liver sinusoidal endothelial cells are a site of murine

JOURNAL OF VIROLOGY, Sept. 2009, p. 8869–8884 Vol. 83, No. 170022-538X/09/$08.00�0 doi:10.1128/JVI.00870-09Copyright © 2009, American Society for Microbiology. All Rights Reserved.

Liver Sinusoidal Endothelial Cells Are a Site of MurineCytomegalovirus Latency and Reactivation�

Christof K. Seckert, Angelique Renzaho, Hanna-Mari Tervo, Claudia Krause, Petra Deegen,Birgit Kuhnapfel, Matthias J. Reddehase,* and Natascha K. A. Grzimek

Institute for Virology, University Medical Center of the Johannes Gutenberg University, Mainz, Germany

Received 30 April 2009/Accepted 11 June 2009

Latent cytomegalovirus (CMV) is frequently transmitted by organ transplantation, and its reactivationunder conditions of immunosuppressive prophylaxis against graft rejection by host-versus-graft disease bearsa risk of graft failure due to viral pathogenesis. CMV is the most common cause of infection following livertransplantation. Although hematopoietic cells of the myeloid lineage are a recognized source of latent CMV, thecellular sites of latency in the liver are not comprehensively typed. Here we have used the BALB/c mouse modelof murine CMV infection to identify latently infected hepatic cell types. We performed sex-mismatched bonemarrow transplantation with male donors and female recipients to generate latently infected sex chromosomechimeras, allowing us to distinguish between Y-chromosome (gene sry or tdy)-positive donor-derived hemato-poietic descendants and Y-chromosome-negative cells of recipients’ tissues. The viral genome was found tolocalize primarily to sry-negative CD11b� CD11c� CD31� CD146� cells lacking major histocompatibilitycomplex class II antigen (MHC-II) but expressing murine L-SIGN. This cell surface phenotype is typical ofliver sinusoidal endothelial cells (LSECs). Notably, sry-positive CD146� cells were distinguished by theexpression of MHC-II and did not harbor latent viral DNA. In this model, the frequency of latently infectedcells was found to be 1 to 2 per 104 LSECs, with an average copy number of 9 (range, 4 to 17) viral genomes.Ex vivo-isolated, latently infected LSECs expressed the viral genes m123/ie1 and M122/ie3 but not M112-M113/e1, M55/gB, or M86/MCP. Importantly, in an LSEC transfer model, infectious virus reactivated from recipients’tissue explants with an incidence of one reactivation per 1,000 viral-genome-carrying LSECs. These findingsidentified LSECs as the main cellular site of murine CMV latency and reactivation in the liver.

In human cytomegalovirus (hCMV) infection, hematopoi-etic progenitor cells of the myeloid differentiation lineage area recognized cellular site of virus latency (for more-recentreviews, see references 75 and 94), and cell differentiation-dependent as well as cytokine-mediated viral gene desilencingby chromatin remodeling is discussed as the triggering eventleading to virus reactivation (for a review, see reference 7).Although hematopoietic stem cell or bone marrow transplan-tation (BMT) is frequently associated with hCMV reactivationand recurrence in recipients after hematoablative leukemia/lymphoma therapy, the incidence of virus recurrence and dis-ease is highest in the combination of an hCMV-negative donor(D�) and an hCMV-positive recipient (R�) (D�R� � D�R� �D�R�), indicating that donor hematopoietic cells are not theonly source of latent hCMV and actually not the predominantsource (34). Rather, the recipients experience reactivation oftheir own virus. Just the opposite is true in the case of solidorgan transplantation, where the reactivating virus is mostlytransmitted with the transplanted organ (D�R� � D�R� �D�R�) (34). Collectively, these risk assessments support thesuggestion that reactivation, in both instances, occurs in la-tently infected tissue cells, that is, within the recipient’s organsand the transplanted donor organ, respectively. Although tis-

sue-resident cells of hematopoietic origin remain candidates,stromal and parenchymal tissue cells come into considerationas additional sites of CMV latency.

