Abstract

   Malignant peripheral nerve sheath tumors (MPNSTs) are highly aggressive
   soft tissue sarcomas with limited treatment options, and new effective
   therapeutic strategies are desperately needed. We observe
   antiproliferative potency of genetic depletion of PTPN11 or
   pharmacological inhibition using the SHP2 inhibitor (SHP2i) TNO155. Our
   studies into the signaling response to SHP2i reveal that resistance to
   TNO155 is partially mediated by reduced RB function, and we therefore
   test the addition of a CDK4/6 inhibitor (CDK4/6i) to enhance RB
   activity and improve TNO155 efficacy. In combination, TNO155 attenuates
   the adaptive response to CDK4/6i, potentiates its antiproliferative
   effects, and converges on enhancement of RB activity, with greater
   suppression of cell cycle and inhibitor-of-apoptosis proteins, leading
   to deeper and more durable antitumor activity in in vitro and in vivo
   patient-derived models of MPNST, relative to either single agent.
   Overall, our study provides timely evidence to support the clinical
   advancement of this combination strategy in patients with MPNST and
   other tumors driven by loss of NF1.
     __________________________________________________________________

   SHP2i and CDK4/6i provide antitumor efficacy through enhancing RB
   function in patient-derived models of NF1-deficient MPNST.

INTRODUCTION

   Malignant peripheral nerve sheath tumors (MPNSTs) are highly aggressive
   heterogeneous soft tissue sarcomas with limited effective treatment
   options, given their relative insensitivity to conventional systemic
   chemotherapy and radiation and their propensity to metastasize. In many
   patients with neurofibromatosis type 1 (NF1), MPNSTs develop from
   within plexiform neurofibromas (pNFs) ([55]1), the benign precursor
   tumors that themselves can be sources of pain, disfigurement, and
   alteration of function ([56]2, [57]3). The most recurrent genomic
   alterations underlying the pathogenesis of MPNST are loss of function
   (LOF) of the RAS-GAP (guanosine triphosphatase–activating protein)
   neurofibromin [NF1, 90%; ([58]4, [59]5)], CDKN2A (60 to 80%), PRC2
   [polycomb repressive complex 2; 70 to 90%; ([60]4–[61]7)], and TP53, as
   well as gain of chromosome 8 [80%; ([62]8, [63]9)], yet molecular
   targeting of these LOF alterations represents a unique challenge and a
   major unmet need for these patients.

   Despite many clinical trials of chemotherapy and targeted agents, there
   has been little advancement in overall survival for patients with
   MPNST, and, thus, novel therapeutic approaches are desperately needed
   ([64]10). Now, only one class of drug, mitogen-activated protein kinase
   kinase (MEK) inhibitors (MEKis), has been approved for reducing
   symptomatic tumor burden in patients with pNF ([65]11–[66]13).
   Aberrantly up-regulated extracellular signal–regulated kinase (ERK)
   signaling, due to loss of NF1 and dysregulated PRC2 function, is a
   critical effector of tumorigenesis, and, thus, pharmacological MEK
   inhibition has also been tested in models of MPNST ([67]11,
   [68]14–[69]16). However, the preclinical responses of MPNST to
   monotherapy with MEKi and other targeted agents are incomplete,
   suggesting a need for improved understanding of the role of ERK and
   other RAS effector pathways ([70]17, [71]18). The incomplete effects of
   MEKi in MPNST prompted us to explore potential combinatorial
   therapeutics using MEKi and agents targeting the adaptively changed
   signaling elements that emerge upon short-term MEK inhibition. We have
   previously reported that the adaptive and acquired response to MEKi in
   MPNST involves the up-regulation of activity of multiple receptor
   tyrosine kinases (RTKs) ([72]19, [73]20).

   SHP2, encoded by PTPN11, positively regulates numerous transmembrane
   signaling pathways, including RTK ([74]21, [75]22), Janus kinase/signal
   transducer and activator of transcription (STAT) ([76]23), programed
   cell death ([77]24), and B and T lymphocyte attenuator (BTLA) ([78]25),
   through exerting scaffolding and/or catalytic functions in the
   regulation of activity of central components in these pathways. SHP2
   facilitates RAS–guanine nucleotide exchange factor–mediated
   RAS–guanosine 5′-triphosphate (GTP) loading and subsequent membrane
   recruitment of RAS and therefore is required for RTK-mediated RAS/ERK
   activation ([79]26). SHP2 phosphatase also acts as a positive regulator
   of RAS by dephosphorylating RAS and promoting its binding to and
   activation of RAF kinases, leading to increased ERK signaling ([80]27).
   Therefore, SHP2 represents a promising target, and inhibition of SHP2
   can target a point of convergence from upstream RTK signaling and,
   together with inhibition of RAS effector pathways, provide an enhanced
   antitumor effect ([81]11, [82]22, [83]26, [84]28, [85]29). Thus, we
   tested combination SHP2i as a strategy to overcome RTK up-regulation
   seen with single-agent MEKi in tumors with loss of NF1 and,
   subsequently, demonstrated that combined MEKi/SHP2i is additive in both
   in vitro and in vivo models of NF1-MPNST ([86]20). We then observed
   more potent single-agent activity of the clinically available SHP2i
   TNO155 ([87]NCT04000529), compared to SHP099, and therefore
   hypothesized that SHP2i, rather than MEKi, may be a critical antitumor
   strategy for NF1-MPNST. As potential clinical therapies for MPNST
   continue to evolve in response to both clinical and preclinical data,
   MEKi may not be the most ideal combination partner for two reasons: (i)
   Many patients who develop MPNST will have already been exposed to MEKi
   as treatment for their precursor pNF, and (ii) SHP2i and MEKi may
   result in overlapping toxicity profiles due to ERK signaling
   suppression as an effect of both agents, leading to poor tolerability
   in people. In the ongoing study of a SHP2i and a MEKi
   ([88]NCT03989115), drugs are administered on an intermittent dosing
   schedule to address toxicity ([89]30).

   Therefore, we sought an alternate partner to combine with SHP2i to
   advance this agent to a clinical trial for people with MPNST, and we
   have focused specifically on cyclin-dependent kinase 4/6 inhibition
   (CDK4/6i). Frequent loss of CDKN2A, inactivation of retinoblastoma
   protein (RB), hyperactivation of CDKs, and the dependency of D-cyclins
   on RAS signaling suggest that CDK4/6i may be an attractive therapeutic
   strategy in NF1-MPNST ([90]31, [91]32). We hypothesized that the
   cytostatic effects of CDK4/6i may be potentiated by drugs targeting
   upstream regulators of RAS, such as SHP2, in MPNST.

   In this study, we demonstrate that TNO155 and the CDK4/6i ribociclib
   produce additive antiproliferative and antitumor effects in cell lines
   and patient-derived xenograft (PDX) models of NF1-MPNST through
   synergistic inhibition of cell cycle and induction of apoptosis. When
   given together, TNO155 attenuates the adaptive response to CDK4/6i,
   produces deeper and more durable responses, and thus potentiates the
   efficacy of single-agent CDK4/6i in a broad collection of preclinical
   models that are a faithful representation of the genomically
   heterogeneous human MPNST. Together, with the clinical testing of
   TNO155 and ribociclib now underway ([92]NCT04000529), this study will
   serve as an important contribution to the advancement of SHP2i
   combinations in the clinic for people with NF1-deficient cancers.

RESULTS

PTPN11 genetic depletion reduces MPNST cell growth and alters MEK/ERK
activity

   Activating mutations in PTPN11, the gene that encodes SHP2, occur in
   about 50% of patients with Noonan syndrome and a subset of leukemias
   ([93]33), whereas, in solid tumors, PTPN11 mutations occur at a low
   frequency. Wild-type (WT) SHP2 is critical for activation of
   RTK/RAS/ERK signaling that drives proliferation and growth of tumor
   cells. To understand the functional role of WT SHP2 in NF1-MPNST
   growth, we knocked down SHP2 by generating doxycycline (Dox)–inducible
   expression of two unique short hairpin RNAs (shRNAs) targeting PTPN11
   (#818, and #5003) in three NF1-MPNST cell lines ([94]Fig. 1 and fig.
   S1) including the two traditional lines ST8814 and NF90.8 and one
   patient-derived MPNST cell line JH-2-002 ([95]34). PTPN11/SHP2 genetic
   depletion significantly reduced MPNST cell growth, as evidenced by
   long-term colony formation assay ([96]Fig. 1A and fig. S1A) and
   short-term real-time cell confluence monitoring, using IncuCyte imaging
   system ([97]Fig. 1B and fig. S1B). SHP2 protein expression and
   phosphorylation at tyrosine-542 were markedly decreased following SHP2
   knockdown after 72-hour treatment with Dox, while the signaling
   intermediates phospho-MEK (p-MEK) and p-ERK exhibited marginal
   reduction ([98]Fig. 1C and fig. S1C), likely due to rebound in ERK
   signaling.

