Unifying Fluorescence Microscopy and Mass Spectrometry for Studying Protein Complexes in Cells.

Among the tested constructs, we found several combinations of fluorescent protein and affinity tags which were optimal for determining sub-cellular localization and purification of the proteins and protein complexes. We applied these tags to further investigate eisosomes, and found several new protein components of the complexes and obtained the new insights into molecular mechanisms regulating eisosome integrity by reversible phosphorylation and de-phosphorylation. Our results indicate that an approach combining fluorescence microscopy/mass spectrometry into a single method provides a unique perspective into molecular mechanisms regulating composition and dynamic properties of the protein complexes in living cells. clues about the cellular processes and the molecular mechanisms of their regulation. Here we demonstrate the usefulness of the combined fluorescent microscopy/mass spectrometry experiments for the studies of the dynamics of eisosome de-phosphorylation. We also show that from the measured changes in sub-cellular localization and abundances of phosphorylation sites we can deduce the existence of a phosphotase system involved in maintaining eisosome integrity during the cell cycle.


Introduction
Fluorescent proteins have become invaluable probes for studying molecular processes in living cells with the light microscopy techniques [1][2][3]. Proteins, organelles and the entire cells can be selectively visualized using a variety of fluorescent proteins fused to the proteins of interest [1][2][3][4][5][6]. Combined with genetics and molecular biology techniques fluorescence microscopy provides an efficient tool for observing molecular phenotypes useful for dissecting the pathways of cell cycle progression and cell response to the internal and external signals [7]. However, understanding the mechanism controlling the properties of proteins in the cells could be a challenging task, frequently requiring a comprehensive characterization of the proteins at the molecular level.
The proteins tagged with a green fluorescent protein (GFP) can be also purified using GFP antibodies. Cheeseman and Desai [8] and Cristea and co-workers [9] have enriched the GFP tagged proteins and protein complexes for further detailed analysis by mass spectrometry (MS).
The MS based methods for protein analysis are fast, sensitive and able to identify both proteins in complex protein mixtures and residues bearing post-translational modifications [10,11]. Thus, the addition of affinity purification and mass spectrometry steps enabled the researchers to study protein interactions and the post-translational modifications in the context of the protein subcellular localization. Juxtaposition of the protein localization, composition of the protein complexes and post-translational modifications frequently yields a unique perspective of the cellular processes and the molecular mechanisms of their regulation [12,13]. Using fluorescent proteins also as affinity probes can be problematic in several instances. First of all, the good quality antibodies against the rapidly increasing number of fluorescent proteins [3,6] are not yet readily available. Furthermore, rising antibodies specifically recognizing fluorescent proteins originating from the same organism but fluorescing at different color can be difficult or even impossible, because such proteins frequently differ by mutations of only few amino acids [1][2][3][4][5][6]. Thus, we seek an alternative approach to the design of the tags suitable for sub-cellular localization and purification of the proteins and protein complexes, which is (1) independent on the availability of antibody to a specific form of a fluorescent protein, (2) suitable for multiplexing, i.e. simultaneous observation of sub-cellular localization of several proteins and affinity purification of the proteins and stably associated protein complexes, (3) flexible and easy to modify to incorporate better versions of fluorescent proteins and affinity tags after they are discovered.
One possible solution which satisfies the stated requirements is to use a modular tag containing a version of a fluorescent protein fused to an affinity epitope. In this case we can decouple requirements to the both modules and optimize the performance of each one independently for fluorescence microscopy and affinity purification experiments. To our knowledge, this possibility was first realized by Thorn and co-workers [14] who have fused 3HA (3 repeats of YPYDVPDYA epitope from hemagglutinin protein) and 13Myc (13 repeats of EQKLISEEDL epitope, corresponding to a stretch of the C-terminal amino acids of the human c-Myc protein) by guest on May 8, 2020 https://www.mcponline.org Downloaded from 5 tags to several variants of fluorescent proteins. The authors have argued that the fusion of the fluorescent proteins to the affinity epitopes may enable fluorescence and immunochemical analysis but did not test this idea. Cheeseman and Desai fused the S-peptide and hexahistidine epitopes to the GFP protein to enable additional tandem purification steps [8]. Su and co-workers also fused a hexa-histidine tag (6xH) to GFP to purify the recombinantly produced proteins [15].
