Bulk and selective autophagy cooperate to remodel a fungal proteome in response to changing nutrient availability

SUMMARY

Cells remodel their proteomes in response to changing environments by coordinating protein synthesis and degradation. In yeast, degradation occurs via proteasomes and vacuoles, with bulk and selective autophagy supplying vacuolar cargo. Although these pathways are known, their relative contributions to proteome-wide remodeling remain unreported. To assess this, we developed a method (nPL-qMS) to pulse-label the methylotrophic yeast Komagataella phaffii (Pichia pastoris) with isotopically labeled nutrients that, when coupled to quantitative proteomics, enables global monitoring of protein degradation following an environmental perturbation. Genetic ablations revealed that autophagy drives most proteome remodeling upon nitrogen starvation, with minimal non-autophagic contributions. Cytosolic protein complexes, including ribosomes, are degraded through bulk autophagy, whereas degradation of peroxisomes and mitochondria use selective autophagy. Notably, these pathways are independently regulated by environmental cues. Our approach expands known autophagic substrates, highlights autophagy’s major role in fungal proteome remodeling, and provides rich resources and methods for future proteome remodeling studies.

Graphical Abstract

eTOC Summary

Telusma et al. develop and deploy quantitative proteomics to measure protein turnover in yeast as it transitions between different nutrient environments. They show that bulk and selective autophagy cooperate to degrade much of the existent proteome in response to nitrogen limitation, which supports synthesis of new, environment-specific proteins.

INTRODUCTION

Organisms have evolved conserved strategies to cope with fluctuations in available nutrient sources, including altering their proteome through a combination of modulated protein synthesis, protein degradation, and cell growth-driven protein dilution^1-4. When facing changes in nutrient availability, including deficiencies, these pathways work collectively to deplete cells of now-superfluous proteins, thereby allowing for the reallocation of resources towards the synthesis of proteins required for viability in the new environmental condition. In yeast, such protein degradation is largely driven by the proteasome, a compartmentalized AAA energy-dependent protease, and the vacuole – a cellular compartment defined by a single membrane and filled with largely nonspecific hydrolases capable of degrading and recycling proteins, lipids, and nucleic acids⁵.

The ability to degrade proteins is a fundamental aspect of an organism’s adaptation to changing conditions and this degradation is carefully regulated at multiple levels to protect against the toxic effects of nonspecific protein turnover^6,7. Indeed, proteasomal degradation requires that protein substrates be recognized and unfolded by a regulatory complex that typically requires that substrates bear a polyubiquitin modification⁸. Since proteasomal substrates must be mechanically unfolded and translocated through a narrow axial channel in the regulatory particle, proteasomal degradation is thought to primarily process single polypeptides sequentially⁸. In contrast, protein complexes, aggregates, condensates, and cellular organelles often rely on the process of macroautophagy, hereafter referred to as autophagy, for targeting to, and subsequent degradation in, the lysosome/vacuole^7,8. Through autophagy, substrates are engulfed by a double-membraned vesicle that later fuses with the vacuole where encapsulated substrates are degraded. In some instances, this encapsulation occurs non-selectively in a process known as bulk autophagy that, in fungi, strictly depends on the transmembrane protein Atg9^9-11. Additionally, cells can utilize receptor proteins to selectively tether specific cargos to the core autophagic machinery, thereby facilitating their degradation while minimally compromising the rest of the cellular content. In yeast, this “selective autophagy” process relies on Atg9, the autophagy scaffolding protein Atg11, and cargo-specific receptor proteins, including Atg30 and Atg32, which target peroxisomes and mitochondria for selective autophagic degradation, respectively^9,12-16.

The methylotrophic yeast Komagataella phaffii, is particularly well-suited to studying proteomic responses to changing nutrients as it can be grown on a variety of carbon sources, including methanol, glucose, and oleate, each of which requires a uniquely tailored proteome^17-20. As such, transitioning between carbon sources necessitates altered levels of key metabolic pathways to efficiently utilize the available carbon compounds, with these alterations often achieved through proteome remodeling. For instance, when provided the reduced C₁-compound methanol as the sole source of carbon and energy, K. phaffii grows robustly using a methanol utilization pathway that is distributed between the cytosol and a specialized membrane-bound organelle known as the peroxisome^21-23. Indeed, growth of K. phaffii in methanol induces a massive proliferation of peroxisomes and associated methanol utilizing enzymes, including alcohol oxidase (Aox1), which is targeted to the peroxisome^24-26. In contrast, upon adaptation to alternative carbon sources, such as glucose, these peroxisomes are rapidly degraded in the vacuole through a selective form of autophagy known as pexophagy^27,28, and the cell concomitantly upregulates glycolysis and other glucose-utilizing pathways²⁹.

Whereas the pathways responsible for degrading specific substrates in response to changing environmental conditions have been identified for a small subset of individual substrates in K. phaffii¹⁶, we lack a clear, systems-wide understanding of the contributions of the ubiquitin-proteasome system (UPS) and of selective or bulk autophagy to proteome remodeling, particularly during times of nutrient adaptation or starvation. Here, we develop a general approach to quantify proteome dynamics by coupling stable isotope pulse-labeling to quantitative mass spectrometry in K. phaffii. We then employ this assay in the context of growth condition perturbations and mutations targeting key genes involved in the aforementioned degradation pathways. The resulting quantitative resource describes time-resolved dynamics of a combined ~3,200 proteins across four growth perturbations in six different genetic backgrounds. Further, we detail exemplar uses of this resource to assess the relative contributions of bulk autophagy, selective autophagy, and non-autophagic processes to cellular adaptation in changing environmental conditions. We find that bulk autophagy drives degradation of most cellular compartments under conditions of nutrient stress, with selective autophagy supporting degradation when cells are grown in steady-state conditions in methanol. We further observe distinct selective autophagy pathways, with pexophagy occurring rapidly relative to mitophagy under our stress conditions. Importantly, we discover that these degradative pathways do not compete for a limited resource, as ablation of either pathway has no impact on the rate of the other.

RESULTS

Protein degradation drives proteome remodeling during glucose adaptation.

K. phaffii grows robustly on a variety of primary carbon sources, including glucose, methanol, and oleate^17,30, and does so by adjusting its proteome composition in each growth medium²⁰. Indeed, when comparing the proteome of cells grown in methanol to that of cells grown in glucose, we identified ~271 proteins whose levels changed more than two-fold using quantitative mass spectrometry (Figure S1A-B), with 190 upregulated in methanol and 81 upregulated in glucose. Specifically, we observed large changes in levels of proteins linked to methanol and glucose metabolism (Figure S1C-D), consistent with the concept of metabolic specialization^31,32. To understand how cells transitioned from a methanol-adapted proteome to one suited for growth on glucose, we assayed the response of methanol-specific proteins to glucose adaptation via nutrient-pulse-labeling coupled to quantitative mass spectrometry (nPL-qMS)^33-35, which we hypothesized would allow us to independently monitor protein synthesis and degradation on a per-protein basis across large swaths of the proteome.

Briefly, we grew wild-type K. phaffii strain GS115 in “light” ¹²C-labeled, ¹⁴N-labeled methanol medium (M+N). At time 0, this M+N medium was rapidly replaced with “heavy” ¹³C-labeled glucose medium that lacked nitrogen (G−N) (Figure 1A; see Methods). Quantitative mass spectrometry of tryptic peptides isolated from these cells revealed that the stable isotope pulse distinguished newly synthesized proteins (¹³C-labeled, post-pulse) from those present pre-pulse (¹²C-labeled, pre-pulse) based on the isotope envelope in their mass spectra (Figure 1B). Notably, throughout the time course, a fixed volume of culture was harvested, which allowed for direct comparison of the pre-pulse abundance over time without the need to consider dilution by growth. We quantified ~3,200 proteins across three replicate time courses and observed that the abundance of many ¹²C-labeled peptides (pre-pulse) decreased as a function of time following the pulse, consistent with protein degradation (Figure 1C). Although most proteins underwent degradation upon this nutrient transition, a minority remained stable (Table S1), including those annotated as cell-wall resident³⁶. This observation aligned with the notion of cell wall inaccessibility to lysosomal and proteasomal degradation and indicated that the protein degradation we observed across the vast majority of the proteome did not simply result from cell lysis.

Figure 1. Stable isotope pulse-labeling coupled to mass spectrometry quantifies protein degradation.