Fig. 1. PTPN11 genetic depletion reduces MPNST cell growth and alters MEK/ERK
activity.

   [99]Fig. 1.
   [100]Open in a new tab

   (A) ST8814 and JH-2-002 transduced with doxycycline (Dox)–inducible
   constructs expressing shPTPN11 #818 or #5003 were treated with vehicle
   or Dox (300 ng/ml) for 2 weeks. Cells were fixed with 10%
   neutral-buffered formalin (NBF) and then stained with crystal violet.
   (B) Cells as in (A) were treated with vehicle or Dox (300 ng/ml) for up
   to 7 days. The phase confluence was monitored by IncuCyte real-time
   imaging system and normalized to corresponding 0-hour scan. (C) Cells
   as in (A) were treated with vehicle or Dox (300 ng/ml) for 72 hours,
   and the indicated proteins were assessed using immunoblot. (D) JH-2-002
   transduced with Dox-inducible shPTPN11 #818 was treated with vehicle or
   Dox (300 ng/ml) for 72 hours, and, then, three replicates were
   collected for RNA-seq, and the fourth replicate was collected for
   immunoblotting validation. RNA-seq result revealed a significant
   decrease in PTPN11 RNA expression [n = 3, P adj = 0, and log[2] fold
   change (LFC) = −4.69] when comparing shPTPN11 v. ctrl. A total of 51
   genes representing MEK/ERK transcriptional output ([101]35) were
   assessed for their expression after PTPN11 knockdown. A volcano plot
   demonstrating LFC of shPTPN11 v. ctrl as a function of −log[10] (P adj)
   is shown. Black dots, not significant (ns; P adj > 0.05); red dots, LFC
   > 0; and blue dots, LFC < 0 (P adj < 0.05). Thirty-seven of the 51
   genes (73%) in the set were significantly transcriptionally
   down-regulated following PTPN11 genetic depletion.

   We have previously found that the level of p-ERK, which has been used
   as a surrogate marker for RAS/ERK pathway activation, is inherently
   susceptible to feedback regulation by dual-specificity phosphatases
   (DUSPs) and therefore not an accurate measure of signaling output
   ([102]35). MEK/ERK-dependent transcriptional output is defined as the
   set of genes whose expression changes significantly after 8-hour MEK
   inhibition ([103]35). In JH-2-002 shPTPN11 #818, treated with Dox or
   vehicle, we first validated protein expression level of SHP2 and other
   signaling proteins (fig. S1D) and performed RNA sequencing (RNA-seq;
   see Materials and Methods). We confirmed significant reduction in
   transcript levels of PTPN11 (log[2] fold change = −4.69 and adjusted P
   = 0) and analyzed the transcriptional profile of shPTPN11 relative to
   defined ERK signaling output ([104]Fig. 1D). A total of 51 genes
   (excluding KIR3DL2, which was not detected via RNA-seq), are included
   in the analysis (table S1). Thirty-seven of the 51 genes (73%) in the
   MEK-dependent set demonstrated significant down-regulation after PTPN11
   depletion (P adj < 0.05), indicating a global response of ERK signaling
   suppression by SHP2 knockdown ([105]Fig. 1D and fig. S1E). Together,
   these observations concluded that PTPN11 genetic depletion reduces
   MPNST cell growth in an ERK signaling–dependent manner.

SHP2i TNO155 alters growth and gene expression in NF1-MPNST cells

   We first characterized the genomic or protein alterations in NF1,
   CDKN2A, PRC2 components SUZ12 and EED, TP53, and others of importance
   in an updated panel of MPNST cell lines ([106]Fig. 2A) ([107]19) to
   confirm that our cell line panel is a faithful genomic representation
   of the highly heterogeneous MPNST seen in patients. We next tested the
   effect of pharmacological inhibition of SHP2 on NF1-MPNST cell
   proliferation, viability, and growth. While SHP099 (a tool compound
   that has not been tested in clinical trials) was only partially
   effective in reducing cell and tumor growth in our initial studies
   ([108]20), we found that the lead clinical compound, TNO155, which is
   now being tested in clinical trials ([109]NCT03114319 and
   [110]NCT04000529), was more potently able to suppress cell
   proliferation and growth in 10 MPNST in vitro models ([111]Fig. 2, B
   and C). SHP099 also demonstrated activity against cellular viability in
   a three-dimensional (3D) microtissue ex vivo model derived from four
   unique MPNST PDXs (fig. S2A) ([112]9). Further investigation in ST8814
   and JH-2-079c cell lines ([113]9) treated with dimethyl sulfoxide
   (DMSO) or 0.3 μM TNO155, using IncuCyte real-time monitoring system,
   demonstrated caspase-3/7 activation induced by TNO155 (fig. S2B).
   Mechanistically, we found that TNO155 inhibited SHP2 phosphatase
   activity as evidenced by p-SHP2 reduction and markedly suppressed ERK
   signaling and downstream cell cycle pathways, in a dose-dependent
   ([114]Fig. 2D) and time-dependent ([115]Fig. 2E) manner. To further
   confirm the effect of TNO155 on ERK signaling and cell cycle
   regulators, we expanded the investigation to a collection of eight
   NF1-MPNST cell lines including three patient-derived lines (JH-2-002,
   JH-2-079c, and JH-2-103). We again observed that TNO155 effectively
   inhibited cell cycle regulators via ERK signaling suppression
   ([116]Fig. 2F). Notably, TNO155 induced RB1 hypo-phosphorylation
   consistently across the cell lines tested, in a dose- and
   time-dependent manner ([117]Fig. 2, D to F).

Fig. 2. SHP2i TNO155 alters growth and gene expression in NF1-MPNST cells.

   [118]Fig. 2.
   [119]Open in a new tab

   (A) Onco-print of key driver genes (NF1, CDKN2A, SUZ12, and EED) in
   MPNST and putative others is shown. (B) Ten NF1-MPNST cell lines were
   treated with increasing doses of the SHP2i TNO155 for 5 days. Cell
   viability was evaluated using the cell counting kit-8 (CCK-8) assay
   measuring metabolic activity. (C) Three NF1-MPNST cell lines were
   treated with dimethyl sulfoxide (DMSO) or TNO155 (0.3, 1, and 3 μM) for
   about 2 weeks. Cells were fixed with 10% NBF and then stained with
   crystal violet. (D and E) ST8814 and JH-2-002 were treated with
   increasing doses (30 to 3000 nM) of TNO155 for 24 hours (D) or 0.3 μM
   TNO155 over a time course (E). (F) Eight NF1-MPNST cell lines were
   treated with DMSO or 0.3 μM TNO155 for 48 hours. Representative
   signaling intermediates in ERK and cell cycle pathways were detected
   using immunoblot. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (G)
   ST8814 and NF90.8 transduced with Dox-inducible NF1-GRD were pretreated
   with vehicle or Dox (300 ng/ml), followed by treatment with DMSO or
   TNO155 (0.3, 1, and 3 μM) for 5 days. Cell viability was assessed using
   CCK-8. (H) ST8814 and NF90.8 transduced with (marked with +)
   Dox-inducible green fluorescent protein (GFP) or V5-tagged NF1-GRD were
   treated with Dox (300 ng/ml) for 24 hours. RAS activity and signaling
   intermediates were detected using immunoblot following the active RAS
   pull-down assay. PD, pull down; WCL, whole-cell lysate. (I) Venn
   diagram showing the number of significant overlapping genes (P adj <
   0.05 and |fold change| > 1.5) between shPTPN11, TNO155, and trametinib
   at 6- and 24-hour time points. (J) Heatmap demonstrating significant
   changes (P adj < 0.05 marked with “+”) in transcriptional output of the
   51 ERK signature genes, derived from RNA-seq analysis of JH-2-002 after
   DMSO, 0.3 μM TNO155, or 20 nM trametinib treatment (6 and 24 hours).