Although hexahistidene tag performs well for isolation of the over-expressed recombinant proteins, it works poorly for affinity purification of low abundant, endogenously expressed proteins [16]. A double affinity tag containing a single Myc epitope and hexahistidine was also used to purify the recombinantly produced fluorescent proteins [6].
Here we describe the design and implementation of the modular fluorescent and affinity tags.
These tags contain a variety of fluorescent proteins, which can be used exclusively for obtaining sub-cellular visualization, and several small epitope tags utilized to perform two-step affinity purification. To test the performance of the produced constructs, we tagged two yeast proteins, Pil1 and Lsp1, the core components of eisosomes, with a variety of modular tags.
Eisosomes are large heterodimeric protein complexes recently discovered in S. cerevisiae [17].
There are ~ 50-100 eisosomes in each grownup yeast cell, distributed uniformly in the characteristic, dotted, pattern at the cell surface periphery. Each eisosome contains ~ 2,000-5,000 copies of Pil1 and Lsp1. It was shown that eisosomes serve as portals of endocytosis in yeast.
The function of eisosome is regulated by reversible phosphorylation [18,19]. 6 Among the tested constructs, we found several combinations of fluorescent protein and affinity tags which were optimal for determining sub-cellular localization and purification of the proteins and protein complexes. We applied these tags to further investigate eisosomes, and found several new protein components of the complexes and obtained the new insights into molecular mechanisms regulating eisosome integrity by reversible phosphorylation and dephosphorylation. Our results indicate that an approach combining fluorescence microscopy/mass spectrometry into a single method provides a unique perspective into molecular mechanisms regulating composition and dynamic properties of the protein complexes in living cells.
Genomic Tagging, In vivo Disruption and In vivo Mutagenesis -The PCR primer pair, 5'-genespecific primer (51 bases) + GAT CCG CTA GCG CTA CCG GTC-3' (sense) and 5'-gene-by guest on May 8, 2020 7 specific primer (51 bases) TAA TAC GAC TCA CTA TAG GGA GAC -3'(anti-sense), was selected to amplify the desired DNA construct coding for a specific modular tag. The amplified DNA was directly used for C-terminal tagging of the protein of interest by a homologous recombination technique [22,23]. In vivo disruption and mutagenesis of the genomic DNA was performed essentially as described in [24].
A typical molecular weight of a modular fluorescent and affinity tag is ~30 kDa, which includes the molecular weight of a fluorescent protein, two affinity epitopes and spacers between the tags and the protein (see DNA sequences in the Supplemental Data).
Yeast Strains -All yeast strains were derived from MATa BY4741 bar1∆::KanMX strain or MATα BY4742 strain. Transformed clones were selected based on the presence of a selectable marker and a fluorescence signal.
Cell Culturing -Cells were cultured in 0.5-1 liter YEPD media (MP Biomedicals) at 30 ℃ to the density (2-4) x 10 7 cells/ml. For the cell-cycle arrest and release experiments, we cultured several liters of yeast cells. Each liter of a culture was incubated in YEPD media containing 50 μg/l α-factor (GenScript Corporation) for 3 hours, spun down and washed 3 times with double deionized water, then cultured in 1 liter of a regular YEPD media containing 50mg/l pronase enzyme (Sigma-Aldrich) added to the media to digest the remaining α-factor [25]. diagram of the combined microscopy/mass spectrometry experiment is shown in Figure 1. The cultured cells were quickly collected by centrifugation of the media in two 500-ml-buckets of a Hermle Z383 centrifuge (Denvile Scientific) at ~ 4000 x g for 5-10 minutes. A small portion of cells was immediately deposited on a glass slide for the fluorescence microscopy experiment performed on a Zeiss LSM510 confocal microscope. Images were later processed using Zeiss LSM Image Browser software.
