Morphological and genetic screens reveal mechanisms of BiDAC-induced plasma membrane protein degradation

doi:10.21203/rs.3.rs-4438596/v1

Download PDF

Article

Morphological and genetic screens reveal mechanisms of BiDAC-induced plasma membrane protein degradation

https://doi.org/10.21203/rs.3.rs-4438596/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The discovery of bifunctional degradation activating compounds (BiDACs) has led to the development of a new class of drugs that promote the clearance of their protein targets. BiDAC-induced ubiquitination is generally believed to direct cytosolic and nuclear proteins to proteolytic destruction by proteasomes. However, pathways that govern the degradation of other classes of BiDAC targets, such as integral membrane and intraorganellar proteins, have not been investigated in depth. In this study we used morphological profiling and CRISPR/Cas9 genetic screens to investigate the mechanisms by which BiDACs induce the degradation of plasma membrane receptor tyrosine kinases (RTKs) EGFR and Her2. We found that BiDAC-dependent ubiquitination triggers the trafficking of RTKs from the plasma membrane to lysosomes for degradation. Surprisingly, functional proteasomes were required for endocytosis of RTKs upstream of the lysosome. Additionally, our screen uncovered a non-canonical function of the lysosome-associated arginine/lysine transporter PQLC2 in EGFR degradation. Our data show that BiDACs may target proteins to proteolytic machinery other than the proteasome and motivate further investigation of mechanisms that govern the degradation of diverse classes of BiDAC targets.

Biological sciences/Chemical biology/Mechanism of action

Biological sciences/Cell biology/Proteolysis/Lysosomes

Biological sciences/Cell biology/Proteolysis/Ubiquitylation

A mechanistic understanding of the relationship between extracellular signaling and cell proliferation has led to the development of pharmacological agents that mitigate the progression of human cancers. This has been exemplified by the development of antibodies and small molecules that block aberrant pro-growth signals generated by overexpressed receptor tyrosine kinase (RTK) proteins at the cell surface. Antibodies that target the extracellular domains of oncogenic RTKs, such as those found in the epidermal growth factor (EGF) receptor family members EGFR (also known as Her1) and Her2, inhibit RTK signaling by competing with extracellular RTK ligands for receptor binding, by inhibiting RTK dimerization, and by inducing RTK internalization and degradation¹. More recently, cell-permeable small molecule kinase inhibitors have been utilized to directly inhibit RTK activity and potentiate the anti-tumor effects of antibodies^2,3.

Extracellular ligand binding induces dimerization of RTK monomers and subsequent cross-phosphorylation of their intracellular domains. This mechanism has motivated the development of therapeutics that inhibit RTK dimerization. In particular, Her2 lacks a ligand binding domain and thus can be activated through dimerization with Her3 or EGFR bound to their respective ligands, neuregulin and EGF⁴. The Her2 antibody pertuzumab, unlike the first therapeutic Her2 antibody trastuzumab, inhibits Her2/Her3 heterodimerization and is used in cases where trastuzumab and kinase inhibitor therapies have lost effectiveness⁵. Similarly, the covalent kinase inhibitor neratinib also triggers Her2 degradation and may prevent dimerization of Her2 with other Her receptors⁶. Thus, therapeutic modalities that inhibit the scaffolding functions of Her2 may overcome the resistance that develops to other classes of Her2 inhibitors.

Recent advances in the development of bifunctional degradation activating compounds (BiDACs) have enabled the conversion of existing small molecules into potent degraders. These molecules recruit E3 ubiquitin ligase complexes to target proteins, resulting in their polyubiquitination and proteolysis by ubiquitin-dependent degradation pathways⁷. This mechanism presents opportunities to target proteins that also have essential scaffolding functions, since in theory, appending polyubiquitin chains to any lysine on the target protein may be sufficient to trigger its degradation. In agreement with this idea, BiDACs targeting BRD4, an epigenetic regulator that recruits transcriptional machinery to histones, triggered BRD4 degradation and exhibited increased potency relative to BRD4 inhibitors in hematological cancer models⁸⁹. This concept was investigated with respect to RTKs using the small molecule lapatinib, which inhibits the kinase domains of Her2 and EGFR. A BiDAC consisting of lapatinib conjugated to a molecule that binds the E3 substrate adaptor VHL triggered degradation of Her2 and EGFR. As with BRD4, degradation of these RTKs resulted in a more sustained inhibition of downstream signaling and a greater impact on cell viability than inhibition alone¹⁰. However, an unbiased comparison of the effects of RTK inhibition and degradation on cell physiology has not been performed.

The consequences of cellular perturbations can be explored using morphological profiling, a method that utilizes high-throughput microscopy and machine learning-based image analysis to establish morphological phenotypes representing cellular states induced by chemical, genetic or environmental perturbations. Prevalent techniques such as Cell Painting utilize a multiplexed panel of dyes to broadly label organelles and other cellular structures^11,12. Machine learning-based image analysis tools are used to extract and quantitatively compare information from microscopy images, and morphological profiles derived from such data have been shown to be expressive, high dimensional representations of cell state analogous to gene expression profiles¹³. Due to its integrative and information-rich nature, morphological profiling presents a promising approach for unbiased discovery of new phenotypes, accessing signals in the data that lie below the threshold of human perception^14,15.

In this study, we asked if BiDAC-dependent degradation of Her2 produces distinct cellular morphologies when compared to other modalities that inhibit Her2 signaling. We find that Her2 BiDACs induced distinctive changes in Her2 localization and levels. Functional genomic analysis of BiDAC-dependent degradation revealed that the localization changes reflected RTK internalization and degradation in lysosomes. In addition, we show that degradation of EGFR partially depends on a novel function of the lysosome-associated amino acid transporter, PQLC2. These data suggest that BiDACs can induce degradation of their targets through multiple ubiquitin-dependent pathways.

Characterization of RTK BiDACs

To investigate the phenotypic consequences of Her2 degradation, we used a previously reported BiDAC that consists of the Her2/EGFR inhibitor lapatinib linked to a VHL binding group (Lap-VHL). We also synthesized a Her2-specific BiDAC by replacing lapatinib with a tucatinib warhead linked to a VHL binding group by an azetidinyl piperidine linker (Tuc-VHL) (Supplementary Fig. 1). Treatment of two Her2-amplified cell lines, SKBR3 and SKOV3, with Lap-VHL resulted in degradation of Her2 and EGFR (Fig. 1a,b). Exposure to Tuc-VHL resulted in selective degradation of Her2 (Fig. 1a,b) and inhibition of Her2 kinase activity (Supplementary Fig. 2). Lap-VHL and Tuc-VHL containing a non-functional VHL binding group (VHL_Inactive) inhibited Her2 kinase activity but failed to induce Her2 degradation (Fig. 1a,b).

Active BiDACs reduced SKBR3 and SKOV3 viability (Fig. 1c,d) and SKBR3 growth (Supplementary Fig. 3) more potently than their inactive counterparts. This difference was most pronounced when cells were treated at doses that induced Her2 degradation (Fig. 1a,b). Thus, active and VHL_Inactive Her2 BiDACs can be used to compare the consequences of Her2 degradation and inhibition.

Her2 loss of function is associated with diverse morphological phenotypes.

To create unbiased morphological profiles of Her2 loss of function, we utilized a modified version of the Cell Painting protocol¹², combining probes for the nucleus, Golgi, mitochondria, and actin with an immunofluorescence probe for Her2 (Fig. 1e). As expected, Tuc-VHL induced a dose and time-dependent decrease in Her2 signal (Fig. 1f). We collected images of cells treated with a panel of Her2 inhibitors to contextualize the effects of BiDAC-based inhibitors. This panel included Her2-binding antibodies (Trastuzumab, Pertuzumab), covalent inhibitors (Neratinib, BDTX-189), inhibitors of downstream kinases PI3K (Alpelisib) and AKT (Ipatasertib), in addition to Tucatinib, Lapatinib and the VHL_Inactive compounds. Given the time-dependency of BiDAC-induced Her2 degradation, cells were imaged across multiple time points and treatment doses.