Longitudinal analysis of viral genome load in the latencymodels of murine CMV (mCMV) infection of neonatal mice(9, 91) as well as of adult mice after experimental BMT (8, 62,64) has demonstrated a high viral latency burden in multipleorgans long after clearance of viral DNA from bone marrowand blood (reviewed in reference 92). These findings supportthe suggestion that there exist two types of latency, namely, atemporary latency in hematopoietic cells and a latency in tissuecells that lasts through life. Accordingly, both types of latencymay coexist early after primary infection, while “late latency” isrestricted to organ sites. As we have shown previously in asex-mismatched murine BMT model, bone marrow cells(BMCs) derived from latently infected donors in the phase oforgan-restricted “late latency” cannot transmit latent or reac-tivated infection to naïve recipients upon intravenous celltransfer (99).

A first hint for mCMV latency in stromal or reticular cellswas presented long ago by Mercer and colleagues (73), whoshowed that infected cells during acute infection of the spleenare predominantly sinusoidal lining cells and that latentmCMV can be recovered from a major histocompatibility com-plex class II (MHC-II) antigen-negative and Thy-1 (CD90)-negative “stromal” cell fraction, which includes endothelialcells (ECs). These findings strongly argued against T and Blymphocytes, macrophages, and dendritic cells (DCs) beingmajor reservoirs of latent mCMV in the spleen, a conclusionsupported by later work of Pomeroy and colleagues (86). Sim-

* Corresponding author. Mailing address: University Medical Cen-ter of the Institute for Virology, Johannes Gutenberg University,Hochhaus am Augustusplatz, 55131 Mainz, Germany. Phone: 49-6131-39-33650. Fax: 49-6131-39-35604. E-mail: [email protected].

� Published ahead of print on 17 June 2009.

8869

on March 27, 2018 by guest

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

ilarly, Klotman and colleagues (54) as well as Hamilton andSeaworth (44) concluded that in kidney transplantation, donorkidney is the source of latent mCMV and that the latent viralgenome is harbored by renal peritubular epithelial cells (53). Afirst hint for mCMV latency in ECs within the liver was pro-vided by in situ PCR images presented by Koffron and col-leagues (59) showing nuclear staining in cells with a microan-atomical localization suspicious of liver sinusoidal ECs(LSECs). For hCMV, ECs, in particular those in arterial vesselwalls, are regarded as a site of latency on the basis of thepresence of viral DNA in cells expressing an EC marker (81;for a review, see reference 48), although other authors did notdetect viral DNA in venous vessel walls (95). As discussed byJarvis and Nelson (48), these data are not necessarily conflict-ing but may rather reflect the diversity of EC subsets at differ-ent anatomical locations (21, 27). As far as we know, andreactivation of productive infection from ECs that carry latentviral DNA is not yet formally proven for any type of EC.

Hepatitis is a relevant organ manifestation of CMV diseasein immunocompromised hosts (65), and hCMV reactivationhas been reported to be the most common cause of infectionfollowing liver transplantation, in particular in a D�R� com-bination (34, 76). In the murine model of immunocompro-mised hosts, viral histopathology in the liver is dominated bythe cytopathogenic infection of hepatocytes, leading to ex-tended plaque-like tissue lesions (41, 84; reviewed in reference45). Occasionally, however, in these studies, infected hepaticECs as well as Kupffer macrophages were detected by virus-specific immunohistology or by in situ virus-specific DNA hy-bridization.

Using cell-type-specific conditional recombination of a flu-orescence-tagged reporter virus in Cre-transgenic mice ex-pressing Cre selectively in hepatocytes under the control of thealbumin promoter, hepatocytes were recently identified as themain virus-producing cell type during mCMV infection. InCre-transgenic mice expressing Cre selectively in vascular ECsunder the control of the Tie2 promoter, the reporter virus wasfound to recombine also in LSECs, which released an amountof virus sufficient for virus spread to neighboring hepatocytes,although the virus productivity of LSECs was low and contrib-uted little to the overall virus load in the liver (97).

