Quantitative Proteomic Analyses of Human Cytomegalovirus-Induced Restructuring of Endoplasmic Reticulum-Mitochondrial Contacts at Late Times of Infection*

Endoplasmic reticulum-mitochondrial contacts, known as mitochondria-associated membranes, regulate important cellular functions including calcium signaling, bioenergetics, and apoptosis. Human cytomegalovirus is a medically important herpesvirus whose growth increases energy demand and depends upon continued cell survival. To gain insight into how human cytomegalovirus infection affects endoplasmic reticulum-mitochondrial contacts, we undertook quantitative proteomics of mitochondria-associated membranes using differential stable isotope labeling by amino acids in cell culture strategy and liquid chromatography-tandem MS analysis. This is the first reported quantitative proteomic analyses of a suborganelle during permissive human cytomegalovirus infection. Human fibroblasts were uninfected or human cytomegalovirus-infected for 72 h. Heavy mitochondria-associated membranes were isolated from paired unlabeled, uninfected cells and stable isotope labeling by amino acids in cell culture-labeled, infected cells and analyzed by liquid chromatography-tandem MS analysis. The results were verified by a reverse labeling experiment. Human cytomegalovirus infection dramatically altered endoplasmic reticulum-mitochondrial contacts by late times. Notable is the increased abundance of several fundamental networks in the mitochondria-associated membrane fraction of human cytomegalovirus-infected fibroblasts. Chaperones, including HSP60 and BiP, which is required for human cytomegalovirus assembly, were prominently increased at endoplasmic reticulum-mitochondrial contacts after infection. Minimal translational and translocation machineries were also associated with endoplasmic reticulum-mitochondrial contacts and increased after human cytomegalovirus infection as were glucose regulated protein 75 and the voltage dependent anion channel, which can form an endoplasmic reticulum-mitochondrial calcium signaling complex. Surprisingly, mitochondrial metabolic enzymes and cytosolic glycolytic enzymes were confidently detected in the mitochondria-associated membrane fraction and increased therein after infection. Finally, proapoptotic regulatory proteins, including Bax, cytochrome c, and Opa1, were augmented in endoplasmic reticulum-mitochondrial contacts after infection, suggesting attenuation of proapoptotic signaling by their increased presence therein. Together, these results suggest that human cytomegalovirus infection restructures the proteome of endoplasmic reticulum-mitochondrial contacts to bolster protein translation at these junctions, calcium signaling to mitochondria, cell survival, and bioenergetics and, thereby, allow for enhanced progeny production.

mised individuals (1)(2)(3). Severe manifestations of HCMV infection in these patients include interstitial pneumonia, hepatitis, meningoencephalitis, gastrointestinal disease, myocarditis, bone marrow suppression, and retinitis. Because of its impact on human suffering and the health care system, the Institute of Medicine has assigned the highest priority to HCMV vaccine development (4).
Like that of most viruses, HCMV growth in permissive cells requires a large supply of energy and substrates from the infected cell (5)(6)(7). To obtain these, HCMV infection causes a global metabolic up-regulation of central carbon metabolic flux (5,6). In particular, HCMV infection induces increased glucose and glutamine consumption during permissive infection of human foreskin fibroblasts (HFFs) (5)(6)(7). Increased aerobic glycolysis allows HCMV to use glucose biosynthetically, wherein most of the acetyl coenzyme A supports fatty acid synthesis needed for membrane formation for its progeny viruses. Thus, HCMV growth reprograms cellular mitochondrial metabolic flux and institutes its own metabolic program.
Because of its protracted growth cycle over several days, productive HCMV infection further requires antiapoptotic activity for efficient progeny production (8). To extend cell viability, HCMV encodes several products to undermine mitochondrial-mediated apoptosis. The HCMV UL36 -38 locus encodes five proteins with antiapoptotic activities, including the predominant UL37 exon 1 protein (pUL37ϫ1), also known as viral mitochondria localized inhibitor of apoptosis (vMIA), the lower abundance UL37 Medium protein and the UL37 glycoprotein, the UL36 protein and the early UL38 protein (9 -12). pUL37ϫ1 blocks mitochondrial-mediated apoptosis by translocating proapoptotic Bax to the outer mitochondrial membrane (OMM) and suppressing its proapoptotic activity (13)(14)(15). In addition, pUL37ϫ1/vMIA has been reported to alter mitochondrial ATP synthesis by decreasing the activity of the mitochondrial phosphate carrier in transfected cells (16). By early times, HCMV infection also encodes an untranslated beta 2.7 RNA that localizes to mitochondria (17,18). Beta 2.7 RNA interacts with the GRIM-19 subunit of mitochondrial complex I and prevents its re-localization from the inner mitochondrial membrane to discrete perinuclear sites (19). Thus, HCMV infection redundantly blocks mitochondrial-mediated proapoptotic signaling pathways.
Mitochondrial activities, including its metabolic pathways and induction of apoptosis, are modulated by contacts with endoplasmic reticulum (ER) subdomains, known as mitochondria-associated membranes (MAM) (20,21). Direct ER-mitochondrial contacts have been best visualized by electron tomography (22). The physical distance between the ER membrane and OMM is ϳ10 nm. The distance between the organelles is larger (ϳ25 nm) when ribosomes are found between the juxtaposed membranes. Protein tethers can be seen between the ER and OMM. Mitofusin 2 (Mfn2) (23,24), mitostatin (25), PACS-2 (26), and a calcium (Ca 2ϩ ) signaling complex (27), have been implicated in stabilizing contacts between the ER and the OMM.
