Cell-specific Labeling Enzymes for Analysis of Cell–Cell Communication in Continuous Co-culture*

We report the orthologous screening, engineering, and optimization of amino acid conversion enzymes for cell-specific proteomic labeling. Intracellular endoplasmic-reticulum-anchored Mycobacterium tuberculosis diaminopimelate decarboxylase (DDCM.tub-KDEL) confers cell-specific meso-2,6-diaminopimelate-dependent proliferation to multiple eukaryotic cell types. Optimized lysine racemase (LyrM37-KDEL) supports D-lysine specific proliferation and efficient cell-specific isotopic labeling. When ectopically expressed in discrete cell types, these enzymes confer 90% cell-specific isotopic labeling efficiency after 10 days of co-culture. Moreover, DDCM.tub-KDEL and LyrM37-KDEL facilitate equally high cell-specific labeling fidelity without daily media exchange. Consequently, the reported novel enzyme pairing can be used to study cell-specific signaling in uninterrupted, continuous co-cultures. Demonstrating the importance of increased labeling stability for addressing novel biological questions, we compare the cell-specific phosphoproteome of fibroblasts in direct co-culture with epithelial tumor cells in both interrupted (daily media exchange) and continuous (no media exchange) co-cultures. This analysis identified multiple cell-specific phosphorylation sites specifically regulated in the continuous co-culture. Given their applicability to multiple cell types, continuous co-culture labeling fidelity, and suitability for long-term cell–cell phospho-signaling experiments, we propose DDCM.tub-KDEL and LyrM37-KDEL as excellent enzymes for cell-specific labeling with amino acid precursors.

Conceptually, cell-specific labeling with amino acid precursors is an extremely powerful approach for proteomic analysis of long-term cell-cell communication. The approach has broad applications in cell biology and facilitates the investigation of previously inaccessible processes. However, the nascent methodology is restricted by suboptimal amino acid processing enzymes that hamper the labeling fidelity, biological significance, and widespread adoption of CTAP. For example, Gauthier et al. (4) note that extracellular amino acid converting enzymes substantially reduce long-term cell-specific labeling fidelity. As a result, daily media exchange is required in order to achieve cell-specific labeling efficiencies of ϳ80%. The constant removal of conditioned media might undermine the biological significance of continuous cell-cell communication because of the "interrupted" co-culture environment. Moreover, although Gauthier et al. (4) employed DDC from Arabidopsis thaliana (DDC A.tha ), the authors note that several cell lines ectopically expressing this enzyme fail to grow efficiently on DAP. As a result, DDC A.tha is suboptimal for the widespread adoption of cell-specific labeling with amino acid precursors.
Given the broad potential application of cell-specific labeling with amino acid precursors, we sought to develop an optimal enzyme pairing capable of conferring high-fidelity, cell-specific, isotopic labeling to multiple cell types. Here we report the screening, engineering, and characterization of optimized DDC and Lyr enzymes for use in cell-specific labeling with amino acid precursors.
Labeling Efficiency-For the triple-labeled comparison between L-lysine and D-lysine, C3H10T1/2 ϩLyr M37-KDEL cells were grown for five passages on 2.5 mM light, medium (ϩ4 Da), or heavy (ϩ8 Da) L-lysine or D-lysine. For MDA-MB-231 ϩDDC M.tub-KDEL /C3H10T1/2 ϩLyr M37-KDEL co-cultures, cells were seeded at 5 ϫ 10 5 cells per cell type in a 10-cm dish, grown in the presence of 2.5 mM heavy (ϩ8 Da) D-Lysine and 5 mM DAP for 10 days, and then separated via FACS (see below). Cells were washed in PBS, lysed in PLC buffer, and resolved via SDS-PAGE, and gel bands were excised. Proteins were reduced with 10 mM dithiothreitol, alkylated with 50 mM iodoacetamide, and digested with 50 ng of trypsin (V5111, Promega). Peptides were eluted in 5% trifluoroacetic acid (TFA), 50% acetonitrile, lyophilized, and resuspended in 0.1% TFA. Following LC-MS/MS (Orbitrap Velos) (see details below), each sample was searched for all three isotopomeric versions of lysine (Mascot, Matrix Sciences, London, UK) in Protein Discoverer 1.4 (Thermo Scientific). Raw intensity values from each label were summed, and the percentage contribution of each label was calculated per peptide.
