Divergent proteome tolerance against gain and loss of chromosome arms.

INTRODUCTION
Aneuploidy, defined as the abnormal gain or loss of chromosomes or chromosome arms, is a hallmark of human genetic syndromes and many cancers1–3. While somatic aneuploidy is rare in healthy human tissues4,5, it is highly prevalent in cancer, with 90% of solid tumors exhibiting aneuploidy3. This contributes to tumor evolution6,7, metastasis8,9, and drug resistance10–12. In lung squamous cell carcinoma (LSCC), for example, chromosome 3q is gained in over 60% of tumors, whereas chromosome 3p is lost in 78%3. It has been recently reported that tumor cells can develop oncogene-like addiction to specific single-arm aneuploidies13. However, how cells tolerate aneuploidy while maintaining essential biochemical processes remains poorly understood.
A key mechanism underlying aneuploidy tolerance is proteomic buffering14,15, the attenuation of altered gene dosage at the protein level. In aneuploid cancer cells, ~30% of gene affected by copy number alterations (CNAs) do not result in proportional protein level changes16–18, suggesting an adaptive proteomic buffering system. Studies in yeast and human cell lines have demonstrated that protein dosage compensation is particularly strong for subunits of protein complexes19–21. Using quantitative proteomics and dynamic SILAC labeling22
23, we previously showed that in Trisomy 21 cells, protein degradation rates increased significantly for chromosome 21-encoded protein complex members20. Similar trends were observed across diverse yeast strains21.
Yet, whether protein degradation-mediated aneuploidy tolerance applies universally to all types of aneuploidies in mammals remains unresolved. While degradation-based buffering has been proposed as a general mechanism24, existing experimental evidence is largely derived from studies of excessive (i.e., the gain-type) aneuploidies20,21, with little focus on loss-type aneuploidies. In fact, if degradation were the primary compensation mechanism, cells with single-arm or whole-chromosome loss would need to degrade vast amounts of functional protein complex members encoded by unaffected chromosomes, which would be metabolically inefficient and contradict the cellular economy principle14,25. Alternatively, cells might reduce degradation rates of proteins encoded on the lost chromosome to preserve complex stoichiometry26, a potentially more economical strategy. However, neither hypothesis has been rigorously tested, likely due to the lack of appropriate chromosomal loss models and precise methods to quantify both protein degradation and abundance. Additionally, a recent review has emphasized the importance of measuring protein turnover in both the gain-type and loss-type aneuploidies27.
To address this gap, we leveraged our previously established isogenic chromosome 3p loss and chromosome 3q gain (hereafter referred to as 3p loss and 3q gain respectively) human lung epithelial cell models3. Using a combination of label-free28–30 and multiplexed data-independent acquisition mass spectrometry (DIA-MS)23,31, we systematically quantified protein abundance, turnover, and thermal stability to dissect proteostasis mechanisms in response to aneuploidy. Our findings strongly suggest that single-arm chromosome gain and loss are tolerated through distinct proteostasis strategies: while protein degradation was selectively accelerated in 3q gain cells, protein synthesis, rather than degradation, was the primary compensatory mechanism in 3p loss cells. These findings were further validated using ribosome footprints and an additional 8p loss isogenic system. Our proteomics data also revealed a unique increase in protein thermal stability in 3p loss cells but not in 3q gain cells. Although further mechanistic studies are warranted, our work provides unique insights into how aneuploid cells maintain proteome balance, offering fundamental and translational implications for cancer biology and therapeutic strategies.

RESULTS

Protein Dosage Compensation Maintains Protein Complex Stoichiometry in 3q Gain and 3p Loss Cells
Aneuploidy-associated gene dosage imbalance raises fundamental questions about how cells maintain proteostasis, particularly in cancers like LSCC3,32. To investigate this, we analyzed proteogenomic data from LSCC patients in the Clinical Proteomic Tumor Analysis Consortium (CPTAC) database, where 86% (93/108) exhibited 3p loss, and over 93% (100/108) had a large region of 3q gain32. We integrated gene copy number, transcriptome, and proteome data from 94 tumors with paired normal adjacent tissues (NATs) (Figure 1A–C, S1A, and Table S1). As expected, mRNA and protein levels of 3q genes were significantly upregulated (P= 3.52e-14, 2.58e-07, two-sided Wilcoxon rank-sum test, applied below unless otherwise specified), while those of 3p genes were downregulated (P= 1.58e-16, 1.27e-06) in tumor-to-NAT comparisons, as compared to other chromosomes. This reflects the direct impact of these aneuploidies on mRNA and protein levels in patients.
To model these chromosome 3 aneuploidies prevalently associated with LSCC and dissect the mechanisms of protein dosage regulation, we leveraged our isogenic lung epithelial cell models3. The 3p loss cells were generated using CRISPR-based engineering, with some clones further adapting by duplicating the wild-type copy of chromosome 3, leading to 3q gain cells (Figure 1D). Along with diploid wildtype (WT) cells, this system allows controlled dissection of proteome remodeling in aneuploidy, independent of genomic variability, a major confounder in patient data20,32 (Figure S1B). We quantified an average of 11,452 transcripts per genotype by mRNA-seq and 8,090 protein groups on average per genotype using DIA-MS, an accurate and reproducible MS approach 33
34 (Table S2). Utilizing this approach, an average Spearman ρ of 0.987 across three biological replicates was achieved (Figure S1C). In addition, mRNA and protein abundances were globally correlated (ρ= 0.62, Figure S1D), demonstrating high-quality proteomic measurements.
In 3q gain cells, median fold changes (FCs) of mRNA and protein levels for 3q genes were 1.46 and 1.34, respectively (Figure S1E), while in 3p loss cells, the median mRNA FC was 0.52, and protein FC was 0.65. We further categorized these changes as cis-effects (within aneuploidy-affected regions) or trans-effects (on disomic chromosomes, Figure S1F), confirming significant enrichment of cis-effects in both 3q gain (20.09% cis vs. 4.72% trans, P= 2.48e-16, Fisher’s exact test) and 3p loss cells (31.58% cis vs. 5.94% trans, P= 1.25e-28) during protein expression. Thus, we obtained a robust transcriptome–proteome dataset in isogenic cells with 3q gain and 3p loss, which confirms that mRNA and protein levels largely follow the expected 3q gain and 3p loss trends.
Protein complex stoichiometry maintenance has been identified as a primary mechanism for protein dosage compensation in cancer aneuploidy (Figure S1G), including in our previous studies 19–21,26,35. To examine this in 3q gain and 3p loss cells, we classified proteins into complex-in and complex-out groups based on their protein complex membership (as annotated in the comprehensive resource of mammalian protein complexes or CORUM database 20,36, Figure 1E–F). While chr3q genes exhibited comparable mRNA fold changes in both groups (Figure 1E), protein levels showed a significant buffering effect in 3q gain cells, with complex-in proteins exhibiting lower median FCs than complex-out proteins (P= 0.00083, Wilcoxon test). To validate this, we performed western blotting analysis on two chr3q-encoded proteins. For 3q-encoded genes, POLR2H (complex-in) showed clear buffering at the protein level, whereas TBC1D23 (complex-out) did not, supporting proteome-wide observations (Figure 1G). Importantly, a similar pattern was observed in 3p loss cells where complex-in proteins were significantly buffered at the protein level (P= 0.00053), but not at the mRNA level (P= 0.80, Figure 1F). Western blotting analysis confirmed 3p-encoded OXSR1 (complex-in) was buffered, while MAP4 (complex-out) was not (Figure 1G).
Together, our data demonstrated that the protein levels encoded by duplicated 3q or deleted 3p arm in our isogenic cell system were both buffered for complex-participating proteins.