Our previous work demonstrated that HCMV infection targets ER-mitochondrial contacts by trafficking of its pUL37ϫ1 and the N-terminal UL37 glycoprotein fragment (28 -34). A highly conserved UL37ϫ1 leader drives ER translocation, MAM association, and subsequent trafficking to the OMM (31,34). pUL37ϫ1 traffics from the ER, where it causes Ca 2ϩ efflux (35), associates with internal MAM lipid rafts (36) in close proximity to sigma 1 receptor (Sig-1R) (32) and translocates to the OMM where it has potent antiapoptotic activity by recruiting Bax to the OMM and inhibiting its proapoptotic activities (10,13,14,37,38). In addition to targeting the MAM by pUL37ϫ1, we found that HCMV infection increases the abundance of cytosolic glucose regulated protein 75 (GRP75), a component of the MAM Ca 2ϩ macromolecular complex (29). Thus, HCMV positions itself to control ER-mitochondrial cross-talk and thereby modulate key mitochondrial functions.
By targeting the MAM, HCMV predictably acquires another level of control over mitochondrial metabolism and mitochondrial-mediated apoptosis. To gain insight into how HCMV infection alters the functionality of ER-mitochondrial contacts to its advantage, we undertook quantitative analyses of the MAM proteome using stable isotope labeling by amino acids in cell culture (SILAC) (39). A late time of infection was chosen for these studies as HCMV infection dramatically alters metabolic flux (5)(6)(7), the cell secretome (40), widespread changes in ER Ca 2ϩ efflux (35), and a mitigated ER stress response (41) at these times; all events which are likely mediated by the MAM or to affect MAM domains. To date, no quantitative or nonquantitative proteomic analysis of the MAM fraction has yet been reported. Although proteomic analyses have identified HCMV virion components (42) and cellular proteins that interact with HCMV products (43,44) as well as defined the HCMV-induced secretome (40), this is the first quantitative proteomic analyses of HCMV-infected HFFs and the effects of virus on ER-mitochondrial contacts. Using this quantitative proteomic approach, we observed dramatic changes in the MAM proteome by late times of HCMV infection, consistent with its global reprogramming of cellular metabolism. The association of chaperones, translation and ER-translocation machineries, Ca 2ϩ signaling complexes, metabolic enzymes, and apoptotic regulatory proteins with the MAM fraction increased by late times of infection. These results suggest that HCMV infection induces a restructuring of ER-mitochondrial contacts by late times. This restructuring will predictably result in increased protein synthesis at the MAM, cell survival and bioenergetics. Further the increased abundance of mitochondrial inner mitochondrial membrane, intermembrane space and matrix proteins in the MAM fraction suggest that HCMV infection increases the stability of ER-mitochondrial contacts by late times of infection.
SILAC-Labeling of HFFs-Primary HFFs (starting at passage 5) were cultured for four cell doublings in custom-media containing 13 C 6 -Arg and 15 N 2 , 13 C 6 -Lys (SILAC-labeled) or 12 C 6 -Arg and 14 N 2 , 12 C 6 -Lys (unlabeled). HFFs were incubated with labeled medium or unlabeled medium for 3 days. Confluent cells were subcultured at ratios of 1:2 or 1:4 and fed unlabeled or labeled medium, respectively. Two roller bottles (769 cm 2 each) of SILAC-labeled HFFs were uninfected or HCMV-infected (BADwt, multiplicity of infection of 3 plaque forming units/cell and maintained in DMEM containing 2% FCS until 72 h post infection (hpi). In parallel, two roller bottles of unlabeled HFFs were uninfected or infected with HCMV as above. At 72 hpi, infected and uninfected cells were washed 6 times with sterile PBS (Invitrogen), harvested, and paired for MAM fractionation as shown in Fig. 1. In the forward experiment, SILAC-labeled, HCMV-infected cells were mixed 1:1 with unlabeled uninfected cells. In the reverse experiment, the unlabeled HCMV-infected cells were mixed 1:1 with SILAC-labeled uninfected cells.
Subcellular Fractionation and MAM Isolation-Each paired mixture was separately fractioned to isolate enriched heavy MAM from bulk ER and mitochondria as previously described (28,47). The Percoll method used for this analysis is one of the most rigorous procedures to obtain banded MAM and mitochondria. It relies on multiple fractionation steps, each used to separate organelles based upon their connectivity, size, shape, and density. Briefly, cells were homogenized and nuclei and debris were removed by pelleting. The supernatant was pelleted again to obtain crude mitochondria and interconnected MAM. The cytosol and microsomes remained soluble. The MAM and mitochondria were further separated by homogenization and resolved by banding in Percoll density gradients. Percoll gradients efficiently separate, based upon density and shape: the MAM is lighter; whereas, mitochondria are denser and band in the lower half of the gradient (see a representative separation in Fig. 1B). Banded MAM and mitochondria were obtained by side puncture of the ultracentrifuge tube from above and below the bands, respectively, and aspiration. Once extracted from the gradients, the MAM band was diluted and pelleted generating the heavy MAM, which pellets at lower centrifugal forces and is enriched in macromolecular complexes that join the ER and mitochondria (29). This banded fraction was used for the quantitative proteomic analyses. All of the fractions were resuspended in a minimal volume (50 -200 l) of sucrose homogenization medium and assayed for protein concentration using a BCA reagent kit (Pierce), according to the manufacturer's recommended protocol.
