Title: A New Quantitative Cell Microscopy Approach Reveals Mechanistic Insights into Clearance of Membrane Substrates by the AAA Protein Msp1

Msp1 is a conserved AAA ATPase in budding yeast localized to the surface of mitochondria where it prevents accumulation of mistargeted tail-anchored (TA) proteins, including the peroxisomal TA protein Pex15. Msp1 also resides on peroxisomes but it remains unknown how TA proteins native to mitochondria and peroxisomes evade Msp1 surveillance. Using new quantitative cell microscopy tools for studying Msp1 function, we observed that a fraction of peroxisomal Pex15, exaggerated by overexpression, is turned over by Msp1. Kinetic analysis and theoretical modeling revealed that mitochondrial Pex15 molecules are all equally susceptible to Msp1. By contrast, peroxisomal Pex15 molecules are converted from an initial Msp1-sensitive to an Msp1-resistant state. Lastly, we show that Pex15 interacts with the peroxisomal membrane protein Pex3, which shields Pex15 from Msp1-dependent turnover. In sum, our work argues that Msp1 selects its substrates on the basis of their solitary membrane existence.


INTRODUCTION
Tail-anchored (TA) proteins are integral membrane proteins with a single C-terminal transmembrane segment (TMS). In the budding yeast Saccharomyces cerevisiae, the majority of TA proteins are captured post-translationally by cytosolic factors of the conserved Guided Entry of TA proteins (GET) pathway, which deliver them to the endoplasmic reticulum (ER) membrane for insertion by a dedicated insertase (Denic et al., 2013;Hegde and Keenan, 2011). TA proteins native to the outer mitochondrial and peroxisomal membranes are directly inserted into these membranes by mechanisms that are not well defined ( Chen et al., 2014a;Papić et al., 2013, and reviewed in Borgese and Fasana, 2011). Gene deletions of GET pathway components (getD) result in reduced cell growth and TA protein mistargeting to mitochondria and cytosolic TA protein aggregates (Jonikas et al., 2009;Schuldiner et al., 2008). Two recent studies identified the ATPase associated with diverse cellular activities (AAA ATPase) Msp1 an additional factor for supporting cell viability in the absence of GET pathway function (Chen et al., 2014b;Okreglak and Walter, 2014). Specifically, they observed that msp1D cells accumulate mislocalized TA proteins in the mitochondria and that double msp1D getD cells have synthetic sick genetic interactions. This sick phenotype is associated with disruption of mitochondrial function and is exacerbated by overexpression of TA proteins prone to mislocalization. Msp1 is a cytosolically-facing transmembrane AAA ATPase which resides on both mitochondria and peroxisomes. Closelyrelated members of Msp1's AAA ATPase subfamily form hexamers that bind hydrophobic membrane substrates and use the energy of ATP hydrolysis to extract them for protein degradation (Olivares et al., 2016). Several lines of evidence are consistent with the working model that Msp1 operates by a similar mechanism: ATPase-dead mutations of Msp1 are unable to complement msp1D mutant phenotypes; mitochondrial mistargeting of TA proteins leads to their enhanced co-immunoprecipitation with ATPase-dead Msp1; cells lacking Msp1 have increased half-lives of mistargeted TA proteins; and lastly, a complementary analysis of the mammalian Msp1 homolog ATAD1 (Chen et al., 2014b) established a conserved role for Msp1 in correcting errors in TA protein sorting.
AAA proteins can share the same intrinsic enzymatic activity but still be subject to different rules of substrate recognition (Olivares et al., 2016). Some insight into Msp1 substrate selectivity comes from negative evidence showing that native mitochondrial TA proteins are inefficient Msp1 substrates (Chen et al., 2014b). Thus, mistargeted TA proteins might contain intrinsic Msp1 recognition determinants, but it is difficult to exclude the possibility that mitochondrial TA proteins are protected from Msp1 recognition by extrinsic mitochondrial factors. In fact, a similar possibility could explain the apparently mitochondria-specific behavior of Pex15 as a substrate.
Thus, there is circumstantial evidence for the existence of substrate-extrinsic factors specific to mitochondria and peroxisomes that block Msp1 access to native TA proteins at these organelles.
Substrate selectivity mechanisms of many AAA proteins have been successfully dissected by bulk cell approaches for measuring substrate turnover. These approaches are resolution-limited, however, when used to study Msp1 in getD cells because TA proteins mistargeted to mitochondria co-exist with a dominant TA population that remains correctly localized in the same cell. Previous studies overcame this issue through two different approaches that increased the ratio of mistargeted to properly localized substrates. In one case, cells were engineered to produce a Pex15 deletion mutant (Pex15 DC30 ) that is efficiently mistargeted to mitochondria because it lacks its native peroxisomal targeting signal (Okreglak and Walter, 2014). A major limitation of this approach, however, is its inherent unsuitability for establishing if native Pex15 is a latent Msp1 substrate because of undefined peroxisomal factors. Second, a cell microscopy pulse-chase approach was used to monitor turnover of mitochondrial signal from transiently expressed fluorescently-labeled wild-type Pex15 made susceptible to mistargeting by deletion of GET3 (Chen et al., 2014b). In this approach, expression of Pex15 was transcriptionally controlled by the inducible GAL promoter in cells expressing wild-type, ATPase-dead, or no Msp1. Comparison of mitochondrial Pex15 clearance following GAL promoter shut-off revealed that cells lacking functional Msp1 had a reduced fractional rate of substrate clearance; however, these cells also had a larger starting population of mitochondrial Pex15. Because cells expressing functional Msp1 did so during both the Pex15 pulse and chase periods, these results could have been explained by Msp1 only blocking its substrates insertion into the membrane. To determine how Msp1 distinguishes between substrates and non-substrates and to determine if it facilitates substrate extraction or blocks insertion, better tools are needed for temporally controlling and accurately measuring Msp1 activity in vivo, which in turn will set the benchmark for future biochemical reconstitutions of Msp1's cell biological function.
Here, we began by devising a synthetic gene system for independent control of Pex15 (either wild-type or Pex15 DC30 ) and Msp1 expression with two drugs as transcriptional inducers.
Next, we coupled this tool to live-cell confocal microscopy and computational image analysis to quantitate Pex15 signal density separately at mitochondria and peroxisomes. Using this pipeline and a protease-protection assay, we determined that de novo induction of Msp1 activity drove efficient clearance of a Pex15 substrate that was fully membrane-integrated at mitochondria. In the process of dissecting the kinetics of Pex15 turnover at mitochondria, we observed that Msp1 can also eliminate overexpressed Pex15 from peroxisomes, but with a different kinetic profile.
Theoretical model fitting revealed that all mitochondrial Pex15 molecules were equally susceptible to clearance by Msp1. By contrast, we found evidence for a temporal maturation process at peroxisomes that converts newly-resident Pex15 molecules from an Msp1-sensitive to an Msp1resistant state. This conclusion received independent support from lifetime analysis of Pex15 molecules tagged with a tandem fluorescent timer. To define the molecular basis for these observations, we hypothesized that the Msp1-resistant state is enabled by Pex15's native interactions with other peroxisomal proteins. Indeed, co-immunoprecipitation analysis showed that the transmembrane protein Pex3 is a Pex15 binding partner, while acute degradation of Pex3 from peroxisomes in situ induced rapid Msp1-dependent turnover of peroxisomal Pex15. Our findings provide a working mechanistic model for how native Pex15 avoids Msp1 surveillance that can be generalized to native mitochondrial TA proteins. This work also validates a new microscopy toolkit applicable to mechanistic dissections of other quality control systems in vivo.