Flow Cytometry-MDA-MB-231/C3H10T1/2 co-cultures were harvested using trypsin, and live cells were discerned using 0.2 g/ml DAPI (Invitrogen). Cells were then sorted on a FACSAria (BD Biosciences) according to their GFP and RFP expression. The mCherry channel was used for better separation of the RFP-expressing C3H10T1/2 cells. Analyses were performed using the software FlowJo (Tree Star Inc., Ashland, OR).
Selected Reaction Monitoring-To empirically identify proteotypic Lyr/DDC peptides suitable for relative isotopic quantification, we transfected C3H10T1/2 cells with Lyr M37-KDEL , DDC M.tub-KDEL , or an empty vector control; lysed them in PLC buffer; and resolved them via SDS-PAGE. Bands were excised between 35 and 50 kDa and digested as described above. Peptides were eluted in 5% TFA, 50% acetonitrile, lyophilized, and resuspended in 0.1% TFA. Samples were analyzed on a TSQ Vantage triple quadruple mass spectrometer (Thermo Scientific) coupled to a NanoLC-Ultra 1D (Eksigent, Dublin, CA). Reversed-phase chromatographic separation was performed on an Acclaim PepMap100 C18 Nano-Trap Column (100 m inner diameter by 2 cm packed with C18 (5-m bead size, 100 Å)) (Thermo Scientific) and an NTCC-360 packed tip column (75 m inner diameter, 15 cm, 3-m particle size) (Nikkyo Technos, Tokyo, Japan) with a 30-min linear gradient of 5% to 50% solvent B (100% acetonitrile ϩ 0.1% formic acid). The TSQ Vantage was operated with a Q1 unit resolution of 0.7 full width at half-maximum, a Q3 of 0.7 full width at half-maximum, an ion spray voltage of 2200 V, and a capillary inlet temperature of 270°C. Peptide fragmentation was carried out in Q2 at 1.5 millitorr, and collision energies for each peptide were predicted (7). Each selected reaction monitoring (SRM) transition had a minimum dwell time of 20 ms, with cycle times of 1.2 s. The raw data files were produced in Xcalibur 2.1 (Thermo Scientific), and all data were processed using Skyline 2.1 (8). The three most intense peptides from each enzyme were selected for use in relative isotopic protein quantification (supplemental Table S1). Representative product ion chromatograms for these peptides are shown in supplemental Fig. S5.
To discern cell-specific labeling fidelity, C3H10T1/2 cells stably transfected with Lyr M37-KDEL and MDA-MB-231 cells stably transfected with DDC M.tub-KDEL were seeded at 5 ϫ 10 5 cells per cell type in a 10-cm dish and directly co-cultured for 7 days. Each day a population of cells were lysed and tryptic peptides were prepared as described above. Unscheduled SRM for both light and heavy (ϩ8 Da) Lyr M37-KDEL and DDC M.tub-KDEL lysine-containing peptides was performed. The top two transitions from each peptide were selected for relative isotopic quantification of each enzyme, light and heavy transitions were summed, and the heavy percentage was calculated per peptide across technical duplicates.
Discovery LC-MS/MS (Data-dependent Acquisition)-All samples were run on an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific) coupled to a NanoLC-Ultra 2D (Eksigent). Reverse-phase chromatographic separation was performed on a 100 m inner diameter ϫ 20 mm trap column packed in-house with C18 (5-m bead size, Reprosil-Gold, Dr Maisch, Ammerbuch-Entringen, Germany) and a 75 m inner diameter ϫ 30 cm column packed in-house with C18 (5-m bead size, Reprosil-Gold, Dr Maisch) using a 120-min linear gradient of 0% to 50% solvent B (100% acetonitrile ϩ 0.1% formic acid) against solvent A (100% H 2 O ϩ 0.1% formic acid) with a flow rate of 300 nL/min. The mass spectrometer was operated in the data-dependent mode to automatically switch between Orbitrap MS and MS/MS acquisition. Survey full-scan MS spectra (m/z 375-2000) were acquired in the Orbitrap with a resolution of 60,000 at m/z 400 and a Fourier transform target value of 1 ϫ 10 6 ions. The 20 most abundant ions were selected for fragmentation using collision-induced dissociation and dynamically excluded for 8 s. For phosphopeptide samples, the 10 most abundant ions were selected for fragmentation using higher-energy collisional dissociation and scanned in the Orbitrap at a resolution of 7500 at m/z 400. Selected ions were dynamically excluded for 8 s. For accurate mass measurement, the lock mass option was enabled using the polydimethylcyclosiloxane ion (m/z 445.120025) as an internal calibrant. For peptide identification, raw data files produced in Xcalibur 2.1 (Thermo Scientific) were processed in Proteome Discoverer 1.4 (Thermo Scientific) and searched against Swiss-Prot human (20012 02 release, 20,662,136 entries) and mouse (2011 03 release, 15,082,690 entries) databases using Mascot (v2.2). Searches were performed with a precursor mass tolerance of 10 ppm, a fragment mass tolerance of 0.5 Da, and a maximum of two missed cleavages. Static modifications were limited to carbamidomethylation of cysteine, and variable modifications used were oxidation of methionine; deamidation of asparagine/glutamine; isotopomeric labeled lysine (ϩ4 and ϩ8); and phosphorylation of serine, threonine, and tyrosine residues. Peptides were further filtered using a Mascot significance threshold of Ͻ0.05, a peptide ion score of Ͼ20, and a false discovery rate of Ͻ0.01 (evaluated by Percolator (10)). Phospho-site localization probabilities were calculated with phosphoRS 3.1 (11). For relative phosphopeptide quantification, heavy/medium ratios were calculated by Proteome Discoverer 1.4 (normalized to protein median). Only lysine-containing peptides were included in the quantitative analysis.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (12) with the dataset identifiers PXD000870 (labeling efficiency) and PXD000871 (cell-specific phosphoproteomics).

Improved DDC for Cell-specific Labeling with Amino Acid
Precursors-DDC converts DAP into L-lysine (5). Currently, only A. thaliana (thale cress) DDC (DDC A.tha ) has been suc-cessfully reported to confer DAP-dependent growth in coculture (4). However, only one cell line has been successfully established with DAP-dependent growth, and the authors note that several other cell lines do not proliferate efficiently with DDC A.tha . Moreover, there is currently no empirical evidence that DDC A.tha is retained within eukaryotic cells, which could lead to compromised labeling fidelity in co-culture. Given the broad potential applications of cell-specific labeling with amino acid precursors, a DDC enzyme that confers efficient, intracellular, DAP-dependent growth in a variety of cell lines is essential.
Given the poor activity of DDC A.tha , we hypothesized that DDC enzymes sourced from alternative species may confer improved DAP-dependent growth to eukaryotic cells. To investigate this, we compared DDC A.tha with orthologous DDC from M. jannaschii (a thermophilic archaea) (DDC M.jan ), H, pylori (a microaerophilic stomach bacterium) (DDC H.pyl ), and M. tuberculosis (a pathological mycobacterium) (DDC M.tub ) (Fig. 1A). These enzymes were screened because of their distinct dimerization profiles and active site loop structures (13) (supplemental Fig. S2). The orthologous enzymes all expressed to similar levels (Fig. 1B), and despite the absence of a predicted signal peptide (supplemental Fig. S3A) (SignalP 4.1 (14)), small traces of each DDC were detected in conditioned media. Notably, DDC M.tub conferred dramatically improved DAP-dependent growth relative to all other enzymes. As a result, DDC M.tub was advanced for further optimization and characterization.
As intracellular DDC is required to ensure true cell-specific labeling, we anchored DDC M.tub in the ER by adding a Cterminal KDEL retention sequence (15) (DDC M.tub-KDEL ). DDC M.tub-KDEL exhibited reduced extracellular levels relative to wild-type DDC M.tub (Fig. 1C) and conferred excellent DAPdependent growth (Fig. 1D). DDC M.tub was found throughout the cell, whereas DDC M.tub-KDEL was retained within the ER (Fig. 1E). These results confirm that DDC can function efficiently within the eukaryotic ER. Stably transfected DDC M.tub-KDEL cells conferred proliferation comparable to that with Llysine when grown on concentrations of DAP greater than 2.5 mM and did not grow on D-lysine at any concentration (Fig. 1F).
To ensure that DDC M.tub-KDEL did not induce an amino acid stress response, stable DDC M.tub-KDEL cells were grown on titrated concentrations of DAP, and the phosphorylation of eIF2␣ on serine 51 (pS51 eIF2␣) was used as an early translational arrest stress marker (16). In accordance with our earlier proliferation studies, DDC M.tub-KDEL conferred non-stressed cell growth at DAP concentrations between 1 and 10 mM (Fig. 1G).