Protein Separation and Mass Spectrometry Analysis-One hundred micrograms of the heavy MAM fraction obtained from the mixture of unlabeled and SILAC-labels cells were further resolved by SDS-PAGE (Fig. 1C). The gel was stained with Bio-Safe Coomassie (Bio-Rad, Hercules, CA) and each lane was cut into 50 serial slices. Proteins in each gel slice were in-gel digested with trypsin. Concentrated peptides from each band were injected via an autosampler (6 l) and loaded onto a Symmetry C18 trap column (5 m, 300 m i.d. ϫ 23 mm, Waters) for 10 min at a flow rate of 10 l/min, 100% A. The sample was subsequently separated by a C18 reverse-phase column (3.5 m, 75 m ϫ 15 cm, LC Packings) at a flow rate of 250 nL/min using an Eksigent nano-hplc system (Dublin, CA). The mobile phases consisted of water with 0.1% formic acid (A) and 90% acetonitrile (B). A 65 min linear gradient from 5 to 60% B was employed. Eluted peptides were introduced into the mass spectrometer via a 10 m silica tip (New Objective Inc., Ringoes, NJ) adapted to a nano-elec-trospray source (ThermoFisher Scientific). The spray voltage was set at 1.2 kV and the heated capillary at 200°C. The LTQ-Orbitrap-XL (ThermoFisher Scientific) was operated in data-dependent mode with dynamic exclusion in which one cycle of experiments consisted of a full-MS in the Orbitrap (300 -2000 m/z, resolution 30,000) survey scan and five subsequent MS/MS scans in the LTQ of the most intense peaks using collision-induced dissociation with the collision gas (helium) and normalized collision energy value set at 35%.
Database Search and SILAC Ratio Measurement-For protein identification and quantification we used Integrated Proteomics Pipeline (IP2) version 1.01 software developed by Integrated Proteomics Applications, Inc. (http://www.integratedproteomics.com/). Mass spectral data were uploaded into IP2 software. Files from each lane were searched against the forward and reverse Uniprot human database (UniProt release 15.4, June 2009, 22697 forward entries) for tryptic peptides allowing one missed cleavage, and possible modification of oxidized methionine (15.99492 Da), heavy arginine (6.0204 Da) and heavy lysine (8.0142 Da). IP2 uses the Sequest 2010 (06_10_13_1836) search engine. Mass tolerance was set at Ϯ 50 ppm for MS and Ϯ 1.5 Da for MS/MS. Data were filtered based on the following Xcorr values (Xcorr Ն 1.9 for z ϭ 1, Ն 2.5 for z ϭ 2, and Ն 3.5 for z ϭ 3). These criteria provided a 3% false discovery rate. Only proteins that were identified by at least two unique peptides were retained for further quantitative analysis. Proteins that were identified with one single peptide were excluded. The 50 bands from each lane were summed in the analysis. Census software version 1.77, built into the IP2 platform (48), was used to determine the ratios of unlabeled and labeled peptide pairs using an extracted chromatogram approach. The distribution of ratios was plotted and correction factors applied to adjust for error in sample mixing. Data were checked for validity by using regression correlation better than 0.98 for each peptide pair.
Protein bands were detected using an ECL detection kit (Pierce). Each blot was stripped as previously described, and reprobed for the detection of other ER, MAM, mitochondrial markers (28,49). Blots were exposed to film (Denville Scientific, Metuchen, NJ) and the exposures were scanned using a Bio-Rad GS-800 Calibrated Densitometer and analyzed using QuantityOne Software. Digital images were generated by using Scan Wizard Pro version 1.21 and processed in Adobe Photoshop version CS5 version 12.0.3.
Confocal Microscopy-Images were acquired with a Zeiss LSM 510 confocal microscope (Intellectual and Developmental Disabilities Research Center) as described (28,30). Excitation wavelengths for the Zeiss LSM 510 microscope were 488 nm, 561 nm, and 633 nm. Images were acquired by sequential excitation through 63x (NA ϭ 1.4) objective. Post-acquisition processing was performed using Adobe Photoshop CS3.

Quantitative Proteomic Analyses of the MAM Proteome in
Uninfected and HCMV-Infected HFFs-The enriched MAM fraction can be reproducibly isolated by banding in Percoll gradients ( Fig. 1B) (47,(51)(52)(53). Using SILAC methodology, we initially identified 1719 unique proteins from proteomic analysis of the MAM fractions of forward (unlabeled uninfected cells and SILAC-labeled, HCMV-infected cells) and reverse (SILAC-labeled uninfected cells and unlabeled, HCMV-infected cells) experiments. Of these, 1254 unique proteins were identified in proteomic analyses of both forward and reverse experiments. Finally, 991 proteins were identified by at least two unique peptide hits in both proteomic sets, increasing the confidence of their authentic association with the MAM fraction. We focused our studies on these most confident 991 proteins. As expected by its identity as an ER subdomain, the MAM fraction predominantly contained proteins from the secretory apparatus (microsomes) and, to a lesser degree, nuclear proteins (nuclear), as the outer nuclear membrane is contiguous with the ER (Fig. 2). Further, consistent with its tight association with mitochondria, the next largest group of proteins was of mitochondrial origin. Surprisingly, a significant number of proteins known to be localized to the cytosol, plasma membrane, and other apparently unrelated subcellular compartments were also identified within the MAM fraction. This latter result suggests the MAM interacts with multiple cellular compartments in

FIG. 1. Experimental design for quantitative proteomics of enriched heavy MAM fraction from uninfected and HCMV-infected HFFs.