(A) Schematic depiction of pulse-labeling experimental design. Media composition pre- and post-pulse, including isotopically labeled species noted. At each timepoint, a fixed volume of culture was added to a fixed reference standard, which was ¹²C,¹⁵N-labeled. (B) Exemplar mass spectra highlighting proteins degraded (top), synthesized (middle), and simultaneously degraded and synthesized (bottom). Pre-pulse, reference, and post-pulse isotope envelopes noted. A subset of timepoints (0, 2, 8, and 26 hours) from wild-type cells are plotted. (C) Clustered heatmap of proteome degradation compared between wild-type, atg9Δ, atg11Δ, pep4Δprb1Δ strains transitioning from growth in M+N to G−N media. Columns represent individual timepoints (see Methods) and rows represent individual proteins. For each peptide, relative abundance value at time n was calculated as the pre-pulse peptide abundance relative to that in the reference isotope channel, normalized to that at time 0. The abundance of each isotope species was estimated from the precursor (MS¹) signal of the M0, M1, and M2 isotopomers using Spectronaut. That is: (pre-pulse_n/reference)/(pre-pulse₀/reference). Protein relative abundance is calculated as the median relative abundance across all uniquely mapped peptides. Timeponts plotted as a linechart above clustermap. (D) Representative fits highlighting the different modes (annotations at right of plots) of degradation captured in C. Kinetic traces are fit as: abundance$(t)={\mathrm{A}}_{\text{deg}}\ast \text{exp}(\text{-}{k_{deg}}^{\ast }t)+\mathrm{B}$ (eq. 1, see Methods), with fit parameters $k_{deg}∕{\mathrm{A}}_{\text{deg}}$ noted above each plot. Uniprot protein ID and protein common names are listed. (E) Distributions of the fit $k_{deg}$ rate (left) and ${\mathrm{A}}_{\text{deg}}$ amplitude (middle) parameters for each protein in each condition as in D. Using these fit parameters, a univariate measure of the fraction of pre-pulse material degraded in 30 hours for each protein was determined, and the distribution of this estimated parameter is plotted for each strain (right).

By carefully inspecting each isotopic envelope, we could additionally monitor protein synthesis, observing proteins that were: 1) exclusively synthesized; 2) simultaneously synthesized and degraded; and 3) degraded in the absence of new synthesis (Figure 1B). We observed that proteins central to methanol utilization were robustly degraded with minimal synthesis, whereas proteins involved in glucose utilization were rapidly synthesized and degraded at rates that increased their levels on balance (Figure S2). Taken together, we found that the vast majority (~90%) of the quantifiable proteome underwent degradation, and that many of the robustly synthesized proteins were related to glucose metabolism.

Atg9-dependent bulk autophagy drives proteome remodeling.

We next assessed the role of autophagy in the large-scale protein degradation we observed by repeating this nPL-qMS experiment in autophagy-deficient atg9Δ cells. Most (2,266/2,549) proteins were stabilized in the Atg9-deletion background, consistent with Atg9-dependent autophagy driving much of the proteome remodeling observed in wild-type cells. Notably, when this experiment was performed in cells lacking vacuolar proteases (pep4Δprb1Δ), the results were similar to that in atg9Δ cells (Figure 1C). To determine the functional consequences of autophagy inhibition, we measured the growth rates of wild-type and atg9Δ cells in our various media. Observed growth rates were virtually indistinguishable between these strains in either methanol (M+N) or glucose (G+N) medium, and they initially grew similarly when shifted from methanol (M+N) to glucose medium lacking nitrogen (G−N) (Figure S3). However, after ~5 hours in G−N medium, growth of atg9Δ cells ceased, whereas growth persisted in wild-type cells for at least an additional 10 hours, highlighting the functional impact of autophagy on nutrient adaptation. This effect was not glucose-dependent, as a similar dependence on Atg9 for cell growth was observed as cells transition from M+N to M−N medium (Figure S3).

De novo protein synthesis is suppressed in autophagy deficient cells

We assessed how nutrient deprivation impacted protein synthesis by examining our mass spectra for the presence of ¹³C-labeled tryptic peptides, which indicated newly synthesized proteins in our nPL-qMS assay. In wild-type cells, protein synthesis patterns reflected glucose adaptation, as proteins involved in methanol metabolism were not synthesized, whereas glycolytic enzymes were robustly synthesized and exhibited broad ¹³C-labeled isotope envelopes (Figure S2A-B). To determine if impaired protein degradation affected the ability of cells to sustain protein synthesis in nitrogen-starved conditions, we monitored protein synthesis in autophagy-deficient (atg9Δ) cells, specifically focusing on proteins utilized in glucose medium. Whereas these proteins were actively synthesized in wild-type cells, their synthesis was greatly suppressed in atg9Δ cells, highlighting a role for autophagy in maintaining translational capacity under stringent conditions - likely through the recycling of intracellular resources. In support of this autophagy-dependent recycling model, we noted a shift of the newly synthesized isotope envelopes towards the mass expected for the pre-pulse nutrients (¹²C, ¹⁴N) in wild-type cells compared to atg9Δ cells (Figure S4A). Notably, we observed that Atg9 was dispensable for de novo protein synthesis as cells adapted to glucose medium in the presence of nitrogen (Figure S4B).

Rates of autophagic degradation vary broadly between protein substrates.

To quantify degradation proteome-wide, we modeled the time-dependent change in protein abundance as a single-exponential decay process (see Methods), producing apparent first-order degradation rate constants (rate, $k_{deg}$), and the fraction of initial material degraded (amplitude, ${\mathrm{A}}_{deg}$) for each protein. In wild-type cells, the half-lives of proteins varied widely, from ~1 hour to over 130 hours (corresponding to $k_{deg}$ ranging from ~0.8 hr⁻¹ to ~0.005 hr⁻¹), and amplitudes ranged from 0.2 to 1. Interestingly, many proteins were only partially degraded, (i.e. ${\mathrm{A}}_{deg}<1$), consistent with proteolytic protection of these proteins via potential mechanisms such as differential subcellular localization^37,38, post-translational modification^39-41, or exhaustion of degradative capacity^42-45. In atg9Δ and pep4Δprb1Δ cells, we observed sharp reductions in the observed degradation rates and amplitudes that collectively had profound impacts on the fraction of initial material degraded during a 30-hour timeframe (Figure 1D-E). We next defined Atg9-dependent substrates as those whose kinetic fit reported that more than 20% of the starting material was degraded after 30 hours in wild-type cells, and that less than 20% was degraded in the same time span in the atg9Δ strain. Although most protein degradation was ablated in these autophagy-deficient cells, a small number of proteins were degraded in an Atg9-independent fashion (Figure 1D). Interestingly, these Atg9-independent substrates separated into two different groups: 1) those whose degradation relied on vacuolar proteases, consistent with macroautophagy-independent vacuolar degradation (Table S2), and 2) those degraded in a Pep4/Prb1-independent fashion, consistent with degradation by the ubiquitin-proteasome system or other non-vacuolar proteases (Table S3). This later group included the proteins Pex1, Pex5 and Pex6, which are thought to be part of the peroxisomal Pex5 receptor recycling/RADAR pathway⁴⁶, with Pex5 being a reported proteasomal substrate when this recycling pathway is blocked⁴⁷. Notably, strain-dependent substrate degradation profiles, and their associated genetic dependencies were largely consistent across replicates, though some replicate-to-replicate variability was observed as expected for a study of this breadth (Figure S5).

Atg11-dependent selective autophagy is active during proteome remodeling.

To assess the relative contributions of bulk and selective autophagy, we next quantified growth rates and protein degradation in atg11Δ cells, as Atg11 is a critical scaffold supporting selective autophagy in yeast^10,11,48. This experiment revealed that: 1) wild-type and atg11Δ cells grew similarly in all media tested (Figure S3); 2) that most protein substrates were degraded similarly in the two strains when cells transitioned from M+N to G−N medium; and 3) that a cluster of proteins required both Atg9 and Atg11 for their degradation in response to this medium change (Figure 1C-E), suggesting that selective autophagy drives their turnover. Indeed, when inspecting canonical bulk autophagy substrates^49,50, such as the cytosolic protein phosphoglycerate kinase (Pgk1), we observed robust degradation in wild-type and atg11Δ cells but not in the atg9Δ strain, consistent with the Atg11-independent nature of bulk autophagy (Figure 1D). In contrast, peroxisomal substrates such as alcohol oxidase and catalase required both Atg9 and Atg11, consistent with their degradation requiring pexophagy, a well-characterized form of selective autophagy^12,28. Notably, some peroxisomal turnover is reported to occur by micropexophagy, wherein the yeast vacuole directly engulfs peroxisomes, without the involvement the double-membraned autophagosome⁵¹. However, as this process is also Atg9-dependent⁵², we were unable to distinguish the relative contributions of micro- and macropexophagy.

Nutrient adaptation promotes distinct compartment-specific degradation patterns.

To further characterize the selectivity of degradation in response to this nutrient shift, substrates were classified based on their cellular compartments using Gene Ontology annotations^53,54 (see Methods). We then compared strain-specific degradation profiles of proteins exclusively localized to the nucleus, cytoplasm, endoplasmic reticulum, Golgi apparatus, mitochondria, peroxisome, and plasma membrane. This analysis revealed that, with the exception of peroxisomes and mitochondria, most proteins in the other cellular compartments we analyzed were robustly degraded in an Atg9-dependent, Atg11-independent fashion, suggesting that bulk autophagy plays a significant role in the degradation of diversely localized proteins (Figure 2A). For example, proteins from the Golgi apparatus and those from the ER, which together form a central hub for the secretory pathway, were largely degraded in an Atg9-dependent, but Atg11-independent fashion, with similar kinetics observed for proteins throughout the secretory pathway (Figure S6A). Likewise, cytosolic ribosomal proteins were largely degraded in an Atg9-dependent, Atg11-independent fashion, consistent with bulk autophagic turnover (Figure S6B). Notably, however, slow degradation of large ribosomal subunit proteins was observed in both atg9Δ and pep4Δprb1Δ cells, indicating contributions of non-vacuolar proteolysis to the turnover of the large subunits specifically. This observation stands in opposition with prior reports of a strict autophagic requirement for degradation of large ribosomal subunits in yeast⁵⁵. Interestingly, we also observed robust Atg9-dependent, Atg11-independent degradation of proteasomal subunits (Figure S7).