   Considering the competitive role of NF1 GAP function and SHP2
   inhibition in reducing the accumulation of RAS-GTP during nucleotide
   cycling, we asked whether NF1 GAP function played a role in response to
   TNO155. Therefore, we reconstituted the expression of NF1-GRD
   (GAP-related domain) in two traditional NF1 LOF cell lines, ST8814 and
   NF90.8, using a Dox-inducible murine retrovirus expression system.
   Reintroduction of NF1-GRD to restore GAP function partially rescued
   cell viability upon TNO155 treatment ([120]Fig. 2G), suggesting that
   NF1 loss confers sensitivity to SHP2i, an observation consistent with
   previous studies ([121]26, [122]36). Relative to the green fluorescent
   protein (GFP) control, reexpression of NF1-GRD reduced RAS-GTP levels,
   indicative of a successful restoration of NF1 GAP function
   ([123]Fig. 2H).

   To compare the transcriptional effects of shPTPN11, TNO155, and the
   MEKi trametinib individually, we conducted RNA-seq analysis of JH-2-002
   treated with DMSO, TNO155, or trametinib at 6- and 24-hour time points,
   and shPTPN11, to measure both direct and adaptive responses to ERK
   signaling inhibition. Analysis of the transcriptional response showed
   that treatment with 0.3 μM TNO155 for 6 and 24 hours led to fewer
   differentially expressed genes compared to the treatment of 20 nM
   trametinib or shPTPN11 ([124]Fig. 2I and table S2), which is in line
   with the immunoblots measuring ERK signaling and cell cycle regulators
   at the same time points (fig. S2C). Specifically, 740 genes were
   significantly altered (P adj < 0.05 and absolute fold change > 1.5)
   following TNO155, while 3301 and 3156 genes were significantly altered
   after trametinib and shPTPN11, respectively. Most of the genes (~80%)
   significantly altered by TNO155 overlapped with those altered by
   trametinib ([125]Fig. 2I). For both treatments, the number of
   significantly altered genes increased over a 6- to 24-hour time course,
   indicating the existence of indirect and adaptive changes to
   transcription after ERK signaling suppression. We also assessed ERK
   signaling using the 51 ERK signature genes described above and found
   most genes in the signature to be significantly down-regulated upon
   drug treatment. This suggests that the TNO155 treatment, while
   affecting fewer genes ([126]Fig. 2I), has a pronounced effect on the
   ERK signature genes, despite less potent inhibition by TNO155 relative
   to trametinib ([127]Fig. 2J, fig. S2D, and table S1), consistent with
   the degree of ERK signaling inhibition by both agents observed in fig.
   S2C.

Loss of function in RB1 confers resistance to TNO155

   A recent study highlights the transcriptional up-regulation of cell
   cycle regulators during the progression from pNF to MPNST ([128]31). We
   also observed a consistently enhanced cell cycle transcriptional
   profile in samples from our NF1 biospecimen repository, using the gene
   category defined as cell cycle regulators ([129]31) and found that
   PLK1, CCNA2, and BIRC5 demonstrated significant up-regulation in MPNST
   compared to pNF ([130]Fig. 3A), suggesting a driver role of cell cycle
   dysregulation in tumor pathogenesis and disease progression. Given the
   reported genomic heterogeneity of NF1-MPNST and the impact this
   heterogeneity has at the protein level ([131]19), we sought to
   determine whether steady-state expression level of a specific molecular
   marker can predict sensitivity to SHP2i treatment. We first
   characterized the signaling protein markers in the updated panel of
   MPNST cell lines ([132]Fig. 3B and fig. S3, A and B). We found that, of
   the nine cell lines in this panel (which also includes the sporadic
   MPNST line STS26T), all exhibited loss of p16 expression
   ([133]Fig. 3B), seven exhibited loss of SUZ12 expression (fig. S3A),
   and all expressed various levels of p53 protein (fig. S3A) despite
   known mutations in some lines ([134]Fig. 2A). The two NF1-MPNST lines
   that retained SUZ12 expression (fig. S3A), NF96.2 and JH-2-079c, harbor
   p.S241R and p.D243Y mutations in EED, respectively ([135]Fig. 2A). We
   examined the protein profiles involved in cell cycle regulation
   ([136]Fig. 3B and fig. S3A) and in AKT and ERK pathways (fig. S3B) to
   determine whether there were some alterations in baseline expression of
   a single unique molecule that explained the lower TNO155 sensitivity
   exhibited in three of the cell lines in [137]Fig. 2B, but there were no
   salient correlations.

Fig. 3. Loss of function in RB1 confers resistance to TNO155.

   [138]Fig. 3.
   [139]Open in a new tab

   (A) Volcano plot from RNA-seq analysis revealing cell cycle
   dysregulation comparing MPNST v. pNF. (B) Selected cell cycle
   regulators were evaluated in the nine NF1-associated and one sporadic
   (STS26T) MPNST cell lines. (C) ST8814 and JH-2-002 transduced with
   Dox-inducible constructs expressing shPTPN11 #5003 or #818 were treated
   with vehicle or Dox (300 ng/ml) for 72 hours, and RB phosphorylation at
   serine-780 or serine-807/811 was assessed. (D) ST8814 parental (Par),
   or stable lines transduced with pLXSN-neomycin-E7 WT or E7 d21-24
   mutant; and JH-2-002 parental, or stable lines transduced with
   pMSCV-puromycin-E7 WT or E7 d21-24 mutant, were treated with increasing
   dose of TNO155 for 5 days. Cell viability was evaluated by using the
   CCK-8 assay. (E) Cells as in (D) were treated with DMSO or 0.3 μM
   TNO155 for 48 hours.

   To further explore signaling mechanisms that may potentiate TNO155
   efficacy, we sought to identify effectors downstream of ERK signaling
   that may attenuate SHP2i dependence. It has been reported that the loss
   of RB1 function results in partial resistance to MEK and RAF inhibition
   in BRAF^V600E mutant melanoma cells ([140]37), and, given the high
   frequency of cell cycle dysregulation in MPNST, we focused on the RB
   tumor suppressor pathway. Notably, all MPNST cell lines on our panel
   have retained RB protein expression ([141]Fig. 3B), and none exhibited
   evidence of genomic alterations in RB1 ([142]Fig. 2A), suggesting the
   presence of functional RB. RB1 plays a major role in the Gap 1 (G[1])
   cell cycle checkpoint that blocks synthesis (S)–phase entry ([143]38),
   and p-RB is associated with RB1 inactivation. RB is sequentially
   phosphorylated by cyclin D–CDK4/6 and cyclin A/E–CDK2 during the cell
   cycle progression. RB serine-780 is the major phosphorylation site by
   cyclin D1–CDK4 in early G[1] phase, and cyclin D1 is required for the
   phosphorylation at this site in vivo ([144]39, [145]40); while RB
   serine-807/811 is mainly phosphorylated by cyclin A/E-CDK2 at late G[1]
   and G[1]-S transition ([146]40), and its phosphorylation is likely to
   prime phosphorylation at other sites of RB ([147]41). In
   PTPN11-depleted cell lines, p-RB was reduced, indicative of enhanced RB
   activity ([148]Fig. 3C), an observation also seen with SHP2
   pharmacological inhibition with TNO155 in [149]Fig. 2 (D to F). These
   results demonstrate that ERK signaling inhibition via PTPN11 genetic
   depletion, or treatment with SHP2i, induces RB hypo-phosphorylation,
   resulting in RB activation. Notably, pRB-S807/811 exhibited more
   substantial decrease upon TNO155 (48 hours; [150]Fig. 2F) or shPTPN11
   after 72-hour Dox exposure ([151]Fig. 3C), compared to pRB-S780,
   suggesting the rebound in cyclin D1–CDK4 complex activity as a
   potential mechanism of signaling adaptation in response to SHP2i or
   knockdown.