To test their applicability across multiple cell types, we transfected DDC A.tha and DDC M.tub-KDEL into eight eukaryotic cell lines and measured cell viability following growth on 5 mM DAP (Fig. 1H). Relative to cells grown on 10 mM L-lysine, DDC A.tha transfected cells typically achieved between 15% and 40% proliferation when grown on 5 mM DAP. In contrast, all cell lines transfected with DDC M.tub-KDEL achieved between 80% and 100% proliferation. Moreover, cells transfected with DDC A.tha failed to grow beyond passage 2, whereas all cells transfected with DDC M.tub-KDEL proliferated on DAP continuously. These observations suggest that DDC M.tub-KDEL is a broadly applicable DAP-processing enzyme for cell-specific labeling with amino acid precursors.
Although DDC from M. tuberculosis conferred superior DAPdependent proliferation to multiple eukaryotic cells, the mechanism underlying this improvement is currently unclear. To investigate whether this phenotype was unique to DDC M.tub or represented a general feature of Mycobacterium DDC enzymes, DDC from the closely related M. leprae (DDC M.lep ) and M. avium (DDC M.avi ) were also expressed in eukaryotic cells. Interestingly, all Mycobacterium DDC enzymes substantially outperformed DDC A.tha (Fig. 1I). Mycobacterium DDC enzymes contain a unique cysteine residue at position 93 that forms a disulfide bond with a cysteine common to all other DDCs at DDC M.tub position 375 (supplemental Fig. S2). As DDC is thought to function as a homodimer (17), we hypothesized that the collectively improved performance of Mycobacterium DDC enzymes could be due to their common intermonomer disulfide bridges. To test this hypothesis, we compared wild-type DDC M.tub to a Cys93Ala mutant DDC M.tub . The DDC M.tub Cys93Ala point mutation severely impaired DDC activity (Fig. 1J) independent of the expression level (Fig.  1K). This observation implies that the intermonomer disulfide bridges found in Mycobacterium DDC enzymes are responsible for their improved activity in eukaryotic cells.
Given its dramatically improved catalytic activity, intracellular localization, and applicability to multiple eukaryotic cells, we propose DDC M.tub-KDEL as an optimal disulfide-homodimer enzyme for cell-specific labeling with amino acid precursors.
Improved Lyr for Cell-specific Labeling with Amino Acid Precursors-Lyr catalyzes the conversion of D-lysine into Llysine (6). P. mirabilis wild-type Lyr (Lyr WT ) contains a putative signal peptide (amino acids 1-30) that has been predicted to facilitate Lyr secretion from eukaryotic cells (supplemental Fig. S3B) (SignalP 4.1 (14)). Gauthier et al. (4) note that wildtype P. mirabilis Lyr is prolifically secreted from eukaryotic cells. Extracellular Lyr converts labeled D-lysine to labeled L-lysine in conditioned media and severely compromises coculture labeling efficiency. Consequently, growth medium has to be changed daily in order to maintain cell-specific labeling efficiency, which undermines continuous co-culture cell-cell communication. As with DDC, a truly intracellular Lyr enzyme is required for high-fidelity cell-specific labeling of continuous co-cultures.
Structural studies suggest P. mirabilis Lyr contains a globular catalytic core (amino acids 37-407) distinct from the putative signal peptide (18). In an attempt to limit extracellular Lyr (while retaining the catalytic activity), we removed amino acids 1-36 from the enzyme (Lyr M37 ) and added a C-terminal KDEL ER retention motif (Lyr M37-KDEL ) ( Fig. 2A). When transfected into C3H10T1/2 (C3) eukaryotic fibroblasts, Lyr M37 and Lyr M37-KDEL enzymes conferred improved proliferation on lower concentrations of D-lysine relative to Lyr WT (Fig. 2B). Only full-length Lyr WT protein could be detected in conditioned media, whereas a smaller, cleaved version of Lyr WT was found in cell lysates. Lyr M37 protein was largely located in cell lysates, although small traces of Lyr M37 could still be detected in conditioned media (Fig. 2C). Thus, although the signal peptide appears to enhance Lyr secretion, overexpressed Lyr can still escape the cell of origin independently of the signal peptide. In contrast, Lyr M37-KDEL was exclusively located in cell lysates and could not be detected in conditioned media. Lyr WT located to the ER (for secretion), Lyr M37 was indiscriminately dispersed through the cell, and Lyr M37-KDEL resided within the ER (for retention) (Fig. 2D).
