Met–HER3 crosstalk supports proliferation via MPZL3 in MET-amplified cancer cells

Receptor tyrosine kinases (RTKs) are recognized as targets of precision medicine in human cancer upon their gene amplification or constitutive activation, resulting in increased downstream signal complexity including heterotypic crosstalk with other RTKs. The Met RTK exhibits such reciprocal crosstalk with several members of the human EGFR (HER) family of RTKs when amplified in cancer cells. We show that Met signaling converges on HER3–tyrosine phosphorylation across a panel of seven MET-amplified cancer cell lines and that HER3 is required for cancer cell expansion and oncogenic capacity in vitro and in vivo. Gene expression analysis of HER3-depleted cells identified MPZL3, encoding a single-pass transmembrane protein, as HER3-dependent effector in multiple MET-amplified cancer cell lines. MPZL3 interacts with HER3 and MPZL3 loss phenocopies HER3 loss in MET-amplified cells, while MPZL3 overexpression can partially rescue proliferation upon HER3 depletion. Together, these data support an oncogenic role for a HER3–MPZL3 axis in MET-amplified cancers. Supplementary Information The online version contains supplementary material available at 10.1007/s00018-022-04149-w.

Receptor tyrosine kinases (RTKs) have been among the earliest and most effective targets for 51 precision medicine in human cancer (Yarden & Sliwkowski, 2001). RTK activation canonically 52 proceeds through receptor homo-oligomerization and trans-autophosphorylation, but many 53 cases of heterotypic signaling between different RTKs, frequently referred to as crosstalk, have 54 been reported in the literature (Paul & Hristova, 2019;Ullrich & Schlessinger, 1990). Epidermal 55 growth factor receptor (EGFR)-family RTKs in particular have become targets for clinical 56 intervention in human cancer. The EGFR family contains four paralogous receptor tyrosine 57 kinases that evolved from a single precursor EGFR homologue and exhibit extensive crosstalk 58 with each other (Yarden & Sliwkowski, 2001). EGFR, human epidermal growth factor receptor 2 59 (HER2) and HER3 are frequently overexpressed in human cancers, and have been shown to 60 induce canonical cancer-associated signals upon their activation by mutation, gene amplification 61 or constitutive ligand presentation ( Unlike EGFR and HER2, their paralogue HER3 does not possess intrinsic kinase activity due to 68 substitutions at critical positions in the kinase domain (Knighton et al., 1993). In order to 69 phosphorylate tyrosine residues in the HER3 cytoplasmic tail to recruit and activate intracellular 70 signaling molecules, it must interact with another functional tyrosine kinase (Kovacs et al., 2015).

71
HER3 exhibits this crosstalk preferentially with HER2 and EGFR, with which it can form 72 heterodimers (Sliwkowski et al., 1994;Yarden & Sliwkowski, 2001  Y. W. Zhang et al., 2013), this benefit was not supported in clinical trials (Spigel et al., 2013). 95 Nonetheless, MET amplification is recognized as a mechanism of resistance following small-96 molecule inhibitor targeting of EGFR mutant lung cancer, and conversely, EGFR and HER3 97 amplification or mutation have been observed as resistance mechanisms to Met  on Met activity in two lung cancer cell lines tested (EBC1 and H1993) ( Figure 1B). Tyrosine 137 phosphorylation of HER3, by contrast to EGFR and HER2, was dependent on Met activity in all 7 138 MET-amplified cell lines tested ( Figure 1B). This supports a preference for HER3 among EGFR 139 family RTKs for Met dependent tyrosine phosphorylation.

