A PEROXO-Tag Enables Rapid Isolation of Peroxisomes from Human Cells.

Peroxisomes are metabolic organelles that perform a diverse array of critical functions in human physiology. Traditional isolation methods for peroxisomes can take more than 1 h to complete and can be laborious to implement. To address this, we have now extended our prior work on rapid organellar isolation to peroxisomes via the development of a peroxisomally localized 3XHA epitope tag ("PEROXO-Tag") and associated immunoprecipitation ("PEROXO-IP") workflow. Our PEROXO-IP workflow has excellent reproducibility, is easy to implement, and achieves highly rapid (~10 min post homogenization) and specific isolation of human peroxisomes, which we characterize here via proteomic profiling. By offering speed, specificity, reproducibility, and ease of use, the PEROXO-IP workflow should facilitate studies on the biology of peroxisomes.

Introduction

Peroxisomes are membrane-bound organelles that perform a diverse array of biological functions in human cells. Defects in peroxisomal biogenesis or associated enzymatic functions can lead to devastating pathologies that manifest early in life (Islinger et al., 2018). With regards to the study of peroxisomes, direct interrogation of isolated peroxisomes can provide insights into peroxisomal physiology beyond that achieved using whole-cell analyses.

We have previously developed multiple workflows that utilize 3XHA epitope tags localized to either mitochondria (Bayraktar et al., 2019, Chen et al., 2016, Chen et al., 2017) or lysosomes (Abu-Remaileh et al., 2017) to enable rapid immunocapture of the respective subcellular compartment. These methods have been successfully used in conjunction with downstream metabolomic analyses to study the dynamics of mitochondria (Bayraktar et al., 2019, Chen et al., 2016) and with metabolomic and proteomic analyses to study the behavior of lysosomes (Abu-Remaileh et al., 2017, Wyant et al., 2017, Wyant et al., 2018) and identify changes that were not readily visible with whole-cell or whole-tissue analyses. Historically, isolating peroxisomes has been one of the more challenging enterprises in organellar purification because peroxisomes can be fragile and share similar biophysical properties (e.g., density) with other organelles, such as mitochondria and lysosomes (Volkl and Fahimi, 1985). In addition, the vast majority of isolation methods for peroxisomes have relied on workflows that can be lengthy and laborious to implement. Indeed, many of the common methods for isolating peroxisomes involve density-gradient centrifugation workflows that can last more than 1 h from the time of homogenization (Antonenkov et al., 2004b, Ghosh and Hajra, 1986, Neat and Osmundsen, 1979, Volkl and Fahimi, 1985), increasing the likelihood for distortion of the native organellar profile (e.g., losing associated peripheral membrane proteins) and damage to the organelle itself. Alternative approaches, such as magnetic-bead-based affinity purification of peroxisomes via endogenous ABCD3, a peroxisomal membrane protein (Luers et al., 1998, Wang et al., 2012), or workflows utilizing zonal free flow electrophoresis (Islinger et al., 2010) or immune free flow electrophoresis (Volkl et al., 1999) all can take more than 1 h from the time of homogenization as well.

To address these issues, we have extended our original epitope-tagging approach (Abu-Remaileh et al., 2017, Chen et al., 2016) to peroxisomes and generated a “PEROXO-Tag” that allows for rapid (~10 min post homogenization), specific, and facile isolation of peroxisomes in a highly reproducible manner. The PEROXO-Tag is a chimeric protein consisting of three HA epitopes attached to the N terminus of monomeric EGFP and a segment of PEX26 fused to the EGFP C terminus, which allows for localization to and insertion into the peroxisomal membrane. Leveraging the high-affinity, high-specificity interaction between the tandem HA epitopes of the PEROXO-Tag and the cognate anti-HA antibody, we show that this PEROXO-Tag and the associated immunoprecipitation (“PEROXO-IP”) workflow can be used on human cells for the rapid immunoisolation of peroxisomes, which we characterize here via proteomic analysis. Our PEROXO-IP workflow also exhibits high reproducibility and allows for straightforward isolation of peroxisomes from cells expressing the PEROXO-Tag. Taken together, this work extends our previous efforts for developing a tool kit of epitope-tag-based handles for the rapid and specific isolation of organelles to peroxisomes.