Collectively, these observations confirm that P. mirabilis Lyr activity is independent of its putative signal peptide and can function in the eukaryotic ER. Thus, unlike Lyr WT (or "modified" Lyr (4)), Lyr M37-KDEL is a truly intracellular D-lysine conversion enzyme.
Cells stably transfected with Lyr M37-KDEL achieved proliferation comparable to that with L-lysine when grown on concentrations of D-lysine greater than 1 mM (Fig. 2E). To ensure that Lyr M37-KDEL did not induce an amino acid stress response, Lyr M37-KDEL cells were grown on titrated concentrations of D-lysine, and the early translational arrest stress marker pS51 eIF2␣ (16) was quantified (Fig. 2F). In accordance with earlier proliferation studies, Lyr M37-KDEL conferred non-stressed cell growth at D-lysine concentrations between 1 and 10 mM. Stable Lyr M37-KDEL cells proliferated on multiple D-lysine heavy isotopes (ϩ4 Da and ϩ8 Da) and achieved labeling efficiencies comparable to those of SILAC when grown in monoculture (Fig. 2G). Given its catalytic proficiency and improved intracellular localization, we propose Lyr M37-KDEL as an excellent D-lysine processing enzyme for cell-specific labeling with amino acid precursors.
Lyr M37-KDEL and DDC M.tub-KDEL Isotopic Labeling Stability in Interrupted and Continuous Co-cultures-The primary application of cell-specific labeling with amino acid precursors is the study of continuous cell-cell communication in proliferating co-cultures. Previous CTAP methodology requires daily media exchange to maintain labeling stability and thus does not permit the study of continuous cell-cell communication.
To examine the ability of our improved cell-specific labeling enzymes to support studies of continuous cell-cell communication, we first tested the influence of continuous versus interrupted co-culture on cell-specific labeling fidelity.
To test the long-term labeling stability of the optimized enzymes in co-culture, we transfected Lyr M37-KDEL into RFP ϩ C3 fibroblasts and DDC M.tub-KDEL into GFP ϩ MDA-MB-231 epithelial tumor cells. C3 ϩLyr M37-KDEL cells were then labeled as heavy (ϩ8 Da) and directly co-cultured with light MDA-MB-231 ϩDDC M.tub-KDEL cells in the presence of heavy D-lysine and light DAP. To investigate the influence of media exchange on labeling efficiency, we maintained cells as either "interrupted" (with daily media exchange) or "continuous" (without media exchange) co-culture. After 10 days, cells were separated into monoculture via FACS (using RFP and GFP) (supplemental Fig. S4), lysed, and resolved via SDS-PAGE, and tryptic peptides were analyzed via LC-MS/MS (Fig. 3A). Light and heavy precursor ions were quantified for each peptide, and a mean labeling efficiency percentage was calculated for each population. Critically, DDC M.tub-KDEL and Lyr M37-KDEL conferred ϳ90% cell-specific labeling fidelity in both interrupted (91.2% light/90.8% heavy) and continuous (89.6% light/88.6% heavy) 10-day co-cultures (Fig. 3B).
As DDC M.tub-KDEL is specifically expressed in light MDA-MB-231 cells and Lyr M37-KDEL is specifically expressed in heavy C3 cells, the isotopic lysine composition of each enzyme provides a discrete proxy for general cell-specific labeling efficiency. To compare the temporal labeling stability of this CTAP method with that of traditional SILAC labeling, we combined heavy C3 ϩLyr M37-KDEL cells with light MDA-MB-231 ϩDDC M.tub-KDEL cells and co-cultured for 7 days in either L-lysine (ϩ0 Da) ϩ L-lysine (ϩ8 Da) (SILAC amino acids) or DAP (ϩ0 Da) ϩ D-lysine (ϩ8 Da) (CTAP amino acid precursors). Temporal labeling efficiencies for the ectopic Lyr M37-KDEL and DDC M.tub-KDEL enzymes were subsequently determined for each co-culture using SRM (Fig. 3C) (supplemental Fig. S5). As expected, light and heavy L-lysine was metabolized by both cell types in the SILAC co-culture and resulted in a rapid loss of cell-specific labeling fidelity (1 to 2 days) (7 days ϭ 47.3% light DDC M.tub-KDEL /44.3% heavy Lyr M37-KDEL ). These observations confirm that SILAC labeling is unsuitable for long-term analysis of cell-specific proteomes in co-culture. In contrast, both interrupted (7 days ϭ 9.5% light DDC M.tub-KDEL /87.0% heavy Lyr M37-KDEL ) and continuous (7 days ϭ 10.1% light DDC M.tub-KDEL /87.1% heavy Lyr M37-KDEL ) CTAP conditions conferred stable cell-specific labeling fidelity throughout their 7-day co-cultures. These results further confirm that optimized DDC M.tub-KDEL and Lyr M37-KDEL enzymes are suitable for the long-term cell-specific quantitative proteomic analysis of uninterrupted, continuous co-cultures.
