K activity driven by PTEN loss. Compared with parental PC3 cells, PC3-sgMLST8 cells grew at a markedly reduced rate (Fig. 4B). Although expression of mLST8^WT restored the growth rate of PC3-sgMLST8 cells, mLST8^Mut1/2 did not. The reduction in cell growth caused by loss of mLST8 was mimicked by loss of SIN1, suggesting that mLST8 is an mTORC2-specific regulatory cofactor (Supplementary Fig. S4A and S4B). Furthermore, PC3-sgMLST8 cells reconstituted with mLST8^Mut1/2 generated tumors that were nearly five times smaller in volume and with a 2.5-fold decrease in final tumor weight (Fig. 4C). pAKT^Ser473 was markedly diminished in PC3-sgMLST8 tumors expressing mLST8^Mut1/2, consistent with inhibition of mTORC2 signaling in vivo, although pS6K1 was similar in both (Fig. 4D). Tumor cell proliferation, measured by Ki-67 staining, was significantly decreased in tumors expressing mLST8^Mut1/2, whereas tumor cell death, as measured by cleaved caspase-3 staining, was modestly elevated (Supplementary Fig. S4C). Collectively, these results show that disruption of the mLST8–mTOR interaction can be used for selective inhibition of mTORC2 in cancer.

Recent structural analyses of the mTOR complexes revealed several intricacies of mTOR complex stability, substrate accessibility, and sensitivity to rapamycin (6, 7, 11–14). However, the role of mLST8 in assembly of the mTORC2 complex has not yet been clearly defined. Here we provide experimental evidence that distinct interaction points within mLST8 contact mTOR, and at the same time, mLST8 binds with the SIN1-RICTOR complex to generate mTORC2. This agrees with previous reports showing that mLST8 does not bind directly to RICTOR (12), but interacts strongly with SIN1, such that SIN1 deficiency abolished mTOR–Rictor interactions (15). We found that mLST8 loss, or even disruption of the mLST8 sequences required for mTOR binding, completely impaired mTORC2 assembly and function in multiple untransformed and cancer cell lines. We propose that mLST8 is the scaffold upon which the molecular components of mTORC2 assemble. As SIN1 is a key regulator of mTORC2 activity (16–18), the stabilizing effect of mLST8 on mTOR and SIN1 becomes critical for mTORC2 function.

Identification of the selective role of mLST8 in mTORC2 has broad implications. mTORC2 has recently been identified as a new effector for oncogenic mutant Ras (19). RICTOR is also amplified in several types of human cancer (20, 21). Because several oncogenic mechanisms converge to activate PI3K, including growth factor overexpression, RTK activation or amplification, PIK3CA mutation, or PTEN loss, mTORC2 is frequently activated across many cancers (16). A growing body of evidence suggests that PI3K-dependent tumors cannot develop in the absence of mTORC2 signaling (10, 21, 22). We have shown herein that targeting mLST8 recapitulates mTORC2-selective inhibition seen in previous studies targeting RICTOR and have identified the discrete interaction points between mTOR and mLST8 that completely abolish mTORC2 signaling when disrupted. Thus, it is possible that drug design strategies aimed at disrupting mTOR–mLST8 interactions would be an effective approach at selective mTORC2 targeting in cancer. Further, currently available mTORC1 inhibitors relieve the potent negative feedback upon the PI3K signaling pathway, thus reviving PI3K signaling and activity of its many downstream effectors that promote tumor cell survival. There is no currently approved mTORC2-selective inhibitor, despite the significant interest in their development. The discoveries herein unveil a new approach for mTORC2-specific inhibitor design, which may hold an impact for patients with mutant Ras and/or PI3K-active cancers.

J. Chen has received speakers bureau honoraria from NCI intramural program site visit, NIH TPM study section, UT San Antonio T-32 retreat keynote speaker, University of Cincinnati, University of Notre Dame, University of Nebraska, and University of Kentucky. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Hwang, J. Chen

Development of methodology: Y. Hwang, L.C. Kim

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Hwang, L.C. Kim, W. Song, D.N. Edwards

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Hwang, L.C. Kim, R.S. Cook, J. Chen

Writing, review, and/or revision of the manuscript: Y. Hwang, L.C. Kim, W. Song, R.S. Cook, J. Chen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Hwang

Study supervision: Y. Hwang, J. Chen

This work was supported by a VA Merit Award 5101BX000134 (to J. Chen), NIH grants R01 CA177681 (to J. Chen), R01 CA95004 (to J. Chen), T32 CA009592 (to L.C. Kim and D.N. Edwards), F31 CA220804-01 (to L.C. Kim), and Susan G. Komen Award PDF #17480733 (to D.N. Edwards).