Results

Description of the PEROXO-IP Workflow

To enable rapid isolation of peroxisomes from human cells, we generated the PEROXO-Tag, a chimeric protein in which three HA epitopes are attached to the N terminus of monomeric EGFP and amino acids 237–305 of human PEX26 are fused to the EGFP C terminus, allowing for localization to and insertion into the peroxisomal membrane (3XHA-EGFP-PEX26 or HA-PEROXO). As a control, we also designed a counterpart protein in which the three HA epitopes were exchanged for three Myc epitopes (3XMyc-EGFP-PEX26 or Control-PEROXO) (Figure 1A). Amino acids 237–305 of human PEX26 have been used previously to localize a protein to peroxisomes (Hosoi et al., 2017). We chose to use this segment of PEX26, rather than full-length PEX26, to minimize unintended biological effects from exogenous expression of the Control-PEROXO and HA-PEROXO genes. Indeed, human PEX26 is a peroxin that interacts with the PEX1-PEX6 complex, as well as PEX14, and it has been reported previously that amino acids 13–48 and 78–85 in the N-terminal region of PEX26 are required for the interaction with the PEX1-PEX6 complex and PEX14, respectively (Tamura et al., 2014).

Figure 1

A Workflow for the Rapid and Specific Isolation of Peroxisomes from Human Cells

(A) The design of the 3XMyc-EGFP-PEX26 (Control-PEROXO) and 3XHA-EGFP-PEX26 (HA-PEROXO) proteins. EGFP, monomeric EGFP; PEX26, amino acids 237–305 of human PEX26.

(B) Schematic of the PEROXO-IP workflow, which allows for rapid (~10 min once cells are homogenized) and specific isolation of peroxisomes from human cells for downstream profiling via approaches such as proteomic analysis. The immunocapture workflow used here is generally similar to that reported previously (Chen et al., 2017).

(C) Immunofluorescence of the indicated T47D cell line. Representative images with epitope-tag (identified by Myc-positive or HA-positive structures) (green), peroxisomal (identified by ABCD3-positive structures) (red), and nuclear (blue) signals are shown. Staining of epitope tags was done despite the presence of EGFP in the respective proteins because cells are sorted for low EGFP levels in our methodology, thus rendering the EGFP signal weak. Insets represent zoomed-in fields. The scale bars in the large and small images represent 5 μm and 1 μm, respectively.

(D) Immunoblot analysis of T47D whole cells and immunoprecipitates from the PEROXO-IP workflow. The names of the protein markers used and the corresponding subcellular compartments appear to the left and right of the immunoblots, respectively. Golgi, Golgi complex; ER, endoplasmic reticulum.

In our PEROXO-IP workflow, we utilize cells lentivirally transduced with a 3XMyc-EGFP-PEX26 construct (Control-PEROXO cells) or a 3XHA-EGFP-PEX26 construct (HA-PEROXO cells). Both Control-PEROXO and HA-PEROXO cells are sorted for low EGFP levels via FACS to reduce the chances of unwanted effects that might be secondary to excessive amounts of the Control-PEROXO or HA-PEROXO proteins, such as mislocalization of the epitope-tagged proteins and perturbations of native cell biology. Utilizing the PEROXO-IP workflow, we are able to achieve rapid peroxisomal isolation in ~10 min post homogenization using a methodology that is generally similar to that described previously (Chen et al., 2017). This is significantly faster than the vast majority of prior isolation methods, which can require more than 1 h to isolate peroxisomes after the starting material has been homogenized. In addition, the peroxisomal isolation buffer used in this workflow (i.e., “KPBS”), which we had developed initially for rapid isolation of mitochondria (Chen et al., 2016), is compatible with mass spectrometry, thus allowing for downstream mass-spectrometric analyses, such as proteomics (Figure 1B).