A machine learning model was trained to describe each image as a vector in a high-dimensional space. The averaged vector representations for each treatment condition were clustered and visualized on a two-dimensional projection (Fig. 1g). Points in proximity to each other suggest treatment conditions that produce similar morphological phenotypes. For SKOV3 cells treated for 48 hours, DMSO controls were localized to the center of the projection (Fig. 1g, gray outline) while inhibitors of downstream PI3K and AKT kinases produced distinctive, dose-dependent phenotypic trajectories in the lower right quadrant (Fig. 1g, purple outlines). In contrast, the BiDAC warheads Lapatinib and Tucatinib induced phenotypes that clustered together at the lower left quadrant, suggesting that these non-covalent Her2 inhibitors do not produce distinct cell states despite their differential specificity for EGFR (Fig. 1g, green outline). The active BiDACs formed a distinct cluster at the top of the projection which was adjacent to the cluster formed by Neratinib, a Her2 covalent inhibitor previously shown to both inhibit Her2 and trigger its degradation⁶ (Fig. 1g, red outline). The covalent inhibitor BDTX-189 appeared at the bottom of the plot away from Neratinib, indicating that inhibitors of the same target can produce distinct phenotypic profiles (Fig. 1g, blue outline). Active BiDACs administered at higher doses were dispersed away from the primary cluster, suggesting that they induced distinct degradation-dependent phenotypes. Lastly, we observed that Pertuzumab and Trastuzumab induced their own phenotypes that were not adjacent to any of the small molecule inhibitors (Fig. 1g, orange and yellow outlines), consistent with the distinct modes by which these antibodies disrupt Her2 activation. Morphological profiles generated for SKBR3 cells, which are more sensitive to Her2 inhibitors, resembled those for SKOV3, indicating that the clustering patterns were not dependent on degree of dependency on Her2 signaling (Fig. 1h).

Distinct, time-dependent clustering of active BiDAC-treated wells was correlated with the time required to achieve Her2 degradation (Fig. 1f). Indeed, distinct active BiDAC phenotypes were evident at some doses after only 6 hours of treatment (Fig. 2a,b, red outlines). To determine the extent to which these phenotypes depend on Her2 staining, we trained models lacking Her2 imaging information. These models showed that active BiDAC clusters were less distinctive and were closer to the VHL_Inactive BiDACs in the latent space of the model (Fig. 2c,d and 2e,f), a result that was observed in both cell lines at early and late time points (Supplementary Figs. 4 and 5). Upon manual examination of the images, we observed that BiDACs induced formation of Her2-postive puncta (Fig. 2g). These results suggest that changes in Her2 levels and localization are the primary features that account for distinct morphological phenotypes induced by active BiDACs.

Identification of machinery that functions in BiDAC-induced RTK degradation.

The unexpected relocalization of Her2 observed after treatment with functional BiDACs suggested that this phenotype is mechanistically related to RTK degradation. To investigate the mechanisms that govern BiDAC-induced RTK degradation in depth, we used a functional genomics approach that employed SKOV3 cells expressing Cas9 and GFP/RFP knocked-in at the genomic loci of Her2 and EGFR, yielding expression of Her2-GFP and EGFR-RFP (RTK-FP). Treatment of RTK-FP cells with Lap-VHL or Tuc-VHL induced the degradation of the expected RTKs (Fig. 3a,b). Treatment with epidermal growth factor (EGF), which binds to and downregulates expression of EGFR, resulted in degradation of EGFR-RFP but not Her2-GFP (Fig. 3c). Lap-VHL-induced degradation of both RTKs was partially blocked by addition of a competing VHL-binding molecule (Fig. 3d), consistent with dependence of RTK-FP degradation on VHL recruitment. Thus, RTK-FPs, like untagged RTKs, are amenable to degradation by BiDACs.

To perform the genetic screen, we infected RTK-FP cells with a whole genome CRISPR library containing 6 single guide RNAs (sgRNAs) per gene. Ten days after infection, cells were treated with 1 µM Lap-VHL, which resulted in a 70% or greater decrease in RTK-FP levels after 24 hours (Fig. 3a,e). Cells that failed to degrade RTK-FPs, defined as those having the 0.5-1% highest RFP or GFP fluorescence levels, were enriched by cell sorting in two separate screens (Fig. 3f). To determine the degree of sgRNA enrichment in the sorted populations, libraries from the sorted and the bulk unsorted populations were prepared from extracted genomic DNA, analyzed by next generation sequencing, and compared using MAGeCK analysis ¹⁶.

Cells that failed to degrade RTK-FPs expressed sgRNAs targeting key E3 complex components, including VHL, the Cullin scaffolding protein Cul2, and the elongin B/C adaptors TCEB2 and TCEB1¹⁷ (Fig. 3g-i). In addition, we detected enrichment of sgRNAs targeting ubiquitin, the ubiquitin-activating enzyme UBA1 (E1) and the Cullin-associated ubiquitin-conjugating enzyme UBE2R2. We also identified core neddylation machinery necessary for activation of cullin E3s, which consists of Nedd8, the heterodimeric NAE1-UBA3 Nedd8 activating enzyme and the UBE2M Nedd8-conjugating enzyme¹⁸. Interestingly, we identified all nine subunits of the signalosome, which functions in the removal of Nedd8 from Cullin scaffolds (Fig. 3i). This finding is consistent with a previous study which demonstrated that perturbation of the signalosome complex prevents BiDAC-mediated degradation by triggering the autodegradation of E3 substrate receptors¹⁹. We were surprised that sgRNAs targeting the proteasome were not enriched in our screens. Instead, we found enrichment of guides targeting multiple V_o and V₁ subunits of the lysosomal ATPase (V-ATPase) and its accessory subunit ATP6AP1²⁰ (Fig. 3j), suggesting that BiDAC-induced degradation of RTKs is mediated by the endolysosomal system.

BiDAC-induced degradation of RTKs requires functional lysosomes and proteasomes.

Lysosomal proteolysis requires acidification of lysosomes by the V-ATPase. To investigate the role of lysosomes in BiDAC-induced RTK degradation, we treated RTK-FP cells with the V-ATPase inhibitor, Bafilomycin-A1 (Baf-A1). Baf-A1 partially inhibited RTK-FP degradation by Lap-VHL (Fig. 4a,b) and Her2-GFP degradation by Tuc-VHL (Fig. 4c). Baf-A1 also blocked the BiDAC-induced degradation of untagged Her2 and EGFR in RTK-FP (Fig. 4d,e) and SKBR3 cells (Fig. 4f). Treatment with the E1 inhibitor MLN-7243 and the proteasome inhibitor Carfilzomib completely blocked RTK-FP degradation (Fig. 4a-c). We asked if the partial inhibition of EGFR-RFP degradation by Baf-A1 was common to EGF, which binds to EGFR and triggers its degradation by the endolysosomal pathway²¹. EGFR-RFP levels in cells treated with EGF were partially stabilized by Baf-A1 (Fig. 4g), suggesting that BiDACs and EGF induce EGFR degradation by similar mechanisms. To determine if BiDACs trigger trafficking of RTKs to lysosomes, we imaged RTK-FP cells expressing the lysosomal marker LAMP1-SNAP. In untreated cells, RTKs were present at plasma membrane projections and a limited number of lysosomes labeled by LAMP1-SNAP (Fig. 4h). Lap-VHL induced relocalization of RTKs to lysosomes, consistent with the BiDAC-dependent changes in Her2 localization observed by morphological profiling (Fig. 2g). Concurrent treatment with Baf-A1 resulted in a large increase in RTK-FP puncta that colocalized with lysosomes (Fig. 4h), likely due to impaired lysosomal degradation of RTK-FPs.

Our results raised the possibility that BiDACs trigger RTK degradation by both lysosome and proteasome-dependent pathways. Alternatively, inhibition of one pathway may impact RTK degradation by the other. To distinguish between these possibilities, we examined how the localization of RTK-FPs changes in response to BiDACs when used in combination with degradation inhibitors. First, we measured surface RTK levels by staining cells with antibodies against EGFR and Her2 (Fig. 5a). Treatment with BiDACs resulted in a decrease in levels of surface RTK-FPs (Fig. 5b-d). This decrease was not blocked by Baf-A1, consistent with internalization of RTK-FPs and delivery to lysosomes under these conditions (Fig. 4h). Interestingly, inhibition of ubiquitination or proteasome activity resulted in surface retention of RTK-FPs (Fig. 5b-d). Similar effects on EGFR-RFP were observed after treatment with EGF (Fig. 5e). To track changes in RTK-FP localization, we imaged cells treated with BiDACs and degradation inhibitors by time lapse microscopy. Lap-VHL, but not Lap-VHL_Inactive, triggered appearance of RTK-FP puncta within 2 hours of treatment, which decreased in abundance over the next 8 hours (Fig. 5f-h). Cotreatment with Carfilzomib or MLN-7243 prevented RTK-FP clearance and puncta formation, resulting in retention of RTK-FPs at the plasma membrane. In contrast, Baf-A1 did not inhibit puncta formation, which increased in abundance over time (Fig. 5g,h). Appearance of Her2-GFP-positive puncta was most evident after Baf-A1 treatment, likely due to dequenching of GFP fluorescence associated with the Baf-A1-induced increase in lysosomal pH (Fig. 5g). Together, these data show that proteasome inhibition prevents degradation of RTKs by blocking BiDAC-induced internalization.

The amino acid transporter PQLC2 regulates EGFR-RFP degradation by a novel mechanism.