Protein Degradation Follows Distinct Patterns in Gain- and Loss-Type Chromosome 3 Aneuploidy
Protein degradation has been implicated as a key dosage compensation mechanism in aneuploidy, particularly for proteins involved in complexes20,21. To determine whether 3q gain cells compensate for increased gene dosage via enhanced degradation, we examined whether complex-in 3q-encoded proteins exhibit accelerated turnover. Conversely, 3p loss cells allow us to test two hypotheses: (i) whether they compensate for gene deletion by broadly accelerating proteome degradation to restore protein complex balance24, or (ii) by selectively reducing degradation rates of cis-encoded complex-in proteins26.
To quantify protein degradation kinetics, we employed pSILAC-DIA, a multiplexed method that integrates pulse-chase stable isotope labeling (pSILAC) with data-independent acquisition mass spectrometry (DIA-MS)23,37. By modeling the ratios of degraded (“Light”) and newly synthesized (“Heavy”) protein fractions, pSILAC-DIA enables precise determination of degradation rates (kdeg) with proteome-wide coverage as we shown23,31,38. Our pSILAC-DIA analysis quantified kdeg for 6,173 and 6,281 proteins in WT and 3q gain cells, respectively, across four time points (Figure 2A, Table S2).
In 3q gain cells, complex-in 3q-encoded proteins were determined with significantly higher kdeg FC values (compared with WT cells) than both complex-out 3q-encoded proteins and complex-in proteins from other chromosomes (P= 0.014 and 0.0040, respectively). This reinforces degradation as a primary mechanism for buffering excess protein dosage in gain-type aneuploidy (Figure 2B). Given previous findings in Trisomy 21 cells and yeast strains20,21, this result aligns with a conserved role for degradation in counteracting protein overexpression in gain-type aneuploidy models.
In 3p loss cells, however, this protein-degradation-centric compensation mechanism failed to manifest. Interestingly, pSILAC-DIA revealed no significant kdeg changes for trans-proteins from all other non-3p chromosomes (Figure 2B), contradicting the expectation that universal degradation reduction underlies buffering in loss-type aneuploidy (i.e., hypothesis [i]). Somewhat surprisingly, our data also did not detect any significant kdeg FC differences for complex-in or complex-out 3p proteins (P= 0.91), contradicting the plausible selective protein degradation regulation in hypothesis (ii). It should be stressed that, in our model, 3p genes in 3p loss cells undergo a twofold reduction in dosage, whereas 3q genes in 3q gain cells exhibit only a 1.5-fold increase, as compared to WT. If selective degradation was a general mechanism for maintaining protein complex stoichiometry, we would have expected a pronounced reduction in kdeg for complex-in 3p proteins to counteract their reduced dosage, as a significant kdeg change was identified in 3q gain cells using the same experimental system.
To corroborate these findings, we analyzed pre-existing (“light”) and newly synthesized (“heavy”) MS signals from pSILAC labeling at the 24-hour time point. We calculated L/(H+L) ratios as a measure of undegraded, pre-existing protein fractions and compared them between 3q gain, 3p loss, and WT cells. In 3q gain cells, this ratio was significantly reduced for complex-in 3q-encoded proteins, confirming enhanced degradation (Figure 2C). In stark contrast, no such compensatory reduction in degradation was observed in 3p loss cells, based on six biological replicates across different labeling time points (Figure 2D, S2A–2C). Together, these results demonstrate that complex-in 3p proteins in 3p loss cells do not exhibit degradation reduction to maintain protein complex stoichiometry, challenging assumptions about either universal or selective degradation-mediated buffering in loss-type aneuploidy.