Figure 2. Selective autophagy drives degradation of some, but not all, organelle-localized proteins.

(A) Proteins were grouped based on annotated subcellular localization (see Methods), and the fit kinetic model for each protein was used to calculate the fraction of initial material degraded in 30 hours in wild-type (WT), atg9Δ, and atg11Δ cells. This value is compared between genetic backgrounds as a density hexplot with marginal distributions shown on top and right axes. Exemplar protein or protein complexes from each compartment are noted in dashed boxes. (B) Clustered heatmap following Figure 1 depicting annotated mitochondrial (purple) and peroxisomal (orange) proteins subject to selective autophagic degradation. A cluster of mitochondrial proteins degraded more slowly than typical mitochondrial proteins are outlined. (C) Degradation profiles and fit (eq. 1, see Methods) and fit parameters ($k_{deg}∕{\mathrm{A}}_{\text{deg}}$) for representative proteins localized to the peroxisome (top) or mitochondria (bottom), with sub-organellular localization in brackets (OMM: outer mitochondrial membrane; IMM: inner mitochondrial membrane). The Kinetic traces are fit following eq. 1 (see Methods). Uniprot protein ID and protein common names are listed. The dashed box highlights a subset of slowly degraded IMM proteins.

Degradation in these compartments contrasted with that in the peroxisomes and mitochondria, including mitochondrial ribosomes (Figure S6C), where Atg11-dependent selective autophagy was prevalent (Figure 2B). Notably, such peroxisomal degradation was rapid relative to that of mitochondria, both when monitoring peroxisomal and mitochondrial substrates by nPL-qMS (Figure 2B-C) or through immunoblotting (Figure S8A, see Methods). Careful inspection of degradation patterns by nPL-qMS and immunoblotting amongst peroxin proteins, which are essential for peroxisome biogenesis, revealed a diverse pattern of genetic dependencies (Figure S8B-D, Table S4). Specifically, proteins strictly localized to the peroxisomal membrane, such as Pex2, Pex3, Pex11, and Pex22, were degraded in a selective autophagy-dependent manner, likely captured through pexophagy. In contrast, many cytosolic proteins implicated in peroxisome biogenesis, including Pex5, which was previously suggested to be a selective autophagy substrate⁴⁹, exhibited genetic dependencies consistent with non-autophagy, non-vacuolar degradation (i.e. Atg9-, Atg11-, and Prb1/Pep4-independent). Finally, other peroxins exhibited Atg9- and Pep4/Prb1-independent degradation, consistent with non-autophagic pathways supporting their turnover.

Additionally, within the mitochondria, nPL-qMS revealed differential degradation rates for various protein complexes, with many members of the F₀/F₁ ATP synthase, cytochrome C oxidase, MICOS, NADH dehydrogenase, and pyruvate dehydrogenase complexes degraded more slowly than other mitochondrial proteins (Figure 2C, Figure S9A-B, Table S5). We also observed Atg9- and Prb1/Pep4-independent degradation of mitochondrial respiratory chain proteins, including Cox4, Cox13, Cox5A, and Coq5 (Figure S9C), which are likely targets of AAA proteases present in the mitochondria^56,57.

Bulk and selective autophagy cooperate to degrade and reprogram entire metabolic pathways.

In K. phaffii, proteins required for methanol utilization are distributed in the cytosol and peroxisome^21,58-60, and we speculated that this difference in cellular localization might necessitate different pathways to clear those substrates. Indeed, whereas steady-state measurements showed that both methanol-enriched cytoplasmic and peroxisomal proteins are expressed at similar levels (Figure S1C), our nPL-qMS revealed they are degraded through different pathways. For instance, peroxisomally localized enzymes were rapidly removed through selective autophagy, whereas those in the cytoplasm were slowly removed through bulk autophagy (Figure S10, Table S4). We also found the glycolytic enzyme triose-phosphate isomerase (Tpi1), often thought to be a cytosolic protein, was subject to selective autophagy. However, the carbon-source dependent localization of Tpi1 in K. phaffi has been reported, with the enzyme localized to the cytosol in glucose-containing medium, but in the peroxisome in medium containing methanol⁶¹. This raises the possibility that Tpi1 harbors a cryptic peroxisome targeting sequence in K. phaffii, as has been reported in other fungal species⁶². Together, these findings suggest a potential regulatory mechanism in which Tpi1 degradation is linked to its metabolic context. Interestingly, no methanol assimilation-related proteins were degraded in atg9Δ cells, highlighting the primacy of autophagy relative to proteasomal degradation in this context.

Basal autophagy is active under steady-state growth conditions.

The extensive proteome remodeling observed in response to changing nutrients prompted us to investigate whether autophagy also actively degraded proteins in the absence of a nutrient perturbation. To assess this, cells were grown under steady-state conditions in M+N medium, and a nPL-qMS assay was performed (see Methods). Under these steady-state conditions, we observed minimal protein degradation, with the vast majority of detected proteins exhibiting proteolytic stability up to 50 hours (Figure 3A, Figure S11A). These stable proteins included markers of bulk autophagy (e.g. phosphoglycerate kinase Pgk1), consistent with a lack of bulk autophagy in this growth medium. Nevertheless, peroxisomal turnover was detected as evidenced by the selective autophagic degradation of peroxiredoxin (Pmp20), catalase (Cta1), and the peroxin Pex11 (Figure 3A), whereas methanol-enriched cytosolic proteins were stable (Figure S11B). Similarly, mitochondrial-resident proteins were largely stable in these steady-state conditions (Figure S11C).

Figure 3. Peroxisomes are degraded in steady-state growth conditions.

(A) Degradation across the proteome compared between cells grown at steady-state in nitrogen-rich methanol media (¹²M+¹⁴N → ¹²M+^14/15N) to those transitioning to glucose media without nitrogen (¹²M+¹⁴N → ¹³G−N). In each condition, the fit kinetic model for each protein was used to calculate the fraction of initial material degraded in 30 hours in wild-type, atg9Δ, and atg11Δ cells. This value is compared between different media conditions post-pulse as a density hexplot with marginal distributions shown on top and right axes. Degradation profiles measured in steady-state growth in methanol are shown for exemplar substrates exhibiting noted genetic dependencies for degradation. Kinetic traces are fit as: abundance$(t)={\mathrm{A}}_{\text{deg}}\ast \text{exp}(\text{-}{k_{deg}}^{\ast }t)+\mathrm{B}$ (eq. 1), with fit parameters $k_{deg}∕{\mathrm{A}}_{\text{deg}}$ noted. Uniprot protein ID and protein common names are listed. (B) The fit rate (left) and amplitude (right) degradation parameters for 15 peroxisome-localized proteins compared between cells transitioning from methanol media to glucose media without nitrogen (¹²M+¹⁴N → ¹³G−N) and those grown in steady state conditions in methanol media (¹²M+¹⁴N → 12_M+14/15_N).

Pexophagy occurred at a slower rate and to a lesser extent in these steady-state growth conditions compared to that observed under nutrient-shift induced conditions (Figure 3B). Surprisingly, under these steady-state conditions, fit amplitudes were ~0.5 for these peroxisomal proteins (i.e. ~50% of these peroxisomal proteins would never be degraded, even under steady-state conditions), suggesting an uncharacterized mechanism by which cells protect subsets of peroxisomal proteins from the basal levels of selective autophagy we observed. Notably, many Atg9-independent substrates (e.g. YscB, Figures 3A) displayed similar degradation kinetics during steady-state growth on methanol and following the M+N to G−N nutrient-shift, indicating that autophagic degradation was more nutrient-shift dependent than non-autophagic degradation (Figure S11).

Pexophagy resistance depends on maternal retention of peroxisomes.

To account for our observation that roughly half of the peroxisomal proteome resisted degradation during steady-state growth on methanol medium, we considered two models: 1) cells uniformly induced pexophagy, but only a subset of peroxisomal proteins (or peroxisomes) within each cell was susceptible to such degradation, or 2) a subset of cells, for example mother or daughter cells, induced pexophagy and degraded their peroxisomes completely, while others did not (Figure 4A). To distinguish between these models, we developed and deployed a microscopy-based assay using a photoconvertible fluorophore genetically fused to the peroxisomal matrix protein catalase 1 (Cta1) (see Methods). Upon photoconversion, which was >90% efficient, we could differentiate pre-existing peroxisomes, which were red, from newly synthesized ones, which were green (Figure 4B, S12). We then tracked the fate of these distinct populations ~8.5 hours after photoconversion when degradation had plateaued as measured by our nPL-qMS.

Figure 4: Maternal retention of peroxisomes is required for partial protection from pexophagy in steady state growth conditions.