We found that sgRNAs targeting the amino acid transporter PQLC2 (encoded by the SLC66A1 gene) were enriched in cells that failed to degrade EGFR-RFP (Fig. 6a). PQLC2 localizes to the lysosomal membrane where it transports the cationic amino acids (CAA) lysine, arginine and histidine from the lysosomal lumen to the cytosol^22,23. Deletion of PQLC2 in RTK-FP cells (PQLC2^KO), confirmed by sequencing its genomic locus (Fig. 6b), did not strongly impact basal EGFR-RFP expression levels (Fig. 6c). However, EGFR-RFP degradation by Lap-VHL was reduced in PQLC2^KO relative to control cells (Fig. 6c,d). Deletion of PQLC2 also inhibited EGFR-RFP degradation in response to EGF (Fig. 6e,f). We reasoned that loss of PQLC2 may block EGFR-RFP internalization, trafficking, or proteolysis in lysosomes. Imaging of PQLC2^KO cells expressing LAMP1-SNAP revealed residual EGFR-RFP signal in lysosomes following treatment with Lap-VHL (Fig. 6g), indicating that loss of PQLC2 inhibits lysosomal proteolysis of EGFR-RFP.

PQLC2 contains two proline-glutamine (PQ) domains that are critical for its CAA transport function (Fig. 7a)²⁴. PQLC2 also recruits the C9orf72-SMCR8-WDR41 (CSW) complex to lysosomes by binding to WDR41, an interaction that is enhanced when cells are starved of CAA²⁵. Two dileucine motifs mediate the localization of PQLC2 to lysosomes²⁶. To determine which functions of PQLC2 are necessary for EGFR-RFP degradation, we expressed doxycycline (dox)-inducible PQLC2-Flag variants in PQLC2^KO cells (Fig. 7b). PQLC2-Flag mutants were localized to puncta that were morphologically similar to those formed by wild type (WT) PQLC2 with the exception of PQLC2 (L288A/L289A)-Flag, which was present at peripheral structures consistent with the plasma membrane (Fig. 7c). In the absence of dox, EGFR-RFP degradation by Lap-VHL was attenuated in most cells, but a small population of cells exhibited more EGFR-RFP degradation due to leaky PQLC2 expression from the dox-inducible promoter (Fig. 7d). Dox induction of WT PQLC2 expression in all cells fully restored EGFR-RFP degradation to that of control cells (Fig. 7d,e). Surprisingly, expression of PQLC2 variants that contain inactivating mutations in their PQ domains (P55L and P201L)²⁴ resulted in complete rescue of EGFR-RFP degradation (Fig. 7f). Similarly, we observed rescue by PQLC2 variants containing mutations that disrupt binding to WDR41 (F49A, Y105A, and Y195A)²⁶, and the CSW complex (Fig. 7g). Lastly, mis-localization of PQLC2 to the plasma membrane through mutation of its dileucine motifs (L127A/L128A and L288A/L289A) resulted in complete rescue (Fig. 7h), though a fraction of dileucine-mutant PQLC2 is still present at lysosomes (see Fig. 8h below). Together, these results indicate that PQLC2 regulates EGFR-RFP degradation by a mechanism that is not governed by its known biological functions.

We observed that WT and mutant PQLC2 migrated at a size of ~ 15–18 kD by SDS-PAGE, at half of its predicted molecular weight of 32 kD (Fig. 7b). We reasoned that PQLC2-Flag undergoes cleavage, which yields a C-terminal fragment that is detected by the anti-Flag antibody. To investigate this possibility, we generated PQLC2^KO cells that expressed Myc and Flag tags at the C or N-terminus (Myc-PQLC2-Flag or Flag-PQLC2-Myc) and analyzed expression of the resulting N and C-terminally tagged versions. C-terminally Flag or Myc tagged PQLC2 migrated at the reduced size (Fig. 7i). However, we failed to detect the N-terminally tagged forms (Fig. 7j). To determine the putative cleavage site, PQLC2-Flag was purified by immunoprecipitation and analyzed using Edman sequencing (Supplementary Fig. 6). Ten amino acids corresponding to residues 172–181 were identified at the N-terminus, indicating potential cleavage after a glutamic acid residue that resides within a predicted intralysosomal loop (Supplementary Fig. 7). Mutation of residues directly adjacent to the putative cleavage site did not impact the size of PQLC2-Flag, and a fragment containing only the residues C-terminal to the cleavage site was not expressed (Fig. 7k). Mutation of a larger segment of the intralysosomal loop produced a protein that migrated at ~ 30 kD (Fig. 7k), confirming that this region contains the cleavage site. Expression of the cleavage-deficient mutant in PQLC2^KO cells rescued BiDAC-dependent EGFR-RFP degradation (Fig. 7l).

We were surprised that the C-terminal cleavage product of full length PQLC2 was sufficient to rescue lysosomal degradation of EGFR-RFP. To confirm that the cleaved form behaves like the putative full- length protein in cells, we determined the identity of its binding partners by co-immunoprecipitation followed by mass spectrometric analysis. Consistent with previous reports²⁵, WT but not PQ domain mutant PQLC2-Flag copurified with WDR41 (Supplementary Fig. 8) and other components of the CSW complex (Supplementary Fig. 9). However, both WT and mutant PQLC2 co-purified with V₀ subunits of the V-ATPase, including AT6V0A1 and ATP6BV0D1 (Supplementary Fig. 10). These data indicate that the cleaved form of PQLC2 can still bind its partners and suggest that the C-terminal product alone can rescue EGFR-RFP degradation, albeit by an unknown mechanism.

PQLC2 functions in the maintenance of lysosomal pH

Given the interaction of PQLC2 with subunits of the V-ATPase, we asked if loss of PQLC2 impacts lysosomal pH. We exploited the pH-sensitive fluorescence lifetime properties of mScarlet fluorescent protein²⁷ by generating PQLC2 null cells expressing mScarlet-LAMP1 (PQLC2^KO-pH), which localizes mScarlet to the lysosomal lumen. Incubation of PQLC2^KO-pH or control-pH cells in buffers of increasing pH containing ionophores resulted in a corresponding increase in mScarlet lifetime (Fig. 8a) across a range of pH 4 to pH 6 (Fig. 8b,c). When cells were cultured in complete media, mScarlet-LAMP1 lifetime was longer in PQLC2^KO-pH than control-pH cells, (Fig. 8d), which indicated that deletion of PQLC2 resulted in a higher lysosomal pH (Fig. 8e). PQLC2^KO-pH cells also contained significantly enlarged lysosomes (Fig. 8f). To determine if the elevated pH and size of PQLC2-deficient lysosomes is associated with a general defect in lysosomal degradation, we tracked lysosome-dependent turnover of LC3 in cells treated with the autophagy inducer Torin-1. Autophagy induction in control cells was accompanied by conversion of LC3-I to LC3-II and degradation of LC3-II over time (Fig. 8g). In contrast, LC3-II levels increased over time in PQLC2^KO cells (Fig. 8h), suggesting that LC3-I underwent conversion to LC3-II but was not efficiently turned over. Thus, loss of PQLC2 impairs lysosomal degradative functions.

We posited that the incomplete EGFR-RFP degradation in PQLC2^KO cells was a consequence of dysregulated lysosomal pH. We therefore asked if expression of the PQLC2 mutants that rescued EGFR-RFP degradation in RTK-FP cells also restored lysosomal pH in PQLC2^KO-pH cells. To detect PQLC2 in live cells, we expressed PQLC2 fused to a C-terminal SNAP tag (PQLC2-SNAP). The tag did not interfere with PQLC2 localization, as WT and cleavage deficient PQLC2-SNAP were localized to lysosomes, and the dileucine motif mutant partitioned between lysosomes and the plasma membrane (Fig. 8i). PQLC2-SNAP expression levels were variable between cells, with some cells expressing only mScarlet-LAMP1 and others expressing both mScarlet-LAMP1 and PQLC2-SNAP (Fig. 8j). We exploited this heterogeneity to determine if re-expression of PQLC2-SNAP restored the pH of individual lysosomes (Fig. 8k). Lysosomes negative for PQLC2-SNAP were of the same pH as those in PQLC2^KO-pH cells (Fig. 8l). Expression of PQLC2 (WT)-SNAP or any of the mutants restored pH to that of lysosomes in control-pH cells (Fig. 8l). The restoration in pH was concurrent with a decrease in lysosome size (Fig. 8m). These data suggest that maintenance of lysosomal pH, a function not previously attributed to PQLC2, may be necessary for degradation of lysosomal substrates such as EGFR.