Selective Protein Synthesis Regulation, Not Degradation, Drives the Buffering Effect in 3p Loss-Type Aneuploidy
How do 3p complex-in proteins compensate their downregulation in 3p loss cells? Protein levels can be regulated by adjusting either synthesis or degradation rates. For example, increased degradation with constant synthesis reduces protein abundance, as seen in 3q gain cells, whereas increased synthesis without degradation changes would result in protein up-regulation37. Since 3p complex-in proteins exhibit dosage compensation in 3p loss cells without changes in degradation, we hypothesized that this buffering effect is driven by an increased synthesis rate (Figure 3A).
To test this, we first estimated steady-state protein synthesis rates (Ksyn) based on kdeg and the estimated protein copies (iBAQ values, see Methods39). In 3q gain cells, the Ksyn FC values (compared with WT) for 3q-encoded proteins showed no difference regardless of their involvement in complexes (P= 1.0, Figure 3B). In contrast, in 3p loss cells, Ksyn FC values were determined to be significantly higher for 3p complex-in proteins than complex-out proteins (P= 0.0025, Figure 3B), supporting our hypothesis. We next examined heavy signals from the SILAC labeling process, which largely reflects newly synthesized proteins. Indeed, this analysis revealed that those 3p-encoded complex-in proteins had significantly higher FC values than complex-out proteins (Figure 3C, S3A). Thus, this difference in newly synthesized protein abundance is not attributable to reduced degradation but rather to a relatively accelerated synthesis rate, explaining the buffering of 3p-encoded complex-in proteins in 3p loss cells.
To orthogonally validate this mechanism, we employed ribosome profiling (ribo-seq), a powerful widely accepted approach for profiling protein translation, in which ribosome-protected fragment (RPF) levels can serve as a proxy for translation rates40. Comparing 3p loss cells to WT, mRNA FCs showed no significant differences between complex-in and complex-out groups (P= 0.82). However, our ribo-seq results suggest that the RPF FCs were significantly higher for complex-in proteins with high reproducibility (P= 0.020, Figure 3D, S3B–S3C, Table S2). This further confirms that increased translation underlies the buffering of 3p complex-in proteins.
To evaluate whether synthesis-based buffering possibly extends to other loss-type aneuploidies, we applied our previously engineered isogenic model of 8p loss (Methods, Figure 3E)41, a frequent single-arm deletion in human epithelial cancers42. Using the identical mRNA-seq and DIA-MS approaches, we first confirmed that mRNA levels of 8p genes were not buffered (P= 0.82, Figure 3F, Table S3), whereas protein levels exhibited significant compensation (P= 0.0072, Figure 3F, Table S3). Notably, protein kdeg values remained unchanged between 8p loss and WT cells (P = 0.86, Figure 3F, Table S3), regardless of their protein complex status for both cis- and trans-encoded proteins, consistent with our findings in 3p loss. Again, based on heavy isotope signals of synthesized proteins during SILAC labeling, we found that 8p complex-in proteins had significantly higher isotopic incorporation rates than complex-out proteins (Figure 3G, S3D), demonstrating that synthesis-driven compensation could also exist in 8p loss cells.
To summarize, our results demonstrate that loss-type aneuploidy compensates for protein complex members via selective upregulation of synthesis rather than degradation, a mechanism that appears conserved in single-arm chromosomal deficiencies such as 3p and 8p loss.

Chromosome 3 Aneuploidy Alters Protein Thermal Stability and Proteomic Networks
To further understand the proteome-level adaptations triggered by chromosome 3 aneuploidy, we conducted thermal proteome profiling (TPP)43,44 combined with DIA-MS. WT, 3q gain, and 3p loss cells were heated gradually from 37°C to 65°C at increments of 1.5°C, generating soluble protein fractions for DIA-MS quantification (Figure 4A, Methods). Following, protein melting temperature differences (ΔTm) relative to WT controls were calculated, showing high reproducibility across replicates (Figure S4A). This Cellular Thermal Shift Assay (CETSA) analysis shows that in 3q gain cells, no significant shifts in thermal stability were observed for proteins encoded on the gained chromosome arm (Figure 4B, Table S4). Strikingly, in 3p loss cells, proteins encoded on the deficient 3p chromosome that participate in complexes exhibited significantly increased melting temperatures compared to both 3p complex-out proteins (P = 0.0033) and complex-in proteins encoded on other chromosomes (P= 0.0083, Figure 4B, Table S4). Western blotting-based CETSA validated these findings: MRPS25, a 3p complex-in protein, exhibited a notable +3.6°C Tm shift, whereas POLR2H, a 3q complex-in protein, remained unchanged (Figure 4C, S4B), revealing an involvement of protein thermal stability specific in 3p loss cells. Previous studies have shown that protein thermal stability changes are associated with protein-protein interaction (PPI) states44,45. We therefore systematically correlated the observed ΔTm values with the known protein interaction numbers in Bioplex 3.046 for each 3p or 3q-encoded protein. Intriguingly, a significant positive correlation between ΔTm and the number of protein interactors was observed specifically in 3p loss cells (Spearman ρ= 0.43, P= 0.010), but not in 3q gain cells (Spearman ρ= −0.03, P= 0.86; Figure 4D). These results suggest that selective enhancement of cis-protein thermal stability might be a distinctive feature of loss-type aneuploidy, potentially mediated by more instance and stronger extent of protein-protein interactions.
Given the relative increase in protein synthesis for 3p complex-in proteins in 3p loss cells (Figure 3), we next examined changes in ribosome expression and stability. First, we analyzed whether proteins involved in ribosomal complexes displayed correlated Tm shifts. Cytoplasmic ribosomal subunits exhibited a higher average Tm shift (+0.5°C) in 3p loss cells compared to 3q gain cells (~0°C, Figure 4E). Second, we integrated multi-omics datasets to perform Gene Ontology (GO) enrichment analysis. We found that transcripts associated with both large and small cytoplasmic ribosomal subunits were downregulated in 3p loss cells. Additionally, the GO term “ribosome biogenesis” was consistently downregulated across transcript and RPF levels (Figure 4F). This aligns with a recent study reporting that genes involved in ribosome biogenesis and translation are downregulated at the transcriptome level in RPE1-derived monosomic cells47, which might be associated with the general stress response caused by the aneuploidy35,48. Third, more interestingly, despite transcriptional downregulation, cytoplasmic ribosomal protein abundance seems to be buffered in 3p loss cells. This buffering was further supported by a remarkable increase in protein half-life, translation rate, and thermal stability for both large and small ribosome subunits (Figure 4F). In contrast, the ribosomal proteome exhibited weaker regulation across molecular layers in 3q gain cells (Figure S4C). Together, these results imply a complex but significant interplay of transcriptional, translational, and post-translational regulation on ribosome homeostasis in 3p loss cells.
Our multi-omic profiling also revealed broader biological adaptations beyond cytosolic ribosomes, particularly in mitochondrial function and cancer signaling pathways. Mitochondrial ribosomal subunits, essential for mitochondrial protein synthesis, exhibited stronger thermal stabilization in 3p loss cells (+1.8°C) than in 3q gain cells (+0.9°C, Figure 4E) and were consistently upregulated across transcriptome, proteome, half-life, synthesis rates, and thermal stability in 3p loss cells. These emphasized the elevated mitochondrial activity in both 3p loss and 3q gain aneuploidy. In addition, mitochondrial respiratory chain complexes I and III were coordinately upregulated at transcriptome and proteome levels in 3p loss cells, with stronger increases in protein half-life and thermal stability than in 3q gain cells (Figure S4C). Among mitochondrial respiratory proteins, those encoded by mitochondrial DNA (mtDNA) showed a relatively pronounced increase in protein abundance in 3p loss cells (Figure S4D), denoting a direct or indirect effect of 3p loss in mitochondrial ribosome stabilization. Furthermore, ECM-receptor interactions, TGF-β signaling and p53 signaling pathways were downregulated in both 3p loss and 3q gain cells (Figure 4E, S4C). However, further investigation is needed to determine the mechanistic consequences and causal links underlying the aneuploidy alternations and signaling adaptations.
In summary, our CETSA data suggest a potential PPI network-driven enhancement of protein thermal stability and the multilayered ribosomal regulation in 3p loss cells, whereas both loss- and gain-type chromosome 3 aneuploidy may commonly influence metabolic and signaling rewiring.