Efficient clearance of a full-integrated substrate from mitochondria by de novo Msp1 induction.
To generate a defined Msp1 substrate population prior to initiation of Msp1 activity, we devised a synthetic gene system for orthogonal drug-controlled expression of Pex15 and Msp1. Briefly, we created a yeast strain genetic background with two transcriptional activator-promoter pairs: 1. the doxycycline (DOX)-activated reverse tetracycline trans-activator (rTA) (Roney et al., 2016) Figure 1A and see below). This was followed by 2 hours of DOX wash-out to allow for mitochondrial maturation of newly-synthesized YFP-Pex15 ( Figure 1A). Using confocal microscopy, we could resolve the relatively faint mitochondrial YFP fluorescence from the much brighter punctate YFP fluorescence (corresponding to peroxisomes, see below) by signal co-localization with Tom70-mTurquoise2 (a mitochondrial marker; Figure 1B (Pex15 DC30 ) that is efficiently mistargeted to mitochondria because it lacks a C-terminal peroxisome targeting signal (Okreglak and Walter, 2014) ( induction, we harvested cells after DOX treatment. Following cell lysis, we isolated crude mitochondria by centrifugation and treated them with Proteinase K (PK). Immunoblotting analysis against a C-terminal epitope engineered on Pex15 revealed the existence of a protected TMScontaining fragment that became PK-sensitive after solubilizing mitochondrial membranes with detergent ( Figure 2D). Taken together, these findings argue that Msp1 can extract a fullyintegrated substrate from the mitochondrial outer membrane and gave us a new tool for mechanistic dissection of Msp1 function in vivo.