   To evaluate whether loss of RB1 function confers resistance to TNO155,
   we designed an experiment to inhibit RB1 function by introducing the
   ectopic expression of human papillomavirus 16 E7. E7 inhibits RB1
   function by binding to RB1, which disrupts the RB1/E2F complex leading
   to release and activation of E2F, and by enhancing RB1 degradation
   ([152]37, [153]42, [154]43). While E2F activation by E7 can also lead
   to RB1 activation through p16 and CDK4/6 ([155]42), this cannot occur
   in our panel of p16-null NF1-MPNST cell lines ([156]Fig. 3B). To test
   the hypothesis that LOF of RB1 confers TNO155 resistance, we generated
   stable cells expressing WT E7 or E7 d21-24 (a mutant incapable of
   binding to RB1), a dominant negative control. We observed that WT
   E7–expressing cells were more resistant to TNO155, relative to parental
   and E7 d21-24 mutant cells, with the median inhibitory concentration
   increase of 5-fold in JH-2-002 and greater than 20-fold in ST8814
   ([157]Fig. 3D). By measuring the direct protein expression of cyclin D1
   and ERK signaling intermediates p-MEK and p-ERK, we found that
   treatment with TNO155 for 48 hours in E7 d21-24 mutant and parental
   cells was more potent than treatment in E7 WT cells ([158]Fig. 3E),
   suggesting that the E7 mutant with disabled RB1 binding restores
   sensitivity to TNO155. Therefore, we conclude that RB pathway is
   required for TNO155 sensitivity, and this evidence served as a
   justification for further exploration of targeting cell cycle/RB
   signaling in NF1-MPNST.

Responses of NF1-MPNST cells to the CDK4/6i ribociclib

   Inactivation of RB function as a consequence of recurrent loss of
   CDKN2A (encoding p16/INK4a; [159]Fig. 3B), hyperactivation of CDK, and
   up-regulated expression of RABL6A [a negative regulator of RB1;
   ([160]31, [161]32)] suggests that small-molecule CDK4/6i may be an
   additional therapeutic strategy in MPNST, particularly as a combination
   strategy. Seven of the 10 NF1-MPNST cell lines displayed limited
   sensitivity to the CDK4/6i ribociclib in vitro, by measuring metabolic
   activity as an indication of cell viability ([162]Fig. 4A). A
   complementary 3D microtissue assay revealed variable levels of
   sensitivity of tumor tissues derived from three individual MPNST PDXs
   to ribociclib as well (fig. S4A). To understand mechanisms accounting
   for the insensitivity in some lines, we carried out a time course
   combined with dose response study over 96 hours. We discovered that,
   while all three CDK4/6is—abemaciclib, palbociclib, and
   ribociclib—effectively inhibited p-RB as a readout of reduction in
   cyclin D–CDK4/6 complex activity in a dose-dependent manner, the three
   CDK4/6is all increased p-CDK4 at threonine-172 (T172), an indicator of
   CDK4 activation (fig. S4B), a phenomenon also recently observed in
   breast cancer cells, which suggests that CDK4/6is stabilize the active
   p-T172 CDK4–cyclin D complex ([163]44–[164]46). We next tried to
   validate this observation and focused on ribociclib in an expansion
   cohort of eight NF1-MPNST cell lines ([165]Fig. 4B and fig. S4C).
   Treatment with 1 μM ribociclib not only effectively reduced p-RB and
   later phase effectors of cell cycle, polo-like kinase 1 (PLK1) and
   cyclin A, but also caused a slight induction of ERK signaling, which,
   in turn, increased cyclin D1, cyclin E, and p-CDK4 T172 expression
   concurrently ([166]Fig. 4B and fig. S4C). This suggests that CDK4/6is
   induce an adaptive signaling response that reactivates the complex
   activity of cyclin D–CDK4/6, likely through the RTK/RAS/ERK pathway due
   to the high dependency of cyclin D expression on ERK signaling
   ([167]47). To validate this hypothesis, we performed a proteomic
   profiling using human RTK phosphorylation array that detects 71 human
   RTK. Among these, tyrosine pan-phosphorylation of ErbB3, insulin-like
   growth factor I receptor (IGF-IR), Axl, and members in Ephrin signaling
   was up-regulated by ribociclib ([168]Fig. 4C and fig. S4D), which may
   further activate downstream RAS and cell cycle signaling, supporting
   the concept of combining CDK4/6i with agents targeting RAS/ERK pathway.

Fig. 4. Responses of NF1-MPNST cells to the CDK4/6i ribociclib.

   [169]Fig. 4.
   [170]Open in a new tab

   (A) Ten NF1-associated MPNST cell lines were treated with increasing
   doses of the CDK4/6i ribociclib for 5 days. Cell viability was
   evaluated by using the CCK-8 assay. (B) Two NF1-MPNST cell lines were
   treated with DMSO or 1 μM ribociclib over a time course. Signal
   intermediates in ERK and cell cycle pathways were assessed. (C)
   JH-2-002 cells were treated with DMSO or 1 μM ribociclib for 24 hours.
   Seventy-one phosphorylated human RTKs were evaluated using human RTK
   phosphorylation array C1. Signal intensity from technical duplicates
   was quantified using densitometry analysis and normalized to ribociclib
   v. DMSO, and, notably, altered RTKs are shown. (D) Cells as in
   [171]Fig. 3E were treated with increasing doses of ribociclib for 5
   days. Cell viability was evaluated by using the CCK-8 assay.

   Previous reports revealed that, in addition to the activation of
   upstream signaling pathways ERK/AKT/Hippo, the expression levels of
   selected cell cycle regulators (including CDK2/4/6, cyclins, p16, p21,
   E2F, p-CDK4 T172, RB, FOXM1, AURKA, p53, and MDM2/4), as well as the
   dissociation of p21 from cyclin D–CDK4 complex, have putative relevance
   to CDK4/6i resistance ([172]45, [173]48). We investigated additional
   explanations for CDK4/6i resistance observed in some cell lines.
   Specifically, we measured protein expression of several cell cycle
   regulators and signaling molecules in ERK/AKT pathways but did not find
   a universal association of baseline expression of any single molecule
   with responsiveness to ribociclib across our NF1-MPNST cell line panel
   ([174]Fig. 3B and fig. S3, A and B). As demonstrated in [175]Fig. 3
   that RB function is critical for TNO155 efficacy, we also observed that
   it may be necessary for ribociclib efficacy as well, as resistance to
   ribociclib results from loss of RB function by WT E7, compared to
   mutant E7 d21-24 ([176]Fig. 4D). This suggests that CDK4/6 inhibition
   could be effective against MPNST in combination with SHP2i.

Combined inhibition of SHP2 and CDK4/6 effectively suppresses MPNST cell
growth

   LOF of RB due to the selective up-regulation of cell cycle signaling in
   MPNST relative to the precursor tumor, pNF, contributes partially to
   the insensitivity of NF1-MPNST to inhibitors of RAS/ERK signaling,
   including SHP2i and CDK4/6i ([177]Figs. 3 and [178]4). Therefore, we
   hypothesized that reinstatement of functional RB by CDK4/6i would
   enhance the treatment efficacy of SHP2i. In addition, SHP2i might
   attenuate the adaptive resistance to CDK4/6i through suppression of
   RTK/RAS/ERK signaling, thereby potentiating and transforming the
   cytostatic effect of CDK4/6i to be cytotoxic. With this in mind, we
   tested the combinatorial effect of TNO155 and ribociclib (referred to
   as the combination) on cell growth via three complementary methods, and
   we found the additive activity of this combination against short- and
   long-term cell growth ([179]Fig. 5, A to D, and fig. S5, A to C).
   Furthermore, the combination of high doses of TNO155 and ribociclib was
   able to overcome RTK-mediated trametinib resistance ([180]Fig. 5A and
   fig. S5B) ([181]19). To identify signaling changes that may lead to the
   additive effect, we investigated ERK and cell cycle pathways using
   eight NF1-MPNST cell lines following 72- and 96-hour treatment with
   TNO155 and/or ribociclib. The combination mitigated ERK signaling
   (p-MEK and p-ERK), cyclin D1 expression, and p-CDK4 T172 induced by
   ribociclib and more potently inhibited RB phosphorylation, E2F1, cyclin
   E/A, and PLK1, in the majority of MPNST cell line models ([182]Fig. 5E
   and fig. S5D). Again, SHP2 knockdown by inducible shPTPN11 alleviated
   ribociclib-induced ERK signaling (fig. S5E) and resulted in greater
   decrease of cyclin A and PLK1 when combined with ribociclib
   ([183]Fig. 5F).