Whereas the downstream consequences of blocking RTK-dependent signaling have been studied in detail, the effects of BiDAC-dependent RTK degradation have not been previously explored. In this study, we compared the consequences of inhibiting Her2 using RTK inhibitors, BiDACs and antibodies. By using unbiased morphological profiling, we were able to identify phenotypic states associated with these perturbations. After selectively removing Her2 staining information from the morphological profile, we observed that BiDACs and their inactive counterparts induced similar phenotypes, suggesting that they have limited additional effects on cell state beyond inhibiting Her2. Thus, the greater potency of BiDACs may be due to a more sustained inhibition of canonical Her2 signaling rather than through loss of Her2 kinase-independent functions¹⁰. Unlike BiDACs, AKT and PI3K inhibitors and anti-Her2 antibodies induced distinct phenotypes. The distinctiveness of the AKT and PI3K inhibitor phenotypes was expected, as these kinases integrate many different growth signals, and their inhibition impacts pathways beyond those regulated by Her2²⁸. Surprisingly, our data indicate that anti-Her2 antibodies have an impact on cell physiology that is distinct from other inhibitory modalities, potentially due to their ability to simultaneously alter Her2 dimerization, localization and expression¹. A potential limitation to using pharmacological perturbations was that our BiDACs also inhibited Her2 kinase activity and triggered degradation of EGFR. Future studies can address this limitation by inducing Her2 degradation with genetically encoded degrons. Taken together, our unbiased profiling of mechanistically distinct Her2 inhibitors reveals a diverse spectrum of phenotypes associated with agents notionally known to target the same protein. Furthermore, the morphological profiling approach can be used to provide mechanistic insight into how BiDACs trigger degradation of their targets.

The finding that BiDAC-mediated ubiquitination targeted RTKs to the endolysosomal pathway was surprising given our current understanding of ubiquitin dependent trafficking of RTKs to lysosomes. Lysosomal EGFR degradation triggered by binding to EGF is thought to require modification of EGFR by monoubiquitin²⁹ or lysine 63 (K63)-linked ubiquitin chains³⁰. Our functional genomics screen identified Ube2R2 as an E2 that functions in BiDAC-dependent degradation of Her2-GFP. Ube2R2 has been shown to ubiquitinate its substrates with lysine 48 (K48)-linked ubiquitin chains in vitro³¹, a linkage which canonically targets ubiquitinated proteins for degradation by the proteasome. Our data suggest that BiDAC-dependent K48 ubiquitination is also sufficient to trigger internalization and delivery of RTKs to lysosomes. A recent study used a genetically encoded split ubiquitin system to label EGFR with K63 or K48 ubiquitin chains. K63 ubiquitination resulted in the expected targeting of EGFR to lysosomes³². Modification by K48 chains triggered EGFR endocytosis but resulted in mistargeting of EGFR to the endoplasmic reticulum. A direct comparison between this study and BiDAC-induced EGFR ubiquitination is not possible since the split ubiquitin system does not allow attachment of linkage specific ubiquitin chains to lysines within EGFR. However, these results raise the possibility that BiDAC-induced K48 chains may recruit machinery that further ubiquitinates EGFR with K63-linked chains, thus promoting its trafficking to lysosomes³³. Another possibility is that the sustained K48 ubiquitination induced by BiDACs is sufficient to promote internalization and sorting of RTKs to lysosomes, even in the absence of K63 ubiquitination. Unexpectedly, we found that inhibition of the proteasome blocked RTK degradation by interfering with BiDAC-induced internalization of RTKs from the plasma membrane, an explanation for why Lap-VHL-dependent Her2 degradation was previously reported to require active proteasomes¹⁰. However, our results do not rule out a role for proteasome-dependent degradation of RTKs after internalization since lysosome inhibition only partially prevented RTK-FP degradation. We suspect that this may be a feature of our experimental system, as treatment with lysosome inhibitor did not fully reverse EGF-dependent degradation of EGFR, a known substrate of the endolysosomal pathway. Further studies are needed to determine the identity and dynamics of ubiquitin chain linkages that govern BiDAC-induced internalization, trafficking, and degradation of plasma membrane proteins.

Our screen uncovered a potential mechanism by which PQLC2 regulates lysosomal homeostasis and consequently, degradation of lysosomal substrates. PQLC2 functions both as a CAA transporter and as sensor of CAA abundance within lysosomes. Reduction of CAA availability stabilizes PQLC2 in a conformation that binds WDR41, which in turn recruits other members of the CSW complex to lysosomes²⁵. The role of the CSW complex is poorly understood, but it has been proposed to activate GTPase activity in proteins that function in vesicular trafficking and autophagy³⁴. By reintroducing PQLC2 mutants into PQLC2-null cells, we found that CAA transport and CSW binding were dispensable for rescue of EGFR-RFP degradation, lysosome size and pH. These results agree with a study using C. elegans which examined the impact of deleting the PQLC2 homolog LAAT-1 on lysosome function²⁴. Loss of LAAT-1 resulted in lysosome enlargement, impaired lysosomal degradation, and delayed development of the mutant worms. Supplementation of LAAT-1-null worms with CAA rescued development but did not impact lysosomal phenotypes, suggesting that while LAAT-1-dependent CAA transport was sufficient to maintain cytosolic CAA levels to support organismal growth, LAAT-1 regulated lysosomal homeostasis through a different mechanism. Our results indicate that PQLC2 regulates lysosomal pH, potentially through its interaction with the V-ATPase at the lysosomal membrane. Interestingly, PQLC2 has optimal CAA transport activity within the pH range found in lysosomes²². Thus, regulation of V-ATPase activity by PQLC2 may be required for lysosomal acidification, which in turn sets the conditions under which PQLC2 optimally transports CAA. We also found that the reintroduced PQLC2 undergoes cleavage within an intra-lysosomal loop, yielding a C-terminal fragment. While we were unable to obtain an antibody to confirm the size of endogenously expressed PQLC2, previous reports support our findings that the dominant form of PQLC2 is expressed as a cleaved fragment of the full-length protein^24,25, including a study in which PQLC2 was fused to a hemagglutinin tag at its genomic locus²⁵. Our results do not rule out the possibility that a minor, undetectable fraction of PQLC2 does not undergo cleavage, as reintroduction of a non-cleaved PQLC2 mutant was sufficient to rescue lysosomal phenotypes and restore EGFR degradation. Additional studies are needed to investigate the relationship between PQLC2 cleavage and its CAA transport and sensing functions.

Cell line generation and culture conditions

All SKOV3 cell lines and SKBR3 cells were cultured at 37°C with 5% CO₂ in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) and Glutamax. Parental SKOV3 and SKBR3 lines were obtained from ATCC. RTK-FP SKOV3 cells (Sigma-Aldrich) were subcloned using FACS to yield a homogenously expressing clonal line.

RTK-FP PQLC2^KO cells were generated by infection of RTK-FP cells with lentivirus containing lentiCRISPR v2 encoding Cas9 and sgRNAs targeting exon 3 or exon 2 of the SLC66A1 gene. Virus-containing media from HEK293T cells was added at a 1:1 v/v ratio to RTK-FP cells with 5 µg/mL Polybrene transfection reagent. The medium was changed after 24 hrs. After an additional 24 hrs of incubation, infected cells were selected by treatment with 2 µg/mL puromycin for 72 hrs to generate a knockout pool. Clonal lines expressing sgRNA #1 or sgRNA #2 were isolated by limited dilution cloning and knockout of PQLC2 in individual clones was confirmed using genomic sequencing. SKOV3 PQLC2^KO cells were similarly generated by infection of SKOV3 cells with LentiCrisprV2 encoding PQLC2 sgRNA #1, selection with puromycin and subcloning. Control cells were generated by infection of RTK-FP or SKOV3 cells with lentivirus containing LentiCRISPR v2 encoding an sgRNA targeting the cell surface protein CD81.

RTK-FP PQLC2^KO cells expressing PQLC2 rescue constructs or LAMP1-SNAP were generated by infection with a dilution series of lentivirus containing a construct encoding PQLC2-Flag or LAMP1-SNAP under control of a dox-inducible promoter. Infected cells were selected by treatment with 10 µg/mL blasticidin for 7 days and the multiplicity of infection (MOI) was determined by CellTiter-Glo assay (Promega). Lines infected with a dilution that yielded an MOI ≤ 0.3 were used for downstream experiments. LAMP1-SNAP-expressing cells were further enriched by FACS. Prior, expression of LAMP1-SNAP was induced by treatment with 500 ng/mL doxycycline for 48 hrs and LAMP1-SNAP was fluorescently labeled by treating cells with 1 µM SNAP-Cell 647-SiR (New England Biolabs) for 30 min at 37°C. Cells were pelleted and washed twice prior to sorting of the 647-SiR-positive population.

PQLC2^KO-pH cells were generated by infection of SKOV3 PQLC2^KO cells with lentivirus containing a construct encoding mScarlet-LAMP1 under control of a dox-inducible promoter, followed by selection with 10 µg/mL blasticidin. mScarlet-LAMP1-expressing clones were isolated by FACS. PQLC2^KO-pH expressing PQLC2 rescue constructs were generated by infection with virus containing a construct encoding PQLC2-SNAP under dox control. PQLC2-SNAP-expressing cells were enriched by labeling cells with SNAP-Cell 647-SiR, as above, followed by FACS.