Differential kinetic signatures of mitochondrial versus peroxisomal Pex15 clearance by Msp1
While performing the previous analysis, we observed that b-estradiol also enhanced YFP-Pex15 signal decay at punctate, non-mitochondrial structures. To test if these punctae corresponded to peroxisomes, we used a strain with mCherry-marked peroxisomes (mCherry-PTS1) and induced YFP-Pex15 expression with a lower DOX concentration (10 µg/ml). Indeed, we saw robust YFP and mCherry signal co-localization with little apparent Pex15 mistargeting to mitochondria ( Figure   3A Figure 3D). Notably, Western blotting of lysates did not show the same degree of YFP-Pex15 turnover in the absence of Msp1 induction as was observed in cell microscopy. We harvested equal volumes of a dividing cell culture for each Western blotting timepoint whereas our cell microscopy monitors levels of YFP-Pex15 in individual organelles, and therefore YFP-Pex15 dilution through peroxisome fission and cell division may explain this discrepancy.
To our knowledge, Msp1-induced turnover of peroxisomal Pex15 had not been reported previously. We found two pieces of evidence that this unexpected phenotype was the product of Pex15 overexpression. First, treatment of pTET-YFP-PEX15 cells with the lower (10 µg/ml) DOX concentration still induced a >10-fold higher YFP fluorescence at peroxisomes relative YFP-Pex15 expressed from its native promoter (Figure 3-Figure Supplement 1B-C). Second, we could detect no difference in natively-expressed peroxisomal Pex15 levels when we compared wild-type and msp1D cells ( Figure 3E). This is unlikely a signal detection problem because we could robustly detect the accumulation of natively-expressed Pex15 DC30 at mitochondria in msp1D cells ( Figure   3E).
Why does Msp1-dependent turnover of peroxisomal Pex15 necessitate excess substrate when the same AAA machine clears mitochondria of even trace amounts of mistargeted Pex15?
In search of an answer to this question, we repeated our analysis at higher temporal resolution and found a major difference between the kinetic signatures of mitochondrial and peroxisomal Pex15 turnover by Msp1 ( Figure 4A and see below). Specifically, while mitochondrial Pex15 turnover showed simple exponential decay (i.e. linear decay after log-transformation), the decay of peroxisomal Pex15 appeared to be more complex, comprising faster and slower kinetic components. We detected no major kinetic differences between Msp1 targeting to mitochondria and peroxisomes that could explain this phenomenon (Figure 1-Figure supplement 1B-C) but found a potential clue from a proteome-wide pulse-chase study showing that while most proteins decay exponentially, some exhibit non-exponential decay that can be explained by their stoichiometric excess over their binding partners (McShane et al., 2016). Since peroxisomal membranes have unique residents that interact with native Pex15, we hypothesized that nonexponential decay of overexpressed peroxisomal Pex15 arises due to the existence of an Msp1sensitive "solitary" Pex15 state and an Msp1-insensitive "partner-bound" Pex15 state. This solitary state would be minimally populated by endogenously expressed Pex15 under steady-state conditions, but a significant fraction of overexpressed Pex15 molecules would be solitary because of stoichiometric excess. By contrast, since mitochondria are unlikely to have Pex15-binding partners, mitochondrial Pex15 would exist in an obligate solitary state and would therefore decay exponentially.
To test this hypothesis, we fit our microscopic YFP-Pex15 decay data against two competing stochastic models, which were previously used to describe proteome-wide protein decay data (see Experimental Methods for modelling details) (McShane et al., 2016). In the 1-state (exponential) model ( Figure 4B, left), we posit that all Pex15 molecules have the same probability of decay (k decay ). In the 2-state (non-exponential) model ( Figure 4B, right), we introduce the probability (k mat ) of nascent Pex15 maturation, alongside distinct probabilities for decay of the nascent (k decay,1 ) and mature (k decay,2 ) Pex15 states. Because the 2-state model contains two additional parameters, it is more susceptible to overfitting than the 1-state model. To avoid this problem, we compared our fits using the Akaike Information Criterion (see Experimental Methods), which provides a parameter-adjusted assessment of fit quality.
To analyze mitochondrial Msp1 substrate turnover, we chose YFP-Pex15 DC30 over wildtype Pex15 to avoid measuring weak mitochondrial signals juxtaposed to strong peroxisomal signals. We also restricted our analysis to the first 60 minutes of b-estradiol treatment because longer Msp1 induction times led to a significant fraction of mitochondria with no detectable YFP signal, which would interfere with turnover fitting ( Figure 4F, inset). From the parameter analysis of the best fits, we found that Pex15 in the nascent state decayed >4-fold faster (k decay, 1 = 2.86 hr -1 ) than Pex15 in the mature state (k decay, 2 = 1.09 hr -1 ).