Fig. 5. Combined inhibition of SHP2 and CDK4/6 effectively suppresses MPNST
cell growth.

   [184]Fig. 5.
   [185]Open in a new tab

   (A) ST8814 and NF90.8 parental (Par) and trametinib-resistant (Res)
   lines and eight NF1-MPNST cell lines were treated with DMSO, TNO155
   (0.3, 1, and 3 μM), ribociclib (1 and 3 μM), and their combination for
   7 to 10 days. Direct cell counting using trypan blue exclusion assay
   was performed by TC20 automated cell counter. (B) Cells as in (A) were
   treated with drugs for 2 to 3 weeks, and colony formation was evaluated
   using crystal violet assay. (C) Area under the curve (AUC) is
   calculated on the basis of the IncuCyte cell confluence monitoring of
   the seven cell lines as shown, treated with DMSO, TNO155 (0.3 μM),
   ribociclib (1 μM), and their combination for 6 days. (D) Two NF1-MPNST
   cell lines were treated with DMSO, TNO155 (0.3 μM), ribociclib (1 μM),
   and their combination for 6 days. Cell confluence was monitored using
   IncuCyte imaging systems. (E) Four NF1-MPNST cell lines were treated
   with DMSO, 0.3 μM TNO155, and/or 1 μM ribociclib for 72 and 96 hours.
   ERK signaling and cell cycle regulators were evaluated using
   immunoblot. (F) Three NF1-MPNST cell lines transduced with shPTPN11
   #818 were pretreated with vehicle or Dox (300 ng/ml) for 72 hours,
   followed by treatment with DMSO or 1 μM ribociclib for additional 72
   hours. Cell lysates were assessed for expression of the indicated
   proteins.

Combination of TNO155 and ribociclib additively inhibits cell cycle and
induces apoptosis

   To deepen the understanding of mechanisms behind the increased efficacy
   of the combination, we performed RNA-seq of JH-2-002 after treatment
   with TNO155 and/or ribociclib for 24 hours. The corresponding samples
   to RNA-seq demonstrated signaling responses in line with our prior
   observations in [186]Fig. 5E (fig. S6A). As expected, 2503 genes were
   differentially expressed upon treatment with the combination compared
   to 1277 and 740 genes that were altered upon ribociclib and TNO155
   treatment alone ([187]Fig. 6A). Furthermore, more than half of the
   genes differentially expressed upon treatment with the combination
   (1444 genes) were differentially expressed in the individual treatments
   ([188]Fig. 6A and table S3). Pathway enrichment analysis against the
   Kyoto Encyclopedia of Genes and Genomes (KEGG) database revealed that
   cell cycle signaling is significantly enriched in the intersection set
   ribociclib and the combination (fig. S6B). Deeper investigation into
   cell cycle signaling demonstrated an additive inhibition of mitotic
   prometaphase ([189]Fig. 6B) and cell cycle checkpoints ([190]Fig. 6C)
   by the combination, relative to TNO155 and ribociclib alone. To
   validate this finding, we carried out cell cycle distribution analysis
   in five NF1-MPNST cell lines (ST8814, NF10.1, JH-2-002, JH-2-079c, and
   JH-2-103) as an aggregate using flow cytometry, and we saw an
   increasing trend of percent cells in G[1] phase and a decreasing trend
   of percent cells in S phase when comparing cells treated with DMSO,
   TNO155, ribociclib, and the combination ([191]Fig. 6D).

Fig. 6. Combination of TNO155 and ribociclib additively inhibits cell cycle
and induces apoptosis.

   [192]Fig. 6.
   [193]Open in a new tab

   (A) Upset matrix plot derived from RNA-seq analysis, showing the
   overlapping numbers of significant genes (P adj < 0.05 and |fold
   change| > 1.5) after 24-hour treatment with 0.3 μM TNO155, 1 μM
   ribociclib, and their combination, and normalized to DMSO control.
   Rows, the sets; columns, intersections between these sets. (B and C)
   Heatmaps demonstrating the more potent transcriptional inhibition of
   mitotic prometaphase (B) and cell cycle checkpoints (C) by combined
   TNO155 and ribociclib relative to either single agent alone,
   highlighting BIRC5, PLK1, CLSPN, and CCNA2 (P adj < 0.05 marked with +
   and LFC). (D) Five NF1-MPNST cell lines were treated with DMSO, 0.3 μM
   TNO155, and/or 1 μM ribociclib for 48 hours, following overnight
   starvation in 0.1% FBS-containing culture medium to synchronize cells.
   Cells were fixed in ice cold 70% ethanol and stained with propidium
   iodide/ribonuclease staining solution (Cell Signaling Technology, no.
   4087) and then analyzed by flow cytometry. (E) ST8814 and JH-2-079c
   were treated with DMSO, 0.3 μM TNO155, and/or 1 μM ribociclib for 48
   hours, and, then, protein lysates were assessed using human apoptosis
   antibody array (R&D Systems, no. ARY009). Signal intensity from
   technical duplicates was quantified by densitometry analysis using
   ImageJ and normalized to DMSO. Data represent means ± SEM. (F) Nine
   NF1-MPNST cell lines were treated as in (E), and the indicated proteins
   involved in apoptosis and ERK signaling were detected using immunoblot.

   As cell proliferation and programmed cell death are closely
   interconnected, we screened for the relative expression of 35
   apoptosis-related proteins under these four conditions. Using one
   traditional and one patient-derived cell lines as a representative
   cohort, we found universally decreased expression of two inhibitors of
   apoptosis proteins (IAPs) survivin/BIRC5 and claspin/CLSPN in ST8814
   and JH-2-079c treated with the combination ([194]Fig. 6E and fig. S6C)
   and increased caspase-3/7 activation (fig. S6D). survivin/BIRC5,
   claspin/CLSPN, PLK1, and CCNA2 are also down-regulated upon the
   combination treatment ([195]Fig. 6, B and C, highlighted), suggesting
   that the regulation occurs at the transcriptional level. In an expanded
   validation cohort of nine NF1-MPNST cell lines, survivin, claspin, and
   PLK1 were more potently inhibited by the combination, relative to
   either single-agent alone ([196]Fig. 6F and fig. S6E). These results
   demonstrate that the combined use of TNO155 and ribociclib prevents
   signaling adaptation to CDK4/6i and converges on the restoration of RB
   function, which, in turn, induces G[1] cell cycle arrest and
   transcriptionally inhibits the IAP survivin and claspin to activate
   caspase-3/7, subsequently activating apoptosis.

Combination of TNO155 and ribociclib is active against MPNST tumor growth in
vivo

   Given our results demonstrating the additive in vitro effects of the
   combination, we hypothesized that SHP2i and CDK4/6i may be the superior
   combination of RAS signaling pathway inhibitors in MPNST, due to better
   or comparable in vivo efficacy compared to SHP2i and MEKi, and have
   more favorable tolerability, as judged by mouse body weight in both
   immuno-compromised and competent models ([197]49–[198]52). To address
   whether this combination is active in in vivo models of MPNST, we used
   a total of six unique PDX models to assess the combination versus
   either single agent, using doses (TNO155, 7.5 mg/kg, twice daily; and
   ribociclib, 75 mg/kg, once daily) lower than those selected in
   tolerability studies ([199]50). TNO155 has unexpectedly potent
   single-agent activity across the six PDXs over 4 weeks ([200]Fig. 7, A
   and B), compared to our previous results using the tool compound SHP099
   ([201]20). We thus defined the top three PDXs that were almost equally
   sensitive to the combination and TNO155 alone as very sensitive (VS;
   [202]Fig. 7A), and the bottom three PDXs that displayed differential
   sensitivity to the combination and TNO155 as partially sensitive (PS;
   [203]Fig. 7B). We directly compared the efficacy of TNO155 alone and
   TNO155/ribociclib combination against the TNO155/MEKi combination. We
   observed the most potent antitumor effects seen with TNO155 and
   ribociclib, relative to TNO155 or TNO155 and trametinib in WU-545, and
   similar responses between the two combinations driven by TNO155
   sensitivity across a panel of the remaining four PDXs (fig. S7A),
   consistent with previous studies in both RTK-driven and KRAS-mutant
   cancers ([204]50).