Phosphoproteomic Analysis of Cell-specific Signaling in Continuous Co-culture-As DDC M.tub-KDEL and Lyr M37-KDEL support the long-term cell-specific labeling stability of both interrupted and continuous co-cultures, we next sought to investigate the cell-signaling implications of daily media exchange. First, we monitored the relative abundance of cellsignaling factors present in conditioned media from both interrupted and continuous 5-day co-cultures. In contrast to interrupted co-cultures, multiple growth factors and cytokines accumulated in continuous co-cultures (Fig. 4A). This soluble signaling potential is not established in interrupted co-cultures and highlights the need for new technologies to study cell-cell communication in co-culture. To investigate the cellspecific signaling consequences of the two approaches, we compared the phosphoproteomic profile of an interrupted MDA-MB-231 ϩDDC M.tub-KDEL light/C3 ϩLyr M37-KDEL medium co-culture with a that of a continuous MDA-MB-231 ϩDDC M.tub-KDEL light/C3 ϩLyr M37-KDEL heavy co-culture (Fig.  4B). An interrupted versus interrupted experiment was also performed to control for stochastic phospho-signaling differences that might have arisen during the 5-day co-culture. Cell-specific phosphoproteomic analysis of C3 cells revealed substantial signaling differences between the interrupted and continuous co-cultures (Fig. 4C) (supplemental Table S2).
As these phosphorylation events would not have been detected using existing interrupted co-culture methodology, this observation suggests that novel biological questions now can be addressed through the enhanced CTAP technology. Given their improved intracellular localization, applicability to multiple cell types, and robust cell-specific labeling stability, we propose DDC M.tub-KDEL and Lyr M37-KDEL as excellent enzymes for the phosphoproteomic analysis of long-term continuous cell-cell communication.

DISCUSSION
Cell-specific labeling with amino acid precursors permits discrete proteomic labeling of proliferating cell types in coculture. In order for this technology to function successfully, two discrete enzymes must confer distinct cell-specific L-lysine precursor proliferation to eukaryotic cells. In this study, we proposed novel lysine-precursor converting enzymes for efficient, stable, and continuous cell-specific labeling with amino acid precursors. Each day, a cell population was lysed and light and heavy versions of three proteotypic peptides from each enzyme were analyzed via selected reaction monitoring. As each enzyme is only ectopically expressed in one cell type, the light:heavy ratio for each peptide provides a discrete proxy for cell-specific isotopic labeling efficiency. SILAC co-cultures rapidly lost cell-specific labeling fidelity (Ͻ2 days), as L-lysine can be metabolized by both cell types. In contrast, both interrupted and continuous CTAP environments maintained cell-specific labeling fidelity over a 7-day co-culture. Error bars ϭ S.E. of technical replicates (n ϭ 3).
A primary finding of this study is that DDC enzymes from the Mycobacterium genus confer excellent DAP-dependent proliferation to eukaryotic cells. Although this study identified DDC M.tub as the optimal enzyme, DDC M.lep and DDC M.avi also conferred improved DAP-dependent growth to eukaryotic cells relative to alternative species. This superior performance  Table S2. appears to be facilitated by an intermonomer disulfide bond that stabilizes the active DDC homodimer in Mycobacterium DDC enzymes.