DISCUSSION
Proteome homeostasis in aneuploid cells has been widely attributed to protein degradation, a mechanism primarily established in gain-type aneuploidy for removing excess protein copies1,2,20,21,27,49,50. Until now, the direct large-scale protein turnover measurement in loss-type aneuploidy has been lacking to the best of our knowledge. Here, using state-of-the-art mass spectrometry-based proteomics, we demonstrate that loss of chromosome arms (3p and 8p) does not trigger widespread degradation changes, but instead engages a distinct synthesis-driven mechanism to maintain the cellular proteostasis. Our results show that proteome buffering in loss-type aneuploidy is primarily mediated by selective upregulation of protein synthesis for complex-in proteins, rather than by broad degradation suppression of trans-proteins24. At first glance, this may seem unexpected. However, coordinated upregulation of cis-encoded protein synthesis ensures synchronized production of multi-subunit complexes, providing a metabolically efficient solution for cells facing single-arm chromosome deletions.
What is the mechanistic basis for the selective upregulation of protein synthesis in loss-type aneuploidy? One possible explanation could be co-translational assembly, a process where protein complex subunits are synthesized and assembled in a coordinated manner. Prior studies indicate that nascent subunits within the same complex exhibit highly synchronized translation rates, which is prevalent in human cells51,52. This means that cells may actively modulate subunit synthesis for protein complexes and PPIs to ensure regulatory synchrony52,53. Supporting this, we observed that 3p complex-in proteins exhibit increased thermal stability in 3p loss cells, which is potentially driven by altered protein-protein interactions (Figure 4D) that reinforce co-assembly. However, the correlation between protein synthesis fold-change and Tm differences was only weakly positive (Figure S4E), suggesting that additional factors—such as ribosome regulation or chaperone-assisted folding—may contribute. Interestingly, in our results, cytosolic ribosomal proteins were transcriptionally downregulated but exhibited increased post-translational stability and extended half-lives, particularly in 3p loss cells. This might suggest a finely tuned translational adaptation that may in turn facilitate the selective synthesis compensation we observed. In addition, it will be interesting to determine the relevance of specific protein post-translational modifications in regulating protein thermal stability and turnover 54
23 in loss-type aneuploid cells.
In summary, our study unveils multi-dimensional and distinct proteome-level mechanisms for cellular tolerance to gain or loss-type aneuploidy arms, a common genome alteration in diseased human cells.

Limitations of the study.
Although our 3p and 8p loss cells exhibited similar proteomic buffering, the exact molecular mechanism triggering increased protein synthesis in loss-type aneuploidy remains unresolved. Whether this mechanism extends to other chromosome losses, tissue types, or high-grade aneuploidies remains to be determined. While we observed coordinated regulation of proteins across the transcriptome, proteome, proteostasis, and thermal stability levels, including organelle-specific trans-regulation such as ribosomal, mitoribosomal, and mitochondrial adaptations, our study does not address whether similar pathways are consistently altered in other biological contexts, such as different cell or tissue types. Although our isogenic model presents a decent system currently for studying arm-level aneuploidy in mammalian cells, we did not determine how long the proteome buffering would remain in our cultured cell models 21 and individual clones as well as how non-Chr3 chromosomes might evolve over the cell culturing process 38. Future high-throughput analyses in diverse loss-type aneuploidy models will be required to investigate these further55. Finally, while our study focuses on tolerance mechanisms through proteomic buffering, it does not resolve the causal relationship between specific aneuploidy and cancer development, a topic that has been extensively debated in the field of cancer aneuploidy research13,56,57.

