Human cytomegalovirus UL37 immediate-early regulatory proteins traffic through the secretory apparatus and to mitochondria

The human cytomegalovirus (HCMV) UL3638 immediate-early (IE) locus encodes multiple, partially overlapping transcripts and products (Fig. 1a, b). Four IE and one early transcripts are encoded in this region (Kouzarides et al., 1988 ; Tenney & Colberg-Poley, 1991a , b ; Goldmacher et al., 1999 ). The largest (3·4 kb) transcript encodes a 487 aa type I integral membrane N-glycoprotein, gpUL37, while the 1·7 kb RNA encodes the UL37 exon 1 protein (pUL37x1) (Kouzarides et al., 1988 ; Al-Barazi & Colberg-Poley, 1996 ). The 162 aa open reading frame (ORF) contained in the UL37 exon 1 sequences is common to both gpUL37 and pUL37x1. Thus, the UL37 proteins have identical amino termini. Downstream of these sequences, pUL37x1 has one unique residue whereas gpUL37 has unique N-glycosylation sites, a transmembrane (TM) domain and a cytosolic tail. UL37 exon 1 sequences are well conserved in multiple clinical strains, indicating that they are probably required for HCMV growth in humans (W. A. Hayajneh, M. M. Lesperance, D. G. Contopoulos-Ioannidis and A. M. Colberg-Poley, unpublished results).

(24K): Fig. 1. (a) HCMV UL3638 transcripts. The RNA map indicates the direction of transcription, exons (filled boxes), introns (shaded boxes), 3' untranslated regions (thin lines) and poly(A) tails (arrows). The start (open arrowheads) and stop (open ovals) sites of translation are also indicated. The bent and open arrows on the scale represent UL3638 promoters and polyadenylation signals, respectively. The panel on the right shows the genes, RNA sizes and kinetic class. E, Early. (b) UL37 proteins, peptides and antibodies. The features of gpUL37 include its hydrophobic leader and TM domains (coils), acidic domain (open box), N-glycosylation sites (branches) and cytosolic tail. The sequences contained in pUL37x1 are indicated. The peptide sequences used for generation of antisera are indicated. A FLAG epitope was attached to the carboxy terminus of the gpUL37 ORF as indicated by the shaded oval.

gpUL37 and pUL37x1 selectively transactivate expression of genes under the control of the cellular heat-shock protein 70 (hsp70) promoter or of HCMV early gene promoters (Colberg-Poley et al., 1992 , 1998 ; Tenney et al., 1993 ; Zhang et al., 1996 ). Products of these HCMV early genes are required for its oriLyt DNA replication (Pari & Anders, 1993 ; Pari et al., 1993 ). pUL37x1 can repress the IE1 and IE2 activation of the HCMV US3 promoter (Biegalke, 1999 ). pUL37x1 and gpUL37 have recently been found to have anti-apoptotic activities (Goldmacher et al., 1999 ). Their common hydrophobic leader sequence is required for anti-apoptotic activity (Goldmacher et al., 1999 ), while their acidic domain is required for HCMV early promoter transactivation (Colberg-Poley et al., 1998 ).

gpUL37 is known to traffic through the endoplasmic reticulum (ER) to the medial-Golgi in non-permissive HeLa cells and in permissive human diploid fibroblasts (HFF) (Al-Barazi & Colberg-Poley, 1996 ; Zhang et al., 1996 ). Conversely, c-myc-tagged pUL37x1 has been shown to traffic to mitochondria (Goldmacher et al., 1999 ). As gpUL37 and pUL37x1 share their amino-terminal sequences, including hydrophobic leader sequences, which conventionally target nascent proteins to the secretory pathway, we wished to define their trafficking patterns further in human cells. To that end, we generated antisera against residues in their common sequences and examined human cells that expressed the desired protein transiently. Co-localization was performed by using optical sections generated by confocal laser scanning microscopy or by immunoblotting of fractionated cells. gpUL37 and pUL37x1 traffic unconventionally both through the secretory apparatus and to mitochondria in human cells.