A, Forward (left) and reverse (right) SILAC approaches. HFFs were grown in light ( 12 C 6 -Arg and 12 C 6 , 14 N 2 -Lys) or heavy ( 13 C 6 -Arg and 15 N 2 , 13 C 6 -Lys) medium for four cell doublings. HFFs, grown in heavy medium, were HCMV (strain BAD wt)-infected for 72 h; whereas, HFFs, grown in light medium, were uninfected for 72 h. Prior to subcellular fractionation, two roller bottles of SILAC-labeled, HCMV-infected HFFs were mixed (1:1) with two roller bottles of unlabeled, uninfected HFFs (forward experiment). In the reverse experiment, SILAC-labeled, uninfected HFFs were mixed (1:1) with unlabeled, HCMV-infected HFFs at 72 hpi. The mixed cells were fractionated to obtain heavy MAM as previously described (28,47). Heavy MAM proteins (100 g), obtained from mixed unlabeled and SILAC-labeled cells, were resolved by SDS-PAGE. Fifty sequential slices of each lane were in-gel digested with trypsin and analyzed by LC-MS/MS analysis. B, Percoll gradients of uninfected and HCMV infected HFFs. Uninfected HFFs and HCMV-infected HFFs (ϳ5 ϫ 10 7 cells/each) were gently lysed by homogenization, pelleted by centrifugation at 10,300 ϫ g for 10 min at 4°C. The supernatant was removed and alternately processed, whereas the pellet was resuspended in 300 l of ice-cold mannitol buffer A, briefly homogenized, layered over 10 ml of a 30% Percoll suspension in mannitol buffer B, and subjected to ultracentrifugation at 95,000 ϫ g for 65 min at 4°C. Banded MAM and mitochondria were recovered from the gradient by needle extraction from above (MAM, 200 l) and below (Mito, 2 ml) the indicated points. C, SDS-PAGE of forward and reverse MAM samples. Heavy MAM from the paired forward or reverse paired HFFs were resolved by SDS-PAGE and stained by BioSafe Coomassie. addition to mitochondria and may indicate that the identified proteins dually localize in the MAM as well. For example, GRP75 localizes in the mitochondrial matrix and has been recently found as the cytosolic component of the MAM macromolecular Ca 2ϩ signaling complex (22,29).
To verify the identity of our subcellular fractions, we examined the relative abundances of known ER-mitochondrial, secretory apparatus, and glycolytic enzymes in the subcellular fractions from uninfected and from HCMV-infected HFFs (Fig.  3A). FACL4, a known MAM marker (26) was faintly detected in the MAM fraction (0.1), but dramatically and specifically increased in the MAM (33.0-fold) despite decreasing in the total cell fraction (0.4-fold) after HCMV infection ( Fig. 3B and 3C). Analogously, mitochondrial markers, COX2 and Prohibitin, were detected in the enriched MAM and mitochondria fractions. The abundance of COX2 increased in the MAM (4.7fold) and in mitochondria (1.8-fold) above the induction observed in total protein (1.2-fold) during HCMV infection. Similarly, Prohibitin increased 4.8-fold and 2.7-fold in MAM and mitochondria, respectively, during HCMV infection despite being marginally induced in total extracts (1.2-fold).
Membrin, an ER-Golgi SNARE protein (54), was detected in the MAM (0.3) but to levels less than in microsomes (0.6). Membrin abundance increased in the MAM (4.9-fold), despite being reduced in total protein lysates (0.5-fold) during HCMV infection. Intriguingly, the abundance of Membrin decreased fivefold in the microsomal fraction (0.2) commensurate with the ϳfivefold increase in the MAM fraction, suggesting a potential relocalization during HCMV infection.
Conversely, the trans-Golgi protein, Golgin 97, was very weakly detected in MAM (0.03) and mitochondria (0.02) at levels ϳ100-fold less than in the cytosol (3.2). Nonetheless, its abundance increased in the MAM (5.5-fold) or mitochondria (2.3-fold) after HCMV infection even though the increase in the total extract was only modestly increased (1.4-fold).
Glycolytic enzymes (PGM 5 and Hexokinase 1) were weakly detectable in the MAM of uninfected HFFs. Nonetheless, after HCMV infection, the relative abundances of both PGM 5 (6.2-fold) and Hexokinase 1 (9.1-fold) increased markedly in the MAM above the values in total extracts (2.5-fold and 0.8-fold, respectively). Together, our results establish the presence of known MAM (FACL4) and mitochondrial (COX2, Prohibitin) markers in the enriched MAM fraction and their strong induction in the MAM after HCMV infection, above that observed with their corresponding changes in total protein. Further, these results show the very low abundance of nonresident proteins (such as the trans-Golgi protein, Golgin 97) in the enriched MAM. Moreover, HCMV infection specifically alters the abundance of some proteins in the enriched MAM fraction (FACL4, COX2, Prohibitin, Membrin, PGM 5, and Hexokinase 1) even though their total abundances are not proportionately increased in the HCMVinfected cell, suggesting their relocalization from other subcellular compartments.
To comprehensively and independently quantify proteins that were altered in the cellular MAM fraction after HCMV infection, we performed differential SILAC strategy where one set of cells for example cells that will be HCMV-infected are metabolically labeled with heavy 13 C 6 -Arg and 13 C 6 , 15 N 2 -Lys whereas cells that will be uninfected are metabolically labeled with light 12 C 6 -Arg and 12 C 6 , 14 N 2 -Lys. HCMV-infected and uninfected cells are then mixed at 1:1 ratio for subsequent MAM fractionation and proteome profiling. Because trypsin cuts proteins after Lys and Arg residues, each peptide from each protein is detected as pairs of light and heavy ion corresponding to 12 C 6 -Arg/ 12 C 6 , 15 N 2 -Lys labeled peptide and 13 C 6 -Arg and 13 C 6 , 15 N 2 -Lys, respectively. The intensity ratio of the heavy to light peptide reflects the amount of the corresponding protein in the SILAC labeled versus unlabeled cells. To further validate our data, we have performed SILAC switch labeling experiment (see Fig. 1 for detailed experimental design). MAM proteome profiling of HCMV-infected versus uninfected cells, derived from two independent experiments using forward and reverse SILAC strategies, exhibited close symmetry, as shown by scatter plot comparison (Fig. 4A). Comparing labeled and unlabeled peptide pairs of a MAMspecific marker protein Ero1␣ (55) between sample sets, we observed a consistent trend of Ero1␣ up-regulation in MAM The known subcellular localization of 991 human proteins identified in heavy MAM proteome in both forward and reverse experiments (with identified peptide number cutoff of at least 2) using UniProt are represented by percent of total. Identified HCMV proteins were excluded from the analysis. The portion designated microsomes includes proteins, which localize in the ER, Golgi apparatus, endosomes, lysosomes, melanosomes, and vesicles using Uniprot.
domains in HCMV-infected cells (Fig. 4B). Fold changes are presented as a ratio of light unlabeled peptides over heavy labeled peptides. Because labeling media, heavy and light, were reversed in the two experiments, changes because of HCMV infection appear inversely related. In the case of Ero1␣, for example, 13-and ninefold increases in MAM abundance of the protein were detected after HCMV infection.