Msp1 selectively clears newly-resident Pex15 molecules from peroxisomes
The 1-state and 2-state models of peroxisomal Pex15 turnover make distinct predictions about the effect of Msp1 expression on the age of Pex15 molecules. Specifically, in the 1-state model, transient Msp1 overexpression in cells with constitutive Pex15 expression should equally destabilize all Pex15 molecules, thus rapidly reducing their mean age over time ( Figure 5B, top left panel). By contrast, in the 2-state model, Pex15 age should be buffered against Msp1 overexpression because of two opposing forces ( Figure 4B and Figure 5B, top right panel): At one end, there would be an increase in k decay,1 leading to less nascent Pex15, which would drive down the mean age over time. However, there would also be an opposing consequence of rapid depletion of new peroxisomal Pex15 by Msp1: the mature population of Pex15 would receive fewer (younger) molecules, which would drive up the mean age over time. Notably, both models predict that transient Msp1 expression would result in a decrease in peroxisomal Pex15 levels, albeit with differing kinetics ( Figure 5B, bottom panels). We simulated Pex15 levels and age following transient Msp1 activation in the 1-and 2-state models with a set of possible half-lives that ranged from our microscopically determined value of 58 minutes to as slow as 143 minutes, as reported in the literature (Belle et al., 2006) ( Figure 5B). Since our half-life value includes decay due to dilution from cell division, it is likely an underestimate of the actual value.
To measure the effect of Msp1 overexpression on the age of Pex15 molecules, we Nterminally tagged natively-expressed Pex15 with a tandem fluorescent timer (tFT-Pex15) (  and Khmelinskii et al., 2012). In this strain background, we marked peroxisomes using mTurquoise2-PTS1 and induced overexpression of Msp1 from a ZD promoter using b-estradiol ( Figure 5A). Live-cell confocal microscopy combined with computational image analysis revealed a progressive reduction in peroxisomal sfYFP signal following Msp1 overexpression consistent with the predictions of both models, though with kinetics more akin to the predictions of the 2state model ( Figure 5B-C, bottom panels). More strikingly, the peroxisomal mCherry:sfYFP fluorescence ratio was insensitive to b-estradiol treatment, consistent with the prediction of the 2state model ( Figure 5B-C, top panels). Collectively, our experimental evidence and theoretical analysis strongly support the existence of a Pex15 maturation process at peroxisomes that converts newly-synthesized Pex15 molecules from an Msp1-sensitive to an Msp1-insensitive state.
Pex3 is a Pex15-interacting protein that protects Pex15 from Msp1-dependent clearance at peroxisomes To gain insight into the molecular basis of Pex15 maturation at peroxisomes, we hypothesized the existence of peroxisomal proteins that interact with Pex15 and whose absence would reveal that natively-expressed Pex15 is a latent substrate for Msp1. The cytosolic AAA proteins Pex1 and Pex6 are two prime candidates for testing this hypothesis because they form a ternary complex with Pex15 (Birschmann et al., 2003). However, we did not observe the expected decrease in YFP-Pex15 levels in pex1D or pex6D cells that would be indicative of enhanced turnover by Msp1 To look for additional Pex15 binding partners, we noted that the Pex1/6/15 complex is a regulator of peroxisome destruction by selective autophagy (Kamber et al., 2015;Nuttall et al., 2014). This process is initiated by Atg36, a receptor protein bound to the peroxisomal membrane protein Pex3 (Motley et al., 2012). Indeed, we found that Pex15 interacts with Pex3 by co-immunoprecipitation analysis ( we could test if Pex3 protects Pex15 from Msp1-dependent turnover, we had to overcome a major technical challenge. Specifically, Pex3 is essential for targeting of numerous peroxisomal membrane proteins, which is why pex3D cells lack functional peroxisomes (Fang et al., 2004).
Since Pex3 is normally turned over very slowly (Figure 6-Figure Supplement 1F and Belle et al., 2006), promoter shut-off is not a suitable method for acutely depleting Pex3. Instead, we developed an auxin-inducible degradation system that rapidly eliminates Pex3 from peroxisomes in situ. First, we appended a tandem V5 epitope tag followed by an Auxin-inducible degron sequence (Nishimura et al., 2009) to the cytosolic C-terminus of Pex3 (Pex3-V5-AID). Next, we overexpressed an E3 ubiquitin ligase from rice (OsTir1) that binds and ubiquitinates the AID to enable degradation of AID fusions by the proteasome. OsTir1 overexpression can by itself stimulate basal protein turnover of AID fusions, but degradation is potentiated by Auxin addition.
Indeed, Western blotting analysis for the V5 epitope revealed that Auxin addition induced rapid we observed that Auxin treatment immediately increased the rate of Msp1-dependent Pex15 turnover ( Figure 6B). Thus, acute removal of the Pex15-interacting protein Pex3 from peroxisomes in situ unmasks Pex15 as a latent substrate for Msp1-dependent destruction.