Transient transfection.
U373 cells (ATCC), HeLa or HFF cells were subcultured on glass slides 1 day prior to transfection. Cells were transfected with 10 µg p414 (gpUL37), p327 (pUL37x1), p816 (gpUL37FLAG) or p370 [HCMV glycoprotein B (gB)] by using LipofectAMINE (Gibco BRL) at a ratio of 3:1 (lipid:DNA) in Optimem (Gibco BRL) (Colberg-Poley et al., 2000 ). The lipid:DNA mixture was removed from the cells after a 4 h incubation and the cells were supplied with Dulbeccos modified Eagles medium containing 10% foetal calf serum. Cells were harvested 48 h after transfection by washing three times with PBS and fixed with methanol (-20 °C, 10 min). Alternatively, cells were fixed with acetone (-20 °C, 10 min) with similar results. Fixed cells were stored at -20 °C until staining. For staining of unfixed cells, cells were harvested by washing with PBS but were stained immediately without fixation.

Epitope-tagging of gpUL37.
The full-length UL37 ORF was tagged at its carboxy terminus with a FLAG marker octapeptide (DYKDDDDK). To generate the recombinant plasmid, UL37 exon 3 sequences encoding amino acids 453487 were amplified by PCR by using primers (nt 4999750017 and 4991349933) into which EcoRI and BamHI restriction enzyme sites were incorporated. The restriction enzyme-cleaved PCR amplification products were cloned into EcoRI/BamHI-cleaved pFLAG-CMV5a vector (Sigma) to generate p801. The upstream UL37 exon 1, 2 and 3 sequences, which encode the amino terminus of gpUL37, were obtained by EcoRI/NruI cleavage of p414 (Colberg-Poley et al., 1992 ) and inserted into p801, thereby generating p816. The identity of the desired recombinant was verified by restriction enzyme cleavage and reactivity of the expressed protein with both Ab1064 and anti-FLAG antibodies.

Anti-UL37 antisera.
Rabbit antisera (Ab1064 and Ab1386) that detect pUL37x1 and/or gpUL37 specifically were generated by Covance Research Products (Denver, PA, USA) following repeated injection of keyhole limpet haemocyanin-coupled peptides into rabbits (Fig. 1b; Table 1). Mouse polyvalent antiserum (m51) against gpUL37 and pUL37x1 was generated at Childrens Research Institute by multiple immunizations of a BALB/c mouse by intralingual injection of 100 µg of the gpUL37 DNA expression vector, p414, which encodes the complete, 487 aa UL37 ORF (Colberg-Poley et al., 1992 ). The mouse was anaesthetized by inhalation of metofane (Mallinckrodt Veterinary) prior to immunization.

Table 1. Primary and secondary antibodies used in these studies

Indirect immunofluorescence staining.
Antibodies used in these studies are listed in Table 1. Cells were stained as described previously (Al-Barazi & Colberg-Poley, 1996 ; Zhang et al., 1996 ) with the following modifications. Rabbit Ab1064 (or preimmune serum, 1:2001:400) or mouse polyvalent m51 (1:101:20) antibodies were used at 37 °C for 1 h or 4 °C overnight. Mouse anti-FLAG M2 antibody (1:180, Sigma) was used to detect FLAG-tagged gpUL37. In order to assess the localization of UL37 proteins, cells were doubly or triply stained with antibodies against organelle markers. Cells were stained with mouse or rabbit antiserum against protein disulphide isomerase (PDI) (1:25, StressGen), an ER protein (Hillson et al., 1984 ), mouse anti-Golgi zone antiserum (1:25, Chemicon) or human autoimmune serum against mitochondria (1:25, ImmunoVision). Secondary antibodies used were fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit IgG (1:50, Southern Biotechnology), FITCgoat anti-mouse IgG (1:50, Jackson ImmunoResearch), Texas red (TR)-conjugated goat anti-mouse IgG (1:50, Kirkegaard and Perry Laboratories), TRgoat anti-rabbit IgG (1:50, Jackson ImmunoResearch) and cyanine 5 (Cy5)-conjugated goat anti-human IgG (1:100, Jackson ImmunoResearch).