An overall trend of MAM proteome enrichment was observed at late times of HCMV infection (Fig. 5A). Over 65.3% (647/991) of identified MAM proteins increased in abundance by 72 hpi. Most of these were induced between two-and fivefold, although a small subset of proteins increased more than 10-fold at late times of HCMV infection (Fig. 5B). Another 33.3% of identified MAM proteins were not altered by HCMV infection, either staying at the same level in the presence or absence of HCMV infection (14.8%, 147/991) or changing in discordant ways between forward and reverse experiments (18.5%, 183/991). Surprisingly, only 1.4% (14/991) of the identified MAM proteins were consistently decreased at late times of HCMV infection, suggesting the importance of the MAM proteome for HCMV infection and progeny production.
We first mined the heavy MAM proteomic data for known MAM proteins (Table I). Multiple MAM proteins including lipid synthetic enzymes, Ca 2ϩ signaling proteins, ER-mitochondrial membrane tethering complexes were detected. Markers with nearly exclusive MAM localization under normal cell conditions, such as ACAT1, ERp57, ERp44, Ero1␣, erlin 1, and erlin 2 were detected in the MAM proteome by SILAC technology. MAM-enriched chaperones, notably BiP/GRP78, calnexin, and calreticulin, were readily detected in both proteomic sets. MAM junction proteins, which physically associate the MAM and mitochondria membranes, were also observed. These include components of the MAM Ca 2ϩ macromolecular bridge, GRP75 and VDAC, as well as Mfn1, which can form heterodimers with Mfn2. VDAC-interacting proteins and proteins involved in ER-to-mitochondria Ca 2ϩ signaling were detected as well as cell stress chaperones involved in regulating Ca 2ϩ homeostasis. Nonetheless, Sig-1R, an MAM lipid raft marker (56), was detected only in one of the replicate proteomic experiments. Intriguingly, some MAM markers such as PACS-2, Autocrine motility factor receptor, inositol 1,4,5-triphosphate receptor type 3 (IP3R), phosphatidylserine synthases 1 and 2, Mfn2, ceramide synthase, and FACL4 were not detected by proteomic analysis. This may reflect their relatively low abundance within the MAM proteome, their low efficiency of retention during MAM fractionation, or a masking of their detection by other proteins. After HCMV infection, chaperones associated with the MAM (20) increased dramatically (Table I). Particularly notable is the increased abundance (10-fold) of GRP78/BiP, known to be required for permissive HCMV infection (57). The induction levels of the MAM proteins were independently verified by the reverse approach.
Electron tomography imaging has visualized ribosomes at ER-mitochondrial contact sites (22). To determine the feasibility of translation and translocation occurring at or near MAM domains, we examined the detection of minimal translational and translocation machineries in the MAM fraction (Table II). Signal recognition particle-dependent translation in the ER minimally requires the signal recognition particle re-  (Table II).
The twelve host cell proteins with the highest increases in MAM abundance during late HCMV infection are listed in Table III. This list of proteins represents all unique peptide hits of host cell MAM proteins detected to be induced by at least 10-fold during HCMV infection in both forward and reverse   experiments. Tellingly, the most induced proteins are cellular chaperones including heat shock proteins, HSP10, HSP60, and GRP75, which are known to participate in various signaling cascades and affect lipid bilayer fluidity when they become membrane-associated (61). In contrast, the MAM proteins decreased the most are listed in Table IV. Notably, collagen and other secreted proteins were dramatically decreased in the MAM fraction of HCMV-infected HFFs at late times of infection. Validation Studies of HCMV Induction of Key MAM Proteins-Proteomic data of key MAM proteins at late times of HCMV infection were validated using Western analyses of subcellular fractions. HCMV induced substantial increases in ER chaperones, BiP/GRP78 (eightfold), calnexin (4.7-fold), and calreticulin (10.7-fold) in the heavy MAM (Fig. 6). More moderate increases of MAM junction proteins, PACS-2 (2.6fold), Mfn1 (5.9-fold), and Mfn2 (2.2-fold), were documented by Western analyses as were increased MAM lipid raft proteins, erlin 2 (5.2-fold) and Sig-1R (fourfold). Components of the MAM Ca 2ϩ signaling complex, VDAC (6.8-fold) and GRP75 (sevenfold) and mitochondrial chaperones, HSP60 (18-fold) and prohibitin (8-fold), were more dramatically increased. In contrast, minor changes were observed for ␣-tubulin, which was used to normalize protein levels in each fraction. HCMV pUL37ϫ1 was detected in the microsomal, MAM and mitochondrial fractions of HCMV-infected HFFs, as we have previously found (28,29,31). The values of induction observed for key MAM components by Western analyses mostly fell within the induction value ranges determined by quantitative proteomics (Fig. 6B). Together, these studies verified the induction of key, known MAM proteins at late times of HCMV infection. These results further suggest that HCMV infection induces MAM restructuring to benefit its growth. Fig. 7 pictorially represents the observed association of glycolytic enzymes to the MAM fraction at late times of HCMV infection. The quantitative proteomics demonstrated increases in glycolytic enzymes (supplementary Table S2) of most glycolysis reactions with values from 1.9-fold (pyruvate kinase) to 14.4-fold (phosphoglycerate mutase). These values were verified by the reverse experiment. Nonetheless, two glycolytic enzymes, triose phosphate isomerase and phosphoglycerolkinase, were not detected with confidence in the MAM fraction but were detected in banded mitochondria (data not shown).