(A) Schematic of photoconversion-based approach to track peroxisomal protein degradation. The photoconvertible mEos2 protein was fused to catalase (Cta1) to fluorescently label the peroxisomal matrix. A brief UV pulse (10 min) irreversibly converts green mEos2-Cta1 to red, thereby marking the pre-existing pool of peroxisomal protein. Following photoconversion, cells continue growing in M+N medium. Four models were proposed to explain the observed ~50% retention of pre-existing peroxisomal proteins: 1a) half of the peroxisomal matrix proteins are indefinitely protected in all peroxisomes, with the unprotected proteins replaced by those newly synthesized; 1b) some peroxisomes are entirely protected, while others are fully degraded and replaced; 2a) daughter cells inherit stable, protected peroxisomes, while mother cells degrade and replace theirs; and 2b) mother cells retain the set of degradation-resistant peroxisomes, while daughter cells receive newly synthesized ones or synthesize peroxisomes themselves. (B) Fluorescence microscopy of wild-type cells expressing mEos2-Cta1 either before (top) or after 10 min irradiation using UV light (bottom). (C) Per-cell quantitation of remnant photoconverted mEos2-Cta1 after 8.5 hours of growth on M+N media. For each strain, protection was quantified per cell as the ratio of red fluorescence to total fluorescence (red / [red + green + 1]), normalized first to the median ratio immediately after photoconversion, and then further normalized to the median value of this ratio in the atg30Δ control cells. Bars depict mean value; error bars denote S.E.M. Statistical significance (two-tailed Student’s t-test) indicated. (D) Fluorescence microscopy images of wild-type, atg30Δ and ipn1Δ cells expressing photoconverted mEos2-Cta1 after 8.5 hours of growth on M+N medium. (E) Magnification of panels from D, as described above, with subset of budding cells outlined in pink (left). As depicted in cartoon (right), note that in wild-type and atg30Δ cells, daughter cells predominantly contain green-labeled (newly synthesized) peroxisomes, while mother cells retain both green and dual-labeled (red + green/yellow) peroxisomes. Note that in wild-type and atg30Δ cells, strictly green-labeled peroxisomes were observed in daughter cells, whereas mother cells contained both green and dually green- and red-labeled (i.e. yellow) peroxisomes. Low-intensity, diffuse fluorescence was observed in the vacuoles of mother cells in wild-type but not in atg30Δ cells. In inp1Δ cells, strictly green-labeled peroxisomes were observed in daughter cells, and no red-labeled peroxisomes were observed in mother or daughter cells, however diffuse green fluorescence was observed in mother cell peroxisomes and vacuoles. A scale bar (5μm) for each set of panels is plotted on the merged image.

In this assay, we observed that the vast majority of wild-type cells exhibited partial Cta1 degradation as evidenced by reduced red florescence (Figure 4C), and we did not find any evidence to support the existence of a substantial population of cells that were degradation-resistant. Additionally, we observed that such degradation was concomitant with the synthesis of new (green) Cta1 that, in some cells, co-localized with old, photoconverted Cta1, indicating that newly synthesized Cta1 could be directed to a pre-existing peroxisome that was partially or fully resistant to degradation. By quantifying these images across thousands of cells (see Methods), we found that in aggregate approximately 40% of the red fluorescence signal remained in wild-type cells relative to that in atg30Δ cells, consistent with our proteomics results, and indicative of pexophagy driving degradation of a significant fraction of this material (Figure 4D).

Interestingly, when inspecting budding cells, we observed that mother cells preferentially retained the photoconverted (red) peroxisomes, with daughter cells either strictly receiving newly synthesized (green) Cta1 from the mother cell or making this protein themselves. We assessed the dependence of the inheritance pattern in cells lacking Inp1, which tethers peroxisomes to the plasma membrane and is required for mother cells to retain a subset of peroxisomes during budding⁶³. In inp1Δ cells, we observed complete degradation of the photoconverted Cta1 (Figure 4C-E), indicating that maternal retention of peroxisomes was required for the observed degradation resistance. A similar Inp1-dependence on peroxisomal protein protection was observed through our nPL-qMS assay (Figure S13).

Nitrogen starvation is the primary driver for the observed proteome remodeling.

Next, we assessed the isolated contributions of changing the carbon source and of removing nitrogen to the proteome remodeling we observed as cells transitioned from M+N to G−N medium. We first measured protein turnover as cells transitioned from M+N to G+N medium (see Methods), finding that much of the proteome was stable, with the overall pattern of degradation, as assessed by principal component analysis, being most similar to cells grown in steady state M+N conditions. The minimal degradation observed was similar to that seen in atg9Δ and pep4Δprb1Δ cells transitioning between M+N to G−N medium (Figure 5A). Moreover, like in steady-state growth conditions, selective turnover of the peroxisomes was observed, and their rate and extent of degradation was significantly reduced relative to that seen as cells adapted to G−N medium (Figure 5B).

Figure 5. Nitrogen starvation is sufficient to promote proteome-wide turnover.

(A) Principal component analysis (PCA) of the fit degradation rate and amplitude parameters for each protein measured in wild-type cells undergoing the media transitions: M+N to M+N (circles), M+N to G+N (squares), or M+N to G−N (triangles), with each mark noting an independent biological replicate (left). PCA as described above, but now used to globally compare degradation in wild-type, atg9Δ, atg11Δ, and pep4Δprb1Δ cells transitioning from growth on M+N to G−N media (right). (B) Schematic of media transitions (top) and fit degradation rate and amplitude parameters for peroxisomal proteins, compared across the different media transition conditions (bottom), with dots denoting individual peroxisomal proteins and boxes marking average values. (C) Proteome-wide degradation compared between cells transitioning from ¹²M+¹⁴N to ¹³G−N media and cells transitioning from ¹²M+¹⁴N to ¹³M−N media. In each condition, the fit kinetic model for each protein was used to calculate the fraction of initial material degraded in 30 hours in wild-type (top), atg9Δ (middle), and atg11Δ (bottom) cells. This value is compared between different media conditions post-pulse as a density hexplot with marginal distributions shown on top and right axes. (D) Different degradation profiles exhibiting noted genetic dependencies for degradation as cells transitioned from ¹²M+¹⁴N to ¹³M−N media. Kinetic traces are fit as: abundance$(t)=\mathrm{A}\ast \text{exp}(\text{-}{k_{deg}}^{\ast }t)+\mathrm{B}$ (eq. 1, see Methods), with fit parameters $k_{deg}$ and ${\mathrm{A}}_{\text{deg}}$ noted above each plot. Uniprot protein ID and protein common names are listed.

To directly test whether the nitrogen starvation signal was the primary driver of our previously observed massive proteome remodeling, we assessed turnover as cells transitioned from M+N to M−N medium. As we previously observed when simultaneously limiting nitrogen and changing the carbon source, large-scale proteome remodeling was induced under nitrogen starvation stress alone, with bulk autophagy contributing substantially to this degradation (Figure 5C). As in the M+N to G−N transition, we could find examples of stable proteins, those degraded via bulk autophagy, selective autophagy, or independent of autophagy (Figure 5D). Selective autophagic turnover of peroxisomes and mitochondria was again observed, with peroxisomes being degraded more rapidly than mitochondria (Figure S14A). Additionally, degradation of specific proteins involved in peroxisome biogenesis, methanol utilization, and the secretory pathway exhibited diverse genetic dependencies (Figure S14B-C).

Rates of mitophagy and pexophagy are independently governed.

The disparate degradation rates we observed suggested that the mitochondrial and peroxisomal degradation pathways could be competing for a single rate-limiting resource – Atg11-associated autophagy initiation complexes, for example – or, instead, could be independently regulated (Figure 6A). To explore this idea, we asked whether inhibiting degradation of either mitochondria or peroxisomes could stimulate the degradation of the other, as would be predicted by the limiting resource model. In K. phaffii, the receptor protein Atg30 marks peroxisomes for autophagic degradation⁶⁴, whereas receptor protein Atg32 targets mitochondria to autophagosomes^13,14. Performing our nPL-qMS assay in atg30Δ cells resulted in a near-complete loss of pexophagy, with minimal impacts on bulk autophagy or mitophagy. Similarly, loss of Atg32 impaired mitophagy but did not affect the turnover of peroxisomes (Figure 6B-D), indicating that a shared resource is not limiting degradative flux of either pathway. Instead, these findings suggested that the pexophagy and mitophagy pathways are independently regulated, potentially at the level of activation of the selective autophagy receptors themselves^65-67.

Figure 6. Selective autophagic pathways are independent regulated.

(A) Two hypothetical models for how peroxisomes and mitochondrial are degraded through autophagy. In the shared limiting resource model (left), the receptors Atg30 and Atg32 compete for the same limiting autophagic machinery or resource such that the presence of one pathway is expected to slow the other. This model predicts that inhibition of either pathway should increase the flux through the unperturbed pathway. In the independent regulatory model (right) the pathways are not in competition and are not predicted to impact one another. (B) Clustered heatmap plotting the relative abundance of peroxisomal (orange) or mitochondrial (purple) proteins over time in wild-type, atg9Δ, atg11Δ, atg30Δ, and atg32Δ cells transitioning from M+N to M−N media. Columns represent individual timepoints, which are plotted above. (C-D) Degradation profiles and fit kinetic models (eq. 1, see Methods) of representative peroxisomal (C) or mitochondrial (D) proteins, with Atp synthase related proteins, which are degraded relatively slowly, outlined. Uniprot protein ID and protein common names are listed, with fit parameters $k_{deg}$ and ${\mathrm{A}}_{\text{deg}}$ noted above each plot.