Confocal laser scanning microscopy.
Analyses were performed with a Bio-Rad MRC1024 confocal laser scanning microscope (Center for Microscopy and Image Analysis, George Washington University), which allows for triple excitation. Triple excitation lines at 488, 568 and 647 nm were used for the excitation of FITC, TR and Cy5, respectively. Emission was measured at 520 (FITC), 615 (TR) and 670 (Cy5) nm. Individual signals were captured sequentially to avoid any spurious overlap of the emission signals. Individual optical sections, obtained by using z-dimensions between 0·1 and 1·0 µm, were examined to determine co-localization of UL37 proteins with cellular organelle markers. Optical sections were obtained by using either the 40x (NA 1·3) lens or the 60x (NA 1·4) lens. The vertical resolution of the 40x lens is 0·60·8 µm (Paddock, 1999 ; R. Rufner, personal communication). Images were generated by using Adobe Photoshop (version 4.0), Bio-Rad plug-ins and Microsoft Publisher 98.

Subcellular fractionation and Western blot analysis of cells expressing gpUL37FLAG.
HFF or HeLa cells were lipofected with 20 µg gpUL37FLAG (p816) or gpUL37 (p414) expression vectors as described above. At 48 h after transfection, the cells were washed three times with PBS and resuspended in MTE buffer (0·27 M mannitol, 10 mM TrisHCl, 0·1 mM EDTA, pH 7·4). The resuspended cells were sonicated briefly and the nuclei and unbroken cells were removed by centrifugation at 700 g for 10 min. Mitochondria were isolated by differential centrifugation (Chomyn, 1996 ). Briefly, a mitochondrion-rich fraction was obtained by centrifugation of the supernatant at 15000 g for 10 min. Mitochondria were separated from lysosomes by sedimentation through a discontinuous sucrose gradient (1·0/1·7 M sucrose, both in 10 mM TrisHCl, 0·1 mM EDTA, pH 7·6) in an SW60 rotor at 39000 g for 10 min. Fractions of the interface (mitochondria) and top band (lysosomes) were collected, diluted in MTE and pelleted by centrifugation at 15000 g for 10 min. Total protein concentrations were determined by the BCA protein assay (Pierce). The purified fractions were then examined by electrophoresis in 10% SDSpolyacrylamide gels and Western blot analysis with antibodies against FLAG (5 µg/ml, M2, Sigma), mitochondria (1:1000, anti-Grp75, StressGen) and lysosomes (1:500, anti-LAMP1, StressGen) (Table 1). The reactive antibodies were detected by using a chemiluminescent immunoblot detection system with CSPD substrate (Tropix).

gpUL37 traffics through the secretory apparatus
gpUL37 traffics through the ER to the medial-Golgi network in permissive HFF cells and in HeLa cells (Al-Barazi & Colberg-Poley, 1996 ; Zhang et al., 1996 ). The HCMV integral membrane glycoprotein gB is also known to traffic through the secretory apparatus; however, gB shows differential distribution in permissive human glioblastoma U373 cells and in HFF cells (Fish et al., 1998 ). In order to determine whether this was also the case for gpUL37, we examined gpUL37 trafficking in U373 cells, with Ab1064. As Ab1064 detects all UL37 proteins that share UL37x1 sequences, we used transient transfection of U373 cells to localize the desired authentic protein. gpUL37 was readily detected in transiently transfected U373 cells by the bright green fluorescence at 520 nm (Fig. 2a, d). Autofluorescence of cells that were lipofected with control p370 DNA was minimal and dull in cells imaged by confocal microscopy and did not produce the bright, specific patterns observed with our antibodies (Ab1064, 1386 and m51) or with the commercially available anti-FLAG antibody (M2, Sigma) (data not shown and see Figs 3 and 4). With anti-PDI and secondary TR antibodies, the ER of U373 cells was stained brightly (Fig. 2b). gpUL37 partially co-localized with PDI in the ER of U373 cells, as indicated by the yellow overlap in the merged optical sections (Fig. 2c).