LSECs represent a unique liver-resident population of anti-gen-presenting cells that bear the capacity to cross-presentantigens to naïve CD8 T cells (68, 69; reviewed in reference57). They constitute the fenestrated endothelial lining of thehepatic sinusoids (15). By separating the sinusoidal compart-ment of the liver from the space of Disse and the liver paren-chyma, LSECs form a boundary surface for sensing of patho-gens and interaction with passenger lymphocytes. Theyperform a scavenger function contributing to hepatic clearanceof bacterial degradation products derived from the gastroin-testinal tract (103; reviewed in reference 57). According to thisphysiological role, antigen presentation by LSECs is associatedwith tolerance induction rather than with triggering an inflam-matory immune response (29, 56, 68, 104).

Here we provide evidence to support the suggestion thatmCMV has chosen the tolerogenic and long-lived LSECs as animmunoprivileged niche for establishing viral latency in theliver.

MATERIALS AND METHODS

Experimental BMT and establishment of latent mCMV infection. Sex-mis-matched but otherwise syngeneic BMT with male BALB/c mice as BMC donorsand female BALB/c mice as BMC recipients was performed as described previ-ously (83). In brief, hematoablative conditioning of 8- to 10-week-old recipientswas achieved by total-body gamma irradiation with a single dose of 6.5 Gy. BMTwas performed 6 h later by infusion of 5 � 106 femoral and tibial donor BMCsinto the tail veins of the recipients. At ca. 2 h after BMT, recipients were infectedwith 105 PFU of mCMV (strain Smith ATCC VR-194, purchased in 1981; thestock virus has been passaged only four times since then) in the left hind footpad.Criteria for the definition of latency were specified previously (for a review, seereference 92) and include the absence of infectivity in key organs of CMVtropism (liver, spleen, lungs, and salivary glands) as well as reduction of viralDNA load in tail vein blood to less than one copy per 104 leukocytes in alongitudinal analysis using M55/gB-specific real-time quantitative PCR (qPCR)(99). Clearance of viral DNA from blood is the criterion that takes longest to befulfilled in the BMT model, usually �8 months (62, 99). Animals were bred andhoused under specified-pathogen-free conditions in the Central Laboratory An-imal Facility (CLAF) of the Johannes Gutenberg University. Animal experi-ments were approved by the Landesuntersuchungsamt Rheinland-Pfalz accord-ing to German federal law § 8 Abs. 1 TierSchG under permission number177-07/021-28.

Quantitation of viral and cellular genomes by qPCR. Quantitation of the viralgene M55/gB, the diploid cellular gene pthrp (71), and the (male) sex-determin-ing region on chromosome Y, gene sry (60), formerly known as the testis-determining gene on chromosome Y, gene tdy (42), was performed by qPCR withprimers and probes specified in the following sections.

(i) M55/gB. For M55/gB-specific qPCR, oligonucleotide 5�-TGCTCGGTGTAGGTCCTCTCCAAGCC-3� (nucleotides [nt] 83,175 to 83,200) (89) (GenBankaccession no. U68299; complete genome) was used as probe gB_Taq_Probe2,oligonucleotide 5�-CTAGCTGTTTTAACGCGCGG-3� (nt 83,137 to 83,156)served as forward primer gB_Taq_For2, and oligonucleotide 5�-GGTAAGGCGTGGACTAGCGAT-3� (nt 83,227 to 83,207) served as reverse primergB_Taq_Rev2, yielding an amplification product of 91 bp.

(ii) pthrp. For pthrp-specific qPCR, oligonucleotide 5�-TTGCGCCGCCGTTTCTTCCTC-3� (nt 177 to 197) (71) (GenBank accession no. M60056) was usedas probe PTHrP_Taq_Probe1, oligonucleotide 5�-CAAGGGCAAGTCCATCCAAG-3� (nt 155 to 174) served as forward primer PTHrP_Taq_For1, and oligo-nucleotide 5�-GGGACACCTCCGAGGTAGCT-3� (nt 255 to 236) served asreverse primer PTHrP_Taq_Rev1, yielding an amplification product of 101 bp.