Characterization of the PEROXO-IP Workflow

Importantly, we found that our PEROXO-IP workflow behaved as anticipated when evaluated via orthogonal methods. As part of our initial characterization of the PEROXO-IP methodology, we utilized a standard cultured human cell line, T47D. Immunofluorescence-based examination revealed co-localization of the Control-PEROXO and HA-PEROXO proteins with the peroxisomal marker ABCD3 (Figure 1C). Importantly, immunoblot analysis of the immunoprecipitates using Control-PEROXO cells (Control-PEROXO IP) and HA-PEROXO cells (HA-PEROXO IP) revealed enrichment of peroxisomes relative to a variety of other organelles and minimal organellar contamination of the Control-PEROXO IP (Figure 1D), demonstrating that our PEROXO-IP workflow can achieve rapid and specific isolation of peroxisomes. We assessed peroxisomes using a variety of different markers, namely, ABCD3, CAT, SCP2, and PEX19. The form of SCP2 we examined is the ~14-kDa product that results from transcription of a downstream promoter for the SCP2 gene and has been used previously to better assess the integrity of isolated peroxisomes (Antonenkov et al., 2004a). PEX19 is a peripheral membrane protein that shuttles between the cytosol and surface of peroxisomes; the percentage of the total cellular population of PEX19 that is actually on peroxisomes can be variable (Colasante et al., 2017), but we did observe the presence of the protein in our peroxisomal isolates from both T47D cells (Figure 1D) and HEK-293T cells (Figure 3A), suggesting that the workflow can even capture non-integral membrane proteins that interact with the peroxisome on a more transient basis, which may be a reflection of the rapidity of the method. It is interesting that, in T47D cells, CAT does not exhibit similar patterns of enrichment as ABCD3 and SCP2, which is unlikely to be a reflection of CAT loss from damaged, isolated peroxisomes as SCP2, a smaller matrix protein, does not behave similarly. Instead, the pattern of enrichment of CAT is likely secondary to the CAT peroxisomal targeting signal type 1 (PTS1)-like sequence (i.e., KANL) being less potent than the standard PTS1 sequence (e.g., SKL), which can result in a proportion of CAT that is cytosolic (Legakis et al., 2002, Otera and Fujiki, 2012); an additional factor contributing to the pattern of CAT enrichment could be the presence of some CAT in a non-peroxisomal organelle, such as mitochondria (Calvo et al., 2016).

Figure 3

Proteomic Characterization of Isolated Peroxisomes

(A) Immunoblot analysis of HEK-293T whole cells and immunoprecipitates from the PEROXO-IP workflow. The names of the protein markers used and the corresponding subcellular compartments appear to the left and right of the immunoblots, respectively. Golgi, Golgi complex; ER, endoplasmic reticulum.

(B) Examination of reproducibility among Control-PEROXO IP and HA-PEROXO IP biological replicates subjected to proteomic interrogation. Scatterplots of the reported exclusive intensities for each protein for each biological replicate of the indicated IP group and correlation matrices demonstrating the Spearman rs value for each indicated comparison are shown. The primary proteomic data shown in Table S1B were used to generate this panel. For all correlation analyses of different replicates in either the Control-PEROXO IP group or HA-PEROXO IP group, p < 0.001 and all correlations were significant after application of the Benjamini-Hochberg procedure (FDR = 5%). For the purposes of plotting and analysis, only proteins with reported exclusive intensities in all four biological replicates of a given IP group were included for that IP group's scatterplot and correlation analyses; accordingly, n = 1,360 proteins for the Control-PEROXO IP group and n = 1,426 proteins for the HA-PEROXO IP group.

(C) Scatterplot of proteins identified by proteomics in the immunoprecipitates from the PEROXO-IP workflow. The primary proteomic data shown in Table S1B were used to generate this panel. The dotted line indicates where the p value = 0.023; all points above this line are those that are significant after application of the Benjamini-Hochberg procedure (FDR = 5%). For the purposes of graphing, only proteins that have a value for log2(HA-PEROXO IP (median)/Control-PEROXO IP (median)) are plotted here (n = 1434 proteins). Regardless, all 45 proteins annotated as peroxisomal per the Gene Ontology (GO) Resource in our primary proteomic data (n = 1,441 proteins) are plotted here and indicated in blue, and, of these 45 peroxisomal proteins, 32 are among the top proteins from our analysis (i.e., have a log2 score >0 and are statistically significant after application of the Benjamini-Hochberg procedure). Proteins with no peroxisomal annotation are in gray.