(97K): Fig. 2. (a)(c) gpUL37 co-localizes with an ER marker in permissive U373 cells. U373 cells transiently expressing gpUL37 were fixed with methanol and stained with rabbit Ab1064 and mouse anti-PDI followed by staining with FITCanti-rabbit IgG and TRanti-mouse IgG antibodies. A confocal microscope optical section is shown using the (a) FITC, (b) TR or (c) FITC/TR channels (z=0·2 µm). Bar, 30 µm. (d)(f) gpUL37 co-localizes with a Golgi zone marker in U373 cells. U373 cells transiently expressing gpUL37 were fixed with methanol and stained with rabbit Ab1064 and mouse anti-Golgi zone antibodies followed by staining with FITCanti-rabbit IgG and TRanti-mouse IgG antibodies. A confocal optical section is shown using the (d) FITC, (e) TR or (f) FITC/TR channels (z=1 µm). Bar, 30 µm. (g)(i) gpUL37 traffics to the cell periphery in HeLa cells. HeLa cells transiently expressing gpUL37 were fixed with methanol and stained with rabbit Ab1064 and mouse m51 antibodies followed by staining with FITCanti-rabbit IgG and TRanti-mouse IgG antibodies. A confocal optical section using (g) FITC, (h) TR or (i) FITC/TR channels is shown (z=1 µm). Bar, 20 µm.

(91K): Fig. 3. gpUL37 traffics to the PM in HFF cells. (a)(f) HFF cells transiently expressing FLAG-tagged gpUL37 (ac) or untagged gpU37 (df) were fixed with methanol and stained with mouse anti-FLAG M2 and rabbit anti-PDI antibodies. The secondary antibodies used were FITCanti-mouse IgG and TRanti-rabbit IgG. Confocal optical sections show FITC (a, d), TR (b, e) or FITC/TR (c, f) channels (z=0·25 µm). Bars, 30 µm. (g) HFF cells transiently expressing gpUL37 were unfixed and stained immediately with mouse m51 antiserum (overnight at 4 °C) and FITCanti-mouse IgG antibodies. The FITC channel is shown (z=0·2 µm). Bar, 30 µm.

(105K): Fig. 4. gpUL37 co-localizes with a mitochondrial marker in HFF cells. (a)(f) HFF cells transiently expressing gpUL37 were fixed and stained with rabbit Ab1064, mouse anti-PDI and human anti-mitochondrion antibodies. The cells were then stained with FITCanti-rabbit, TRanti-mouse and Cy5anti-human IgG. Confocal optical sections show the FITC (a), TR (b) or Cy5 (c) channels (z=0·1 µm). Merged images of FITC/TR (d), FITC/Cy5 (e) and FITC/TR/Cy5 (f) are shown. (g)(l) HFF cells transiently expressing gpUL37 were stained with rabbit Ab1064, mouse anti-Golgi zone and human anti-mitochondrion antibodies. The cells were then stained with FITCanti-rabbit, TRanti-mouse and Cy5anti-human IgG. Confocal optical sections show FITC (g), TR (h) and Cy5 (i) channels (z=0·5 µm). Panels show merged images of FITC/TR (j), FITC/Cy5 (k) and FITC/Cy5/TR (l) channels. (m)(o) Control HFF cells expressing HCMV gB were fixed with methanol and stained with Ab1064 and mouse anti-Golgi zone antibodies followed by staining with FITCanti-rabbit and TRanti-mouse IgG. Confocal optical sections show the FITC (m), TR (n) and FITC/TR (o) channels (z=0·5 µm). Bars, 30 µm.

In order to determine whether gpUL37 was translocated from the ER into the Golgi apparatus, we examined U373 cells with an anti-Golgi zone antibody (Fig. 2e). Partial co-localization of gpUL37 with the Golgi marker in U373 cells was observed, as indicated by the yellow overlap in the merged optical sections (Fig. 2f). gpUL37 is also found outside the ER and Golgi, as shown by the green colour in the merged optical sections (Fig. 2c, f). These results are consistent with our previous results of gpUL37 trafficking through the ER to the medial-Golgi in HeLa and HFF cells (Al-Barazi & Colberg-Poley, 1996 ; Zhang et al., 1996 ). As in our previous experiments with Ab1601, directed against the gpUL37 carboxy terminus (Al-Barazi & Colberg-Poley, 1996 ; Zhang et al., 1996 ), gpUL37 was not detected in the nucleus by immunofluorescence analysis with Ab1064. Staining of gpUL37 with a third antibody (Ab1386), against the N-glycosylation domain, showed a similar pattern of staining in the ER and Golgi (data not shown). These results indicate that gpUL37 traffics through the ER to the Golgi apparatus in permissive U373 cells.