DISCUSSION

It has long been appreciated that nutrient limitations profoundly influence protein degradation pathways⁴, including the ubiquitin-proteosome system and bulk and selective autophagy. However, in most instances, our understanding of the relative contributions of these different pathways to overall proteolytic flux has been limited, in part due to limitations of the techniques employed. For example, western blotting^50,55,68 and fluorescence assays^69-72 only provide targeted measurements of a handful of substrates. They additionally rely on the specificity of affinity reagents or the ability to genetically integrate tags without impacting target function to accurately quantify physiological protein levels. Mass spectrometry enables near proteome-wide tag-free measurements⁷³ but, as typically implemented, can only measure steady-state protein levels, which obscures proteins whose synthesis is balanced with degradation. For example, we observed significant peroxisomal degradation (and balanced synthesis) under steady state (M+N) growth conditions, which would have been overlooked with traditional western, fluorescence, or mass spectrometry-based approaches.

Alternative approaches to identify proteins targeted by different proteolytic pathways include proximity assays such as IP-MS⁷⁴ or enzymatic proximity labeling⁷⁵, with each producing a catalog of potential substrates. These proximity-based methods in isolation cannot, however, distinguish between co-localization that leads to degradation and that which does not, and these methods fail to provide turnover rates, degradation kinetics, or the fraction of a given substrate that is degraded under specific conditions, each of which can be critical for assessing proteome remodeling.

Within the autophagy field, overall autophagic activity levels are often assayed using reporters, such as the ratio of lipidated to non-lipidated Atg8-family proteins, proteolytic cleavage of a marker protein fused to an autophagic substrate, or quantitation of autophagosomes via transmission electron microscopy or fluorescent puncta associated with autophagosome formation⁷⁶. Notably, these assays cannot assess the specific substrates targeted by autophagy and, in the absence of additional modulations aimed at revealing “autophagic flux”, it can be challenging to distinguish an increase in formation of autophagosomes from a failure of autophagosome/lysosome fusion^77,78. Moreover, none of these assays provide a kinetic rate for autophagy generally, let alone on a per-substrate basis across the proteome.

Our nutrient-pulse-labeling/quantitative mass spectrometry (nPL-qMS) approach provides a global view of protein degradation, allowing one to measure both degradation rates and the total fraction of proteins degraded on a per-substrate basis. As such, it can be used as a quantitative, functional readout of which proteins are degraded in response to an environmental change, but it cannot directly ascribe the proteolytic pathway utilized. To solve this problem, we have coupled our approach to genetic ablation of key genes required for various forms of autophagy. In K. phaffii, this method was effective, as the degradation of many proteins exhibited switch-like responses to these manipulations, with protein degradation completely ablated in the cognate knockout strain. This switch-like behavior contrasted with that observed in cultured human cells where we have observed many protein substrates that could flexibly access either lysosomal or proteasomal degradation pathways⁷⁹.

In designing our method, we were concerned that our isotope labeling scheme, which requires three mass-distinguishable labeling patterns to independently monitor protein synthesis and degradation would limit protein coverage. However, when coupled with advances in instrumentation⁸⁰ and data analysis⁸¹, we successfully detected and quantified ~3,200 proteins of the K. phaffii proteome (Figure 1, S1), which consists of ~5,000 encoded proteins⁸², only a subset of which are expected to be expressed in any particular condition. This extensive coverage facilitated detailed tracking of subcellular compartments and their dynamics, allowing for deep comparisons between the various nutrient adaptation and genetic backgrounds we compared. Moreover, the resulting measurements constitute a rich resource to interrogate the regulation and dynamics of protein turnover in different conditions. These data are available through an interactive web application (https://sites.mit.edu/jhdavislab/datasets/) where users can inspect the degradation profiles of these proteins and assess how these profiles change in atg9Δ or atg11Δ cells.

To interrogate the pathways supporting adaptive response to nutrient stress, we subjected K. phaffii to a stringent stress condition by changing the carbon source in the growth medium and simultaneously depriving cells of nitrogen. Our aim was to induce stress responses, which we expected would lead to dynamic alterations in protein degradation and gene expression, as well as nutrient scavenging and cellular metabolism. Our findings emphasize that when K. phaffii adapts to nitrogen deprivation, either with or without a concomitant change in the carbon source, it initiates massive remodeling of the proteome, with a large but incomplete fraction of the proteome targeted for degradation. The remodeling is predominately driven by autophagy, as evidenced by the substantial stabilization of the proteome in atg9Δ cells and the observed Atg11-dependence on some autophagy substrates (Figure 1). Notably, ablation of bulk autophagy resulted in a severe growth defect (Figure S3), likely a result of the cell’s inability to recycle nutrients in the absence of Atg9-dependent bulk autophagy, which we observed when carefully inspecting the isotope envelopes of newly synthesized proteins in either wild-type or atg9Δ cells (Figure S4).

Bulk autophagic degradation targeted both cytosolic substrates and those associated with the secretory pathway, as evidenced by the similar degradative rates and genetic dependencies of proteins in the lumen or membrane of the ER, Golgi, and vesicles transiting between these compartments (Figure 2). Nuclear-localized proteins exhibited a more complex degradation dependence, with different proteins displaying genetic dependencies consistent with bulk autophagy, selective autophagy, or non-autophagic degradation. Further highlighting the complex interplay between the UPS system and autophagy in this compartment^83,84, we observed autophagy-dependent degradation of proteasomal subunits upon nitrogen starvation (Figure S7).

In conditions of acute nutrient stress, K. phaffii employs both bulk and selective autophagy to restructure its proteome, with these pathways cooperating to target distinct subsets of the proteome. Our study specifically highlights the simultaneous activation of two selective autophagy pathways, mitophagy and pexophagy, with pexophagy degrading substrates rapidly relative to mitophagy (Figure 2). Additionally, we noted variations in the degradation kinetics of mitochondrial proteins with a key subset of proteins, including F₀/F₁ ATP synthase subunits, being degraded more slowly than other mitochondrially localized proteins degraded via mitophagy (Figure S9). The simple topological organization of the mitochondria did not explain the observed degradation patterns, as we observed similar rates of degradation for proteins localized to the outer membrane or to the matrix. This lack of a topological dependence suggested that some protein complexes in the mitochondria may be more stable to vacuolar proteolysis, or otherwise less susceptible to autophagic degradation. In contrast to that in the mitochondria, selective autophagic degradation of peroxisomal proteins was more uniform, with both membrane-associated and matrix-localized proteins undergoing relatively rapid selective autophagic degradation.

A recent study in cultured human cells proposed that in some conditions, selective autophagy pathways are in competition and, specifically, that upregulated pexophagy limits aggrephagy and parkin-dependent mitophagy⁸⁵. The authors further posited that these pathways competed for limiting quantities of active ULK1, as pharmacological stimulation of ULK1 rescued the pexophagy-induced limitation in mitophagy. Interestingly, in our assays, selective autophagy pathways in yeast undergoing metabolic adaptation do not appear to be in competition for a limiting pool of initiation factors, as ablation of either pexophagy or mitophagy had no impact on the degradative capacity of the remaining pathway. These data are instead consistent with a model in which the cognate selective autophagy receptors can be independently regulated – for example through phosphorylation. Furthermore, our data highlights strict receptor specificity for organelle degradation, unlike in human cells, where multiple autophagy receptors (e.g., p62, NBR1, OPTN, and TAX1BP1) can target overlapping cargo types⁸⁶. This apparent divergence suggests that competition among selective autophagy pathways may be more characteristic of higher eukaryotes, reflecting increased regulatory complexity or shared component multifunctionality. Comparative studies are needed to determine whether this is a conserved or context-dependent feature.

Whereas the bulk autophagic degradation we observed largely depended on nitrogen limitation, pexophagy was observed even under steady-state growth in methanol-containing medium, where peroxisomes are continuously synthesized (Figure 3). Why might the cell simultaneously synthesize and degrade these proteins? We hypothesize that the observed turnover acts to maintain a healthy pool of peroxisomes, as the peroxisome-resident proteins are known to undergo oxidative damage as a byproduct of methanol metabolism, which produces significant levels of hydrogen peroxide^58,87. However, if stochastic metabolism-associated damage were to trigger degradation of these proteins via pexophagy, one would expect simple first order kinetics, with all of the pre-pulse material eventually degraded. Instead, we observed that ~50% of the pre-pulse peroxisome-associated proteins were stable, implying that they are either protected from damage, or otherwise sequestered from the degradative machinery. Using a fluorescence microscopy-based assay that allowed us to distinguish new versus old pools of the peroxisomal protein catalase, we observed maternal retention of a subset of peroxisomes that appeared degradation resistant (Figure 4,S12). Additionally, we found a strong genetic dependence on the maternal retention factor Inp1 in this maternal-retention based protection (Figure S13). This led to a model of peroxisomal degradation resistance wherein the existing population of peroxisomes is largely retained in mother cells, while newly synthesized peroxisomes are preferentially delivered to daughter cells. Interestingly, the retained pre-existing peroxisomes are in-part resistant to degradation, and they remain functional for matrix protein import.