Plasmids

CRISPR guide RNA (sgRNA) sequences targeting the SLC66A1 gene were obtained from the Avana library. The oligonucleotide sequence preceding the protospacer is 5’ caccgCATCTCTTACAGACCTACA 3’ for sgRNA #1 and 5’ caccgGAGACTCCTGCAACCTCAT 3’ for sgRNA #2. For the sgRNA sequence targeting the CD81 gene, the oligonucleotide sequence preceding the protospacer is 5’ caccg tggcttcctgggctgctacg 3’. Nucleotides in lowercase show the overhangs that allow cloning the oligonucleotides into the BsmBI site of lentiCRISPR v2.

The dox-inducible PQLC2 lentiviral expression construct was generated by inserting PQLC2-myc-flag (OriGene Technologies) between the NheI and NotI sites of the custom dox-inducible lentiviral expression vector pSHUSHV2, encoding a blasticidin resistance cassette and the reverse tetracycline-controlled transactivator. PQLC2 mutants were generated by site directed mutagenesis followed by insertion into pSHUSHV2. PQLC2 constructs containing N and C-terminal Myc or Flag tags were generated by incorporating sequences encoding these tags into the PCR amplification primers. PQLC2 containing a C-terminal SNAP tag was generated by Gibson cloning, followed by insertion into pSHUSHV2. LAMP1-SNAP was generated by amplification of the LAMP1 cDNA sequence (Origene Technologies) and the SNAP DNA sequence (New England Biolabs) followed by joining using Gibson assembly and insertion into pSHUSHV2. mScarlet-LAMP, previously described²⁷, was amplified and inserted into pSHUSHV2.

Western blotting

Cells were harvested by scraping or trypsinization and washed with PBS after pelleting by centrifugation. Cell pellets were lysed in 1% SDS/water and solubilized by sonication. Cell lysates were boiled at 100 ºC for 10 min, and protein concentrations were determined using Bicinchoninic acid (BCA) assay (Thermo Fisher Scientific). 20–40 µg of lysate was combined with Laemmli buffer containing 2.5% v/v ß-Mercaptoethanol (BME), heated at 65C for 10 min and separated by SDS-PAGE on 4–20% Criterion TGX gels (Bio-Rad) at 150V for 75 min. The proteins were transferred to 0.22 µm nitrocellulose membranes (Bio-Rad) using the Trans-Blot Turbo transfer system (Bio-Rad). Membranes were blocked in 5% BSA/PBST for 30 min at room temperature followed by overnight incubation at 4 ºC in primary antibodies diluted in 5% BSA/PBST. The following day, membranes were washed 3 times for 5 minutes in PBST, and incubated in secondary anti-mouse IRDye 680RD or anti-rabbit IRDye 800CW antibodies (LI-COR) for 30 minutes at room temperature. After three washes in PBST, the membranes were imaged on an Odyssey Imaging System (LI-COR).

The following primary antibodies were diluted 1:1000 v/v, unless otherwise noted, in 5% BSA/PBST and used for immunoblots in this study: anti-Her2 (Cell Signaling Technology cat no. 4290S), anti-EGFR (Cell Signaling Technology no. D38B1), anti-GAPDH diluted 1:5000 (Cell Signaling Technology no. D7166S), anti-Phospho-Her2 (Tyr1196) (Cell Signaling Technology no. 6942S), anti-Flag (Sigma-Aldrich no. F1804), anti-Myc-Tag (Cell Signaling Technology no. 2272S), anti-WDR41 (Novus Biologicals no. 83812), anti-ß-actin (Cell Signaling Technologies no. 3700s) and anti-LC3 (MBL International Corporation no. PM036).

Microscopy

Live RTK-FP cells expressing LAMP1-SNAP were imaged using a spinning disk confocal microscope (Nikon) equipped with a 60x 1.4na oil objective, in an environmental chamber set to 37 ºC and 5% CO₂. Cells were seeded at a density of 10,000 cells per well in a glass bottom #1.5H-N 96-well plate (Cellvis) containing 100 µL of complete phenol-red free RPMI 1640 per well. 500 ng/mL of dox was added to the media to induce expression of LAMP1-SNAP. After a 48 hr incubation, cells were treated with 1 µM Lap-VHL or Lap-VHL_Inactive and 100 nM Bafilomycin-A1 (Selleck chemicals) for 24 hrs. To label LAMP1-SNAP immediately prior to imaging, cells were treated with 1 µM SNAP-Cell 647-SiR for 30 min at 37 ºC, and subsequently washed three times with complete media while adhered to the imaging chamber. After an additional 30 min incubation, the media was replaced with fresh media containing 1 µg/mL Hoechst 33342 Trihydrochloride Trihydrate (Invitrogen) to stain nuclei.

Time lapse imaging of live RTK-FP cells was performed using a Opera Phenix Plus System (Revvity Health Sciences) set to 37 ºC and 5% CO₂. Cells were seeded as above and treated with 1 µM Lap-VHL in combination with degradation inhibitors. To label nuclei, cells were treated with 50 nM SiR-DNA (Cytoskeleton, Inc.) and single plane images across 20 fields per well were acquired in confocal mode every 30 min. At each time point, the number of puncta in each well was determined using Harmony imaging and analysis software (Revvity) and normalized by the number of SiR-DNA-positive nuclei in the well.

For immunofluorescence microscopy analysis of PQLC2-Flag expression, RTK-FP cells were seeded as above and incubated for 48 hrs in the presence of 500 ng/mL dox. Cells were washed three times with PBS and fixed with 4% paraformaldehyde in PBS for 15 min at room temperature. Following fixation, cells were washed three times with PBS, and permeabilized with 0.1% Triton X-100 in PBS for 5 min at room temperature. Cells were blocked in 1% BSA in PBS for 30 min and stained for 2 hrs with a 1:500 v/v dilution of anti-Flag monoclonal antibody (M2, Sigma-Aldrich) in blocking solution. After 3 washes with blocking solution, cells were stained for 1 hr with anti-Mouse IgG Fragment conjugated with Alexa Fluor 647 (Cell Signaling Technologies) diluted 1:5000 v/v in blocking solution, washed 3 times, and imaged on the Opera Phenix Plus System.

Analysis of SKBR3 cell proliferation was performed using an Incucyte SX5 live-cell imaging and analysis instrument (Sartorius). 5000 cells per well were seeded in a 96-well tissue culture plate and incubated for 24 hrs prior to drug treatment and start of imaging. Images were taken in the phase channel every 6 hrs using the 10X objective over a total period of 120 hrs, and percent confluence over time was determined using the Incucyte image analysis software.

Flow cytometry

Flow cytometric analysis was performed using the LSRFortessa X-20 cell analyzer (BD Biosciences). For analysis of total RTK-FP levels, cells were collected from 96-well plates by trypsinization, transferred to V-bottom plates and run using the high throughput sampler. For staining of surface RTKs, trypsinized cells were transferred to V-bottom plates and pelleted at 300 x g for 5 min. Cells were resuspended in 50 µL of a solution containing 0.5% v/v BSA in PBS, 2 µg/mL anti-EGFR Alexa Fluor 405-conjugated antibody (R&D Systems) and 2 µg/mL anti-Her2 Alexa Fluor 647-conjugated antibody (R&D Systems). After a 30 min incubation at room temperature, 100 µL of 0.5% BSA/PBS was added to each well and the plates were centrifuged at 300 x g for 5 min. Plates were washed once with 150 µL of 0.5% BSA/PBS, and the cells were resuspended in FACS buffer containing 2% FBS in PBS for analysis.

Geometric mean fluorescence intensity (MFI) was calculated using FlowJo analysis software (FlowJo, LLC). Each MFI value was normalized by subtracting the background MFI determined from analysis of parental SKOV3 cells and divided by the value obtained from untreated cells within each treatment condition.

Mass spectrometry

To prepare lysates for immunoprecipitation, two replicate samples of RTK-FP PQLC2^KO cells, RTK-FP PQLC2^KO cells expressing PQLC2-Flag or PQLC2^KO cells expressing PQLC2(2x PL)-Flag were treated with 500 ng/ml dox for 48 hrs. After pelleting, cells were washed in cold PBS and lysed in cold RIPA buffer containing 50 mM Tris-HCl pH 7.5, 150 mM sodium chloride, 1 mM EDTA, 1% Triton X-100, 1% Sodium deoxycholate and 1X Protease inhibitors (Thermo Fisher Scientific). Cells were rotated in lysis buffer for 30 min at 4ºC, and lysates were cleared by centrifugation at 20,000 x g for 15 min. Supernatants were transferred to a new tube, and protein concentrations were quantified using BCA assay.

For each replicate, 10 mg of lysate was combined with 40 µL of anti-Flag M2 magnetic beads (Sigma-Aldrich) slurry and rotated for 2 hrs at 4ºC. Beads were separated from lysates using a magnetic tube rack, washed 5 times with cold RIPA buffer, and 2 times with cold wash buffer containing 150 mM Sodium chloride, 10 mM Tris HCl pH 7.5 and 0.5 mM EDTA. Bound proteins were eluted in 40 µL of a 100 µg/mL solution of 3X Flag peptide (Sigma-Aldrich) resuspended in TBS, by incubation for 15 min at room temperature.