gpUL37 traffics to the plasma membrane (PM)
In order to determine whether gpUL37 is translocated detectably from the Golgi apparatus to the cellular PM, we examined HeLa cells expressing gpUL37 by using Ab1064 (Fig. 2g) and m51 (Fig. 2h). Mouse m51 polyvalent antiserum (Fig. 2h) detected gpUL37 at the cell periphery, while rabbit Ab1064 (Fig. 2g) detected gpUL37 in vesicles close to the PM. These results suggest that gpUL37 might traffic to the cell surface from the Golgi apparatus.

In order to establish whether this pattern of gpUL37 trafficking was detectable with another antibody, HFF cells transiently expressing gpUL37FLAG were stained with anti-FLAG (Fig. 3a) and anti-PDI antibodies (Fig. 3b). gpUL37FLAG was readily detected in transiently transfected HFF cells by anti-FLAG antibodies (Fig. 3a). The pattern of gpUL37FLAG distribution, with bright internal staining and patches localized at the PM, again suggested translocation of gpUL37 from the secretory apparatus to the PM (Fig. 3a, c). These results are similar to the gpUL37 staining of the cell periphery observed following m51 staining of transfected HeLa cells (Fig. 2h, i). A significant proportion of the gpUL37FLAG, similar to authentic gpUL37 (Fig. 2c; Al-Barazi & Colberg-Poley, 1996 ), co-localized with PDI, as observed by the yellow overlap in the merged optical sections (Fig. 3c). The specificity of the gpUL37FLAG staining by anti-FLAG antibody was verified by staining of control HFF cells that expressed authentic, untagged gpUL37 (p414). HFF cells that expressed gpUL37 without the FLAG tag were not stained by the anti-FLAG antibody (Fig. 3d). As expected, HFF cells were stained brightly by anti-PDI antibody (Fig. 3e). Nonetheless, no detectable overlap was observed in the merged optical sections of control cells (Fig. 3f).

In order to determine unequivocally whether gpUL37 traffics to the external surface of HFF cells, unfixed HFF cells expressing gpUL37 were stained with mouse m51 antiserum (Fig. 3g). Bright surface fluorescence was observed in unfixed cells that expressed gpUL37 (Fig. 3g). This bright fluorescence was not observed in control HFF cells expressing gB when stained with m51 antiserum (data not shown).

gpUL37 traffics to mitochondria in HFF cells
gpUL37 has anti-apoptotic activity, but is not known to traffic to mitochondria (Goldmacher et al., 1999 ). In order to determine whether gpUL37 traffics to mitochondria, as other anti-apoptotic proteins do, we examined HFF cells that expressed gpUL37 with a mitochondrial marker (Fig. 4). For these experiments, we performed triple labelling to examine the overlap of gpUL37 with either PDI (Fig. 4af) or a Golgi marker (Fig. 4gl). This analysis allowed us to visualize the overlap of mitochondrial and ER or Golgi compartmentalization and to determine whether mitochondrial overlap signals could result from the localization of gpUL37 in either the ER or the Golgi apparatus. gpUL37 transiently expressed in HFF cells was readily detected by Ab1064 (Fig. 4a, g). The cellular ER (Fig. 4b), the Golgi zone (Fig. 4h) and mitochondria (Fig. 4c, i) were clearly stained by their respective antibodies. In the optical sections shown, gpUL37 co-localized predominantly with the mitochondrial marker, as indicated by the aquamarine colour in the merged optical sections (Fig. 4e, k). The patterns of gpUL37 distribution in the optical sections from the transfected cells shown (Fig. 4a, g) were similar to those observed for the mitochondria in the same optical sections (Fig. 4c, i). Staining of control HFF cells that transiently expressed gB with Ab1064 showed only dull, non-specific staining of the lipofected cells (Fig. 4m). Moreover, no detectable co-localization of Ab1064-stained proteins with the Golgi zone marker (Fig. 4n) was observed in the merged optical sections of the negative-control cells (Fig. 4o). Taken together, our results suggest that gpUL37 traffics to mitochondria, in addition to trafficking through the secretory apparatus to the PM. Moreover, co-localization of gpUL37 with the mitochondrial marker is not attributable to its known trafficking through the ER or Golgi apparatus, as these are distinguishable.