(D) Violin plots of the organellar enrichments for peroxisomes, mitochondria, or ER. Top proteins shown in Table S1C were used to generate this plot, but only proteins that were localized by the GO Resource solely to the indicated organelle and not to the other two organelles were used to represent the corresponding organelle. Proteins were chosen in this way to minimize the confounding effects of including proteins that could have localization to more than one of the three organelles. Peroxisomes (n = 15 proteins), mitochondria (n = 159 proteins), ER (n = 56 proteins). The black dotted line indicates the median, blue dotted lines indicate the quartiles, and the violin plots extend from the minimum to maximum of each dataset. The actual numerical value of the median is shown below each corresponding violin plot. ∗∗∗p < 0.001.

(E) Identities and general classifications of the 32 previously annotated peroxisomal proteins found among the top proteins. The corresponding NCBI gene symbols are shown.

For all proteomic analysis in Figure 3, see Table S1 for additional data and details.

We next examined whether introduction of the Control-PEROXO or HA-PEROXO proteins perturbed native features of peroxisomes, such as levels of known peroxisomal proteins or the morphology and distribution of peroxisomes. Using immunoblot analysis of whole cells, we did not observe any noticeable difference in the levels of standard peroxisomal markers such as ABCD3 and CAT or the peroxisomal proteins PEX19, SCP2, and endogenous PEX26 in Control-PEROXO or HA-PEROXO cells versus those transduced with the corresponding empty vector (Figure 2A). In addition, immunofluorescence-based examination of peroxisomes revealed no clear changes to peroxisomal morphology or distribution (Figure 2B).

Figure 2

Examination of the Effects of the Control-PEROXO and HA-PEROXO Proteins on Peroxisomes

(A) Whole-cell immunoblot analysis of the indicated T47D cell line. The names of the proteins appear to the left. RPTOR and β TUBULIN are loading controls.

(B) Immunofluorescence of the indicated T47D cell line. Representative images with peroxisomal (identified by ABCD3-positive structures) (red) and nuclear (blue) signals are shown. The scale bar represents 5 μm.