Sample preparation and mass spectrometric analysis was performed by Creative Proteomics (Shirley, New York, USA). Eluted proteins were separated on an 12% SDS-PAGE gel and visualized by silver staining. To prepare proteins for in-gel digestion, each gel slice was cut into 1 mm³ cubes and combined with 1 mL of a solution containing 1:1 v/v ratio of 30 mM Potassium ferricyanide and 100 mM Sodium thiosulfate until destaining was complete. The supernatant was removed, and the gel slices were combined with 200 µL water for 10 min followed by 1 mL of 50 mM Ammonium bicarbonate for 30 min. Gel slices were subsequently dehydrated by incubation in 500 µL of Acetonitrile for 30 min. After removal of Acetonitrile, the gel slices were rehydrated with a solution containing 10 mM Dithiothreitol (DTT), and incubated at 56ºC for 1 hr. The DTT solution was removed, and the gel slices were incubated in 500 µL of Acetonitrile for 10 min at room temperature, followed by incubation in a solution containing 50 mM Iodoacetamide at room temperature in the dark. After removal of Iodoacetamide and incubation in 500 µL of Acetonitrile for 10 min, gel slices were incubated in trypsin digestion solution for 45 min on ice, followed by incubation at 37ºC overnight. The supernatant was transferred to a new tube, and the slices were incubated in a 1:2 v/v 50 mM Ammonium bicarbonate:Acetonitrile solution for 1 hr at 37ºC. This procedure was repeated, and all peptide extractions were combined and acidifed with 0.1% Trifluoroacetic acid. After lyophilization, the digested peptides were resuspended in 20 µL of 0.1% Formic acid for LC-MS/MS analysis. Peptides were

separated on a UltiMate 3000 nano UHPLC system (Thermo Fisher Scientific) using a PepMap C18, 100Å, 100 µm × 2 cm, 5 µm trapping column and a PepMap C18, 100Å, 75 µm × 50 cm, 2 µm analytical column. 1 µg of sample was run at a flow rate of 250 nL/min using a gradient of 2–8% of buffer B (0.1% formic acid in 50% Acetonitrile) for 3 min, 8–20% Buffer B for 56 min, 20–40% buffer B for 37 min, and 40–90% buffer B for 4 min.

Samples were analyzed on a Q Exactive HF mass spectrometer (Thermo Fisher Scientific, USA) equipped with an ESI nanospray source. The full scan was performed between 300-1,650 m/z at the resolution 60,000 at 200 m/z. The automatic gain control target for the full scan was set to 3 x 10⁶. The MS/MS scan was operated in Top 20 mode using the following settings: resolution, 15,000 at 200 m/z; automatic gain control target, 1 x 10⁵; maximum injection time, 19 ms; normalized collision energy, 28%; isolation window, 1.4 Th; charge state exclusion, unassigned, 1, > 6; dynamic exclusion, 30 s.

Six raw MS files were analyzed and searched against the Homo sapiens protein database using Maxquant software (version 1.6.2.6). The parameters were set as follows: the protein modifications were carbamidomethylation (C) (fixed), oxidation (M) (variable); the enzyme specificity was set to trypsin; the maximum missed cleavages were set to 2; the precursor ion mass tolerance was set to 10 ppm; MS/MS tolerance was set to 0.5 Da.

Edman sequencing

RTK-FP PQLC2^KO cells expressing PQLC2-Flag were treated with 500 ng/mL dox for 48 hrs, and lysates were prepared as described above. 60 mg of lysate in RIPA buffer was combined with 150 µL of anti-Flag M2 magnetic bead slurry (Sigma-Aldrich) and rotated for 2 hrs at 4ºC. Beads were separated from lysates using a magnetic tube rack, washed 5 times with cold RIPA buffer, and bound proteins were eluted by boiling beads in 2X Laemmli buffer for 10 min. The supernatant was transferred to a new tube, combined with 2.5% v/v ß-Mercaptoethanol, and proteins were separated by SDS-PAGE using a 12.5% Bis-Tris gel (Bio-rad) at 150V for 75 min. The proteins were transferred to 0.22 µm PVDF membranes (Bio-Rad) using the Trans-Blot Turbo Transfer System (Bio-Rad). The membranes were stained with Coommassie Gel Code Blue at room temperature for 10 min, and destained several times with 50% methanol. A PVDF band corresponding to the expected size of PQLC2-Flag was cut from the membrane and further destained with 50% methanol in a tube, followed by two washes with water. 10 cycles of Edman sequencing analysis was performed (Creative Proteomics) using an ABI Procise 494HT protein sequencing system (Thermo Fisher Scientific).

Morphological profiling

Cells were seeded inn384-well PhenoPlates (Perkin Elmer) using a Fluent Liquid Handler (Tecan). Prior to seeding, cells were counted using a Vi-Cell XR cell counter (Perkin Elmer), and their concentration was adjusted to generate the required stock concentrations that varied depending on the final time point of analysis. 50 µL of cell suspension was added to each well at a density of 2500 cells per well for the 2 hr and 6 hr time points, 1500 cells per well for the 24 hour time point, 1000 cells per well for the 48 hour time point, and 500 cells per well for the 72 hour time point. After seeding, the plates were left at room temperature for 30 min to allow uniform attachment of cells to the plate. The plates were subsequently transferred to an incubator for 16 hrs.

Cells were treated with experimental compounds using the Echo 555 Acoustic Liquid Handler (Beckman Coulter). Specifically, low dead volume Echo source plates containing 10 µL of compound per well were prepared and stored at -80°C. Source plates were limited to a maximum of three thaws at room temperature and centrifuged at 1000 rpm for 5 min prior to dosing. After dosing, Phenoplates were returned to an incubator for the duration of the 2–72 hr treatments. The dose order for small molecules corresponds to the following concentrations: 0, 0.05 µM; 1, 0.1 µM; 2, 0.35 µM; 3, 1 µM; 4, 3.5 µM; 5, 10 µM. The dose order for antibodies corresponds to the following concentrations: 0, 0.3 µg/mL; 1, 1 µg/mL; 2, 3.3 µg/mL; 3, 10 µg/mL; 4, 33 µg/mL; 5, 100 µg/mL.

After incubation in drugs, the cells were stained with 0.5 µM of MitoTracker Deep Red FM (Invitrogen) in cell culture media by adding 10 uL of a 3 µM MitoTracker solution to each well using a Multidrop Combi Reagent Dispenser (Thermo Fisher). The cells were then incubated at 37°C for 30 min. During this time, the cell staining solution was prepared as follows: 1% BSA solution was made by dissolving fresh Bovine Serum Albumin (BSA) powder in Hank’s Balanced Salt Solution (HBSS) followed by filtration through a 0.22 µm filter. Next, a 0.1% v/v Triton X-100 solution was prepared by mixing 1% BSA/HBSS with a 10% Triton X-100 stock solution. Dyes from the Cell Painting panel were manually added at the following concentrations to generate the post-fixation permeabilization and staining solution containing 1 µg/mL Phalloidin Conjugates CF430 (Biotium), 1 µg/mL Hoechst 33342 Trihydrochloride Trihydrate (Invitrogen) and 1.5 µg/mL Wheat Germ Agglutinin (WGA) Alexa Fluor 555 (Invitrogen). To stain Her2, Trastuzumab (Herceptin, Genentech) and Pertuzumab (Perjeta, Genentech) were labeled using the DyLight 488 Antibody Labeling Kit (Life Technologies Corporation) and added to the staining solution at a concentration of 1 µg/mL. Lastly, the staining solution was combined with 0.05% v/v sodium azide to prevent contamination during staining. After staining mitochondria, the plates were transferred with lids removed to a BioTek EL406 Microplate Washer Dispenser with BioStack (Agilent). 20 µL of a 16% paraformaldehyde (PFA) solution was added to each well using a syringe pump to yield a final concentration of 4% PFA. Cells were fixed for 20 minutes at room temperature and PFA was removed using a wash manifold on the EL406. Excess PFA was removed by washing each well twice with 60 µL of HBSS. Cells were subsequently incubated in 20 µL of the permeabilization and staining solution for 30 min at room temperature followed by two washes with HBSS. Upon completion of staining, all wells were filled with 60 µL of an HBSS solution containing 0.05% v/v sodium azide, and the plates were sealed before imaging. Each stained 384-well plate was imaged using a Opera Phenix Plus System (Revvity Health Sciences) by acquiring 5-channel fluorescence and digital phase contrast images across 9 fields in the center of each well with a 20X objective.

For generation of deep learning image embedding, images were down sampled to 512 x 512 x 6, with each channel corresponding to an image acquisition channel. Separate deep convolutional neural networks were trained using all images for each cell type. All networks were a modified ResNet³⁵ backbone trained using ArcFace loss³⁶, with each image’s class label corresponding to a combination of its timepoint x dose x perturbation. This training objective encourages the network to minimize differences in representation due to batch effects between technical replicates. For downstream analysis, the 512-dimensional final activation vector of the network was used as the image embedding. All image embeddings were further normalized by subtracting the mean of all DMSO embeddings per plate. For analysis in the absence of Her2 staining, a separate network was trained without including the Her2 channel in the input images.