Subcellular fractionation and Western blot analysis of HFF or HeLa cells expressing gpUL37FLAG
In order to determine independently whether gpUL37 traffics to mitochondria, we used subcellular fractionation of transiently transfected cells to purify mitochondria and examined these for the presence of gpUL37FLAG protein (Fig. 5). gpUL37FLAG (~250 kDa) was detected by immunoblotting of mitochondria purified by sucrose gradient centrifugation from transiently transfected HFF cells (Fig. 5a, lane 4). Conversely, no reactive protein was detected in mitochondria from HFF cells expressing untagged gpUL37 (lane 3), indicating the specificity of the reactivity of the FLAG antibody with gpUL37FLAG. Lysosomal fractions from transiently transfected HFF cells expressing either gpUL37FLAG (lane 2) or gpUL37 (lane 1) did not react detectably with the anti-FLAG antibody, demonstrating the absence of detectable levels of gpUL37 in the lysosomal fractions. In order to determine the purity and identity of the fractions, the Western blot was then reacted with anti-Grp75 antibody (Fig. 5b). The reactivity of fractionated mitochondria from transfected HFF cells with anti-Grp75 antibody (lanes 3 and 4) showed the presence of mitochondria in the expected fractions. Conversely, the reactivity of lysosomal fractions from transfected HFF cells expressing gpUL37 (lane 1) or gpUL37FLAG (lane 2) with the anti-Grp75 antibody was minimal, indicating that the lysosomal and mitochondrial compartments were well separated in the discontinuous sucrose gradients. The identity of the top band as lysosomes was verified independently by reactivity with anti-LAMP1 antibody, a marker for lysosomes (data not shown).

(24K): Fig. 5. gpUL37FLAG protein is detected in purified mitochondria from transiently transfected HFF and HeLa cells. (a)(b) HFF cells were transfected with p414 (lanes 1 and 3) or p816 (lanes 2 and 4). Forty-eight h after transfection, cells were lysed and mitochondria were purified on discontinuous sucrose gradients. Proteins from purified lysosomes (lanes 1 and 2, 5 µg per lane) and banded mitochondria (lanes 3 and 4, 15 µg per lane) were separated by electrophoresis in a 10% SDSpolyacrylamide gel, blotted and reacted with M2 anti-FLAG antibody (a) or anti-Grp75 antibody (b). (c)(d) HeLa cells were transfected with p414 (lanes 1) or p816 (lanes 2) and mitochondria were purified on sucrose gradients. Proteins (15 µg per lane) were resolved in 10% SDSpolyacrylamide gels and reacted with M2 anti-FLAG antibody (c) or anti-Grp75 antibody (d).

Similar results were obtained in transiently transfected HeLa cells (Fig. 5c, d). gpUL37FLAG (~85 kDa) was detected in purified mitochondria from HeLa cells expressing gpUL37FLAG (Fig. 5c, lane 2) but not in HeLa cells expressing untagged gpUL37 (lane 1). The presence of mitochondria in these subcellular fractions was demonstrated by the reactivity of the fractionated proteins with anti-Grp75 antibody (Fig. 5d, lanes 1 and 2). Thus, gpUL37 traffics detectably to mitochondria in HFF cells and in HeLa cells.