Although DDC M.tub-KDEL provided excellent DAP conversion for cell-specific labeling with amino acid precursors, we note that growing stable DDC cells on DAP (EC 50 ϭ 0.6 mM) is 15-fold less efficient than growing cells on L-lysine (EC 50 ϭ 0.04 mM). It is currently unclear why higher concentrations of DAP are required. Previous work has suggested that the rate of DAP uptake could be rate-limiting (19). However, as we observed identical performance between DDC M.tub (un-localized) and DDC M.tub-KDEL (ER-localized), DAP localization from the eukaryotic cytoplasm to the ER does not appear to be a rate-limiting step. Future efforts to improve DDCϩDAP efficiency could focus on improving DAP uptake into the eukaryotic cytoplasm (e.g. by modulating amino acid transceptors (20)). Another explanation could be poor DDC enzyme performance. Although DDC M.tub substantially outperformed all other DDC enzymes tested, future efforts to improve DAP-dependent proliferation could also expand beyond the multiple DDC orthologs described here. However, a C-terminal KDEL motif should be added to ensure intracellular enzyme retention.
Our data confirm that P. mirabilis Lyr effectively converted D-lysine to L-lysine in eukaryotic cells. Moreover, we show that by removing the putative signal peptide (M37) and anchoring the enzyme in the ER (via KDEL motif), Lyr M37-KDEL can act as a suitable intracellular enzyme for cell-specific labeling with amino acid precursors. Lyr M37-KDEL confers efficient cell proliferation on concentrations of D-lysine greater than 1 mM and can convert isotopically labeled medium (ϩ4 Da) and heavy (ϩ8 Da) D-lysine into labeled L-lysine. As a result, when combined with DDC M.tub-KDEL , Lyr M37-KDEL can be used for triple-labeled comparisons between co-cultures.
Although Lyr M37-KDEL provided sufficient D-lysine conversion for cell-specific labeling, we note that growing Lyr cells on D-lysine (EC 50 ϭ 0.38 mM) is around 10-fold less efficient than growing cells directly on L-lysine (EC 50 ϭ 0.03 mM). Again, it is currently unclear why higher concentrations of D-lysine are required. As with DDC, one explanation could be enzyme performance. Although P. mirabilis Lyr is clearly suitable for cell-specific labeling (as shown in Figs. 3B and 3C), it is possible that orthologous Lyr enzymes could confer more efficient D-lysine-dependent growth. Our success with orthologous DDC screening suggests that future studies might benefit from testing enzymes from alternative species. However, if undertaking such a screen, one should consider removing any putative signal sequences and adding a C-terminal KDEL motif for intracellular retention. Another explanation for the discrepancy might be poor intracellular import of D-lysine. However, as we observed substantially improved activity of intracellular Lyr M37 and Lyr M37-KDEL (Fig. 2B) relative to extracellular Lyr WT , our data suggest that intracellular D-lysine-to-L-lysine conversion is actually more effective than extracellular conversion. Crucially, when the DAP concentra-tion was Ͼ5 mM and that of D-lysine was Ͼ2.5 mM, neither DDC M.tub-KDEL nor Lyr M37-KDEL experienced growth-limiting proliferation relative to traditional growth conditions. Thus, as long as sufficient DAP and D-lysine are supplied in the growth media, the performance of DDC M.tub-KDEL or Lyr M37-KDEL should remain suitable for cell-specific labeling with amino acid precursors.
Given the broad applicability of the revised enzymes, the reported methodology now extends the study of cell-cell communication to a diverse selection of different cell types. To this end, we have deposited DDC M.tub-KDEL and Lyr M37-KDEL expression constructs in Addgene (DDC M.tub-KDEL : 51529 and Lyr M37-KDEL : 51530) for widespread application by the proteomic community.
The optimized enzyme pairing described here provides stable proteomic labeling of specific cell types in continuous co-culture. However, we did observe a modest reduction in labeling efficiency for co-cultures relative to pre-labeled monocultures. Temporal analysis of labeling fidelity indicated that this decrease occurred during the first 2 days of co-culture. The small reduction in labeling efficiency could have been due to multiple factors. For example, secreted proteins can be taken up by cells of a different label during normal paracrine signaling (21). Moreover, although we experienced decreased extracellular enzyme secretion from live cells, it is possible that dead cells could release active enzymes into co-culture media. Furthermore, gap-junctions between adjacent cells can transport small molecules (Ͻ1 kDa) (22) and could theoretically leak labeled amino acids between confluent co-cultures. Such events could represent a biological limit to longterm cell-specific labeling in co-culture, and interfering with these processes (i.e. blocking gap-junctions) might undermine the biological significance of co-culture experiments. However, as illustrated by our triple-labeled phosphoproteomic co-culture comparison (Figs. 4B and 4C), the superior labeling stability of Lyr M37-KDEL and DDC M.tub-KDEL capably enables high-fidelity cell-specific experiments from continuous co-cultures.