141
As the HER3 kinase domain possesses impaired intrinsic tyrosine kinase activity, tyrosine 142 phosphorylation of HER3 is normally attributed to its heterodimerization with EGFR or HER2 upon 143 binding its ligand neuregulin, thus utilizing EGFR or HER2 kinase activity to induce HER3-144 dependent signaling (Yarden & Sliwkowski, 2001 To test whether HER3 phosphorylation proceeded through one of these mechanisms in MET-156 amplified cells, we treated the Met-dependent cell lines in our panel with PHA (0.5 m); the EGFR 157 small molecule kinase inhibitor, gefitinib (1 m); the HER2 small molecule kinase inhibitor, 158 lapatinib (1 m); the Src-family broad-spectrum kinase inhibitor dasatinib (0.1 m); or DMSO as 159 control. Tyrosine phosphorylation of HER3 in all cell lines studied was decreased following 160 inhibition of the Met kinase using PHA but not with any of the other inhibitors tested, indicating 161 that Met kinase activity is primarily responsible for HER3 tyrosine phosphorylation in these MET-162 amplified cells ( Figure 1C). Furthermore, the observation that HER2 phosphorylation was 163 sensitive to lapatinib but not PHA in Snu5, Okajima, KatoII and OE33 cells indicates that the Met-164 HER3 signaling axis is separated from HER2 activity in these MET-amplified cancer cells ( Figure  165 S1). These data demonstrate that HER3 phosphorylation is primarily regulated by Met in a MET-166 amplified setting, and the consistent observation of this across seven independently derived cell 167 lines further indicates that tyrosine phosphorylation of HER3 by Met is under strong selection in 168 MET-amplified cancers.

HER3 depletion impairs proliferation and reduces tumour growth in Met-dependent cancer cells 192
Recurrence of a Met-HER3 crosstalk axis across multiple MET-amplified cancer cell lines further 193 suggests that HER3 may act as a ubiquitous signal transducer downstream from Met in this 194 context. To test whether HER3 contributed to oncogenic transformation in MET-amplified cancer 195 cells, we depleted HER3 using two independent shRNA hairpins in MET-amplified EBC1, H1993 196 and KatoII cells ( Figure 2A). HER3 depletion decreased cell proliferation in all three MET-amplified 197 cell lines tested when compared to controls ( Figure 2B). HER3 knockdown also impaired the 198 ability of KatoII cells to form colonies in soft agar, indicating that impaired cell expansion also 199 affects oncogenic phenotypes canonically associated with Met-dependent transformation 200 ( Figure 2C) (Lai et al., 2014). In H1993 cells, which do not form colonies in soft agar under normal 201 conditions (data not shown), HER3 knockdown impaired colony-forming capacity in 2D cell 202 culture ( Figure 2D). This stood in contrast to our observations in non-MET-amplified HeLa cells, 203 in which HER3 knockdown did not exert any effect on clonogenic capacity in 2D cell culture 204 ( Figure S2A). This led us to test whether HER3 was required for MET-amplified tumour formation 205 in-vivo.

207
To evaluate if HER3 depletion impacted tumorigenicity we subcutaneously injected immune-208 deficient NSG mice with H1993 and KatoII cells harbouring stable knockdown of HER3, or control 209 cells infected with an empty shRNA expression vector (pLKO). Tumours formed rapidly from 210 KatoII control cells, whereas tumours grown from HER3-depleted cells formed with delayed 211 kinetics and were smaller than control tumours at all time points ( Figure 2D and S2B). Tumours 212 from HER3-depleted cells did not re-express HER3 protein, demonstrating that knockdown 213 remained stable through the length of the experiment ( Figure S2C). Subcutaneous tumours from 214 H1993 control cells similarly grew more rapidly than those grown from HER3-depleted cells at

242 243
Gene expression analysis of HER3-dependent transcripts across MET-amplified cell lines 244 We predicted that depletion of HER3 would affect proliferation in MET-amplified cells by 245 reducing the phosphorylation of signaling pathway proteins canonically activated downstream of 246 Met and HER3. To test this, we analyzed protein extracts from HER3-depleted EBC1, H1993 and 247 KatoII cells for phosphorylation of the activation loop residues (T202/Y204) on ERK1/2 for MAPK 248 pathway activity and markers of mTORC2 activity (S473) on the Akt signaling transduction protein 249 to test PI3K pathway activation. Phosphorylation of these molecules and the activation of their 250 downstream transcriptional targets, along with tyrosine phosphorylation of the dimerization 251 domain at position 705 in the transcription factor STAT3, have been shown to be critical for 252 proliferation, colony formation and tumourigenesis in a panel of MET-amplified gastric cancer 253 cell lines, including KatoII (Lai et al., 2014). However, we observed no significant change in these 254 signaling pathways in HER3-depleted cells ( Figure 3A). While Met-dependent HER3 signaling has 255 been reported to suppress apoptosis, we did not observe a significant increase in apoptosis by
LGR6 was 295 unaffected and BHLHE41 was decreased in expression in H1993 and KatoII but not EBC1 cells.