pUL37x1 traffics through the secretory apparatus and to mitochondria
pUL37x1 has anti-apoptotic activity and c-myc-tagged pUL37x1 is known to traffic to mitochondria (Goldmacher et al., 1999 ). We examined co-localization of authentic pUL37x1 and a mitochondrial marker in HFF cells. pUL37x1 is predicted to react with Ab1064, since it shares its amino-terminal residues 2740 with gpUL37. In order to determine whether pUL37x1, which has a hydrophobic leader, traffics through the secretory apparatus, we also examined its co-localization with PDI and Golgi zone markers in transfected HFF cells (Fig. 6). pUL37x1 was detected readily with Ab1064 in transfected HFF cells (Fig. 6a, g). The ER of HFF cells was stained brightly by anti-PDI antibody (Fig. 6 b). pUL37x1 co-localized partially with PDI, as seen by the yellow overlap in the merged optical sections (Fig. 6c). The mitochondria of the same cell were stained brightly by the anti-mitochondrion antibodies (Fig. 6d). pUL37x1 co-localized partially with the mitochondria in the cells, as seen by the aquamarine overlap in the merged optical sections (Fig. 6e). The ER-localized pUL37x1 (yellow overlap) is distinguishable from the mitochondrion-localized pUL37x1 (aquamarine overlap) in the triply merged optical sections (Fig. 6f). Nonetheless, some overlap of the two compartments containing pUL37x1 (white overlap) was observed (Fig. 6f). In order to determine whether pUL37x1 is translocated from the ER to the Golgi apparatus, we examined its co-localization with a Golgi marker (Fig. 6gi). pUL37x1 was stained brightly in transfected HFF cells (Fig. 6g), as was the Golgi apparatus (Fig. 6h). pUL37x1 co-localized partially with the Golgi marker (yellow overlap) in transiently transfected HFF cells (Fig. 6i). In this merged optical section, pUL37x1 (green) was detected in other compartments outside as well as within the Golgi apparatus (yellow). Taken together, our results suggest that pUL37x1, similar to gpUL37, traffics through the ER to the Golgi apparatus and to mitochondria in permissive HFF cells. These results are in contrast to the findings of Goldmacher et al. (1999) , who found complete co-localization of c-myc-tagged-pUL37x1 with mitochondrial markers, as we detected pUL37x1 in both the secretory apparatus and mitochondria.

(90K): Fig. 6. pUL37x1 traffics through the secretory apparatus and to mitochondria. (a)(f) HFF cells transiently expressing pUL37x1 were fixed with methanol and stained with rabbit Ab1064, mouse anti-PDI and human anti-mitochondrion antisera. The cells were then stained with FITCanti-rabbit, TRanti-mouse and Cy5anti-human IgG. Confocal optical sections show the FITC (a), TR (b), FITC/TR (c), Cy5 (d), FITC/Cy5 (e) and FITC/TR/Cy5 (f) channels (z=0·2 µm). (g)(i) HFF cells transiently expressing pUL37x1 were stained with rabbit Ab1064 and mouse anti-Golgi zone antibodies. The cells were then stained with FITCanti-rabbit and TRanti-mouse IgG antibodies. Confocal optical sections show the FITC (g), TR (h) and FITC/TR (i) channels (z=0·5 µm). Bars, 30 µm.

Despite being non-nuclear proteins, gpUL37 and pUL37x1 can regulate nuclear gene expression either positively (Colberg-Poley et al., 1992 , 1998 ; Tenney et al., 1993 ; Zhang et al., 1996 ) or negatively (Biegalke, 1999 ). Nuclear regulation by gpUL37 and pUL37x1 is likely to be distinct from that of HCMV major IE nuclear-regulatory protein IE1, as their amino and carboxy termini do not traffic detectably to the cell nucleus (this study; Zhang et al., 1996 ). IE1 activates its own promoter through 18 bp repeats and NF-κB elements and does not bind TBP, Tef-1 or Oct-1 (Sambucetti et al., 1989 ; Hagemeier et al., 1992 ; Lukac et al., 1997 ). Conversely, IE2 can interact with TBP, TFIIB and TAF130 (Caswell et al., 1993 ; Hagemeier et al., 1992 ; Lukac et al., 1997 ). As gpUL37 and pUL37x1 are detectable by multiple antisera in the ER, Golgi apparatus, PM and mitochondria but not in the nucleus, we predict that their roles in nuclear gene regulation do not involve direct interactions with nuclear transcription complexes, but rather they involve signalling from the ER, Golgi apparatus, PM or mitochondria.