RESOURCE AVAILABILITY

Lead contact
Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Yansheng Liu (yansheng.liu@yale.edu).

Materials availability
This study did not generate new unique reagents.

Data and code availability
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository58 with the dataset identifier PXD062437. RNA-seq and ribosome footprints data were uploaded to the GEO repository59 and are available on GEO (GSE293644). Original western blot image data has been deposited at Mendeley and are publicly available as of the date of publication. The DOI is listed in the key resources table. All data are publicly available as of the date of publication.

This paper does not report original code.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR METHODS

Experimental Model and Study Participant Details

Isogenic cell model generation and cell culture
All engineered aneuploid cell lines with chromosome 3p or chromosome 8p loss were generated as previously described3,41. The clones were engineered in human immortalized lung epithelial cells called AALE66. In brief, chromosome arm deletion were generated by transfecting CRISPR-Cas9 to cut near the centromere along with a plasmid containing an artificial telomere, puromycin selection marker, and 1kb DNA sequence homologous to chromosome 3p or chromosome 8p3,41. Some cells with chromosome 3p loss underwent adaptation and transitioned to a state of chromosome 3q gain3. To separate cells with 3p loss and 3q gain, we performed single cell cloning on the adapted cell population. Single-cells were plated using fluorescence-activated cell sorting without any markers or antibodies and plated in 50% conditioned media. Genomic DNA was isolated from each clone for low-pass whole genome sequencing. Copy number calls were made using ichorCNA67. AALE cell culture was maintained at 5% CO2 and 37°C in small airway cell basal medium (#CC-3119, SABM) supplemented with Supplements and Growth Factors (#CC-4124, SAGM).

Method Details

RNA sequencing
Cells from the chromosome 3 isogenic aneuploidy model (3p loss/3q gain/WT) were washed three times with PBS, snap-frozen, and then lysed with the QIAShredder columns (#79654, QIAGEN) according to the manufacturer’s instructions. For each cell line, three biological replicates were processed. Total RNA was isolated with the Qiagen RNeasy Mini Kit (#74104, QIAGEN), including the DNA digestion step using the RNase-Free DNAase set kit (#79254, QIAGEN). PolyA isolation and sequencing library preparation was performed using the NEBNext Ultra Directional RNA Library Prep Kit (#E7760, NEB). Each set of samples were pooled and sequenced in one lane of HiSeq2500 RR (Illumina), 100 basepair paired end. For the transcriptome analysis of chromosome 8 isogenic aneuploidy model (8p loss/8p WT), total mRNA samples were prepared as described above with three biological replicates, and the following library preparation was conducted in Yale Center for Genomic Analysis (YCGA). Sequencing was conducted on NovaSeq X Plus platform (Illumina).
RNA sequencing from chromosome 3/chromosome 8 aneuploidy model was processed using the nf-core/rnaseq pipeline (v3.14.0)68. The RNAseq were also aligned to the human genome GRCh38. A simple TPM (transcripts per million) cutoff of 1 was applied to retain possibly expressed genes at the transcriptomic level22.

Ribosome Footprints
3p loss cells and 3p WT cells were subjected to ribosome profiling. For each cell line, three biological replicates were processed. Cells were cultured under standard conditions until they reached 80% confluency. They were then treated with 10 μg/μL cycloheximide for 5 minutes at 37°C. Following treatment, cells were harvested and lysed. Ribosome-protected fragments (RPFs) were generated and purified using RiboLace Pro (Immagina Biotech; catalog # RL00P-12)69. The RPFs underwent end repair using the Small RNA Library Prep Kit (NEB; catalog # E7300S) and size selection using the RNA Clean & Concentrator-5 Kit (Zymo; catalog # R1015) to enrich RNA fragments sized smaller than 200 nt. RPFs were reverse transcribed and tagged by Small RNA Library Prep Kit (NEB; catalog # E7300S) following the recommended conditions. RPF libraries were then amplified by Phusion High-Fidelity PCR Kit (NEB; E0553L). Further size selection and purification of RPF libraries was conducted using Mag-Bind TotalPure NGS beads (Omega Bio-tek; catalog # M1378–01) at a beads-to-product ratio of 1.3:1. The NGS libraries were quantified using the sparQ Universal Library Quant Kit (Quantabio; catalog # 95210–100) and analyzed using a TapeStation (Agilent) before being sequenced with NovaSeq X Plus (Illumina) in YCGA or NextSeq 1000 system in Quintara Biosciences. The target sequencing depth for the RPF libraries was 200 million reads per library. The analysis of ribosome footprint data was performed using RiboDoc (v0.9.1.1)61 on Podman (v3.4.4). RPFs ranging from 25 to 45 nucleotides were aligned to the human genome (GRCh38) within the analysis pipeline. For identification and removal of human rRNA contaminants, the SILVA database was utilized70. The obtained ribosome footprint counts were first normalized based on the total counts in each sample. Then, the normalized ribosome footprint counts for each gene were used to calculate RPF FC values.