For cell segmentation and image intensity analysis, all channels other than the Hoechst nuclei channel were grouped as “cyto” channels. Cyto channels were combined by first linearly rescaling each individual channel image to have its 3rd to 97th percentile pixel intensity values fall in the range [0, 1], with all outliers clipped to 0 or 1. The channels were then combined by taking the per pixel maximum value across all rescaled cyto channel images. The nuclei channel and composite cyto channel were then down sampled to 512 x 512 and processed using the Cellpose model³⁷ to derive individual cell segmentation masks. For each channel, the median intensity value per cell was then computed.

t-SNE plots were generated by averaging the image embeddings across the timepoint x dose x perturbation groups for each plate. The two dimensional t-SNE embeddings were generated using a multicore implementation of the Barnes-Hut t-SNE algorithm^38,39. The perplexity value was manually tuned for each dataset.

Functional genomic screens

For the EGFR-RFP and Her2-GFP BiDAC screens, RTK-FP Cas9 cells were seeded in 5-layer flasks (Corning) at a density of 11 million cells per flask. After 72 hrs, the cells were infected with the Avana sgRNA library⁴⁰ at a multiplicity of infection of 0.3 and treated with 5 µg/ml of Polybrene transfection reagent (Fisher Scientific). The media was changed after 24 hrs, and cells were incubated for an additional 24 hrs. 48 hrs after infection, cells were treated with 2 µg/ml of puromycin to initiate selection. 48 hrs after addition of puromycin, the puromycin-containing media was replaced with fresh media lacking puromycin and the cells were allowed to recover for an additional 72 hrs. After recovery and 9 days following infection, cells were treated with 1 µM of Lap-VHL and incubated for 24 hrs in BiDAC. Prior to FACS, 60 x 10⁶ cells were removed to serve as a non-sorted control. For enrichment by FACS, cells were harvested by trypsinization in TrypsinLE and resuspended in FACS buffer composed of PBS containing 2% FBS. The top 1% RTK-FP-expressing cells were sorted using a Bigfoot Cell Sorter (Thermo Fisher Scientific), comprising a total of 750 x 10³ cells. After collection, cells were pelleted and frozen at -80°C.

To prepare genomic DNA, sorted cells were mixed with 5 x 10⁶ non-infected RTK-FP cells to increase recovery, and the sorted and unsorted cell pellets were resuspended in 100 µL of TE buffer (10 mM Tris-HCl pH 8.0, 10 mM EDTA) per 5 x 10⁶ cells. 900 µL of Lysis Buffer containing 10 mM Tris-HCl pH 8.0, 10 mM EDTA, 0.5% SDS and 0.2 mg/mL Proteinase K (Thermo Fisher Scientific) was added to 100 µL of cells. Cell lysates were incubated at 56°C and vortexed vigorously every 15 min, followed by an additional overnight incubation. After the incubation, 5M NaCl was added to lysates to yield final concentration of 0.2M NaCl. A phenol chloroform extractions was performed using MaXtract tubes (Qiagen) by mixing lysates with an equal volume of Phenol:Chloroform:Isoamyl alcohol 25:24:1 (Life Technologies), followed by centrifugation at 1500 x g and extraction by an equal volume of chloroform. The aqueous phase was treated with 25 µg/mL of RNAse A (PureLink, Life Technologies) for 1 hr at 37°C with periodic mixing followed by a second phenol chloroform extraction. The aqueous phase was combined 1:30 v/v with 3M Sodium acetate, pH 5.2 and the DNA was precipitated with 2.5 volumes of ethanol for 1 hr at room temperature. Precipitated DNA was pelleted by centrifugation at 4,000 x g for 1 hr at 4°C in a swinging bucket centrifuge. The DNA pellet was washed 3 times with 75% ethanol, allowed to dry and resuspended in 200 µL of water per 5 x 10⁶ lysed cells. DNA was further solubilized by incubation for 1 hr at 55°C followed by passage 10–20 times through a 28 gauge needle. DNA purity was assessed using a NanoDrop spectrophotometer and its concentration was determined using Qubit Fluorometric Quantification (Thermo Fisher Scientific).

Genomic sgRNA libraries were prepared by PCR amplification of all DNA extracted from sorted cells and 70 µg of DNA extracted from the unsorted control samples using common primers. 4 µl of the PCR products was used as a template for another round of PCR with barcoded primers. The products were purified using SPRIselect beads (Beckman Coulter), and their purity and concentration was determined using the 2100 Bioanalyzer instrument (Agilent). The libraries were sequenced at a read depth of 60 x 10⁶ reads per sample using a NovaSeq 6000 System (Illumina). MAGeCK was used to normalize the read counts by the median sgRNA count values, calculate fold changes and determine statistical significance¹⁶.

Fluorescence Lifetime Imaging Microscopy (FLIM)

Cells were seeded at a density of 10,000 cells per well in a Cellvis 8-well chambered coverglass and treated with 500 ng/ml dox for 48 hrs. To image SNAP-tagged PQLC2, cells were treated with 1 µM SNAP-Cell 647 SiR for 30 min and washed 3 times with complete RPMI 1640 medium. The medium was replaced again following a 30 min incubation with complete RPMI 1640 lacking phenol red. During imaging, cells were incubated at 36ºC and 5% CO₂ using a stage top incubator (Okolab).

For pH calibration experiments, cells were imaged immediately after a 5 min incubation in high potassium imaging buffer containing 140 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, 5 mM glucose and 10% Carmody’s buffer⁴¹ to set the desired pH. Ionophores consisting of 10 µM nigericin and 2 µM monensin was also added to the imaging buffer to equalize proton gradients across cellular membranes.

Fluorescence lifetime imaging was performed on an SP8 point scanning confocal microscope (Leica) equipped with a time-correlated single photon counting (TCSPC) add-on (PicoQuant GmbH, Berlin, Germany). mScarlet was excited with the 561 nm output of the white light laser at an average power of approximately 10–20 µW and a repetition rate of 40 MHz. Emission was collected with a 100x/1.4 NA oil objective (HC PL APO CS2, Leica), and photon arrival times were determined with the built-in HyD detectors and the PicoHarp 300 TCSPC module. Spectral detection was used to restrict collection wavelengths to 590–700 nm, and a notch filter was included to block 488/561/633 nm. Scanning was performed with the built-in Leica non-resonant scanner with a total time per field of view of 20 s. Images were binned to pixel sizes of 200 nm x 200 nm to obtain adequate photons for lifetime determination. All imaging was performed with the confocal pinhole set to 1 Airy unit at 580 nm.

For two color imaging with SiR-SNAP647, we took two sequential images of each field of view. We imaged mScarlet with the 561 nm output of the white light laser collection from 587 to 625 nm. We imaged SiR647 SNAP with the 640 nm output of the white light laser (average power 0.5 µW) with collection from 660 to 788 nm. An additional notch filter (470/640 nm) was included to block the 640 nm illumination. To reduce motion artifacts from lysosomal movement, we alternated between the two illumination colors between lines. Because the PicoQuant software did not support between line switching, we used a custom parser to process the raw TCSPC output and reconstruct an image. Parser code is available on GitHub (https://github.com/AndrewGYork/tools/blob/master/picoquant_tttr.py). For two color images, the total acquisition time per field of view was approximately 40 seconds.

TCSPC images were used to calculate mean arrival time at each pixel either within the PicoQuant SymPhoTime software (one color imaging) or with custom parser code (two color imaging). Fluorescence intensity images were generated by summing the photons at each pixel over the fluorescence lifetime decay.

To segment fluorescence lifetimes and intensity images, we first performed instance segmentation of lysosomes using the mScarlet-LAMP1 emission channel and the “StarDist2D” model trained on the “2D_versatile_fluo” dataset⁴². The model distinguishes individual lysosomes within 2D images by recognizing their star-convex shape. Prior to segmentation, images were preprocessed by a normalization to adjust the pixel intensity distribution to a standard range and by up sampling to adjust image features to the scale of those used in the training model. After segmentation, the masks were down sampled back to the original size. The segmentation output was used to extract lysosome area as the total number of pixels in the masks scaled by the camera pixel resolution, mean fluorescence values in both the mScarlet-LAMP1 and PQLC2-SNAP channels, and the mean fluorescence lifetime in the mScarlet-LAMP1 channel. These quantifications were performed using Scikit Image’s region properties function⁴³. For fluorescence lifetime quantification, only those pixels with a photon count above 5 were used to calculate the mean florescence lifetime value.

pH calibration (ionophore) data were fit to a four parameter logistic function to extract a calibration curve, where tau is the lifetime, min and max are the minimum and maximum lifetime values observed (at low and high pH respectively), and b is the slope factor (similar to the Hill coefficient).

$$tau=\text{max}+ \frac{\text{min}- max}{1+ {\left(\frac{pH}{pKa}\right)}^{b}}$$

Median lifetime of all lysosomes in each condition was converted to pH using the corresponding calibration curves. For the PQLC2-SNAP rescue experiments, the PQLC2^KO-pH pH calibration curve was used to determine the pH of lysosomes lacking PQLC2-SNAP signal and the Control-pH calibration curve was used to determine the pH of PQLC2-SNAP-positive lysosomes.