By using a functional screen for anti-apoptotic activities encoded by HCMV, Goldmacher et al. (1999) found that pUL37x1 and gpUL37 have anti-apoptotic activities and that UL37 amino-terminal sequences, including the hydrophobic leader, are required for this activity. They found that pUL37x1 traffics to mitochondria, where it binds the adenosine nucleotide transporter and blocks release of cytochrome c from the mitochondria. The trafficking patterns of gpUL37 were not studied. In the present study, with antisera against authentic UL37 proteins, we found that both pUL37x1 and gpUL37 traffic to mitochondria. Our results also differ from those of Goldmacher et al. (1999) in that we observed pUL37x1 trafficking through the secretory apparatus.

Trafficking of pUL37x1 through the ER and Golgi apparatus was anticipated because of its hydrophobic signal sequence. Classically, hydrophobic signal sequences target proteins to the secretory pathway (Walter & Johnson, 1994 ). Indeed, we found that pUL37x1 and gpUL37, which share leader sequences, traffic through the ER and Golgi apparatus (this paper; Al-Barazi & Colberg-Poley, 1996 ).

Trafficking of gpUL37 to mitochondria was observed in both HFF cells and HeLa cells by using confocal microscopy and subcellular fractionation. Thus, our results were verified by using two different approaches and two different antibodies. It is noted that the molecular mass of gpUL37FLAG (~250 kDa) in mitochondria of transiently transfected HFF cells is considerably greater than the mass obtained (~8385 kDa) following its immunoprecipitation by rabbit antiserum against the gpUL37 N-glycosylation domain (Al-Barazi & Colberg-Poley, 1996 ). In contrast, the molecular mass of gpUL37FLAG protein (~85 kDa) purified from mitochondria of transiently transfected HeLa cells corresponded well to its predicted mass. The difference in molecular mass noted in HFF cells may result from preferential reactivity of Ab1525, which was generated against the N-glycosylation domain, to partially glycosylated gpUL37 rather than to the fully glycosylated species, whereas the FLAG antibody against the carboxy-terminal tag used in these experiments is not impeded from reacting with the fully glycosylated gpUL37 species. Alternatively, gpUL37 purified in the mitochondrial subcellular fractions may be additionally modified covalently in HFF cells.

It is known that the unconventional trafficking of pUL37x1 to mitochondria requires its hydrophobic leader (Goldmacher et al., 1999 ). This sequence requirement is unexpected, as transport to mitochondria usually requires mitochondrial targetting sequences (von Heijne, 1986 ; Ni et al., 1999 ). The amino termini of mitochondrial precursor proteins possess a leader sequence that can be recognized by the mitochondrial import apparatus. The signal does not share any primary sequence identity but shows a bias towards positively charged amino acids, provided mostly through arginine residues (von Heijne, 1986 ). Mitochondrial leader sequences also share the ability to form an amphiphilic α helix. Thus, there are structural and charge components required to target proteins to mitochondria. pUL37x1 and gpUL37 lack any charged amino acids in their amino-terminal leader sequences (Kouzarides et al., 1988 ). However, just downstream of the hydrophobic signal sequence are multiple positively charged residues that could target gpUL37 and pUL37x1 to mitochondria. Alternatively, the hydrophobic leader may target gpUL37 and pUL37x1 to mitochondria by the use of bridging pores between mitochondrial and ER membranes (Ardail et al., 1993 ). Mitochondrion-associated membranes comprise a pre-Golgi compartment and are localized in discrete regions of the cell that do not correspond to the bulk of ER or mitochondria (Rusiñol et al., 1994 ). Regions of the ER are in contact with outer mitochondrial membranes at the contact sites between inner and outer mitochondrial membranes (Ardail et al., 1993 ). Thus, UL37 proteins may traffic to the ER by virtue of their hydrophobic leaders and translocate to mitochondria and/or through the Golgi apparatus to the cellular PM.

The authors thank Dr Robyn Rufner for her advice on the confocal imaging, Dr Kristen Hoffburr for her advice on mitochondrial purification and Dr Victor Goldmacher for providing us with his results prior to their publication. This work was supported, in part, by a Grant-in-Aid from the American Heart Association (to A.C.-P.), Childrens Research Institute Discovery Funds and the Board of Lady Visitors at Childrens National Medical Center.

Footnotes

^b Present address: Laboratory of Immunobiochemistry, Center for Biologics Evaluation and Research, US Food and Drug Administration, Bethesda, MD 20892, USA.