Total proteome analysis by DIA mass spectrometry
Cells involved in chromosome 3 isogenic cell model (3q gain/3p loss/chr3 WT) and chromosome 8 isogenic cell model (8p loss/ chr8 WT) were harvested and digested as previously described23. In brief, cells were washed three times by precooled PBS, released from dishes using trypsin (Lonza, #CC-5012) and snap-frozen in liquid nitrogen. The cell pellets were immediately lysed by adding 10 M urea (Sigma-Aldrich, #U5218) containing complete protease inhibitor cocktail (Roche, #4693116001). Then the cell pellets were ultrasonically lysed by sonication at 4 °C for 2 min using a VialTweeter device (Hielscher-Ultrasound Technology) and then centrifuged at 20,000 g for 1 h to remove the insoluble material. The protein concentration of supernatant was determined using BioRad Bradford assay (#5000201), and around 200 μg protein was transferred to a clean Eppendorf tube. The protein was reduced using dithiothreitol (DTT, Sigma-Aldrich, #D0632) at 56 °C for 1 h at the final concentration of 10 mM, and then alkylated using iodoacetamide (IAA, Sigma-Aldrich, #I1149) in the dark for 45 min at room temperature with the final concentration of 20 mM. After alkylation, 100 mM NH4HCO3 (Thermo Fisher Scientific, # 09830) was added until the concentration of urea is below than 1M. Sequencing grade porcine trypsin (Promega, #V5111) was then added at a ratio of 1:20 overnight at 37°C. The resulting peptide mixture was desalted with a C18 column (MarocoSpin Columns, NEST Group INC). The final peptides amount was determined by Nanodrop (Thermo Scientific). About 2 μg peptide digests were loaded onto mass spectrometry for the following DIA analysis.
For the total proteome analysis of chromosome 3 isogenic cell model, an Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Scientific) coupled to a nano-electrospray ion source (NanoFlex, Thermo Scientific) was used as the liquid chromatography-mass spectrometry (LC-MS) system with the data acquisition controlled by Xcalibur (Thermo Scientific). LC separation was carried out on EASY-nLC 1200 systems (Thermo Scientific, San Jose, CA) using a self-packed analytical PicoFrit column (New Objective, Woburn, MA, USA) (75 μm× 50cm length) with C18 material of ReproSil-Pur 120A C18-Q 1.9 μm (Dr. Maisch GmbH, Ammerbuch, Germany). A 240-min measurement with buffer B (80% acetonitrile containing 0.1% formic acid) from 4% to 37% and corresponding buffer A (0.1% formic acid in H2O) during the gradient was used to elute peptides from the LC at a flow rate of 300 nL/min. The DIA-MS method consisted of one MS1 scan and 33 MS2 scans of variable windows as previously described33,71. The MS1 scan range was 350–1650 m/z, and the MS1 resolution was 120,000 at m/z 200. The MS1 full scan AGC target value was set to be 2.0E6, and the maximum injection time was 50 ms. The MS2 resolution was set to 30,000 at m/z 200 with the MS2 scan range 200–1800 m/z. The MS2 AGC target value was set to be 1.5E6, the maximum injection time was 52 ms, and the normalized HCD collision energy was 28%. Both MS1 and MS2 spectra were recorded in profile mode. For the total proteome analysis of 8p WT and 8p loss cells, a 150-min measurement with buffer B from 4% to 37% and corresponding buffer during the gradient was used to elute peptides from the LC, and DIA-MS parameters remained unchanged.
DIA-MS data analyses were conducted using Spectronaut (v16.2). The total proteome data were directly searched against the SwissProt protein database (February 2018, 20,258 entries). Carbamidomethylation at cysteine was set as a fixed modification. The possibilities of oxidation at methionine and acetylation at the protein N-terminals were set as variable modifications. Overall, both peptide- and protein-FDR (based on Qvalue) were controlled at 1%, and the data matrix was filtered by Qvalue. The data extraction was performed by Spectronaut with default settings and a Q value cut-off of 1% at both peptide and protein levels. All the other Spectronaut settings for identification and quantification were kept as default.

Cell doubling time determination
The cell doubling time measurement was conducted on a hemocytometer by triplicate cell counting at three time points (24, 60 and 84 h), and the first measurement was performed after 24 hours of cell seeding. The doubling time was estimated using the following equation20:

Where time is the delta time between two time points, the FCN is the cell number at the final time point and ICN is the cell number at the initial time point. As a result, the average doubling times for WT, 3p loss and 3q gain cells were 40.8 h, 54.2 h and 42.8 hours respectively. And the doubling time for 8p WT and 8p loss cells is determined to be very close at 38.5 h.

Pulsed SILAC experiment and multiplex-DIA mass spectrometry
Cells from the chromosome 3 and isogenic cell model (3q gain/3p loss/chr3 WT) and the chromosome 8 isogenic cell model (8p loss/8p WT) were subjected to pSILAC measurement. SILAC 1640 medium (Thermo Fisher, #88365) lacking L-arginine and L-lysine was firstly supplemented with 10% dialyzed FBS (Thermo Fisher, # 26400044) and the same penicillin/streptomycin mix. The Heavy L-Arginine-HCl (13C6, 15N4, purity >98%, #CCN250P1), and L-Lysine-2HCl (13C6, 15N2, purity >98%, #CCN1800P1) were purchased from Cortecnet and spiked into the culturing medium as described previously23. SILAC medium was further supplemented with excess L-proline (Sigma-Aldrich, #P5607) at a concentration of 400 mg/L to prevent the potential arginine-to-proline conversion 38,72. Before SILAC labeling, cells were seeded at 50%−60% confluency and incubated in normal light medium for 24 h, at 5% CO2 and 37 °C overnight. For pSILAC labeling, five time points, including 0 h and four labeling points 1, 4, 12, and 24 h were applied. After pSILAC labeling for each time point, cells were washed three times with precooled PBS and snap-frozen in liquid nitrogen immediately. Then 10 M urea containing complete protease inhibitor cocktail (Roche, #4693116001) was added to each dish, and cells were collected by scraping. The protein extraction, digestion and DIA-MS were performed as employed in the total proteome analysis. Specifically, the pSILAC assays were performed with six independent biological replicates for 3q gain, 3p loss, and chr3 WT cells, and three independent biological replicates for 8p WT and 8p loss cells.
DIA-MS data analyses were performed using Spectronaut (v14.1) on an optimized pSILAC-DIA workflow31. For pSILAC, the spectral library was generated, based on DIA measurements on relevant pSILAC samples. The default settings for the Pulsar search of Spectronaut were used. For modification in labeling, SILAC labels (“Arg10” and “Lys8”) were specified in the second channel. Oxidation at methionine was set as variable modification, and carbamidomethylation at cysteine was set as a fixed modification. For the targeted data extraction and subsequent identification and quantification, the spectral library-based search was chosen. In the search setting, the Inverted Spike-In (ISW) workflow was used31. Both peptide and protein FDR cutoff (Qvalue) were controlled at 1%. For the data analysis in chromosome 8 isogenic cell model, Spectronaut (v16.2) was used, and the other parameters remained unchanged.