DISCUSSION
Errors in TA protein targeting by the GET pathway pose a constant threat to mitochondrial health.
Two recent studies revealed that yeast Msp1 (ATAD1 in humans), a AAA membrane protein resident on the surface of mitochondria and peroxisomes, is part of a conserved mechanism for preventing mistargeted TA proteins from accumulating in mitochondria (Chen et al., 2014b;Okreglak and Walter, 2014). At the same time, this pioneering work raised an important question about Msp1's substrate selectivity: What distinguishes TA proteins mistargeted to mitochondria from TA proteins native to mitochondria and peroxisomes?
Here, we answer this question as it pertains to Pex15, a native peroxisomal TA protein known to be an Msp1 substrate when mistargeted to mitochondria. As our starting point, we used a new microscopy methodology to show that de novo induction of Msp1 activity clears a fullyintegrated Pex15 variant from mitochondria ( Figure 7). This result solidifies the working model in the literature that Msp1 is a mechanoenzyme capable of extracting its substrates from the membrane and should encourage biochemical reconstitution of Msp1's function using liposomes with fully-integrated TA proteins. We were also able to reveal that peroxisomal Pex15 is a latent Msp1 substrate at peroxisomes. The key starting observation that led us to this conclusion was that Pex15 overexpressed at peroxisomes was turned over by an unusual non-exponential process, which depended on Msp1 induction. By model fitting of these data and comparative analysis with the exponential decay of mitochondrial Pex15, we found evidence for a Pex15 maturation mechanism unique to peroxisomes. By positing that this mechanism converts newly-resident peroxisomal Pex15 from an initial Msp1-sensitive state to an Msp1-resistant state, we were able to account for the non-exponential decay kinetics ( Figure 7). Moreover, we validated a key prediction of this mechanism by showing that Msp1 selectively removes peroxisomal Pex15 from the young end of the age distribution.
The molecular details that enable Pex15 maturation into an Msp1-resistant state remain to be worked out. However, our evidence strongly argues that peroxisomal membrane protein Pex3 is a critical component of this process. Pex3 has been previously shown to play a role in the insertion of peroxisomal membrane proteins (Fang et al., 2004). Thus, it is possible that loss of Pex3 function leads to indirect loss of another membrane protein that itself blocks Msp1-dependent turnover of Pex15. We cannot formally exclude this possibility but we find it unlikely for two reasons. First, we showed that Pex3 co-immunoprecipitates with Pex15. Thus, in principle, Pex3 is well-positioned to either occlude an Msp1 binding site on Pex15 or make Pex15 structurally more resistant to mechanodisruption. Second, our engineered Pex3 degradation system is very rapid and its activation causes a near-instantaneous increase in Msp1-dependent Pex15 turnover rate. Thus, for this to be an indirect effect, it would have to be mediated by a membrane protein whose basal turnover was comparably fast. Settling this issue is an important future goal that will be facilitated by assaying Msp1 function with purified components. More broadly, a simple extension of our working model for Msp1 substrate selectivity leads to the intriguing hypothesis that native mitochondrial TA proteins are shielded from Msp1 by their binding partners. The new microscopy methodology we have described here will facilitate testing of this idea in the near future.
It is useful to compare the function of Msp1 with that of FtsH, its closest relative in Escherichia coli. This AAA protease is known to destroy excess copies of membrane-integral components of several complexes, including the bacterial Sec translocon (Kihara et al., 1995;Westphal et al., 2012). The topological diversity of FtsH substrates argues that they are selected as substrates by exposing generic degron features such as hydrophobic patches or locally unstructured regions in their uncomplexed states. Thus far, only TA proteins have been defined as bona fide Msp1 substrates but we speculate that Msp1, analogous to FtsH, can also recognize other membrane proteins that exist in a solitary state for prolonged periods of time. In this light, it will be interesting to test if the human Msp1 homolog ATAD1 protects mitochondria from accumulating peroxisomal membrane proteins in certain Zellweger Spectrum disorders (Muntau et al., 2000).
Lastly, our work provides a broad perspective for seeing the role of Msp1 as a quality control factor. All quality control mechanisms face the challenge of substrate conditionality: a nonsubstrate at one location or point in time can become a substrate at another. For certain ERassociated degradation (ERAD) mechanisms this challenge comes in the form of distinguishing recently translocated glycoproteins in the ER (an unfolded species) from glycoproteins that have failed to fold in the ER despite prolonged residence (reviewed in Shao and Hegde, 2016). To achieve accurate substrate selection, specific ERAD factors query the age of unfolded proteins for the presence of glycans that have become trimmed with time to serve as degradation signals (Clerc et al., 2009;Quan et al., 2008). Interestingly, the glycosidase responsible for glycan trimming must be under tight expression control because its overexpression results in premature destruction of nascent glycoprotein (Wu et al., 2003). By comparison, we have shown that overproduction of Msp1 inappropriately destabilizes nascent peroxisomal Pex15 molecules, presumably by outpacing their normal maturation into an Msp1-resistant state. Consequently, peroxisomes overexpressing Msp1 eventually lose competency for matrix protein import (data not shown), a process that depends on Pex15 (Matsumoto et al., 2003). Thus, our work reveals how diverse quality control mechanisms set time delays for substrate selection to balance tight surveillance against hypervigilant destruction.