Fig. 7. The combination of TNO155 and ribociclib is active against MPNST
tumor growth in vivo.

   [205]Fig. 7.
   [206]Open in a new tab

   (A and B) Five-to-6-week-old female NRG mice bearing six individual
   NF1-MPNST PDXs were treated with vehicle, ribociclib (75 mg/kg, once
   daily), TNO155 (7.5 mg/kg, twice daily), or their combination by oral
   gavage for 4 weeks. Average tumor volume of four to five mice per arm
   was plotted over the time course of treatment days. VS, very sensitive;
   PS, partially sensitive. (C) Tumors of each arm from WU-386 were
   harvested 4 hours after last dose of 4 weeks on drugs and fixed in 10%
   NBF. The Ki-67 expression was assessed using immunohistochemistry
   (IHC). (D and E) Tumors of each arm from WU-225, WU-386 were harvested
   4 hours after last dose, 4 weeks on drugs (D); or 24 hours after last
   dose, 3 days on drugs (E). The indicated proteins involved in ERK and
   cell cycle signaling were detected using immunoblot. (F) Onco-print of
   key driver genes in MPNST and putative others is shown. For NF1 and
   SUZ12, both germline (G) and somatic (S) mutations are shown. (G)
   Volcano plot demonstrating LFC of VS/PS as a function of −log[10] (P
   adj). Blue dots represent 227 genes that were significantly altered (P
   adj < 0.05 and |LFC| > 0.585) when comparing the RNA-seq data of VS
   (JH-2-031, WU-225, and JH-2-002) v. PS (WU-545, JH-2-079c, and WU-386).
   (H) The mice bearing JH-2-031 were on initial treatment as in (A) for
   about 4 weeks and left untreated for another 4 weeks before
   rechallenging with the combination for additional 2 weeks. (I) The mice
   bearing JH-2-002 as in (A) were continuously on treatment for 10 weeks.
   (J) Tumors of each arm from JH-2-002 and JH-2-031 were harvested 4
   hours after last dose of 10 weeks and fixed in 10% NBF. The Ki-67
   expression was assessed using IHC.

   In addition, we investigated the pharmacodynamic (PD) activity of the
   combination to identify potential molecular response features
   underlying the therapeutic response. We biochemically characterized ERK
   and cell cycle signaling, proliferation, and apoptosis at 4 hours after
   4-week treatment ([207]Fig. 7, C and D, and fig. S7, B to D), and 24
   and 48 hours after 3-day treatment ([208]Fig. 7E and fig. S7E). Ki-67
   expression, a marker of proliferation, was more potently reduced by the
   combination compared to TNO155 or ribociclib, as evidenced by
   immunohistochemistry (IHC) staining at the end points of treatment
   ([209]Fig. 7C and and fig. S7B). Early and end-point in vivo PD studies
   were in agreement with what we saw in vitro that the combination
   relieved the adaptive response to ribociclib and led to greater
   suppression of cell cycle regulators, such as p-RB, cyclin A, PLK1, and
   survivin ([210]Fig. 7, D and E, and fig. S7, D and E). Moreover,
   cleaved caspase-3 increased in WU-225 (fig. S7C), and the IAP survivin
   markedly decreased in WU-386 ([211]Fig. 7E and fig. S7E), in tumors
   treated with the combination compared to those with either drug,
   indicating enhanced apoptosis elicited by the combination.

   To ask whether genomic or expression biomarkers exist to predict de
   novo sensitivity to TNO155, we next analyzed whole exome and targeted
   gene sequencing ([212]Fig. 7F) and RNA-seq data ([213]Fig. 7G) derived
   from VS PDXs (JH-2-002, JH-2-031, and WU-225) and PS PDXs (JH-2-079c,
   WU-545, and WU-386). The VS PDX JH-2-031 with activating PIK3CA Q546K
   mutation and WU-225 with TSC2 in frame deletion retained sensitivity to
   TNO155 ([214]Fig. 7F). Several putative molecules relevant to drug
   resistance, such as Wnt family member 2B (WNT2B) and neurotrophic
   receptor tyrosine kinase 2 (NTRK2), were significantly up-regulated (P
   adj < 0.05 and |fold change| > 1.5) in PS compared to VS models
   ([215]Fig. 7G, fig. S7F, and table S4). Although we did not observe a
   difference in sensitivity between TNO155 alone and TNO155 and
   ribociclib at 4 weeks of treatment in the VS PDX, there is a sufficient
   concern that single-agent treatment in this aggressive malignancy would
   be insufficient for anything longer than an immediate short-term
   response. To investigate this further, we retreated the PDX JH-2-031
   cohorts of TNO155 single-agent and the combination with TNO155 and
   ribociclib after a drug holiday ([216]Fig. 7H and fig. S7G), or treated
   the PDX JH-2-002 for an extended course with TNO155 alone or the
   combination ([217]Fig. 7I). We observed combination benefit regardless
   of intermittent ([218]Fig. 7H) or continuous dosing of therapy over a
   10-week treatment ([219]Fig. 7I), indicating that the combination
   resulted in more profound and/or more durable antitumor effects, which
   was supported by a significant difference in Ki-67 ([220]Fig. 7J) and
   cyclin D1 (fig. S7G), and that treatment-emergent resistance to TNO155
   may limit the duration of response.

DISCUSSION

   Our collective data demonstrate that cell proliferation and tumor
   growth of NF1-MPNST are SHP2/PTPN11-dependent. Using both SHP2 genetic
   ablation and pharmacological inhibition, we observe effective reduction
   in MPNST cell proliferation through ERK pathways and downstream cell
   cycle suppression. Reconstitution of NF1 GAP function mediates
   resistance to TNO155, suggesting that NF1-null MPNST are sensitive to
   SHP2i and further that NF1 LOF may be an important genomic biomarker
   for patient selection as clinical trials of SHP2i advance. TNO155
   treatment causes a transcriptional response which significantly
   overlaps with that induced by MEKi, suggesting that the effects of
   SHP2i are largely mediated through ERK, although other RAS effector
   pathways may exert secondary roles. This redundancy in pathway
   inhibition suggests the potential for overlapping clinical toxicity,
   and so alternate combination strategies that enhance the antitumor
   effects of SHP2i and maintain tolerability may be important for trial
   design.

   Consistent with previous studies ([221]28), cell cycle signaling is
   significantly up-regulated in MPNST versus pNF and results in LOF of
   RB, suggesting the role for CDK4/6i as part of a combination strategy.
   NF1-MPNST cells are insensitive to CDK4/6i, likely at least, in part,
   due to adaptive resistance resulting from induction of RTK/ERK
   signaling and reactivation of cyclin D–CDK4/6 complex activity. While
   we did not find any correlation between baseline expression markers and
   sensitivity to TNO155 and/or ribociclib, our data suggest that
   functional RB is critical for both TNO155 and ribociclib sensitivity.
   TNO155 and ribociclib activity converges on restoration of biologically
   active RB, otherwise inactivated by up-regulation of CDK and
   down-regulation of p16, and on greater inhibition of later-phase cell
   cycle regulators PLK1 and cyclin A. Mechanisms of anti-proliferation
   and antitumor effects involve an additive impact on global
   transcription as well as an additive repression in gene transcription
   related to cell cycle signaling. The combination leads to greater G[1]
   cell cycle arrest and also more potently inhibits two apoptotic
   regulators, survivin and claspin, and induces downstream caspase
   activation to promote apoptosis.