Protein synthesis also plays an important role in remodeling the proteome to meet changing metabolic demands. In scenarios where cells transition from methanol to glucose, we observed that they suppress methanol-specific protein synthesis while activating synthesis of glucose-specific proteins (Figure S2). This shift suggests that cells efficiently reallocate resources to support new metabolic state rather than maintaining obsolete pathways, thereby facilitating a smooth transition between nutrient sources. Under starvation conditions, we observed that such protein synthesis and autophagy were tightly linked as de novo protein synthesis was significantly repressed in atg9Δ cells (Figure S4). Moreover, the observed shift in the mass of newly made proteins in atg9Δ cells compared to wild-type cells implied that cells had recycled amino acids or metabolites thereof when autophagy was active. Note that while we did observe degradation of a subset of large ribosomal subunits in autophagy-deficient cells (Figure S6), we do not believe ribosomal degradation is the primary cause of ablated protein synthesis. Rather, we propose that the inability to recycle amino acids and other macromolecules through autophagy is a more significant factor, particularly under starvation conditions, where cells depend on these recycled resources to sustain protein production and enable metabolic adaptation.

Taken together, this work provides a rich, quantitative resource to better understand how K. phaffii balances synthesis and degradation during steady-state growth, and how the proteome is remodeled in response to nutrient changes, all with single-protein resolution. Moreover, it provides the requisite experimental tools and analytic approaches to quantitatively assess the roles of selective and bulk autophagy in supporting cellular adaptation to changing environments, including exposures to toxins, pathogens, and genotoxic, thermal, or oxidative stress. Given the central role of proteome remodeling in human diseases including as cancer, neurodegeneration, viral infection, we expect that this platform will be readily adapted to investigate how cells alter degradation pathways in response to disease relevant stresses.

LIMITATIONS OF THE STUDY

A key limitation of this study is the inability to monitor low-abundance or difficult to extract proteins, including regulatory factors that may mediate adaptive responses. For example, autophagy-related proteins, including those involved in selective autophagy, are generally expressed at low levels, and often associated with membranes. In general, these proteins could not be reliably quantified in this nPL-qMS assay, limiting our ability to directly monitor their dynamics or degradation. This constraint is especially relevant given our observation that mitochondria and peroxisomes are degraded via distinct Atg11-dependent pathways, likely regulated at the level of receptor phosphorylation⁸⁸, which we were unable to measure. Without direct measurement of the selective autophagy machinery, we cannot fully resolve the molecular basis of this pathway specificity or its regulation under different nutrient conditions. Finally, we have focused on degradation on relatively long timescales (i.e., hours) thus this study does not inform our understanding of rapidly turned-over proteins, including those known to be proteasomal substrates.

STAR METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS.

The GS115 K. phaffii (P. pastoris) strains used are as follows.

Strain name	Description	Genotype	Source
st_JD714	wild-type	GS115 plB3::HIS4	Subramani lab
st_JD715	atg9Δ	GS115 atg9Δ::ZeocinR plB::HIS4	Subramani lab
st_JD716	atg11Δ	GS115 atg11Δ::ZeocinR plB::HIS4	Subramani lab
st_JD719	atg30Δ	GS115 atg30Δ::ZeocinR plB::HIS4	Subramani lab
st_JD720	atg32Δ	GS115 atg32Δ::ZeocinR plB::HIS4	Subramani lab
st_JD717	pep4Δprb1Δ	GS115 pep4-, prb1-, his4- (SMD1163) plB::HIS4	This work
st_JD1038	inp1Δ	GS115 inp1Δ::ZeocinR plB::HIS4	This work

Komagataella phaffii strains used in this study. Columns detail strain name within the Davis lab repository; description as used in this study;S1 genotype; and strain source. All strains are available upon request to Lead Contact.

METHOD DETAILS.

Cell culturing and stable isotope pulse-labeling.

All strains were cultured at 30°C in 125 mL baffled flasks shaken in a rotary water bath (Brunswick G-76) at 225 RPM. Growth media (typically 5-25 mL) were defined as follows:

¹⁴N Glucose Fed Medium (G+¹⁴N):

0.67% [w/v] yeast nitrogen base (YNB) without amino acids or ammonium sulfate (Research Products International Y20040); 2.0% [w/v] glucose; 0.5% [w/v] ammonium sulfate (Fisher Scientific BP212-212).

¹⁵N Glucose Fed Medium (G+¹⁵N):

0.67% YNB; 2.0% glucose; 0.5% ¹⁵N-labeled ammonium sulfate (Cambridge Isotope Laboratories NLM-713-25).

¹³C Glucose Nitrogen Starvation Medium (¹³G−N):

0.67% YNB; 2.0% ¹³C glucose (Cambridge Isotope Laboratories CLM-1396-10).

¹⁴N Methanol Fed Medium (M+¹⁴N):

0.67% YNB; 0.5% [v/v] methanol (Fisher Chemicals A412-4); 0.5% ammonium sulfate.

¹⁵N Methanol Fed Medium (M+¹⁵N):

0.67% YNB; 0.5% methanol; 0.5% ¹⁵N-ammonium sulfate.

¹³C Methanol Nitrogen Starvation Medium (¹³M−N):

0.67% YNB; 0.5%¹³C methanol (Cambridge Isotope Labs CLM-359-5).

¹⁵N Oleate Fed Medium (O+¹⁵N):

0.67% YNB; 0.2% [v/v] oleic acid (Thermo Fisher 031997.22); 0.5% ¹⁵N-ammonium sulfate.

Cell culturing for fluorescence microscopy-based analyses.

To investigate why half of the peroxisomal proteome remained protected during steady-state growth on methanol, we used a photoconversion-based labeling strategy to distinguish pre-existing peroxisomes from newly synthesized ones and to track their fate over time. Specifically, the peroxisomal matrix protein (Cta1) was fused to the photoconvertible fluorescent protein mEos2, which undergoes an irreversible shift from green to red fluorescence upon UV exposure. Cells were grown in YPD medium at 30 °C until in exponential phase, then shifted to methanol medium and grown for 24 hours while maintaining exponential growth. For irradiation, cells equivalent to ~10 OD₆₀₀ units when concentrated to 1 mL were harvested, resuspended in 1 mL of methanol medium, and plated onto an 8.5 cm² tissue culture dish. Photoconversion of mEos2-Cta1, expressed under control of the AOX1 promoter, was induced by irradiation using a UV LED black light (Elworks Black Light 150w led lightweight, 395–400 nm) positioned 1.5 cm above the open dish. Irradiation was performed at 4 °C with shaking at speed 4.75 on a Roto-Mix platform. After irradiation, the cells were transferred to a 250 mL flask and diluted with 50 mL of methanol medium. The culture was then incubated at 30 °C with shaking at 250 RPM for 8.5 hours. To optimize photoconversion conditions, various irradiation times (1, 5, 10, and 20 minutes) were tested. A 10-minute irradiation was selected as it yielded high photoconversion efficiency with minimal phototoxicity, and robust cell growth was observed after treatment. For non-irradiation conditions, cells were resuspended in 50 mL of methanol medium and imaged at timepoints 0 and 8.5 hours. The studies were conducted in wild-type, atg30Δ, and inp1Δ cells.

Steady-state protein level measurements.

For steady-state protein levels comparison between methanol and glucose, cells were separately grown in both ¹⁴N- and ¹⁵N-labeled G+N, M+N, and O+N media, and 5 mL of each culture were harvested at ~0.5 OD₆₀₀ via centrifugation 21,000 x g for 5 minutes. Following removal of supernatant, cells pellets were then washed with 500 μL of buffer A [20 mM Tris HCl (pH 7.5), 100 mM NH₄Cl, 10 mM MgCl₂, 0.5 mM EDTA, 6 mM 2-mercaptoethanol] and pelleted as before prior to storage at −80°C. During preparation of cells pellets, each ¹⁴N-labeled sample was resuspended in 2 mL of buffer A. A reference standard was generated by resuspending pellets bearing ¹⁵N-labeled cells in 2 mL of buffer A and mixing equal volumes each resuspension (G+¹⁵N; M+¹⁵N; O+¹⁵N). 50 μL of this mixture was then added to 50 μL of each of the ¹⁴N-labeled pellets, and these samples were then prepped for mass spectrometry as described below.

Pulse-labeling time courses.

For steady-state pulse-labeling experiments in the presence of nitrogen, cells were initially grown at 30°C in pre-pulse medium to an OD₆₀₀ of ~1.75, at which point they were pulse-labeled by the addition of an equal volume of pulse medium, which had been pre-warmed to 30°C. For starvation studies in the absence of nitrogen, nitrogen-containing pre-pulse grown cells were first pelleted at 3,000 x g for 5 min, washed with pre-warmed YNB buffer, pelleted at 3,000 x g for 5 min, and resuspended in pre-warmed starvation medium. At each timepoint, 1 mL of culture was collected, harvested via centrifugation (21,000 x g, 5 min) at 4°C, washed with cold buffer A, and pelleted via centrifugation as above. Cell pellets were then rapidly frozen at −80°C prior to further analysis. In parallel, cells were grown in a ¹⁵N-labeled variant of the pre-pulse medium, harvested at an OD₆₀₀ of ~1, and prepared and frozen as above.