The rest of the sampled cells were frozen by dripping cells in a 50 ml falcon tube filled with liquid nitrogen. The pellets of cells were stored at -80 ºC until the proteins and protein complexes were affinity purified according to the procedures described below.
Tandem Affinity Purification of the Protein Complexes. Tandem affinity purification of proteins was performed essentially as described previously [20]. The detailed protocol can be also found in the in the Supplemental Data. Briefly, the fist purification step was performed by adding ~5- Gel Electrophoresis -The eluted proteins were separated by SDS-PAGE using 4-20% gradient gels (Pantera S, B-Bridge) or 4-20% or microGels (Life-Gels, Life Therapeutics, Australia). The gels were stained with a colloidal coomassie stain (GelCode, Pierce). The bands were exercised, digested with 5-7 μl of a 1 pmol/µl of trypsin solution in ammonim bicarbonate. The tryptic peptides were extracted according to a standard in gel digestion/peptide extract procedure [10].
The details of the protocol could be also found in the Protocol II in the Supplemental Data. be used separately or sequentially to interrogate the same samples deposited on the MALDI target plate, which can be loaded into either of the mass spectrometers [20]. and label the first isotopes of ion peaks in the spectra. We set the major setting of the program, such as, peak centroid to 6, signal-to-noise ratio to 1.5 and resolution to 10,000. This procedure usually results in detecting 50-300 ion peaks in the single spectrum measured in the range 500-4000 m/z with intensities above signal-to-noise ratio 1.5 [20]. The detected values were saved as text files.
The monoisotopic values from the lists can be directly used to identify proteins in simple protein mixtures. We used the XProteo search engine [20] to identify proteins based on tryptic mapping.
Searching for the S. cerevisiae proteins in NCBI non-redundant data base, version 06/10/16 was performed with the following typical settings -protein mass: 0-300 kD, protein pI: and ion's activation time (30ms). These method files can be later executed by native Xcalibur software controlling the ion trap. The v-MALDI-IT is capable of fast acquisition of high quality MS/MS spectra (~3 sec per spectrum) of peptides presented in the sample at a femtomole level [20].
The MSMS spectra were converted to DTA format using the in house written "DTA converter" program utilizing an original subroutine written by Yates [32].  Xcalibur program controlling the trap [20]. The interpretation and assignment of fragments in the fragmentation spectra of phosphopeptide was performed manually PROWL computer tools available from Brian Chait's laboratory at the Rockefeller University.

Isotopic Differentiation of Interactions as Random or Targeted (I-DIRT) techniquewas
implemented exactly as described in [30] to distinguish between the specific and non-specific interactors in the affinity purified complexes (see also Supplemental Data).

Development of a Modular Fluorescent and Affinity Tags -
The produced DNA vectors were used as templates for PCR-based genomic tagging of proteins with a variety of modular fluorescent and affinity tags (see Supplementary Fig.1 and Table1). We tagged two yeast by guest on May 8, 2020 14 proteins, Pil1 and Lsp1, i.e. the core components of eisosomes [17,18] by incorporating different DNA cassettes at the C-termini of the protein genomic sequences. Then we experimented with cells expressing the tagged proteins as a model system to optimize the performance of the modular fluorescent and affinity tags. The schematic diagram of the combined fluorescence microscopy/mass spectrometry experiment is shown in Figure 1.

Optimization of the Performance of Fluorescent and Affinity Tags-
The characteristic distribution of the eisosome particles at the cell periphery [17] was obtained in all cases of tested modular fluorescent and affinity tags (Supplemental Fig. 2). However, not all of the fluorescent probes performed well. We found that GFP(S65T), mCitrine, mTFP1 provide the brightest fluorescent signals. The mCherry protein produced a good signal when we used 568 nm wave length of an excitation laser. The fluorescence signals of EYFP and mOrange proteins were weak and many cells expressing the proteins did not produce fluorescence. An erratic performance of these proteins may be due to the several reasons, including long folding time compared to a typical cell cycle period in yeast (~90 min) and high sensitivity to the pH in micro-environment [3][4][5].
Interference of the affinity and the fluorescent probes could be also additional factor, which we discuss bellow.