TABLE II Minimal required translation and translocation machinery is MAM associated and induced by HCMV infection at late times
Furthermore, the glycolytic pyruvate kinase enzyme detected in our proteomics showed peptides specific for the M2 isoform. Pyruvate kinase M2 has been demonstrated to promote glycolytic flux as well as allow the buildup of intermediate metabolites to support anabolic metabolism, a process linked to rapidly dividing cancer cells (62,63) and also critical for productive HCMV infection (5,6).
In addition to the glycolytic enzymes, all the tricarboxylic acid (TCA) cycle enzymes, except for malic enzyme, were increasingly associated with the MAM after HCMV infection. Notably the induction of enzymes ranged from 7.4-fold (isocitrate dehydrogenase) to 17.6-fold (aconitase). In addition, the electron transport chain (ETC) complexes I, II, III, and IV, and the ATP synthase complex localized in the inner mitochondrial membrane were detected and increased (Ͼ7.9-to 15.5-fold) in the MAM fraction after HCMV infection. Furthermore, we observed up-regulation of proteins involved in glutaminolytic pathways and anaplerosis. We detected inductions of malate shuttle proteins (oxoglutarate/ malate carrier protein, malic enzyme, and malate dehydrogenase) and glutamate intake proteins (Aralar, aspartate aminotransferase, glutamate/OH carrier, and glutamate dehydrogenase), but did not detect inhibitors of glutamate intake (such as aminooxyacetate or ribulose-5-phosphate). Enzymes needed to convert glutamine to ␣-ketoglutarate to enter the TCA cycle (glutaminase and glutamate dehydrogenase) were also detected. Finally, the quantitative proteomic analyses detected an increase in the abundance of lipid synthetic or degradative enzymes listed in supplementary Table S3. Together, these studies place critical metabolic and synthetic pathways needed for HCMV infection at the MAM, where they can be coordinated by ER-mitochondrial crosstalk.
To determine whether HCMV infection induced relocalization of selected MAM proteins, we compared the localization of HSP60, Hexokinase 1, and Sig-1R in uninfected HFFs and HCMV-infected HFFs by immunofluorescence assay (Fig. 8). Although HSP60 and Hexokinase 1 increased in abundance in the HCMV-infected cell, they did not appear to markedly relocalize during HCMV infection. In stark contrast, Sig-1R relocalized from a primarily nuclear localization in uninfected cells to a cytosolic membrane localization during late times of HCMV infection. Together with the results above, these findings show that HCMV selectively induces relocalization of some MAM proteins such as Sig-1R. DISCUSSION Currently, the only licensed antiviral drugs for HCMV target its DNA replication. However, other regulated cellular metabolic reactions occurring during infection have emerged as important to HCMV progeny production and maturation, which may serve as future therapeutic targets. These events include the uncoupling of respiration from glucose metabolism, the concomitant use of glutamine for anaplerosis to maintain cycling of oxidative phosphorylation reactions in the TCA cycle, and the biosynthesis of lipids and membranes. The current studies analyzed HCMV alterations of the MAM, which regulates Ca 2ϩ -responsive mitochondrial enzymes and is enriched in lipid synthetic enzymes. Importantly for these studies, the MAM can be physically fractionated by exploiting their tight association to mitochondria, and distinguishable banding patterns in Percoll gradients (29,47). Using SILAC methodology we probed the MAM compositional changes induced by HCMV infection during late stages of infection, when the cell has been reprogrammed into a virus factory.
The MAM is increasingly being appreciated as a key control of mitochondrial mediated apoptotic signaling, calcium signaling from the ER to mitochondria and consequent regulation of cell bioenergetics (20). Proteomic analyses of enriched MAM allowed for increased sensitivity in detecting MAM components in primary human fibroblasts and the changes induced in the MAM by HCMV infection. Although some MAM markers have been identified, a more comprehensive, unbiased analysis of the MAM proteome has not yet been reported. To our knowledge, this is the first proteomic analysis of enriched MAM, and uses some of the most rigorous methodology currently available to fractionate the MAM and quantify the changes induced by HCMV infection. The detected changes during infection provide insight into which MAM proteins could be playing important roles during HCMV infec- tion. These studies have provided valuable insight into the MAM and alteration during HCMV infection and will serve as a foundation to test the role of individual components during HCMV infection. Functionally, the physical association of the ER and mitochondria regulates important cellular processes such as lipid biosynthesis and nonvesicular transfer of lipids to mitochondria (52,53,64,65). Transfer of ceramide synthesized in the MAM to mitochondria can induce channel formation and apoptosis (66). The MAM accommodates internal lipid rafts, detergent-resistant membranes, which are enriched in Sig-1R (56,67,68). Erlin-1 and erlin-2, associated with IP3R degradation, are also present in internal lipid rafts (69 -71). Importantly, the ER can transmit Ca 2ϩ signals to mitochondria without increasing bulk cytosolic Ca 2ϩ concentrations above physiological thresholds. An MAM macromolecular complex composed of IP3R type 3 (IP3R3)-GRP75-VDAC generates high Ca 2ϩ microdomains at ER-mitochondrial contact sites and thereby enables the function of the mitochondrial un-iporter (72). The MAM is enriched in Ca 2ϩ binding chaperones, calreticulin, Sig-1R, and ERp57 (20,53,56,65,73). Sig-1R, a ligand-dependent MAM chaperone, increases ER Ca 2ϩ efflux by stabilizing IP3R3 (74). IP3R-mediated Ca 2ϩ release is essential in maintaining cellular bioenergetics (21). Nonetheless, if unchecked, Ca 2ϩ release can induce mitochondrial-mediated apoptosis (20). IP3R-mediated Ca 2ϩ signaling activates apoptosis by inducing cytochrome c release from mitochondria (75). Increased mitochondrial Ca 2ϩ has also been associated with increasing electron transport, increased reactive oxygen species production, and opening of the mitochondrial permeability transition pore (76).