   Last, we assessed a total of six PDX models for the efficacy of TNO155,
   ribociclib, and their combination. We found that three PDX models have
   a near complete response to TNO155 single-agent with no apparent
   difference induced by the combination in the short-term, while the
   other three PDX models are PS and have more complete responses with the
   addition of ribociclib. Overall, the combination is notably active in
   all in vivo PDX models of NF1-MPNST tested. RNA-seq analysis of the
   MPNST PDX revealed differential expression of multiple putative
   resistance-relevant genes involved in Wnt/β-catenin/yes-associated
   protein 1 (YAP)/transcriptional co-activator with PDZ-binding motif
   (TAZ) signaling pathways in the three PS models, suggesting that these
   pathways may mediate resistance to TNO155, and agents targeting these
   pathways may be another potential viable partner with SHP2i, which
   remains to be explored.

   Clinical trial efforts have focused on dysregulated RAS effector
   pathways that are critical to MPNST tumorigenesis but, to date, have
   been unsuccessful in people with advanced MPNST. This study suggests
   that, despite past failures using downstream inhibitors of RAS
   signaling, SHP2 may represent a critical node to inhibit MPNST
   tumorigenesis due to the loss of NF1 and that combinations of SHP2i
   with inhibitors of other critical effectors could have greater success
   in the clinic. Our extensive preclinical data demonstrate that
   clinically available inhibitors of SHP2 and CDK4/6 in combination and
   at well-tolerated dosing regimens result in deep and durable
   preclinical, antitumor responses in mouse models with similar efficacy
   as MEKi and SHP2i. In addition, this combination is already under
   investigation in an ongoing trial ([222]NCT04000529), and data to
   support the clinical safety profile of TNO155 and ribociclib are
   expected to result soon. Single-agent TNO155 has been reported to be
   well-tolerated in patients with solid tumors, with the majority of
   reported adverse events being grade 1 or 2 and consistent with
   anticipated on-target effects of SHP2 inhibition ([223]53). Our
   evidence is in favor of the advancement of novel SHP2i into clinical
   trials for the treatment of patients with MPNST.

   Both SHP2i and CDK4i are considered pro-senescence cancer therapies,
   which disable cell proliferative capacity and trigger an inflammatory
   process that ultimately eliminates the tumor cells ([224]49). Here, we
   have provided a large body of evidence demonstrating the synthetic
   lethality of TNO155 and ribociclib in both clinically relevant in vitro
   and in vivo models of MPNST. However, these models are limited to the
   tumor cell–intrinsic signaling response and use immunocompromised mice,
   which preclude the evaluation of the combination on antitumor immunity.
   As such, further studies on treatment efficacy and the role of the
   combination in modulation of tumor immune microenvironment are needed
   in immunocompetent and humanized mouse models of MPNST, aiming to
   identify novel therapeutic opportunities based on mechanistic
   rationale.

   We have generated strong preclinical evidence in support of this
   combination, suggesting that a clinical study evaluating the efficacy
   and tolerability of TNO155, alone or in combination with ribociclib,
   should be pursued in patients with MPNST. Future studies will identify
   potential genomic or expression biomarkers that predict treatment
   efficacy to facilitate selection of patients who will have clinical
   benefit. With the clinical testing of TNO155 and ribociclib now
   underway in patients with other solid tumors, these studies are timely
   and relevant and will serve as an important contribution to the
   advancement of novel SHP2i combinations in clinical use. Last, acquired
   resistance to targeted therapy is common and inevitable. Exploration of
   resistance mechanisms will identify novel combination partners with
   SHP2i and move SHP2i-centric therapeutics forward for optimal clinical
   application.

MATERIALS AND METHODS

Experimental design

   We recently reported that the adaptive response to MEKi in MPNST
   involves the up-regulation of activity of multiple RTKs ([225]19,
   [226]20). Therefore, we hypothesized that SHP2 inhibition might be a
   strategy to overcome the global RTK up-regulation seen with
   single-agent MEKi, and we subsequently demonstrated that combined
   MEKi/SHP2i is additive in both in vitro and in vivo models of NF1-MPNST
   ([227]20). Here, we focus on SHP2i as a primary strategy to inhibit
   MPNST proliferation and tumor growth and identify CDK4/6i as a
   potential partner agent to be tested in combination with SHP2i, thus
   avoiding the overlapping toxicity of two agents that both inhibit
   p-ERK. Conversely, as loss of CDKN2A is a common feature in MPNST, we
   hypothesized that CDK4/6i may be active and further potentiated by
   drugs targeting upstream regulators of RAS, such as SHP2. In vitro
   experiments were performed at least two times. For in vivo studies,
   five mice per arm were randomly assigned to either agent alone or the
   combination and followed for at least 4 weeks. The investigators were
   not blinded to group allocation during data collection and analysis.

Cell lines, antibodies, and reagents

   Four patient-derived NF1-MPNST cell lines (JH-2-002, JH-2-031,
   JH-2-079c, and JH-2-103) and eight PDXs [JH-2-002, JH-2-031, JH-2-079c,
   WU-225, WU-386, WU-545, MN-2, and MN-3-002 (the last two for 3D tumor
   microtissue)] were generated in our laboratories at Johns Hopkins (JH),
   Washington University (WU; St. Louis), or University of Minnesota (MN)
   from biospecimens collected during surgical resection from patients
   with NF1 ([228]8, [229]34). Material was collected under Institutional
   Review Board–approved protocols (JH, no. J1649; WU, no. 201203042; and
   MN, no. STUDY00004719). All patients provided written informed consent.
   All cell lines used in these experiments were verified by short-tandem
   repeat profiling for cell line authentication at Johns Hopkins
   University Core Facility, tested negative for Mycoplasma contamination,
   and passaged in vitro for fewer than 3 months after resuscitation. All
   growth media were supplemented with 10% fetal bovine serum (FBS), 2 mM
   l-glutamine, and 1% penicillin-streptomycin. Trametinib-resistant cell
   lines were maintained in complete growth medium and 20 nM trametinib
   ([230]19).

   TNO155, ribociclib, and trametinib were provided under a material
   transfer agreement with Novartis Institute for Biomedical Research.
   Palbociclib, abemaciclib, and SHP099 were purchased from SelleckChem.
   Drugs for in vitro studies were dissolved in DMSO to yield 10 or 1 mM
   stock solutions and stored at −20°C. The sources and catalog numbers of
   cell lines, plasmids, antibodies, and reagents are detailed in table
   S5.

Immunoblotting

   Cells were disrupted on ice in radioimmunoprecipitation assay lysis
   buffer. Protein concentration was determined with a Pierce
   bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific,
   no. 23227). Equal amounts of protein were separated by
   SDS–polyacrylamide gel electrophoresis, transferred to nitrocellulose
   membranes, immunoblotted with specific primary and secondary
   antibodies, and detected by chemiluminescence with the
   electrochemiluminescence (ECL) detection reagents, Immobilon Western
   chemiluminescent horseradish peroxidase (HRP) substrate
   (MilliporeSigma, no. WBKLS0500), or Pierce ECL Western blotting
   substrate (Thermo Fisher Scientific, no. 32106). The membranes were
   imaged using ChemiDoc touch imaging system (Bio-Rad).

Cell proliferation assay

   Cells were seeded in 96-well plates at 2000 cells per well. A dose
   range of the compound indicated was prepared by serial dilutions and
   then added to the dishes containing adherent cells. Cells were
   incubated with drug for the indicated time. Cell growth was quantitated
   using the Cell Counting Kit-8 (Dojindo). For each condition, three
   replicates of each concentration were measured. Relative survival in
   the presence of drugs was normalized to the untreated controls after
   background subtraction. Graphs were generated using Prism 9 on the
   basis of the average of three replicates.

Active RAS pull-down assay

   Cells were seeded in 10-cm dishes. The following day, the 70 to 80%
   confluent cells were collected, and GTP-bound RAS was quantified using
   an active RAS detection kit (no. 8821) from Cell Signaling Technology
   according to the manufacturer’s instructions.