(iii) sry. For sry-specific PCR, oligonucleotide 5�-AGTTGGCCCAGCAGAATCCCAGC-3� (nt 162 to 184) (42) (GenBank accession no. X55491) was used asprobe Tdy_Taq_Probe1, oligonucleotide 5�-AAGCGCCCCATGAATGC-3� (nt113 to 129) served as forward primer Tdy_Taq_For1, and oligonucleotide 5�-CCCAGCTGCTTGCTGATCTC-3� (nt 216 to 197) served as reverse primerTdy_Taq_Rev1, yielding an amplification product of 104 bp.

Quantitation was performed on the model 7500 real-time PCR system (Ap-plied Biosystems, Foster City, CA) by using dually labeled probes containing afluorescent reporter at the 5� end, a quencher at the 3� end (6-carboxyfluoresceinreporter and 6-carboxytetramethylrhodamine quencher system; Operon Biotech-nologies, Cologne, Germany), and the TaqMan universal PCR master mix (Ap-plied Biosystems). All reactions were performed in a total volume of 20 �l,containing 10 �l of 2� TaqMan universal master mix, a 1 �M concentration ofeach primer, and a 0.25 �M concentration of the probe. Two-step PCR wasperformed with the following cycler conditions: an initial step for 10 min at 95°Cto activate DNA polymerase followed by 50 cycles of 15 s at 95°C and a combinedprimer annealing/extension step for 1 min at 60°C, during which time data werecollected. Standard curves for quantitation were established by using gradednumbers of linearized plasmid pDrive_gB_PTHrP_Tdy (101) as the template.

DNA isolation from host tissues and cells. (i) Isolation of DNA from latentlyinfected liver tissue. To remove circulating intravascular leukocytes, livers fromlatently infected mice were perfused by standard procedures with Gey’s balancedsalt solution (GBSS; 137 mM NaCl, 5 mM KCl, 1.6 mM CaCl2 � 2H2O, 0.9mM MgCl2 � 6H2O, 0.3 mM MgSO4 � 7H2O, 0.2 mM KH2PO4, 1.7 mMNa2HPO4 � 2H2O, 2.7 mM NaHCO3, 5.5 mM glucose, 50 mM HEPES, pH 7.4)and were then homogenized in a model MM300 mixer mill (Qiagen, Hilden,Germany) with a steel ball (0.118 in.) at 30 Hz for 3 min. Twenty-five milligramsof liver homogenate was used for DNA isolation with the DNeasy blood andtissue kit (catalog no. 69506; Qiagen) as described previously (113).

(ii) Isolation of DNA from MACS-enriched cells. Cells of defined cell surfacemarker phenotypes were enriched by positive (immuno)magnetic cell sorting(MACS) as described below, and DNA was then extracted with the DNeasy

8870 SECKERT ET AL. J. VIROL.


http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

blood and tissue kit (Qiagen) as outlined previously (113), according to themanufacturer’s spin column protocol from step 1c onward (87).

(iii) Isolation of DNA from MACS-enriched LSECs for limiting dilution anal-ysis. The DNeasy blood and tissue kit (Qiagen) was used for DNA extraction(see above), except that 500 ng of double-stranded poly(dA-dT) carrier DNA(catalog no. 27-7870; Amersham Biosciences, Piscataway, NJ) was added to eachLSEC preparation sample. DNA was eluted in 200 �l elution buffer. The isolatedDNA was precipitated with 500 �l of absolute ethanol and 20 �l of 3 M sodiumacetate, pH 4.5, overnight at �20°C. Finally, precipitated DNA was dissolved in5 �l Tris buffer, pH 8.0, and the entire sample was used for M55/gB-specificreal-time quantitation.

Virus reactivation in liver tissue explants from latently infected mice. Liverswere perfused as described above and were transferred into minimal essentialmedium plus GlutaMAX (Invitrogen, Paisley, United Kingdom) supplementedwith 10% fetal calf serum (PAA, Pasching, Austria) and 1% penicillin-strepto-mycin (Invitrogen). They were then immediately cut into pieces with an edgelength of 2 to 4 mm, and these were plated separately into 24-well plates (catalogno. 353047; BD Biosciences) with 1 ml medium per well. At defined intervals, 300�l of cell culture supernatant was transferred onto semiconfluent mouse embryofibroblasts for viral plaque assay (PFU assay) by the technique of centrifugalenhancement of infectivity as described in greater detail elsewhere (references64 and 83 and references therein). In all explant cultures, the sampled mediumwas replaced with 300 �l of fresh medium.