Note that when monitoring proteome remodeling as cells transitioned from M+N to G+N medium, cells were grown in “light” methanol medium with ¹⁴NH₄ and moved them to “heavy” G+N medium bearing a 1:1 mixture of ¹⁴NH₄:¹⁵NH₄. This isotope pulse metabolically scrambles the nitrogen, resulting in partially labeled peptides that could be readily distinguished from the “light” precursors and from our reference standard grown in 100% ¹⁵NH₄.

Mass spectrometry sample preparation.

After mixing collected samples with their cognate reference sample as described above, the mixture was immediately lysed by addition of trichloroacetic acid (TCA) to a final concentration of 13% (v/v). Samples were quickly frozen using liquid nitrogen and stored at −80°C for at least 1 hour or overnight if processed the following day. Sample preparation largely followed Sun et al. beginning with thawing samples and precipitating proteins by incubation in 13% TCA on ice for at least 2 hours⁸⁹. Samples were then centrifuged for 30 min at 4°C at 13,000 x g, and the supernatant discarded. The TCA precipitates were then washed with cold (4°C) acetone (500 μL) twice and dried at room temperature. Reduction of protein disulfide bonds was performed in 100 mM ammonium bicarbonate with 10 mM DTT at 65°C for 10 min, and protein alkylation was performed with 20 mM iodoacetamide at 30°C for 30 min in the dark. Proteins were then digested with 0.2 ug of trypsin/lysC (Promega) overnight at 37°C. Digested peptides were desalted using Pierce C18 columns or Thermo Fisher SOLA SPE plates following manufacturers’ protocols, dried in a speedvac, and either stored at −80°C or resuspended in 20 μL sample buffer (4% acetonitrile, 0.1 formic acid) prior to analysis by LC-MS/MS.

LC-MS/MS.

Following desalting, tryptic peptides were separated using an Ultimate 3000 UHPLC system (Thermo Scientific) consisting of a trap column (C18, 75 μm ID x 2 cm, particle size 3 μm pore size 100 Å; Thermo Fisher Scientific #164946) and an analytical column (75 μm ID x 50 cm, particle size 2 μm pore size 100 Å; Thermo Fisher Scientific #ES903). Peptides (2 μL, ~1.5 μg) were loaded onto the trap column at a flow of 5 μL/min in a total volume of ~20 uL. Peptides were washed on the trap column in loading buffer and then resolved on the analytical column using a 130 min linear gradient of 4%-30% buffer B (ACN, 0.1% FA) at a flow rate of 300 nL/min. Peptides were ionized by nano spray electrospray ionization and analyzed using an Thermo Scientific Orbitrap Exploris 480 (all M+N to M−N time courses; M+N to G−N pep4/prb1Δ time course) or HF-X (all other time courses) mass spectrometer as described below.

For data-dependent acquisition (DDA) runs, MS1 scans ranged from 350-1400 m/z and were collected at a mass resolution of 120,000 with AGC target and maximum injection time (maxIT) set at 3 x 10⁶ and 50 ms, respectively. The top 20 most abundant precursor ions were selected for subsequent MS2 scans and fragmented using 25% normalized collision energy (NCE). The fragment analysis (MS2) was performed at a resolution of 15,000, with an isolation mass window size of 2.2 m/z, AGC target and maxIT set at 1 x 10⁵ and 100 ms, respectively.

Data independent acquisition (DIA) MS1 scans followed those above as described for the DDA acquisitions. On the HF-X, MS2 scans were acquired using 25 windows spanning a mass range of 400-1250 m/z (Table S7); precursors were fragmented using 28% NCE, and product ions were analyzed at a mass resolution of 30,000, with AGC and maxIT set at 1 x 10⁶ and 70 ms. On the Orbitrap Exploris 480, MS2 scans were collected using 10 m/z wide precursor windows overlapped by 2 m/z overlaps spanning the range 390-1390 m/z, with fragments generated using 28% NCE and analyzed at a resolution of 30,000 over a fixed 200-2,000 m/z range. Normalized AGC and maxIT were set at 2,000% and 70 ms, respectively.

DDA data analysis.

Raw DDA files were initially converted to the mzML format using MsConvert and subsequently searched within the trans-proteomic pipeline⁹⁰ using the Comet algorithm⁹¹ against Komagataella phaffii Uniprot database for taxon ID 644223. The search parameters include 15 ppm precursor mass tolerance, 0.02 Da fragment mass tolerance, oxidized methionine as a variable modification, and carbamidomethyl cysteine as a static modification. The resulting peptide identifications were validated and scored using PeptideProphet and iProphet⁹². The resulting PepXML files were used to create a non-redundant spectral library in Skyline⁹³. Raw DDA MS files were imported and searched against the spectral library for extraction of MS1 chromatographic peak areas. Skyline was used to qualitatively inspect peak shape, peak area, extracted chromatograms were inspected to ensure co-elution of light and heavy peaks.

DIA data analysis.

Raw DIA files were processed using Spectronaut software (version 15) using DirectDIA mode⁹⁴. In brief, a spectral library was initially generated directly from DIA file by Spectronaut Pulsar, which incorporated a DDA library built from our DDA data. The settings for Pulsar and library generation were the same for both DIA and DDA files and were as follows: specific enzyme set to Trypsin/P, LysC, peptide length ranged 7 to 52; max missed cleavages=0; N-terminal M enabled; labeling included ¹⁵N at all nitrogens (2 channels utilized); fixed modification of carbamidomethyl Cys; variable modifications of Met oxidation and N-terminal acetylation; FDRs set to 0.01 and calculated at the PSM, peptide, and protein level; minimum fragment relative intensity 1%; 3-6 fragments kept for each precursor. For DIA analysis, Spectronaut search parameters were set as follows: mutation with NN predicted fragments to generate decoy, machine learning performed per run, precursor PEP cutoff set to 0.2; precursor q-value cutoff of 0.01; single hit definition by stripped sequence. Using the report output, we exported MS1 intensities separately for channel1 (unlabeled) and channel2 (¹⁵N-labeled) for each identified precursor. More specifically, the report file included: PG.Genes, PG.ProteinNames, EG.PrecursorId, PEP.MS1Channel1 and PEP.MS1Channel2.

Fluorescence microscopy imaging and image processing.

Cells were imaged using a Zeiss Plan Apochromat 63x/1.4 Oil DIC objective mounted on an Axioskop 2 mot plus microscope (Zeiss) equipped with Axio Cam HRm camera and HBO 100 mercury lamp. Exposure times were identical for green and red channels across all experiments (100 ms), whereas DIC images were acquired with auto exposure. Under irradiation conditions, cells were immediately imaged after 10 minutes photoactivation treatment (time = 0) and at 8.5 hours post photoactivation treatment. For imaging, cells were pelleted, washed with sterile water, mixed with low melting agarose, and placed on a glass slide with cover slip and imaged. Green fluorescence from mEos2-Cta1 was acquired using a GFP filter, and red fluorescence using a Rhodamine filter. Images were processed on AxioVision software and representative results from experiments conducted in triplicate are shown.

Immunoblot studies.

To evaluate relative rates of mitophagy and pexophagy, we assessed the degradation of Idh1, a mitochondrial matrix protein, and Aox1, a peroxisomal matrix protein, each expressed under the control of the AOX1 promoter, which is known to be fully repressed in the presence of glucose.

Idh1 was tagged with GFP and analyzed by immunoblotting using an anti-GFP antibody (Clontech, JL8, #632381) to detect full-length Idh1-GFP. Pexophagy was evaluated under the same conditions using an anti-Aox1 antibody (generated in-house).

To minimize overexpression due to the strong AOX1 promoter, cells were grown for only 4 hours in M+N before being shifted to G−N. Samples were collected at 0, 3, 6, 12, 24, and 48 hours and TCA-precipitated. Total protein extracts were prepared from each time point, separated by SDS-PAGE, and subjected to immunoblotting.

A similar approach was used to evaluate the degradation of peroxins, however these were driven by their native promoters. Cells were grown in M+N for 16 hours and then shifted to G−N. Samples were collected at 0, 6, and 24 hours and proteins were TCA-precipitated. All antibodies used for this assay were produced in-house.

QUANTIFICATION AND STATISTICAL ANALYSIS.

Quantitation of steady-state protein abundance.

For steady-state protein level comparison of glucose-grown and methanol-grown K. phaffii, six biological replicates were processed as described above. DIA data was acquired and analyzed as previously described in Spectronaut. ¹⁴N (channel1) and ¹⁵N (channel2) MS1 intensities were separately exported for each detected peptide.