We also investigated the performance of the affinity probes for purification of proteins and protein complexes. To carefully compare purification yields that can be achieved using different epitopes, we incorporated several affinity tags, i.e. 3xFlag, 4x-StrepTag II (4 repeats of by guest on May 8, 2020 https://www.mcponline.org

Downloaded from
WSHPQFEK peptide) and octahistidine 8xH (HHHHHHHH) in tandem at the C-terminus of Lsp1-GFP(S65T) fusion protein. The protein was purified from equal amount of yeast cells, but applying different tandem purification schemes. The results show that tandem purification using 4xStrepTagII-8xHis yields more protein after the two-step purification procedure ( Supplementary Fig. 3A). However, the two-step tandem purification using a StrepTagII-8xHis tag usually resulted in noisier "pull outs" affected by endogenously biotinylated yeast proteins.
Although the addition of increasing amounts of avidin to the cell lysates helped to compete out some of the biotinylated proteins from the Strep-Tactin resin [26], there was always a substantial amount of impurity proteins in the background. In general, the amount of impurity proteins was less when we used 3xFlag-6xHis and 3xMyc-8xHis tags resulting in higher signal/noise ratios of the enrichment procedure.
We also compared the purification yield of Lsp1 fused to the different modular tags, -mCherry-3xFlag-6xH, -mCherry-3xMyc-8xH, -mCherry-3xStrepTagII-8xH and mCherry-4xStrepTag II-8xH. The highest purification yield was again achieved using a tandem purification scheme 4xStrepTagII-8xH albeit with higher impurity content. The 3xMyc-8xH tandem purification produced less protein under the conditions identical to 3xFlag-6xH tandem purification, but still provided enough protein for staining and MS analysis. The efficiency of 3xStrepTagII dropped drastically compared with 4xStrepTagII ( Supplementary Fig. 3B).
Compatibility of fluorescent and affinity probes is another important factor for designing the by guest on May 8, 2020 https://www.mcponline.org Downloaded from modular tagging system. For example, 4xStrepTag II affinity tag strongly compromised and, in some cases, entirely quenched fluorescence when it was fused with a short or a long linker to the N-or C-terminus of GFP(S65T) or mCherry tag. The effect was reproducible in cases when other then Pil1 and Lsp1 proteins were tagged with the same tags. Interestingly, decreasing the number of StrepTag II repeats from 4 to 3 or 1 reduced this problem, but also drastically reduced the protein purification yield (Supplementary Fig. 3B). Perhaps, binding of one of the StrepTag II repeats to a fluorescent protein may unfavorably affect its structure and result in suppressing fluorescence [27].
Presently, we have selected a pair of modular tags, GFP-3xFlag-6xH and mCherry-3xMyc-8xH, among a variety of tested constructs ( Table 1). The two modular tags enabled us to simultaneously track two proteins in a cell and to co-purify the proteins and associated complexes without noticeable cross-interference of the probes (Fig. 2). Using these tags we have performed the combined fluorescence microscopy mass spectrometry experiments to investigate the compositions and the dynamic properties of eisosomes. Figure 3 shows the schematic diagram of several strategies that can be readily implemented using the combined fluorescence microscopy/mass spectrometry experiments. In one strategy (Fig. 3A), we use the detected pattern of protein sub-cellular localization as a clue of whether the identified proteins could be associated in the protein complexes. We start by tagging the protein of interest and first Analysis of the gel bands corresponding to affinity purified proteins of the eisosome complexes (see Fig.2B, Fig.4 and Supplementary Fig. 3) consistently indicates the presence of three major proteins, Pil1, Lsp1 and Mrp8. The first two proteins, Pil1 and Lsp1 are the core eisosome proteins [17]. The third protein, frequently detected in the purified complexes is Mrp8, a putative mitochondrial ribosomal protein (see, for example, in Saccharomyces Genome Database http://www.yeastgenome.org). This protein has been also identified as a possible interacting