Preparation of hepatocyte suspensions. Hepatocytes were isolated frommouse livers as described previously (113), except that the collagenase buffer wassupplemented with 0.05% NB 4 standard-grade collagenase (Serva, Heidelberg,Germany).

Preparation of nonparenchymal liver cell suspensions. Nonparenchymal butliver-resident cells (NPLCs), which account for about 35% to 40% of total livercells (31, 112) and for about 6% of the total liver mass (13), were isolated frommouse livers as described previously (50, 58) with the following modifications.Briefly, mice were euthanized, and the portal vein was dissected and cannulatedwith a 26-gauge needle. Thereafter, the posterior vena cava was opened to allowoutflow of the perfusion solution. Perfusion was performed for 10 s with 0.05%collagenase A (Roche, Mannheim) in GBSS solution with a flow rate of 5ml/min. The liver was then excised, and connective tissue and the gallbladderwere removed. The organ was mechanically separated using forceps, followed bya 45-min incubation in GBSS buffer supplemented with 0.05% collagenase A and2,000 U DNase I (Roche, Mannheim) in a rotary water bath at 37°C and 200 rpm.The resulting cell suspension was then passed first through a sterile steel meshand thereafter through a sterile 100-�m nylon cell strainer (BD Biosciences) andwas washed two times with GBSS at 400 � g for 10 min. For density gradientseparation, the pellet containing the NPLCs was resuspended in 3 ml GBSS andadded to a working solution of 30% [wt/vol] iodixanol (Optiprep; Axis-Shield,Norway) for a final concentration of 17% iodixanol (final density, 1.096 g/ml) toremove debris and enrich the NPLCs. To avoid exsiccation of the cells, thesuspension was overlaid with 2 ml GBSS and centrifuged at 400 � g for 15 minwith the brake turned off. The layer of low-density cells at the interface washarvested, and the NPLCs were washed with GBSS at 400 � g for 10 min. Afterbeing washed, the pellet was resuspended in 10 ml GBSS for cell counting.

Separation and phenotype analysis of NPLCs. The following MicroBeads andantibodies were used for MACS and fluorescence-activated cell sorting (FACS):anti-fluorescein MicroBeads (catalog no. 130-048-701; Miltenyi Biotec, BergischGladbach, Germany), rat anti-mouse CD146 (LSEC-antigen) MicroBeads (cat-alog no. 130-092-007; Miltenyi Biotec), rat anti-mouse CD4 (L3T4) MicroBeads(catalog no. 130-049-201; Miltenyi Biotec), rat anti-mouse CD11b MicroBeads(catalog no. 130-049-601; Miltenyi Biotec), hamster anti-mouse CD11c Mi-croBeads (catalog no. 130-052-001; Miltenyi Biotec), rat anti-mouse CD45RMicroBeads (catalog no. 130-049-501; Miltenyi Biotec), allophycocyanin (APC)-conjugated rat anti-mouse MHC-II antibody (Ab) (clone M5/114, catalog no.130-091-806; Miltenyi Biotec), rat anti-mouse CD16/CD32 Ab (anti-Fc� III/IIreceptor, clone 2.4G2, catalog no. 553142; BD Biosciences), fluorescein isothio-cyanate (FITC)-conjugated rat anti-mouse CD31 Ab (clone 390, catalog no.558738; BD Biosciences), FITC-conjugated rat anti-mouse CD106 Ab (clone429, catalog no. 553332; BD Biosciences), rat anti-mouse LSEC-antigen CD146Ab (clone ME-9F1, catalog no. 130-092-026; Miltenyi Biotec), R-phycoerythrin(R-PE)-conjugated rat anti-mouse CD31 Ab (clone 390, catalog no.MCA1364PE; AbD Serotec, Kidlington, United Kingdom), Alexa Fluor 647-conjugated rat anti-mouse-specific ICAM-3-grabbing nonintegrin-related pro-tein R1 (anti-mSIGN-R1) Ab (clone ER-TR9, catalog no. MCA2394A647; AbDSerotec), and acetylated low-density lipoprotein labeled with Dil dye (Dil-AcLDL; catalog no. L3484; Molecular Probes, Leiden, The Netherlands). LSEC-

specific Abs ME-9F1–FITC and ME-9F1–biotin, used earlier in this study, werea generous gift from A. Hamann, Berlin, Germany.