We observed a dramatic alteration of the MAM proteome at late times of HCMV infection. This conclusion is supported by the augmented congregation of proteins at the MAM, which regulate Ca 2ϩ cross-talk between ER and mitochondria, glutamate uptake and utilization for anaplerosis, metabolic flux of both glycolysis and oxidative phosphorylation, along with an impressive concentration of heat shock proteins. The broad induction of MAM constituents during infection indicates an important role for ER-mitochondrial cross-talk at late times of infection, predictably for the maintenance of cell viability and increased energy production for progeny production. The altered MAM proteome during HCMV infection could be because of the recruitment of crucial proteins to ER-mitochondria junctions, which were either absent or under-represented in MAM of uninfected cells. The induction of membrane as-sociation of soluble cytosolic proteins such as the glycolytic enzymes suggests that protein scaffolding at MAM domains also helps mediate the critical functional changes in ER and mitochondria observed during HCMV infection.
Of the proteins increased in the MAM, heat shock proteins had the highest fold induction caused by HCMV infection. Cellular chaperones are known to be induced by HCMV infection, as well as by ER stress, however their dramatic accumulation at MAM sites was unexpected. Previous evidence supports unique functions for membrane-tethered chaperone proteins, beyond assisting protein-folding (61). HSP10, HSP60, and GRP75, which plays a role in Ca 2ϩ signaling when tethered at the MAM membrane, may likely engage in unique cell signaling functions at MAM domains. Membrane fluidity and microdomain organization are decisive in the transduction of stresses into signals for activation of HSPs (61). Association of certain HSPs with membranes can reestablish fluidity and bilayer stability and restore membrane functionality during stress and modulate thereby cellular responses to stress.
ER stress is known to increase MAM junctions twofold (26). HCMV induces specific components of ER stress at late times of infection, but truncates the full unfolded protein response (41,77,78). This allows the virus to utilize the increased levels of chaperones for the production and assembly of virion components, without a widespread block on nascent polypeptide translation. We detected a prominent trend of induction of MAM proteins around two-to fivefold during late times of HCMV infection, consistent with an increase in MAM junctions because of HCMV-induced ER stress. Importantly, however, we also detect MAM components increased well beyond this modest level of induction, suggesting functionally directed restructuring of MAM subdomains for the functional benefits to HCMV growth, not merely an increase in the number of ER-mitochondria contacts or of the tightness of the ER-mitochondrial junctions.
HCMV infection alters the localizations, shapes, and sizes of secretory organelles at late times of infection during its assembly (79). Viral proteins were also detected in the MAM of HCMV-infected cells but not in uninfected cells. This finding may indicate trafficking or recruitment of other HCMV proteins, in addition to pUL37ϫ1, to the MAM. Alternatively, the identified HCMV proteins may result from virions exiting the nucleus through the outer nuclear membrane, which is contiguous with the ER membrane whose subdomains were used for these studies. Importantly to the enrichment of the MAM compartment, the trans-Golgi marker was very low in abundance in the MAM (1/100 th ) of the levels in the cytosol. Furthermore, we did not detect the highly abundant HCMV proteins that traffic to other membrane-bound compartments including the nucleus (pp150/pUL32 and UL53), secretory apparatus (gB, gH, gL, and gO), or endosomes (US27, UL33, and US28) as well as the assembly compartment (pp28/ UL99). Because the analyses were performed at late times of FIG. 7. Increased levels of the glycolytic, TCA cycle, and ETC pathways associated with ER-mitochondrial contacts at late times of HCMV infection. The relative fold-inductions for proteins involved in glycolysis, TCA cycle, and ETC detected in association with the MAM fraction from HCMV-infected HFFs above uninfected cells (numeric values are shown in supplementary Table S2) are represented by the areas of corresponding filled black circles. The open circles represent glycolytic enzymes not detected by this quantitative proteomic analysis. Proteins with multiple isoforms are also represented by a smaller, gray-filled circle (isoform with lowest induction) and surrounded by a darker outline (isoform with the largest induction). Constituent members of the ETC and ATP synthase complexes are represented only by solid blocks, as these proteins were observed to have increased expression levels of at least sevenfold, but typically over 10-fold higher than in uninfected cells. Glycolysis hexokinase (HK), glucose-6-phosphate isomerase (GPI), phosphofructokinase (PFK), fructose-biphosphate aldolase (ALDO), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), phosphoglycerate mutase (PGM), alpha-enolase (ENO), pyruvate kinase (PK), lactate dehydrogenase (LDH), pyruvate dehydrogenase (PDH), citrate synthase (CS), aconitase hydratase (ACO), isocitrate dehydrogenase (IDH), 2-oxoglutarate dehydrogenase (OGDH), succinyl-CoA ligase (SUCL), succinate dehydrogenase (SDH), fumarate hydratase (FH), malate dehydrogenase (MDH). infection, the undetected HCMV virion proteins, localized to the secretory apparatus, are of much greater abundance than pUL37ϫ1, which we reliably detect in the MAM fraction. Consistent with the proteomic results, pUL37ϫ1 has been detected in the ER, MAM, and mitochondria but not in the nucleus, Golgi or plasma membrane (29,31,50). Together, these findings argue for the enrichment of the MAM in the banded fraction and against a simple mixing of cellular membranes during late times of HCMV infection.