partner of Pil1 and Lsp1 proteins in high-throughput immunopurification [28] and two-hybrid experiments [29]. The analysis of the localization of Mrp8 (see Fig. 4A) indicates that it does not co-localize with the eisosomes, but rather resides in the cytoplasm of the cells. Nevertheless, a small portion of Pil1 and Lsp1 proteins can be co-purified with the Mrp8-GFP-3xFlag-6xH ( Fig   3A). To investigate whether Mrp8 protein interacts with Pil1 and Lsp1 protein in the specific or nonspecific manner we implemented the I-DIRT technique [30]. We affinity purified the Pil1 protein On-bead digestion of the complexes co-purified with Pil1-GFP-3xFlag-6xH followed by mass spectrometry [20] resulted in identification of several additional proteins, Tef1, Pma1 and Ygr130c (Supplemental Fig.6). Tef2 and Pma1 are the impurities and are frequently found in control samples. However Ygr130c was not identified in the control. The pictures of the protein sub-cellular localization from the www.yeastgfp.ucsf.edu data base [31] suggested a dotted pattern of localization at the cell periphery. Intrigued by this, we implement the strategy for localization-driven exploration of the composition of protein complexes (Fig. 3A) and tagged by guest on May 8, 2020

Localization-driven exploration of the composition of protein complexes -
https://www.mcponline.org

Downloaded from
Ygr130c protein with a modular -GFP-3xFlag-6xH tag and investigated the protein sub-cellular localization and the protein interacting partners. Figure 4B confirms eisosome-like, distribution of the protein tagged with GFP-3xFlag-6xH at the cell periphery. On-bead digestion of the co-purified proteins followed by mass spectrometric analysis of the tryptic peptide mixture produced several top candidates: Ygr130c, Ymr031c, Pil1, Lsp1 and Ymr086w (Supplemental Fig. 7). Figure 4B also shows that these proteins were also  technique [24], we accidentally found that removal of 114 amino acids from the C-terminus of Pil1 protein causes Lsp1 to re-localize into the cytoplasm and into several bright spots at the cell periphery (Fig. 5B). This effect is similar to the one observed after complete deletion of Pil1 protein from the cells [17]. It confirms the importance of Pil1 protein, especially the C-terminus, for proper localization of eisosome particles. We decided to use this phenomenon to deliberately change Lsp1 localization and to examine whether this, in turn, will change localization of the other proteins.
Ygr130c, Ymr031c and Ymr086w changed their localization together with Lsp1 in the cells in which the Pil1 protein was truncated at the C-terminus (Fig. 5B). This result suggests that the new discovered proteins physically interact at least with Lsp1, and thus, are the integral components of eisosomes. The biological function of Ygr130c, Ymr031c and Ymr086w and a possible involvement into endocytosis remains to be investigated. Eisosomes can exhibit the dynamic behavior. Eisosome assembly and organization is controlled through reversible phosphorylation of its core proteins [18]. It was shown that at least several kinases, Pkh1 and Pkh2 are involved in eisosome phosphorylation. However, there is no information indicating the existence of phosphatase system involved in eisosome's dephosphorylation.
We measured the dynamics of eisosome de-phosphorylation. First of all, we noticed that eisosomes of the MATa (BY4741) cells disassemble in response to the treatment with the alphamating pheromone (α-factor). At the same time, treatment of the control of MATα (BY4742) cells with α-factor did not affect the distribution of Eisosome particles ( Supplementary Fig. 13A and 13B). Analysis of the mass spectra of Pil1 and Lsp1 purified from α-factor treated or nontreated haploid MATa BY4741 revealed a number of striking differences (Supplementary Fig.   13C and 13D). The abundance of a Pil1 phosphorylated peptide detected in the mass spectra at  Supplementary Figs. 14,15). Drastic increase of the phosphorylation level after the treatment of cells with alpha-factor was noticed only at the reported phosphorylation sites. However, it was recently shown that eisosomes can be phosphorylated at the larger number of sites [18].