(i) Immunomagnetic cell sorting. NPLC subpopulations expressing the cellsurface markers CD4, CD31, CD106, CD11b, CD11c, CD45R, and CD146 wereenriched from total NPLCs by two sequential runs of automated MACS with theautoMACS system (Miltenyi Biotec) or, in the case of sterile isolation of cells, byone run of manual magnetic cell sorting. In brief, up to 107 cells were resus-pended in 90 �l MACS buffer (2 mM EDTA in phosphate-buffered saline [PBS]containing 0.5% [wt/vol] bovine serum albumin) and mixed with 10 �l of thecorresponding MicroBeads. For separation of more than 107 cells, MACS bufferand MicroBeads were adjusted accordingly. After 15 min of incubation in thedark at 4°C, followed by washing and resuspension of the respective cells inMACS buffer, immunomagnetic sorting was performed by using the Posseldseparation program (Miltenyi Biotec) for automatic two-column separation or byusing LS columns for manual separation.

(ii) Two-color cytofluorometric analysis. To determine the purity of isolatedLSECs, the cells were incubated with Dil-conjugated AcLDL for 1 h at 37°C.After incubation, the cells were washed and saturated with 1 �g CD16/CD32monoclonal Ab per million cells to block Fc receptor binding sites. After awashing step, cells were labeled with the FITC-conjugated Ab anti-LSEC antigenCD146. Stainings were carried out in FACS buffer (10 mM EDTA and 20 mMHEPES in PBS containing 0.4% [wt/vol] bovine serum albumin and 0.003%[wt/vol] NaN3). All labeling procedures were performed on ice to minimizereceptor capping or receptor internalization. The analysis was performed with aFACSort (BD Biosciences) using CellQuest 3.3 software for data processing.Fluorescence channel 1 (FL-1) represents fluorescein fluorescence (green), andFL-2 represents Dil fluorescence (red).

(iii) Cytofluorometric cell sorting. For the cytofluorometric purification ofCD31� CD146� cells, density gradient-enriched NPLCs (see above) were satu-rated with 1 �g CD16/CD32 monoclonal Ab per million cells to block Fc recep-tor binding sites and were labeled in the combination CD31–R-PE and CD146-FITC. For the cytofluorometric separation of MHC-II� CD146� and MHC-II�

CD146� cells, MACS-enriched LSECs were likewise blocked against Fc receptorbinding and were labeled in the combination MHC-II–APC and CD146-FITC.Cytofluorometric separation of SIGN-R1� CD146� cells from density gradient-enriched NPLCs (see above) was performed after blocking of Fc receptor bind-ing and labeling of the NPLCs in the combination CD146-FITC and SIGN-R1–Alexa Fluor 647. All stainings were carried out in FACS buffer. The cell sorterused was a FACSVantage cell sorter (BD Biosciences). The excitation wave-lengths were 488 and 633 nm. Band pass filters for 525, 575, and 660 nm wereused to measure FITC, PE, and both APC and Alexa Fluor 647 fluorescence,respectively. Sort gates were set on living cells in the forward-versus-side scatterplot as well as on positive FL-1 (FITC) and FL-2 (R-PE) or positive FL-1 (FITC)and FL-4 (APC, Alexa Fluor 647), respectively. Recovered cells were collected inpolystyrene tubes.

Cultivation of MACS-enriched LSECs. MACS-purified LSECs were resus-pended in Dulbecco’s modified Eagle medium plus GlutaMAX (Gibco) supple-mented with 10% fetal calf serum (PAA, Pasching, Austria) and 1% penicillin-streptomycin (Gibco), and were incubated for 24 h in collagen-coated six-wellplates at a density of 3 � 106 LSECs per six-well culture. For collagen coating ofsix-well plates, collagen R solution (Serva, Heidelberg, Germany) was diluted1:10 with sterile water, and 2 ml of the solution was placed into each six-well vial,followed by drying at 25°C.