Using the exported MS¹ intensities, ¹⁴N/¹⁵N ratios were generated for each identified precursor and the relative protein abundance was reported as the median relative peptide abundance across all peptides assigned to each protein. Additionally, for each growth condition, quantifiable proteins were normalized to their respective actin levels. Normalized log₂ protein abundances were compared using a Student’s t-test, and the resulting p-values were corrected using the Benjamini-Hochberg multiple hypothesis correction procedure. A volcano plot was generated in Python, incorporating the adjusted p-values and log₂-fold changes, with proteins exhibiting more than 2-fold changes with corrected p-values < 0.05 considered significant. Proteins involved in glucose metabolism and methanol metabolism were extracted from the statistically significant group and their relative abundances were compared.

nPL-qMS data quantification.

For each peptide, its relative abundance at a timepoint was determined by normalizing the MS¹ pre-pulse peptide (unlabeled) peak to the MS¹ spike (labeled) peak. Relative protein abundances were defined as the median abundance across all peptides assigned to each protein. To quantify degradation on a protein-by-protein basis, individual protein abundance at different timepoints were fit to a single exponential function:

$$abundance(t)=A_{deg}{e}^{-k_{deg}\ast t}+B$$ eq. 1

where, ${\mathrm{A}}_{\text{deg}}$ and $k_{deg}$, and $\mathrm{B}$ represent the amplitude, rate constant for degradation, and an offset, respectively. Note that in nearly all fits, ${\mathrm{A}}_{\text{deg}}+\mathrm{B}=1$, and with the exception of the most stable proteins, the data is well fit when B is fixed at $1\text{-}{\mathrm{A}}_{\text{deg}}$. The fit quality was assessed using the sum of squared residuals, with poorly fit proteins manually inspected and discarded if the data was of low quality. We next calculated the fraction of initial material degraded at t=30hrs (θ₃₀) and defined proteins undergoing degradation as those for which at least 20% of their initial material was degraded (θ₃₀>0.2) in 30 hours. Any protein with less than 20% fraction degraded (θ₃₀<0.2) was defined as “stable”. This analysis was performed for the following backgrounds: wild-type, atg9Δ, atg11Δ, atg30Δ, atg32Δ, pep4Δprb1Δ.

We then compared the extent of degradation across wild-type, atg9Δ, atg11Δ, pep4Δprb1Δ strains to define different modes of degradation as follows:

Atg9-Pep4/Prb1-dependent: wild-type & atg11Δ θ₃₀>0.2; atg9Δ & pep4Δprb1Δ θ₃₀<0.2.

Atg9-Atg1-Pep4/Prb1-dependent: wild-type θ₃₀>0.2; atg9Δ & atg11Δ & pep4Δprb1Δ θ₃₀<0.2.

Atg9-Atg11-Pep4/Prb1-independent: wild-type & atg9Δ θ₃₀>0.2; atg11Δ & pep4/prb1Δ θ₃₀>0.1. Atg9-Atg11-independent,Pep4/Prb1-dependent: wild-type & atg9Δ & atg11Δ θ₃₀>0.2;

pep4Δprb1Δ θ₃₀<0.2.

Gene ontology analysis of quantified proteins.

We used gene ontology (GO) annotations at the cellular localization category to assign subcellular localization data to proteins we analyzed, focusing on the terms: nucleus, cytoplasm, endoplasmic reticulum, mitochondria, Golgi apparatus, peroxisome, vacuole, plasma membrane and cell wall. To characterize the mode of degradation of a specific compartment, we restricted our analysis to the subset of proteins that were strictly assigned to a single GO compartment. For instance, most quantified Golgi and the ER proteins were annotated to both compartments, but only those bearing a single annotation were analyzed when computing the organelle-scale mode of degradation. We categorized each compartment’s mode of degradation by inspecting the annotated proteins and assigned the mode of degradation as: 1) “non-degraded” if fraction remaining at T= 30 is less than 0.2 across WT, atg9Δ, atg11Δ, pep4Δprb1Δ for the majority of the annotated proteins; 2) “bulk autophagy” if the majority of the annotated proteins were Atg9-Atg11-Pep4/Prb1-dependent in their degradation; 3) “selective autophagy” if the majority of the annotated proteins were Atg9-Atg11-Pep4/Prb1-dependent in their degradation. Note that even in compartments designated as autophagy targets, we observed non-autophagic degradation of a subset of proteins within each compartment.

Quantification of fluorescence microscopy-based data.

Peroxisome fluorescence intensity was quantified using the perox-per-cell software (version 0.0.6_win_rc1) as described⁹⁵, using default settings. Prior to analysis with the software, single images containing DIC, red, and green channels were split into two images: DIC/red and DIC/green.

After generating the excel files with perox-per-cell, peroxisome intensity values were obtained by subtracting the "cytosolic peroxisome signal" from the "total peroxisome channel signal." Quantification of red and green fluorescence signals was performed at the single-cell level, treating each cell as an independent biological replicate. The photoconversion efficiency was calculated for each cell as the ratio of the red fluorescence signal divided by the sum of the red and green fluorescence signals, according to the formula:

$$photoconversion\_efficiency=\begin{array}{c}\hfill red\hfill \\ \hfill (red+green)\hfill \end{array}$$

The red signal remaining at 8.5 hours was determined by calculating the ratio of red fluorescence to total fluorescence including a small stabilizing factor (+1) to avoid division by 0 errors, using the formula:

$$red\_signal\_remaining=\begin{array}{c}\hfill red\hfill \\ \hfill (red+green+1)\hfill \end{array}$$

Each strain’s red fraction at 8.5 hours was then normalized to its corresponding mean red fraction observed at time 0. Finally, this value was normalized to that of the Δatg30 strain to enable a comparison across strains. Dilution due to growth and mEos2 signal decay from oxidation within peroxisomes were not specifically accounted for, as normalizing to the Δatg30 strain inherently includes these variables.

Statistical comparisons of between strains were performed using two-tailed unpaired t-tests, with Welch’s correction to account for unequal variances. Data are presented as mean ± standard error of the mean (SEM) unless otherwise stated, with statistical significance noted in figure.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-GFP	Clontech	#632381
anti-Aox1	This paper	N/A
Chemicals, peptides, and recombinant proteins
Yeast nitrogen base without amino acids or ammonium sulfate	Research Products International	Y20040
Ammonium sulfate	Fisher Scientific	BP212-212
15N-labeled ammonium sulfate	Cambridge Isotope Laboratories	NLM-713-25
13C-labeled glucose	Cambridge Isotope Laboratories	CLM-1396-10
13C-labeled methanol	Cambridge Isotope Laboratories	CLM-359-5
Oleic acid	Thermo Fisher	031997.22
Methanol	Fisher Chemicals	A412-4
Trypsin/LysC sequencing grade protease	Promega
Deposited data
Kinetic fits	This paper	https://sites.mit.edu/jhdavislab/datasets/
Raw mass spectrometry data	MassiVE	ftp://massive-ftp.ucsd.edu/v07/MSV000096185/
Experimental models: Organisms/strains
Komagataella phaffii: Strain: wild-type GS115 plB3::HIS4	Subramani lab	st_JD714
Komagataella phaffii: Strain: atg9Δ GS115 atg9Δ::ZeocinRplB::HIS4	Subramani lab	st_JD715
Komagataella phaffii: Strain: atg11Δ GS115 atg9Δ::ZeocinRplB::HIS4	Subramani lab	st_JD716
Komagataella phaffii: Strain: atg30Δ GS115 atg9Δ::ZeocinRplB::HIS4	Subramani lab	st_JD719
Komagataella phaffii: Strain: atg32Δ GS115 atg9Δ::ZeocinRplB::HIS4	Subramani lab	st_JD712
Komagataella phaffii: Strain: GS115 pep4-, prb1-, his4-(SMD1163) plB::HIS4	Subramani lab	st_JD717
Komagataella phaffii: Strain: inp1Δ GS115 atg9Δ::ZeocinRplB::HIS4	Subramani lab	st_JD1038
Software and algorithms
Skyline 22.2.2.501	University of Washington; MacCoss lab	https://skyline.ms/project/home/software/Skyline/begin.view
Spectronaut version 15	Biognosys	https://biognosys.com/software/spectronaut/
Python version 3	Python Software Foundation	https://www.python.org/
Axios Vision 4.9.1.0	Carl Zeiss Microscopy	Axiovision-le_se64_491_SP2.exe
Image Studio 6.0	Li Cor Bio	https://www.licorbio.com/image-studio
Perox-per-cell v0.0.6_win_rc1	UCSD	https://github.com/AitchisonLab/perox-per-cell
Other
SOLA SPE plates	Thermo Fisher	60309-001
C18 columns	Pierce	PI89873
uHPLC trap column	Thermo Fisher Scientific	164946
uHPLC analytical column	Thermo Fisher Scientific	ES903
Ultimate 3000 UHPLC	Thermo Scientific
Orbitrap Exploris 480	Thermo Scientific
Orbitrap QE HFX	Thermo Scientific
UV LED black light 150w	Elworks Black Light
Axioskop 2 mot plus	Zeiss

Highlights.

• Our nPL-qMS method enables global tracking of protein turnover during cellular adaptation.

• Bulk autophagy drives massive proteome remodeling and supports de novo protein synthesis.

• Selective autophagy targets mitochondria and peroxisomes via independent pathways.

• In methanol media, mother cells use Inp1 to shield peroxisomes from pexophagy.