Yeast strain construction
All S. cerevisiae strains were constructed using standard homologous recombination methods (Longtine et al., 1998) and are listed in the strain table. Cassettes for fluorescent protein tagging at genes' endogenous loci were generated by PCR amplification from the pKT vector series (Sheff and Thorn, 2004). Tandem fluorescent timer-tagged PEX15 strains were made by integrating a PCR product consisting of the PEX15 promoter, an sfYFP gene with no stop codon, an mCHERRY gene with no stop codon, and the PEX15 coding sequence and terminator followed by a geneticin resistance cassette into the URA3 locus of the parent strain. terminator. This cassette was integrated into the strain's URA3 locus. PEX3-FLAG was generated by integrating a 3×FLAG tag sequence, CYC1 terminator, and a nourseothricin resistance cassette before the native PEX3 stop codon.

Protease protection of YFP-Pex15 DC30 -V5 at mitochondria
Yeast culture pTET-YFP-PEX15DC30-V5 cells were pre-grown to mid-log phase (OD600 = 1) in 100 mL YEPD and then diluted to OD600 = 0.1 in 1 L YEPD. Cells were grown with shaking at 30 °C to an OD600 of 1 and then treated with 50 µg/ml doxycycline (Sigma) for 4 hours at 30 °C with shaking.
Cells were harvested by centrifugation.

Protease protection
100 µg of crude mitochondria were subjected to treatment with 5 µg Proteinase K (Roche) or mock treatment in the presence or absence of 1% Triton X-100 (Sigma) on ice for 60 minutes.
Phenylmethanesulfonyl fluoride (PMSF) (Sigma) was added to each sample to a final concentration of 2 mM to inhibit Proteinase K and samples were incubated 10 minutes on ice.
Samples were mixed with boiling SDS-PAGE sample buffer and subjected to SDS-PAGE and immunoblotting analysis as described earlier.

Live-cell imaging of tagged Pex15 and Msp1
Yeast culture Cells were inoculated into 2 mL complete synthetic media with glucose (0.67% yeast nitrogen base, 2% glucose, 1×CSM (Sunrise Sciences)) and grown overnight at 30 °C on a roller drum. The following morning, cells were back-diluted to a cell density of 0.05 OD 600 units/mL in fresh media and grown to mid-log phase (density 0.5-1 OD 600 units/mL) for imaging, with drug treatments as indicated in experiment schematics. b-estradiol (Sigma) was used at 1 µM for all experiments; doxycycline was used at concentrations indicated in figure legends. Cells in culture media were applied directly to the well of a concanavalin A (MP Biomedicals)-coated Lab-Tek II chambered coverglass (Thermo Fisher) and incubated 5 minutes at room temperature to adhere. Culture media was removed and adhered cells were immediately overlaid with a 1% agarose pad containing complete synthetic media with glucose and supplemented with drugs when applicable. The agarose pad was overlaid with liquid media for timelapse imaging experiments.

Confocal fluorescence microscopy
Live-cell imaging was performed at 25 °C on a TI microscope (Nikon) equipped with a CSU-10 spinning disk (Yokogawa), an ImagEM EM-CCD camera (Hamamatsu), and a 100× 1.45 NA objective (Nikon). The microscope was equipped with 447 nm, 515 nm and 591 nm wavelength lasers (Spectral) and was controlled with MetaMorph imaging software (Molecular Devices). Zstacks were acquired with 0.2 µm step size for 6 µm per stack. Camera background noise was measured with each Z-stack for normalization during timelapse imaging.

Sample size estimation and experimental replication details
For quantitative microscopy experiments, the number of cells present in each sample was manually counted in brightfield images and indicated in the associated figure legend. Each experiment was repeated as indicated in the associated figure legend. Replicates represent technical replicates in which the same strains were subjected to repeats of the entire experiment, often on different days.

Image post-processing and organelle segmentation
First, all fluorescence images were normalized to background noise to compensate for uneven illumination and variability in camera background signal. To identify peroxisomes and mitochondria, organelle marker images were processed by an object segmentation script. Briefly, images were smoothed using a Gaussian filter and then organelle edges were identified by processing each slice with a Canny edge detector (Canny, 1986)

Fluorescence intensity analysis
Following organelle segmentation, total fluorescence intensity for Pex15 was determined in each segmented object by summing intensities in the corresponding pixels for YFP fluorescence images (and mCherry images for mCherry-sfYFP-Pex15 and mCherry-sfYFP-Pex15 DC30 in Figure 3).
Fluorescence density was calculated by dividing total pixel intensity by object volume in pixels.
Background was calculated empirically by measuring Pex15 fluorescence intensity in peroxisomes and/or mitochondria in cells lacking fluorescently labeled Pex15, and the mean background density was subtracted from each segmented object's fluorescence density. Because Pex15 fluorescence density was roughly log-normally distributed, mean and standard error of the mean were calculated on logarithmically transformed fluorescence densities when applicable. Plotting was performed using R and the ggplot2 package. See www.github.com/denic_lab/Weir_2017_analysis for tabulated data and analysis code. As indicated by figure legends, error bars represent standard error of the mean when present.