Quantitative Proteomic Profiling of Isolated Peroxisomes

To better characterize the peroxisomes isolated using our PEROXO-IP workflow, we analyzed whole-cell and immunoprecipitate (IP) material using quantitative proteomics and data-independent acquisition. To increase yield for the proteomic work, we scaled up the input and transitioned the PEROXO-IP workflow to HEK-293T cells. We confirmed that our PEROXO-IP workflow still resulted in substantial enrichment of peroxisomes relative to other subcellular compartments under these conditions (Figure 3A). The PEROXO-IP workflow also demonstrated excellent reproducibility across either Control-PEROXO IP or HA-PEROXO IP biological replicates, as evidenced by the overlap of measured abundances in scatterplots and by correlational analysis exhibiting Spearman rs values >0.84 in all comparisons (Figure 3B). Of the 1,441 proteins in our primary proteomic data, 288 proteins display greater signal in the HA-PEROXO IP samples than in the Control-PEROXO IP samples (i.e., the background) and are statistically significant after correcting for multiple hypothesis testing; these 288 proteins are considered the “top proteins” (see Table S1). Thirty-two of these top proteins are annotated as peroxisomal per the Gene Ontology (GO) Resource; for context, the GO Resource has a total of 141 proteins assigned to peroxisomes in humans. To account for factors such as absent or low expression of genes encoding for certain peroxisomal proteins in HEK-293T cells and/or poor behavior of certain peroxisomal proteins with regards to proteomic analysis, we examined all 1,441 proteins of the primary proteomic data, which represent proteins found in IP and/or whole-cell samples, and found 45 peroxisomal proteins. We thus determined the general coverage of the PEROXO-IP workflow to be ~71% (32/45) (Figure 3C and see Table S1). Importantly, the peroxisomal proteins ABCD3, CAT, and SCP2 are present among the top proteins, corroborating the results of our immunoblot analysis (Figures 3A and 3C). Although we could detect PEX19 in our peroxisomal isolates via immunoblotting (Figure 3A), it is not present in our proteomic data (see Table S1), which may reflect a combination of low protein abundance and/or the performance of this protein in our proteomics. Mitochondria and endoplasmic reticulum (ER) proteins can be commonly seen in density-gradient centrifugation preparations of peroxisomes (Gronemeyer et al., 2013, Zhou et al., 2015); as such, we interrogated the top proteins for signs of those two organelles and found 182 mitochondrial proteins and 66 ER proteins per the GO assignments (see Table S1C). This mitochondrial signal is potentially a result of native mitochondrial-peroxisomal contacts (Schrader et al., 2019), transfer of mitochondrial material to peroxisomes via mitochondrial-derived vesicles (Braschi et al., 2010, Neuspiel et al., 2008), and/or the contribution of mitochondria to the de novo synthesized pool of peroxisomes (Sugiura et al., 2017). Similarly, this ER signal could be present because of peroxisomal-ER contacts (Schrader et al., 2019) and/or the contribution of ER to peroxisomes during de novo biogenesis (Islinger et al., 2018, Sugiura et al., 2017). Importantly though, by using an organellar enrichment metric (i.e., HA-PEROXO IP signal/HA-PEROXO whole-cell signal) for the top proteins that also behave as ideal markers (i.e., strictly annotated to only one organelle in the set of peroxisome, mitochondria, or ER), we found that our PEROXO-IP workflow enriched for peroxisomes to a significantly greater degree than for mitochondria or ER (Figure 3D and see Table S1C). We chose proteins only assigned to one of the three organelles to minimize the confounding effects of including proteins that could have multiple localizations. Reassuringly, the median organellar enrichment for peroxisomes is ~12.7-fold greater than for mitochondria and ~14-fold greater than for ER, which strongly corroborates the results from our immunoblot analysis (Figure 3A and see Table S1C) and demonstrates that, despite the presence of mitochondrial and ER proteins in the IP material, mitochondria and ER were not enriched nearly to the same degree as peroxisomes were. With regards to the 32 peroxisomal proteins belonging to the group of top proteins, we examined each of their associated functions and found a diverse range of molecular players involved in classic processes associated with peroxisomes, such as peroxins (e.g., PEX3, PEX11B, PEX13, PEX14), metabolic proteins involved in the oxidation of very long-chain fatty acids (e.g., ACOX1, HSD17B4) or synthesis of ether lipids (e.g., AGPS, GNPAT), and antioxidant defense proteins (e.g., CAT, GSTK1) (Figure 3E). Taken together, these data demonstrate that our PEROXO-IP workflow can be used to rapidly isolate peroxisomes for downstream interrogation of the peroxisomal proteome.

Discussion

Building upon our prior work utilizing epitope tags for rapid isolation of mitochondria (Bayraktar et al., 2019, Chen et al., 2016) and lysosomes (Abu-Remaileh et al., 2017), we have now developed a PEROXO-Tag and PEROXO-IP workflow that allows for rapid (~10 min post homogenization), specific, and facile isolation of peroxisomes from cultured human cells in a highly consistent manner. Introduction of the PEROXO-Tag into cells did not noticeably alter the native features of peroxisomes, such as the whole-cell levels of various peroxisomal proteins and the morphology and distribution of peroxisomes. Proteomic analysis of peroxisomes isolated using our PEROXO-IP workflow identified proteins that are known to be peroxisomal components that participate in a diverse range of biological processes.