Author contributions:

Morphological profiling experiments and data modeling were performed by Q.J. and M.S. under the supervision of Z.C. and B.Y.F. Lifetime microscopy experiments and image processing were conducted by J.L.R. and A.E.Y.T.L. M.E.F. and K.J. performed design and chemical synthesis of BiDAC degraders. S.V. performed western blot analysis of RTK and LC3 expression. A.N. generated RTK-FP cells and oversaw collaborations with C4 Therapeutics. K.B. conceived the project and conducted all other experiments. K.B. and B.F. wrote the manuscript and made the figures. D.S. contributed to discussions and supervised the project.

Acknowledgements

We would like to thank Dr. Stewart Fisher (C4 Therapeutics, Inc.) for his expertise regarding generation of BiDAC degraders. We thank Dr. Daniel Gottschling and Dr. Hannah Baddock (Calico Life Sciences, LLC) for providing comments and suggestions for the manuscript. Calico Life Sciences LLC provided funding for all resources for this work.

FAUVEL B, Yasri A (2014) Antibodies directed against receptor tyrosine kinases. mAbs 6:838–851
Huang L, Jiang S, Shi Y (2020) Tyrosine kinase inhibitors for solid tumors in the past 20 years (2001–2020). J Hematol Oncol J Hematol Oncol 13:143
Pottier C et al (2020) Tyrosine Kinase Inhibitors in Cancer: Breakthrough and Challenges of Targeted Therapy. Cancers 12:731
Diwanji D et al (2021) Structures of the HER2-HER3-NRG1β complex reveal a dynamic dimer interface. Nature 600:339–343
Agus DB et al (2002) Targeting ligand-activated ErbB2 signaling inhibits breast and prostate tumor growth. Cancer Cell 2:127–137
Zhang Y et al (2016) Neratinib induces ErbB2 ubiquitylation and endocytic degradation via HSP90 dissociation in breast cancer cells. Cancer Lett 382:176–185
Békés M, Langley DR, Crews CM (2022) PROTAC targeted protein degraders: the past is prologue. Nat Rev Drug Discov 21:181–200
Winter GE et al (2015) Phthalimide conjugation as a strategy for in vivo target protein degradation. Science 348:1376–1381
Lu J et al (2015) Hijacking the E3 Ubiquitin Ligase Cereblon to Efficiently Target BRD4. Chem Biol 22:755–763
Burslem GM et al (2018) The Advantages of Targeted Protein Degradation over Inhibition: a RTK Case Study. Cell Chem Biol 25:67–77e3
Bray M-A et al (2016) Cell Painting, a high-content image-based assay for morphological profiling using multiplexed fluorescent dyes. Nat Protoc 11:1757–1774
Cimini BA et al (2023) Optimizing the Cell Painting assay for image-based profiling. Nat Protoc 18:1981–2013
Haghighi M, Caicedo J, Cimini BA, Carpenter AE, Singh S (2022) High-Dimensional Gene Expression and Morphology Profiles of Cells across 28,000 Genetic and Chemical Perturbations. Nat Methods 19:1550–1557
Chandrasekaran SN, Ceulemans H, Boyd JD, Carpenter AE (2021) Image-based profiling for drug discovery: due for a machine-learning upgrade? Nat Rev Drug Discov 20:145–159
Eulenberg P et al (2017) Reconstructing cell cycle and disease progression using deep learning. Nat Commun 8:463
Li W et al (2014) MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens. Genome Biol 15:554
Cardote TAF, Gadd MS, Ciulli A (2017) Crystal Structure of the Cul2-Rbx1-EloBC-VHL Ubiquitin Ligase Complex. Structure 25:901–911e3
Gong L, Yeh ETH (1999) Identification of the Activating and Conjugating Enzymes of the NEDD8 Conjugation Pathway*. J Biol Chem 274:12036–12042
Sievers QL, Gasser JA, Cowley GS, Fischer ES, Ebert BL (2018) Genome-wide screen identifies cullin-RING ligase machinery required for lenalidomide-dependent CRL4CRBN activity. Blood 132:1293–1303
Wang L, Wu D, Robinson CV, Wu H, Fu T-M (2020) Structures of a Complete Human V-ATPase Reveal Mechanisms of Its Assembly. Mol Cell 80:501–511e3
Tomas A, Futter CE, Eden ER (2014) EGF receptor trafficking: consequences for signaling and cancer. Trends Cell Biol 24:26–34
Jézégou A et al (2012) Heptahelical protein PQLC2 is a lysosomal cationic amino acid exporter underlying the action of cysteamine in cystinosis therapy. Proc. Natl. Acad. Sci. U. S. A. 109, E3434-3443
Jain A, Zoncu R (2021) Picking the arginine lock on PQLC2 cycling. Proc. Natl. Acad. Sci. U. S. A. 118, e2112682118
Liu B, Du H, Rutkowski R, Gartner A, Wang X (2012) LAAT-1 is the Lysosomal Lysine/Arginine Transporter that Maintains Amino Acid Homeostasis. Science 337:351–354
Amick J, Tharkeshwar AK, Talaia G, Ferguson SM (2019) PQLC2 recruits the C9orf72 complex to lysosomes in response to cationic amino acid starvation. J Cell Biol 219:e201906076
Talaia G, Amick J, Ferguson SM (2021) Receptor-like role for PQLC2 amino acid transporter in the lysosomal sensing of cationic amino acids. Proc. Natl. Acad. Sci. 118, e2014941118
Lazzari-Dean JR, Ingaramo MC, Wang JCK, Yong J, Ingaramo M (2022) mScarlet fluorescence lifetime reports lysosomal pH quantitatively. 10.5281/zenodo.6363342
Hemmings BA, Restuccia DF (2012) PI3K-PKB/Akt Pathway. Cold Spring Harb Perspect Biol 4:a011189
Haglund K et al (2003) Multiple monoubiquitination of RTKs is sufficient for their endocytosis and degradation. Nat Cell Biol 5:461–466
Huang F et al (2013) Lysine 63-linked polyubiquitination is required for EGF receptor degradation. Proc. Natl. Acad. Sci. U. S. A. 110, 15722–15727
Liwocha J et al (2024) Mechanism of millisecond Lys48-linked poly-ubiquitin chain formation by cullin-RING ligases. Nat Struct Mol Biol 31:378–389
Ubiquiton-An inducible linkage-specific polyubiquitylation tool - PubMed. https://pubmed.ncbi.nlm.nih.gov/38103558/
Marx C, Held JM, Gibson BW, Benz CC (2010) ErbB2 trafficking and degradation associated with K48 and K63 polyubiquitination. Cancer Res 70:3709–3717
Tang D et al (2020) Cryo-EM structure of C9ORF72–SMCR8–WDR41 reveals the role as a GAP for Rab8a and Rab11a. Proc. Natl. Acad. Sci. 117, 9876–9883
He K, Zhang X, Ren S, Sun J (2015) Deep Residual Learning for Image Recognition. Preprint at https://doi.org/10.48550/arXiv.1512.03385
Deng J et al (2022) ArcFace: Additive Angular Margin Loss for Deep Face Recognition. IEEE Trans Pattern Anal Mach Intell 44:5962–5979
Stringer C, Wang T, Michaelos M, Pachitariu M (2021) Cellpose: a generalist algorithm for cellular segmentation. Nat Methods 18:100–106
Ulyanov D (2024) DmitryUlyanov/Multicore-TSNE
van der Maaten L (2014) Accelerating t-SNE using Tree-Based Algorithms. J Mach Learn Res 15:3221–3245
Doench JG et al (2016) Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat Biotechnol 34:184–191
Carmody WR (1961) Easily prepared wide range buffer series. J Chem Educ 38:559
Schmidt U, Weigert M, Broaddus C, Myers G (2018) Cell Detection with Star-convex Polygons. 11071:265–273
van der Walt S et al (2014) scikit-image: Image processing in Python. PeerJ 2, e453

There is NO Competing Interest.

Download PDF

Version 1

posted

You are reading this latest preprint version

Morphological and genetic screens reveal mechanisms of BiDAC-induced plasma membrane protein degradation

Status:

Version 1

Abstract

Figures

Introduction

Results

Characterization of RTK BiDACs

PQLC2 functions in the maintenance of lysosomal pH

Discussion

Methods

Declarations

Author contributions:

Acknowledgements

References

Additional Declarations

Supplementary Files

Status:

Version 1