Lentivirus-based inducible shRNA-mediated knockdown cells

   shRNAs targeting PTPN11 #5003 and #818 [oligo sequences listed in table
   S6; ([231]29)] were subcloned into Tet-pLKO-puro vector (Addgene, no.
   21915). The lentiviruses encoding shPTPN11 were packaged in human
   embryonic kidney (HEK) 293T cells. The target cells were infected
   overnight by filtered viral supernatant in the presence of polybrene (8
   μg/ml; Millipore, no. TR-1003-G). The infected cells were selected with
   puromycin (2 μg/ml) for 1 week before the analysis of the knockdown
   effects.

Retrovirus-based gene expression cell system

   The human NF1-GRD ([232]54) was amplified using the genomic
   deoxyribonucleic acid (gDNA) from HEK293T as polymerase chain reaction
   template and subcloned into TTIGFP-MLUEX vector harboring Tet-regulated
   promoter (primer sequences are listed in table S6). The retroviruses
   encoding the rtTA3, NF1-GRD, E7, or E7 d21-24 were packaged in
   Phoenix-AMPHO cells. The medium containing virus was filtered with
   0.45-μm polyvinylidene difluoride filters followed by incubation with
   the target cells for 8 hours. The cells were cultured in virus-free
   medium for 2 days and then selected with hygromycin (300 μg/ml),
   puromycin (2 μg/ml), or neomycin (800 μg/ml) for 1 week. For
   TTIGFP-MLUEX Dox-inducible system, the positively infected cell
   populations were further sorted using a FACSMelody cell sorter (BD
   Biosciences) with GFP as a marker after overnight exposure to Dox (1
   μg/ml), and the sorted positive cells were cultured and expanded in
   medium without Dox but with antibiotics at a maintaining dose until the
   following assays.

Colony formation assay

   NF1-MPNST cells were treated with DMSO, TNO155 (0.3, 1, and 3 μM),
   and/or ribociclib (1 and 3 μM) for 2 to 3 weeks. Cells were washed with
   phosphate-buffered saline (PBS), fixed with 10% neutral-buffered
   formalin, and then stained with 0.1% crystal violet for 30 min.

p-RTK array

   Proteome profiler human RTK phosphorylation array C1 (RayBiotech, no.
   AAH-PRTK-1) was used to evaluate 71 p-RTKs as per manufacturer’s
   instructions. Briefly, cells were rinsed with cold PBS and lysed in
   lysis buffer, and 500 μg of lysates were incubated with blocked
   membranes overnight at 4°C. Membranes were subsequently washed,
   incubated with diluted HRP detection antibody and chemiluminescent
   reagent mix, and imaged on the ChemiDoc touch imaging system.

Apoptosis antibody array

   NF1-MPNST cells were treated with DMSO, 0.3 μM TNO155, and/or 1 μM
   ribociclib for 48 hours, and, then, 500 μg protein of lysates were
   assessed using human apoptosis antibody array (R&D Systems, #ARY009).

Cell cycle analysis by flow cytometry

   Modulation of the cell cycle was determined in NF1-MPNST cells treated
   with DMSO, TNO155, ribociclib, or their combination for 48 hours,
   following incubation with culture medium containing 0.1% FBS for 24
   hours to synchronize cells. After trypsinization, cells were fixed in
   ice-cold 70% ethanol for at least 30 min and were then labeled with
   propidium iodide/ribonuclease staining solution (Cell Signaling
   Technology, no. 4087) and further incubated for 15 min at 37°C. Last,
   cells were analyzed using the FACSCelesta Cell Analyzer (BD
   Biosciences). Data analysis was performed using FlowJo 10.8, and cell
   cycle distribution was assigned by using the implemented models.

In vivo mouse studies

   Nonobese diabetic Rag gamma (NRG; no. 007799) female mice were
   purchased from the Jackson Laboratory. All mouse experiments were
   approved by the Institutional Animal Care and Use Committee at Johns
   Hopkins. Minced tumor fragments from donor mice were implanted
   subcutaneously close to the sciatic nerves of 6- to 8-week-old NRG
   female mice. Drug treatment was started when tumor size reached roughly
   100 to 200 mm^3. Mice were randomized into treatment groups by an
   algorithm that arranges animals to achieve the best-case distribution
   to ensure that each treatment group has similar mean tumor burden and
   SD. Vehicle, trametinib (0.15 mg/kg for JH-2-079c and 0.075 mg/kg for
   others, once daily, Novartis; dissolved in 0.5% hydroxypropyl methyl
   cellulose and 0.2% Tween 80 in water, pH 8), TNO155 (7.5 mg/kg, twice
   daily, Novartis; dissolved in 0.5% methyl cellulose and 0.1% Tween 80
   in water), ribociclib (75 mg/kg, once daily, Novartis; dissolved in
   0.5% methyl cellulose in water), or their combinations were
   administrated by oral gavage on the basis of mean group body weight,
   with treatment schedule of 5 days on/2 days off. The end point of the
   experiment for efficacy studies was considered 4 weeks on treatment
   (extended on the basis of depth of response) or the longest tumor
   diameter of 2 cm as per our approved protocol, whichever occurred
   first. Tumors were measured twice weekly by calipering in two
   dimensions, and tumor volume was calculated by L × W^2(π/6), where L is
   the longest diameter and W is the width.

Whole exome sequencing

   Whole-exome sequencing FastQ files were trimmed by using Trimmomatic v
   0.39 ([233]55) and aligned against reference sequence hg38 via
   Burrows-Wheeler Aligner–Maximal Exact Match (BWA-MEM) ([234]56).
   Duplicated reads were marked by using PICARD MarkDuplicates function.
   Genomic analysis toolkit (GATK) V4.2 base quality score recalibration
   was also used to process BAM files. For PDX sequence data, XenoSplit
   was used to filter out mouse-derived reads using mouse (GRCm38) and
   human (hg38) reference genomes. Somatic single-nucleotide variants
   (SNVs) and small indels were detected using VarScan2 ([235]57),
   Strelka2 ([236]58), MuTect2 ([237]59), and Pindel ([238]60). Variant
   filtering and annotation were done by using Variant Effect Predictor
   ([239]61). Common variants found in the 1000 Genomes minor allele
   frequency (MAF) and GnomAD MAF > 0.05 were filtered out.

   For NF1, SUZ12, and CDKN2A germline and somatic deletion, BAM files
   from normal blood or tumor were used for the germline or somatic
   deletion calling, respectively, by using GATK 4.2. The main steps
   included the following: (i) CollectReadCounts function was used to
   extract the counts of reads based on the IDT xGen Exome Research Panel
   v1.0 bed file; (ii) DetermineGermlineContigPloidy function was used to
   determine the baseline contig ploidy for germline samples given the
   counts data; (iii) GermlineCNVCaller was used to call copy number
   variants in germline samples; and (iv) the B-allele frequency (BAF) was
   called by Hatchet 1.2.1 and normalized read ratio of tumor to normal,
   and BAF data were used to call somatic gene deletion. Onco-print of
   selected genes was manually constructed using Adobe Illustrator.

Bulk RNA-seq library preparation, sequencing, and analysis

   Primary tumor samples derived from 7 MPNST v. 31 pNF ([240]Fig. 3A),
   PDX samples from six patients ([241]Fig. 7G), and 30 samples
   (triplicate in 10 conditions) from JH-2-002 cells were used for
   RNA-seq. The library was prepared using a TrueSeq stranded total RNA
   kit with Ribo-Zero for ribosomal RNA depletion and was sequenced by
   NovaSeq6000 S4 (150–base pair paired-end reads) with targeted coverage
   of 60 million total reads per sample.

   FastQ files were aligned to GRCh38. RNA reads were quantified using the
   Salmon algorithm ([242]62). Raw counts were normalized, and the DESeq2
   package was used to call the differentially expressed genes ([243]63).
   Exploratory data analysis, including UpSet plots
   ([244]https://upsetplot.readthedocs.io/en/stable/), Venn Diagram
   creation, and heatmaps of expression values, were performed using the
   Python package MAGINE ([245]64), and ggplot2 was applied to draw the
   volcano plots ([246]65). Enrichment analysis was performed with respect
   to the KEGG_2021_human gene set through MAGINEs API with enrichR
   ([247]66).

Statistical analysis

   Two-way analysis of variance (ANOVA) was used to calculate statistical
   significance. Analyses were considered statistically significant if
   adjusted P < 0.05.

Acknowledgments