There appear to be multiple potential mechanisms underlying the increased abundance of proteins in the MAM. The most common mechanism, notable for the heat shock proteins, appears to be increased total expression during HCMV infection and a proportional increase in the MAM abundances. Examples of these include GRP78, GRP75, and HSP60. HCMV infection is known to induce a modified ER stress response (77,78,80). The ability of HCMV to modulate the cellular ER stress response is evident by the finding that Sig-1R increases in the MAM despite the fact that HCMV causes ER stress. Sig-1R is a sensor that relocalizes from the MAM to bulk ER when cells undergo chronic ER stress responses (74). Sig-1R relocalizes from the periphery of the nucleus (observed in uninfected cells) to the MAM in about 50% of HCMV-infected HFFs (Fig. 8), suggesting a finely tuned regulation of its location in the infected cell. The abundance of GRP78 is increased during HCMV infection and is essential for HCMV assembly (57). Herein, we found that a significant fraction of the increased GRP78 in HCMV-infected cells localizes in the MAM, suggesting the functional importance of the MAM during HCMV infection. Our finding of GRP78 enrichment in the MAM of infected cells is consistent with its known enrichment in the MAM of uninfected cells (74). GRP75 is also documented to increase in abundance in the heavy MAM fraction of HCMV-infected cells (29) and our current results solidify that finding using quantitative proteomics and Western analyses. Nevertheless, the increase in GRP75 was greater in the MAM than in total abundance suggesting that in addition to a corresponding increase with total abundance, GRP75 may also be relocalized during HCMV infection.
In contrast, the abundances of other cellular proteins selectively increase in the MAM fraction even though there is little to no change in their abundances in total cellular extracts. These changes in their relative abundance in the MAM suggest a targeted relocalization of these proteins from other subcellular compartments to the MAM during HCMV infection. Examples of these include calreticulin whose marginal increase in total fraction is considerably less than the increases in microsomes or heavy MAM fractions. Similarly, the increases in calnexin and PACS-2 abundances in the MAM fraction are greater than in their corresponding increases in total extracts. Consistent with this potential mechanism is the finding that calnexin is a known cargo for PACS-2 (81) and its relocalization to the MAM may be mediated by PACS-2. Finally, relocalization of glycolytic enzymes from the cytosol to ER-mitochondrial contacts appears to underlie at least in part their increased abundance in the MAM. Intriguingly, membrane association of glycolytic enzymes has been found in pancreatic cancer cells (82). Thus, HCMV may selectively retarget key cellular proteins to the MAM and thereby affect its functions during infection.
Mining our proteomic results for the MAM Ca 2ϩ signaling complex components, we gain some understanding of the structural organization of MAM domains during late times of HCMV infection. The OMM porin, VDAC, is known to be involved in more than one macromolecular complex at the MAM. In one of these complexes, it associates with cytosolic GRP75 and IP3R of the ER membrane to form a Ca 2ϩ signaling complex bridging the ER and mitochondria (27). Here, we observed a dramatic increase in GRP75 abundance in the MAM, which we had previously shown with proteinase K digestion to be cytosolic GRP75 (29). VDAC can also form complexes with ANT proteins of the inner mitochondrial membrane to regulate mitochondrial membrane permeability, Bax association, and cytochrome c release. Importantly, VDAC can directly bind ANT of the inner mitochondrial membrane when it is in a low conductance, cationically selective, tensed conformation. This tensed pore structure allows transfer of extramitochondrial Ca 2ϩ into the intermembrane space of the mitochondria. HCMV pUL37ϫ1/vMIA causes ER Ca 2ϩ efflux (35), interacts with Bax and ANT (10,13,14) but blocks the proapoptotic activity of Bax at the OMM (13,37,38). Hexokinase, which was up-regulated in the MAM of HCMV-infected cells, can bind to this complex to promote mitochondrial oxidative phosphorylation, and malate shuttling through VDAC to the cytosol (83). As cytochrome c stably binds ANT in this macromolecular complex configuration, we expected to detect cytochrome c, which resides within the intermembrane space of the mitochondria, by proteomic analyses of the MAM. Consistently, the HCMV-specific induction of cytochrome c levels directly matched induced ANT levels in our experiments (supplementary Table S3). However, we did not detect other components of lipid raft-associated apoptosomes in our MAM fractions, which should have been detected if the cytochrome c in the MAM fraction induced apoptosis. Together, these data suggest that the majority of MAM associated VDAC is engaged in activities allowing Ca 2ϩ signaling to mitochondria, and increased metabolic flux and blocking of proapoptotic signaling.
One important finding of these studies was the increased MAM association of several metabolic enzymes during HCMV infection. Strikingly, we detected HCMV-specific induction of MAM association with cytosolic glycolytic enzymes. Because the cytosol is not pelleted with the MAM and mitochondria during early fractionation steps, but remains in the supernatant, it is unlikely that significant carryover to the Percoll gradient and subsequent contamination of the banded MAM, unless these components were membrane tethered. A similar phenomenon of membrane-tethering of these enzymes has been characterized in rapidly dividing cancer cells (82). Revealingly, a Ca 2ϩ responsive aspartate-glutamate carrier, Aralar, which supplies mitochondria with glutamate for oxidative phosphorylation reactions, emerges as one of the most enriched proteins at MAM domains at 72 hpi. We found a large (12-to 20-fold) increase in MAM-associated Aralar 2 during HCMV infection, which when compared with the mean 20-fold induction seen for HCMV viral proteins in MAM (which are absent in uninfected cells), exposes a significant induction of this protein during late times of HCMV infection, corresponding with the requirement of glutamate for anaplerosis at late times of infection (7).
Our current studies revealed exceptionally few decreased MAM proteins. All of the downregulated MAM proteins were secretory proteins, destined for the extracellular matrix. It appears that HCMV infection specifically reduces the known ability of MAM to support transport of secretory proteins to the Golgi (53). In contrast, global protein secretion from peripheral ER to the cell surface is not diminished during HCMV infection. Previous studies on the HCMV secretome demonstrate an increase in other secreted proteins including cytokines, as well as angiogenic and wound healing proteins from HCMV-infected cells at 72 hpi (40). Taken together, these results argue that HCMV infection selectively diminishes specific MAM functional activities whereas augmenting others.