296
None of the genes elevated in expression upon HER3 knockdown in KatoII cells were differentially 297 expressed in EBC1 or H1993 cells ( Figure S4). NYNRIN is a gene with a putative RNA-binding 298 function, but it has not been characterized biochemically to our knowledge (Mahamdallie et al.,

308 MPZL3 is required for proliferation downstream of HER3 in MET-amplified cells 309
We tested the ability of MPZL3 to rescue HER3-dependent proliferation by overexpressing MZPL3 310 in EBC1 cells with or without knockdown of HER3. Empty-vector control and MPZL3-311 overexpressing EBC1 cells were then transfected with pooled siRNA targeting ERBB3 or control 312 duplexes ( Figure 4A). Cells were monitored for copy number for seven days, and HER3-depleted 313 cells were compared to control-siRNA-treated cells for vector-control (pLKO) and MPZL3-314 overexpressing (MPZL3) cells, respectively ( Figure 4B).  Figure 4B). This supports 318 that MPZL3 overexpression can partially overcome the cell expansion defect following HER3 319 knockdown in MET-amplified cells. 320 321

337 338
To further test the contribution of MPZL3 to MET-amplified cells, we depleted MPZL3 using 339 shRNA in cell lines in our panel. KatoII cells were impaired in their ability to form colonies in soft 340 agar, as observed for HER3 knockdown ( Figure 4C-E). We monitored MPZL3 depletion in H1993 341 MET-amplified cell lines by clonogenic assay under adherent conditions and again observed a 342 dependence on MPZL3 similar to that seen for HER3 ( Figures 4F-H). These data support a model 343 whereby

374
To induced fusion of the FLAG tag to the respective bait protein ( Figure 5B). This indicated that 382 MPZL3 and HER3 could interact directly in the absence of Met. 383 384 Next, we tested whether we could uncouple the HER3-MPZL3 interaction by mutating the single, domain also strongly impaired co-immunoprecipitation of HER3 and MPZL3 upon their co-398 overexpression in HEK293T cells (16% of control) ( Figure 5E). and erlotinib, and treatment of gefitinib-resistant tumours with the Met inhibitor crizotinib is 477 effective in the clinic for these patients (Gainor et al., 2016). While this has been suggested to 478 depend on activation of the PI3K signaling pathways including the Akt pathway, in our panel of 479 MET amplified cells, PI3K activity-associated phosphorylation of Akt remained intact following 480 inhibition of Met kinase with small molecule inhibitors. We similarly did not observe a significant 481 increase in the apoptotic marker annexin-V in EBC1, H1993 and KatoII MET-amplified cells upon 482 depletion of HER3, or changes in known Akt-dependent gene expression (data not shown), 483 supporting the idea that Met-dependent activation of pro-survival signaling downstream of the 484 Akt pathway remained intact. This, in turn, indicated that the contribution of HER3 to Met-485 dependent signaling proceeded through a novel HER3-dependent mechanism, and did not 486 impact the survival pathways canonically associated with oncogenic HER3 signaling.

488
The highest-confidence and most differentially expressed candidate genes downstream of HER3 489 were genes whose expression decreased upon HER3 knockdown, including MPZL3, a gene interaction with HER3. 508 509 Our observations suggest that MET amplification co-opts HER3 and its binding partner MPZL3 to 510 support oncogenic cell proliferation via a previously-unreported mechanism. This was supported 511 by our observation that MET gene amplification predicted higher MPZL3 expression than cell lines 512 without any RTK amplification in the CCLE. While MPZL3 has been reported to act as a tumour 513 suppressor in cutaneous squamous cell carcinoma through its activity promoting differentiation 514 Antibodies and reagents 533 Antibody 148 was raised in rabbit against a C-terminal peptide of human Met ( Quantitative data are presented as the means ± SEM. Statistical significance was assessed using 661 a two-tailed Student's t test, and ordinary one-way ANOVA with Tukey's correction for multiple 662 comparisons, unless otherwise indicated, using Prism software. Significance is as follows: p > 663 0.05, not significant (ns); * p ≤ 0.05; * * p ≤ 0.01; * * * p ≤ 0.001; * * * * p ≤ 0.0001. Data distribution was 664 assumed to be normal, but this was not formally tested. P values and the number of experiments 665 used for quantification and statistical analysis are indicated in the corresponding figure legends.