CETSA and DIA mass spectrometry
Cells from the chromosome 3 isogenic cell model (3q gain/3p loss/chr3 WT) were subjected to pSILAC analysis. The intact cell CETSA experiment was performed as previously described44. Cells were collected, washed with PBS twice, and resuspended in 100 ml of PBS to obtain a single-cell suspension with a concentration of 40 million cells/mL. The single cell suspension was evenly divided into 15 aliquots, and heated at 37 °C, 39 °C, 41 °C, 43 °C, 45 °C, 47 °C, 49 °C, 51 °C, 53 °C, 55 °C, 57 °C, 59 °C, 61 °C, 63 °C and 65 °C for 3 min in a 96-well thermocycler (BIO-RAD C1000 Touch Thermal Cycler), followed by 3 min cooling at 4 °C. The cells were lysed after heat treatment by adding 2x kinase buffer to a final concentration of 50 mM HEPES (Sigma-Aldrich, #H4034), pH 7.5, 5mM beta-glycerophosphate (Sigma-Aldrich, #G6376), 0.1 mM sodium orthovanadate (Sigma-Aldrich, #450243), 10mM MgCl2 (Sigma-Aldrich, #M8266), 2mM TCEP (GOLDBIO, #TCEP25) and 1x protease inhibitor cocktail, followed by the three freeze-thaw cycles with liquid nitrogen. The cell debris and aggregated protein pellet were removed by centrifugation at 20,000 g for 20 min at 4°C. The supernatant was collected and quantified by BioRad Bradford assay (# 5000201). A bovine protein, alpha-1-acid glycoprotein (Sigma-Aldrich, #G3643), was added to each sample to a final concentration of ~ 1 ‰ as an external standard to normalize signal between different mass spectrometry injections. The proteins were subjected to in-solution digestion as described in the total proteome sample preparation. The trypsin was added at a ratio of 1:10. Around 2 μg peptides of each sample were loaded onto mass spectrometry for the following DIA analysis, using the same 150-minute gradient MS method as employed in the total proteome analysis.
All DIA data analysis following CETSA experiment were performed using the DirectDIA function of Spectronaut (v 15.1). The DIA runs were all directly searched against the SwissProt protein database. The possibilities of oxidation at methionine and acetylation at the protein N-termini were set as variable modifications, whereas carbamidomethylation at cysteine was set as a fixed modification. The cross-run normalization was disabled. After performing DirectDIA quantification, protein abundance data were adjusted based on the external bovine protein abundance in each sample (see above) to match the theoretical relative ratio, which was determined by the external protein amount added to each sample. Subsequently, the adjusted DIA intensity values were recalculated—specifically, for each temperature point, the percentage of supernatant-derived peptides that were loaded onto the MS was determined (this percentage was denoted as coefficient A). The DIA intensity at each temperature point was then divided by coefficient A to estimate the initial total amount of each protein in the supernatant. Next, fold changes at each temperature were calculated against the respective 37 °C and as input for the TPP package (v3.24.0) to identify melting points63. The filtering criteria for normalization were modified as this: fold change at the tenth temperature should be between 0.4 and 0.6, fold change at the fourteenth temperature should be between 0 and 0.3, and fold change at the fifteenth temperature should be between 0 and 0.2. The melting temperature difference was then calculated. We only considered proteins quantified with at least 29 temperature points across two biological replicates in the same genotype for downstream analysis.

Western Blot
Protein samples were isolated from one million cells using RIPA lysis buffer (Santa Cruz Biotechnology, #sc-24948) and their concentration was determined by Bradford colorimetric assay. For each western blot lane, 20ug of protein were loaded in 4–12% Bis-Tris polyacrylamide gels (Invitrogen, #NP0321BOX) and ran using MES-SDS buffer (Invitrogen, #B0002) for 90 minutes. Next, proteins were transferred to a nitrocellulose membrane using the iBlot2 semidry system (ThermoFisher, #IB23002) and blocked in TBS buffer (0.05% tween20, 5% non-fat dry milk) for 1hr at RT. Primary antibody incubation was conducted overnight at 4°C followed by 2-hours secondary antibody incubation at RT. The antibodies and respective dilutions used were: OXSR1 (Proteintech, #15611–1-AP, 1:500); MRPS25 (Novus Biologicals, #NBP2–42628, 1:200); MAP4 (Proteintech, #11229–1-AP, 1:500); POLR2H (Proteintech, #15086–1-AP, 1:500); TBC1D23 (Proteintech, #17002–1-AP, 1:500); Cofilin (Cell Signaling, #5175, 1:1000); Anti-mouse IgG HRP-linked (Cell Signaling, #7076, 1:3000); Anti-rabbit IgG HRP-linked (Cell Signaling, #7074, 1:3000). Finally, luminescence signal was developed using SuperSignal™ West Pico PLUS Chemiluminescent Substrate (ThermoFisher, #34577) and imaged in a Chemidoc MP imaging system (Bio-Rad). To analyze the protein levels, each band intensity was determined using imageJ software and subsequently normalized to the endogenous control levels (Cofilin).
To validate changes in thermal stability using Western blotting, three million living cells were collected, washed and resuspended in PBS. The cell suspension was divided equally into six different aliquots in PCR tubes at a 40,000 cells/uL concentration. Next, cell suspension was exposed to a temperature gradient of 37.0°C, 42.6°C, 48.2°C, 53.8°C, 59.4°C, and 65.0°C for 3 min in a 96-well thermocycler, followed by 3 min cooling at 4 °C. Next, an equal volume of 2x kinase buffer was added followed by three flash freeze-thaw cycles in liquid nitrogen in order to lysate the cells. Finally, cell debris was removed by centrifugation at 20,000g for 20min at 4°C and the supernatant was collected for western blot analysis.