What is the dynamics of de-phosphorylation of the detected sites? We have sampled yeast cells at 0, 20, 40, 60, and 80 minutes after release from the alpha-factor for fluorescence microscopy and mass spectrometry experiments. Surprisingly, a substantial portion of Pil1 and Lsp1 proteins remain in the cytoplasm of the cells for almost one hour after removing the pheromone. After one hour, the fluorescence of the cytoplasmic fraction of the protein quickly, within 10-15 minutes, dropped to the level observed in the cells with intact eisosomes (Fig. 6A). This process well correlates with de-phosphorylation of Pil1 and Lsp1. Analysis of the mass spectra of the affinity purified proteins indicates that the abundance of the peptides phosphorylated at S230 and T233, also drops abruptly at ~60-80 min after release from the α-factor block ( Fig. 6C and 6D).
Noticeably, the relative intensities of gel bands corresponding to the phosphorylated forms of the proteins also decrease at the same time.
The observed dynamics revealed a rather abrupt change in phosphorylation level of both proteins at around 60 min of cell cycle progression. We speculate that some unknown phosphatase is activated at that stage of the cell cycle, which dephosphorylates Pil1 and Lsp1. Intriguingly, this by guest on May 8, 2020 process coincides with the biogenesis of new eisosomes particles [34], the process in which new eisosomes are created in the new budding cells. It is tempting to hypothesize that both processes, biogenesis and protein accumulation in the cytoplasm of the cells, are related and controlled by reversible protein phosphorylation and dephosphorylation. Identification of a phosphatase system and the mechanisms of its activation could be the goal for the future projects.
Using an in vivo mutagenesis technique [24], we produced several yeast cell strains containing a combination of the different mutant versions of Pil1 and Lsp1 proteins tagged with fluorescent and affinity probes (see the complete, 3-by-3, mutation matrix in Supplementary Fig. 16). Figure 6E shows images of Pil1 and Lsp1 co-localization the cells from the diagonal of the mutation matrix. Noticeably, simultaneous mutations of S230 and T233 in both proteins to either to aspartic acid or alanine residues produced enhanced phenotypes. Mutations S230D and T233D caused partial disassembly of eisosome particles, or, perhaps, inability to assemble eisosomes particles during eisosome biogenesis causing accumulation of the mutant proteins in the cytoplasm of the cells. Mutations S230A and T233A caused formation of fewer and brighter eisosomes as was described previously by Walther et. al [18]. They showed that mutations of residues S45, S59, S230, and T233 to either aspartic acid or alanine residues in Pil1-GFP along have a profound effect on the integrity of eisosomes. Our experiments revealed the synergetic effect of simultaneous mutations Pil1-GFP-3xFlag-6xH and Lsp1-mCherry-3xMyc-8xH proteins, which had a stronger effect on eisosomes then the effects caused by mutation of the sites in a single protein (See Fig. 6E and Supplementary Fig.16).
by guest on May 8, 2020

Discussion
We have combined fluorescence microscopy and mass spectrometry techniques for studying the composition and dynamic properties of protein complexes in the cells. To combine both techniques we designed and tested a variety of modular tags containing a fluorescent protein for visualization and two small epitope tags for two-step affinity purification. The modular construction of the tag allowed us to decouple requirements to the fluorescence and affinity probes, and optimize the performance of each module independently for fluorescence microscopy and affinity purification experiments (see Table 1). However, several modular tags exhibited strong coupling causing compromised performance of either fluorescent protein or an affinity tag. It is one of our future goals to explore the cause and possible applications of this phenomenon.
Our results confirm the usefulness of the produced reagents and the combined fluorescence microscopy/mass spectrometry approach for studying composition and dynamic properties of eisosomes [17,18,34]. The designed method enabled us to find the new protein components of eisosomes and gain the new insights into the molecular mechanisms regulating eisosome assembly and disassembly by reversible phosphorylation and de-phosphorylation. In our future work, we will investigate the applicability of the method for studying proteins from different organisms, localized to different compartments and present at low amounts in the cells. Our recent results with yeast anaphase promoting complexes (APC) [20] (~5000-1000 copies per cell [31]) and the securin-separase complexes [33] (~ 100-10 copies per cell, unpublished data) indicate a significant potential of the developed modular tags for studying the proteins exhibiting a dynamic pattern of localization and which are present at exceedingly low abundance in the cell.