Simulation of protein age and turnover
To stochastically model peroxisomal Pex15 levels and age following transient MSP1 expression we used a Gillespie algorithm approach (Gillespie, 1977). In brief, this approach cycles through the following steps: 1. Model the expected time until the next "event" takes place (import, degradation, or maturation of a Pex15 molecule) by summing event rates and drawing from an exponential distribution based on the summed rate constant, 2. Age all simulated Pex15 molecules according to time passage, 3. Determine which of the possible events took place by weighted random draws based on each event's probability of occurring, 4. Execute that event, and then repeat these steps until the simulation's time has expired. Based on our observation that Pex15 turnover in the absence of Msp1 occurs with exponential decay kinetics ( Figure 4F), we established starting conditions by drawing 1000 ages from an exponential distribution with half-life indicated in Figure 5B. For the rest of the simulation we used this rate constant to predict import of new molecules and as a steady-state degradation rate constant (and as k decay,2 in 2-state simulations).
We treated this vector of 1000 ages as a single peroxisome containing 1000 Pex15 molecules (this is likely an over-estimation of Pex15 amounts in many cases, but over-estimating Pex15 levels improved statistical robustness of the analysis and did not alter simulation mean outcomes). When simulating steady state 2-state behavior using the calculated k mat value, we found that ~60% of the elements existed in the "unstable" form at steady state (data not shown) and therefore used this as a starting value. For 2-state simulations we randomly drew 600 of the vector elements to be "unstable" at the start of the simulation, weighting probabilities of each draw using an exponential distribution with k mat as the decay rate constant. After validating that our starting conditions represented a stable steady state by simulating without perturbing rate constants, we began the reported simulations with k decay set to 2.38 hr -1 , the best linear fit for turnover from the first 4 time points (for 1-state simulations), or with k decay,1 (for 2-state simulations) set to the calculated value from Figure 4F. Simulations ran for 4 hours of simulated time and values for particle age and abundance were recorded at every simulated minute. 100 simulations were performed with each set of parameters and the mean particle age and abundance at each minute were calculated across the 100 simulations. Finally, we modeled maturation of sfYFP fluorescence and mCherry fluorescence based on established maturation half-times (Hansen and O'Shea, 2013;Khmelinskii et al., 2012, respectively) and calculated the mean population tFT ratio at each minute. We normalized these data to the value at the simulation's starting point. See the Denic Lab Github repository for this paper for analysis code details.