During the course of this project, a separate study by Xiong et al. was published describing the development of a rapid immunopurification scheme for mitochondria, lysosomes, and peroxisomes using a twin-strep-tag localized to the respective organelles (Xiong et al., 2019). The peroxisome-localization sequence used in their work is based on PEX3, whereas ours is based on PEX26. However, compared with the study by Xiong et al., our work goes into substantially greater depth characterizing our Control-PEROXO and HA-PEROXO proteins and our peroxisomal isolates. From a broader perspective, no direct comparison was done between our 3XHA-based approaches for mitochondria (Chen et al., 2016, Chen et al., 2017) and lysosomes (Abu-Remaileh et al., 2017) and the twin-strep-tag-based approaches, so it is difficult to assess how the two epitope-tagging strategies generally compare. Also, of note, the work by Xiong et al. does not report usage of a control IP, which is essential for assessing background binding of materials to beads and is concerning with respect to their metabolomic interrogation of lysosomes. Indeed, from multiple studies (Abu-Remaileh et al., 2017, Bayraktar et al., 2019, Chen et al., 2016, Wyant et al., 2017, Wyant et al., 2018), including the work shown here, we have found the control IP to be critical for bead-based affinity purification of organelles when combined with polar metabolomic, lipidomic, or proteomic analyses.

In summary, we believe that the PEROXO-IP workflow described here will be useful for the isolation and study of peroxisomes given the rapidity, specificity, ease of use, and reproducibility of the methodology. The ability to rapidly isolate peroxisomes can be helpful for the study of proteins by better preserving the dynamic interactions of peripheral membrane proteins with the peroxisomal membrane. Rapidity of isolation can also help preserve the native proteomic profile of peroxisomes, given that these organelles can be fragile once released from cells (Volkl and Fahimi, 1985). The ease of use and speed of isolation should also facilitate scaling up studies on peroxisomes in cultured cells and can thus be useful for examining peroxisomes under multiple genetic, pharmacologic, or environmental perturbations, particularly since the workflow does not require the generation of density gradients. In addition, translation of this methodology to an in vivo mammalian system should be possible using a strategy we have implemented previously for mitochondria (Bayraktar et al., 2019) and thereby enable rapid isolation of peroxisomes from specific cell types without the need for cell sorting. We thus believe that the PEROXO-Tag and PEROXO-IP workflow described in this study will have utility for studying peroxisomes and can help elucidate the more dynamic processes associated with these organelles during various states of cellular function.

Limitations of the Study

It should be noted that we did not perform electron microscopy on isolated peroxisomes in this work. However, we have assessed the integrity and relative enrichment of peroxisomes isolated using our PEROXO-IP methodology via quantitative proteomic analysis and immunoblot analysis with a combination of transmembrane and matrix peroxisomal markers and a variety of markers for non-peroxisomal subcellular compartments. In addition, the PEROXO-IP methodology does have limitations that users should be aware of, as detailed below.

Yield of the PEROXO-IP Workflow

The PEROXO-IP workflow has been optimized for speed of isolation and not for yield, although that has not limited us from performing the desired analyses detailed in this manuscript. Indeed, to allow for increased yield without compromising the rapidity of isolation for applications like proteomics, we have simply increased the material input for each immunoprecipitation. However, we used HEK-293T cells, which are highly proliferative and easy to culture, to obtain the cellular material (i.e., ~140 million cells per immunoprecipitation) needed for our proteomic profiling of peroxisomes. It is important, though, to realize that other cell types may not be as amenable to scaling up for the purposes of increasing material input. An additional issue to consider with regards to yield is that different cell lines will have varying abundances of peroxisomes, which can influence the required amount of input as well.

Specificity of the PEROXO-IP Workflow

In this work, we demonstrate that the PEROXO-IP workflow can specifically isolate peroxisomes in T47D and HEK-293T cells. However, because of the rapidity and relatively mild isolation conditions of our PEROXO-IP workflow, inter-organellar contacts that involve peroxisomes are likely well preserved. Thus, depending on the number and strength of inter-organellar contacts that peroxisomes may have in a given cell line, the amount of contamination in the isolated peroxisomes can vary across cell types.

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, David M. Sabatini (sabatini@wi.mit.edu).

Materials Availability

The lentiviral Control-PEROXO (Addgene #139059) and HA-PEROXO constructs (Addgene #139054) can be obtained through Addgene.

Data and Code Availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifiers being PRIDE: PXD018880 and PRIDE: PXD018918.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.