QUANTIFICATION AND STATISTICAL ANALYSIS

Annotation of protein complex members
The assignment of proteins and their corresponding genes to complexes was determined based on the core set of CORUM 3.0, as curated and published36.

Protein degradation rate calculation
The growing cells are respectively maintained in a steady state37, which means for all the proteins the net change is zero during the SILAC labeling period73. This principle enables us to calculate the protein turnover rate by monitoring the abundance of light and heavy peptide forms across several time points. To estimate protein degradation rate and half-life time, we applied a modeling approach consistent with previous studies20,38,74,75.
Specifically,
The intensity of light and heavy isotope incorporated precursors was extracted from the Spectronaut quantitative results.

Relative isotope abundance () was then determined by calculating the intensity ratio between light channel and the sum of both light and heavy channel74, i.e.

The rate of loss of the light isotope () was determined by modeling RIA onto an exponential decay model, i.e.,

In addition, a basic assumption is that at time 0, the heavy intensity is 0 ().

Here a nonlinear least-squares estimation was performed to fit the model. The values of precursors were weighted as the value of protein20,38. Only the precursors with monotonically increasing H/L ratio during the SILAC labeling period were accepted for the following analysis, a strict filter strategy as previous studies76.
Finally, the protein degradation rate is calculated using the next equation:

Specifically, in the degradation rate analysis of chromosome 3 isogenic models, among all six biological replicates, only protein degradation rates that were measured in at least three biological replicates in both 3q gain and WT (or 3p loss and WT) were included for downstream analysis. In the degradation rate analysis of chromosome 8 isogenic models, only protein degradation rates that were measured in all three biological replicates were considered for downstream analysis.

Protein synthesis rate estimation
The protein abundance changes as a function of time can be given by the following simple relationship37,77:

Where and are the protein synthesis rate and protein degradation rate respectively, and the is the absolute amount of protein at a given time. At a steady state, dP/dt=0, and the above equation is reduced to

i.e., in a given cell, protein synthesis rate equals protein degradation rate times protein amount. The protein amount [P] is represented by the iBAQ values, a proxy for protein copy number78,79. The iBAQ values for pSILAC samples were calculated as follows: for each precursor ion, the DIA intensities of the light and heavy channels were summed. The median intensity of all precursor ions within each sample was then used for cross-sample normalization. After normalization, the median intensity of all precursor ions corresponding to each protein was used as the protein-level DIA intensity. To convert DIA intensity to iBAQ values, a linear equation was fitted to describe the relationship between log2-transformed DIA intensities and log2-transformed iBAQ values, using the DIA intensities and iBAQ values extracted directly from Spectronaut for the relevant non-labeled samples. Finally, the iBAQ values for pSILAC samples were inferred based on the fitted linear equation and their corresponding DIA intensities.

Lung squamous cell carcinoma data analysis
Gene level data, RNA expression data, and total proteome data of lung squamous cell carcinoma (LSCC) patients were downloaded from a published dataset of CPTAC32. Copy number alterations (CNAs) were derived from 108 LSCC tumors. The analysis of the comparison between cancer and paired normal adjacent tissues (NAT) utilized CNA, mRNA, and protein from 94 tumor/NATs pairs, and only proteins with valid mRNA FCs and protein abundance FCs across all the 94 sample pairs were included.

Protein interaction and interactor numbers
The interacator number of each protein was downloaded from Bioplex 3.0 (https://bioplex.hms.harvard.edu/interactions.php), a published human protein interactome dataset46. The bait lists from HEK293T and HCT116 cells, are merged based on the gene symbol. Only the proteins with an interactor number greater than 1 in both lists were retained. Subsequently, the interactor number of each protein was averaged and utilized for the subsequent analysis.

GO functional enrichment analyses
Protein-specific annotation was downloaded from DAVID Bioinformatics Resources 6.865,80. The 1D enrichment analysis was performed in Perseus v1.6.14.062.

Data visualization and other bioinformatic analysis
Most data visualization was performed in R81. The following R packages were used to and visualize the data: ‘ggplot2’ (boxplots, scatterplot, and volcano plots), ‘pheatmap’ (heatmaps), ‘LSD’ (heatscatter plots). The melting curves (Figure 4C) were generated using GraphPad Prism version 8.0 (GraphPad Software, San Diego, California USA). All networks were visualized in Cytoscape v3.10.164. All boxplots were generated using the geom_boxplot function in ggplot2 package of R (center line, median; box limits, upper and lower quantiles; whiskers, Min to Max values). All the P-values in the boxplots in the main or supplementary figures were derived by the two-sided Wilcoxon rank-sum test.

Supplementary Material
1Document S1.
Figures S1–S4.

2Table S1. Transcriptome and proteome analysis of LSCC tumor and paired normal adjacent tissues in CPTAC database, related to Figures 1 and S1.

3Table S2. Transcriptome analysis, proteome analysis, protein turnover analysis and ribosome profiling in engineered chromosome 3 isogenic aneuploidy cell model, related to Figures 1–4 and S1–S4.

4Table S3. Transcriptome analysis, proteome analysis and protein turnover analysis in engineered chromosome 8 isogenic aneuploidy cell model, related to Figures 3 and S3.

5Table S4. Thermal stability investigation in chromosome 3 isogenic aneuploidy model, related to Figures 4 and S4.