Pex3-GFP-AID fluorescence microscopy
Yeast cultures were grown overnight in synthetic medium to logarithmic phase, treated with 3indoleacetic acid (Auxin, 1 mM) (Sigma) or DMSO vehicle as indicated, concentrated, and imaged at room temperature on an Axiovert 200M microscope body (Carl Zeiss) equipped with a CSU-10 spinning disk (Yokogawa) and 488 nm and 561 nm lasers (Coherent), using an oil-immersion 100× 1.45 NA objective (Carl Zeiss). Images were acquired using a Cascade 512B EM-CCD detector (Photometrics) and MetaMorph acquisition software (Molecular Devices).
Dynabeads were washed 4 times with IP buffer and bound proteins were eluted with 20 µl 1 mg/mL 3×FLAG peptide (Sigma) in IP buffer. Immunoblotting analysis was performed as described above.    Beginning at the top left, fluorescence channels for mitochondria (cyan) and peroxisomes (magenta) are split into separate Z-stacks. After 3-dimensional Gaussian smoothing through the Z-stack, each individual slice is analyzed using a Canny edge detector (Canny, 1986) to identify object edges. Next, enclosed areas are filled to generate a mask  The 515 nm laser power was increased relative to the experiment in Figure 1 and therefore AUs are not comparable between these experiments.
Crude mitochondria were isolated from TET-YFP-pex15 DC30 -V5 cells (see Experimental Methods for details) and subjected to Proteinase K (PK) or mock treatment in the presence or absence of 1% Triton X-100 (TX). Samples were resolved by SDS-PAGE and analyzed by immunoblotting with the indicated antibodies. Immunoblotting with an a-V5 antibody visualized bands at the predicted molecular weight for both full-length YFP-Pex15 DC30 -V5 (top) and a smaller protease-resistant fragment (bottom). The asterisk denotes a PKresistant fragment that appears only in detergent-solubilized mitochondria.
Immunoblotting was performed against the surface-exposed mitochondrial outer membrane protein Tom70 and the mitochondrial inner membrane protein Sdh4 to assess accessibility of different mitochondrial compartments to PK.   A. Experimental timeline of a pulse-chase analysis similar to the one described in Figure 1A but with 10 µg/ml DOX. This experiment was performed twice with similar results.
B. Representative confocal micrographs from the experiment described in part A. Each image represents a maximum intensity projection of a Z-stack. Red cell outlines originate from a single bright-field image acquired at the center of the Z-stack. Scale bar, 5 µm.   combined with a new experiment to monitor YFP-Pex15 turnover at high temporal resolution. YFP signal density at mitochondria (red) or peroxisomes (purple) is plotted after normalization to the 0 hour timepoint, with lines directly connecting timepoints. Error bars represent standard error of the mean. These data are reproduced in parts D and F.
B. Schematics of the two competing models for Pex15 turnover. In the 1-state model, newlysynthesized Pex15 is first targeted and inserted into the peroxisome membrane and then degraded by a simple exponential decay process that occurs with the rate constant k decay . In the 2-state model, there is an additional exponential maturation process that converts Pex15 from a nascent state to a mature state at a rate defined by k mat . In addition, this model includes the new exponential decay constant k decay,2 for the mature Pex15 state that is distinct from the k decay,1 of the nascent state.
C. Experimental timeline of the staged expression experiment for monitoring Msp1dependent turnover of mitochondrial Pex15 DC30 with high temporal resolution. 50 µg/ml DOX was used for induction.
D. Quantitation of mitochondrial YFP-Pex15 DC30 fluorescence from the experiment described in part C. YFP signal density at mitochondria was determined as described in Figure  E. Schematic of the staged expression experiment for monitoring Msp1-dependent turnover of peroxisomal Pex15 with high temporal resolution. 10 µg/ml DOX was used for induction. This experiment was performed three times with similar results.
F. Quantitation of peroxisomal YFP-Pex15 fluorescence from the experiment described in part E. YFP signal density at peroxisomes was determined as described in Figure   A. Quantitation of the fraction of mitochondria lacking detectable YFP fluorescence from the experiment described in Figure 4A,C-D. The Y axis represents the fraction of TET-YFP-pex15DC30 mitochondria with fluorescence values less than two standard deviations above the mean autofluorescence background. Model fitting was disrupted if >5% of organelles (dashed red line) did not contain detectable signal, and therefore model fitting was only performed through the one hour timepoint for YFP-Pex15 DC30 turnover. All TET-YFP-PEX15 peroxisomes analyzed in Figure 4A,E-F contained detectable YFP fluorescence (data not shown).
B. Quantitation of peroxisomal fluorescence densities in additional experimental replicates for the experiment described in Figure 4A,E-F. Replicate 1 represents the same data shown in Figure 4A,E-F. Error bars represent standard error of the population mean. See Figure   4E-F for additional details.
C. Fit quality and parameters from model fits shown in Figures 4D and 4F   A. Tandem fluorescent timer schematics. Pex15 is N-terminally tagged with a superfolding YFP (sfYFP) followed by mCherry. sfYFP matures with a half-time of ~10 minutes (Hansen and O'Shea, 2013), whereas mCherry matures with a half-time of ~40 minutes (Khmelinskii et al., 2012).
B. Theoretical fraction of sfYFP molecules and mCherry molecules that are fluorescent at indicated times following synthesis as well as the ratio of mCherry:sfYFP fluorescence signal. Curves calculated based on maturation rates in A.

Figure 6: Pex3 protects Pex15 from Msp1-induced turnover.
A. Wild-type and msp1D cells containing the doxycycline-inducible promoter driving YFP-Pex15 expression (pTET-YFP-PEX15) and expressing Pex3-AID were grown for 3 hours in the presence of 10 µg/ml doxycycline (DOX) before they were washed and grown for 2 hours in drug-free media. Following this period of substrate pre-loading, half of the cells were exposed to 1 mM Auxin, followed by time-lapse imaging of both cell populations using a spinning disk confocal microscope. This experiment was performed three times with similar results.
B. Quantitation of peroxisomal YFP-Pex15 fluorescence from the experiment described in part A. YFP signal density at peroxisomes was determined as described in Figure    On the left, mistargeted Pex15 inserts into mitochondria and is then recognized by Msp1 for extraction. On the right, following insertion into peroxisomes, nascent Pex15 can be recognized by Msp1 in principle but in practice this requires either Pex15 and/or Msp1 to be present above their usual levels. Otherwise, normal Msp1 recognition is slow relative to the faster "maturation" process involving Pex3 interaction with Pex15, which blocks Msp1 recognition. Pre-culture β-estradiol