Immunofluorescence analysis of MACS-enriched LSECs. MACS-purifiedLSECs were cultivated for 24 h on fibronectin-coated coverslips in 24-well platesat a density of 8 � 105 cells per well in the presence of 1.5 �g BODIPYFL-conjugated AcLDL (catalog no. L3485; Molecular Probes) per ml. Afterthree washes with PBS, cell nuclei were stained by incubation of the coverslipsfor 5 min with the DNA-binding, blue-fluorescing dye Hoechst 33342 (catalogno. H3570; Molecular Probes) (1:5,000 in PBS). Finally, cells were washed threetimes in PBS, and the coverslips were mounted with GelMount 295 aqueousmounting medium (catalog no. G0918; Sigma-Aldrich, Steinheim, Germany).

Construction of plasmids. Recombinant plasmids were constructed accordingto established procedures, and enzyme reactions were performed as recom-mended by the manufacturers. Synthetic transcripts were prepared according tothe instructions given in the MEGAscript SP6 kit (Ambion, Austin, TX) tech-nical manual (3).

(i) Plasmid pDrive_M86 for the synthesis of M86 in vitro transcripts. A DNAfragment of M86/MCP (19, 89) was amplified by PCR from DNA derived froma virus stock of mCMV-BAC (bacterial artificial chromosome-cloned mCMVMW97.01) (111). Oligonucleotides M86_For1 (5�-AGGACACGCTGCTGGACAAG-3�), representing map positions 126,827 to 126,808, and M86_Rev1 (5�-TACTCGTGCGCGAAGATGGG-3�), representing map positions 124,361 to

VOL. 83, 2009 CYTOMEGALOVIRUS LATENCY IN THE LIVER 8871


http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

124,380 (GenBank accession no. U68299; complete genome) (89), served asforward and reverse primers, respectively. The 2,467-bp amplification productcontaining an M86 sequence fragment from map positions 124,361 to 126,827was inserted into the pDrive cloning vector (Qiagen, Hilden, Germany) by meansof UA-based ligation to generate the 6,320-bp plasmid pDrive_M86. For use asa template in the in vitro transcription, plasmid pDrive_M86 was linearized bydigestion with SphI.

(ii) Plasmid pSP64_IE1_IE3_poly(A) for the synthesis of immediate early 1(IE1)-IE3 duplex in vitro transcripts. Plasmid pSP64_IE3_poly(A) was gener-ated by subcloning the HindIII/XbaI restriction fragment of plasmid pBlue-ie3(62) in the vector plasmid pSP64_poly(A) (Promega, Madison, WI). An 831-bpie3 fragment consisting of a part of exon 2, intron 2, exon 3, and a part of exon5 was amplified from plasmid pSP64_IE3_poly(A) with primers IE3-SalI_For(5�-AAAGTCGACCGGCCGTCACTTGG-3�), representing map positions5,876 to 5,889 (GenBank accession no. 330543), and IE3-BamHI_Rev (5�-AAAGGATCCCGTGTGGCCTGTACC-3�), representing map positions 8,529 to8,543. The SalI and BamHI restriction sites are underlined. After digestion of theamplificate with BamHI and SalI, plasmid pSP64_IE1_IE3_poly(A) was gener-ated by cloning of the restriction fragment into the BamHI- and SalI-digestedplasmid pSP64_IE1_poly(A) (62).

Isolation of total RNA from LSECs of latently infected mice. RNA was iso-lated using the RNeasy minikit (catalog no. 74106; Qiagen) as described in themanufacturer’s animal cell protocol (88). Cultivated LSECs were lysed directly inthe cell culture well by adding lysis buffer. Lysates were collected with a cellscraper and homogenized by using QIAshredder spin columns (catalog no.79656; Qiagen). Freshly isolated LSECs were pelletized, resuspended in 350 �lof lysis buffer, and homogenized by using QIAshredder spin columns. RNA wasbound to the RNeasy Mini spin column, followed by digestion of contaminatingDNA with DNase using the RNase-free DNase set (catalog no. 79254; Qiagen).After two washing steps, bound RNA was eluted twice in 45 �l of RNase-freewater. To ensure that all contaminating DNA was removed, a second DNasedigestion was performed. The RNA was cleaned by using the RNeasy minikitaccording to the RNA cleanup protocol (Qiagen) (88). Finally, the RNA waseluted in 50 �l RNase-free water.

Analysis and quantitation of viral transcripts. Quantitation of viral transcriptsfrom genes m123 (ie1), M122 (ie3), M112-M113 (e1), and M55 (gB) was per-formed by one-step reverse transcriptase (RT) qPCRs (RT-qPCRs) with primersand probes specified previously (100). For the IE1-IE3 duplex RT-qPCR, thedually labeled probe ie1-taq1 consisted of a 6-carboxyfluorescein reporter atthe 5� end and a Black Hole Quencher (BHQ1a) at the 3� end. Probe ie3-taq1consisted of a Cy5 reporter at the 5� end and a Black Hole Quencher(BHQ3a) at the 3� end. For quantitation of viral transcript M86, which codesfor the major capsid protein (MCP), oligonucleotide 5�-TCGGCCGTGTCCACCAGTTTGATCT-3� (nt 125,996 to 126,020; GenBank accession no.U68299; complete genome) (89) was used as probe M86_Taq_Probe1. Oligo-nucleotide 5�-GGTCGTGGGCAGCTGGTT-3� (nt 125,977 to 125,994)served as forward primer M86_Taq_For1, and oligonucleotide 5�-CCTACAGCACGGCGGAGAA-3� (nt 126,041 to 126,023) served as reverse primerM86_Taq_Rev1, yielding an amplification product of 65 bp.

Quantitations were performed on a model 7500 real-time PCR system (Ap-plied Biosystems) by using dually labeled probes containing a fluorescent re-porter at the 5� end and a quencher at the 3� end (Operon Biotechnologies). Thecorresponding in vitro transcripts IE1, IE3, early 1 (E1), gB, and MCP weregenerated as outlined above and described previously (62, 100) and were used asstandards for quantitation. Standard titrations ranged from 106 to 10 in vitrotranscripts, with 100 and 10 transcripts being measured in duplicate and tripli-cate, respectively. IE1 plus IE3 (duplex RT-qPCR), E1, gB, and MCP transcriptsin LSECs from latently infected mice were quantified from triplicate 1/25 (4%)aliquots of the yield of total RNA from each sample, and the number of tran-scripts in 500 ng of total RNA was extrapolated. Reactions for gB and MCPtranscripts were performed in a total volume of 25 �l, containing 5 �l of 5�Qiagen OneStep RT-PCR buffer, 1 �l of Qiagen OneStep RT-PCR enzyme mix,668 �M of each deoxynucleoside triphosphate, 0.76 �M of each primer, 0.26 �Mprobe, an additional 1.5 mM concentration of MgCl2, and 0.132 �M ROX(5-carboxy-X-rhodamine) as the passive reference. The reaction mixtures for IE1plus IE3 transcripts in the duplex RT-qPCR and for E1 transcripts in thesingleplex RT-qPCR were modified in that the primer concentrations were 0.4�M and 1 �M, respectively. Reverse transcription was performed for 30 min at50°C. The cycle protocol for cDNA amplification started with an activation stepfor 15 min at 95°C and was followed by 50 cycles of denaturation at 94°C for 15 sand a combined primer annealing/extension step at 60°C for 1 min. The efficacyof the RT-qPCRs was �90% throughout. In the case of unspliced gB and MCP

transcripts, the reaction was also performed in the absence of RT to excludeamplification from template DNA instead of RNA.

Virus reactivation from latently infected cells. For identifying cell types capa-ble of virus reactivation, LSEC-enriched (positive MACS) and LSEC-depleted(MACS flowthrough) NPLCs derived from three latently infected XY-XX chi-meras were transplanted into naïve female recipients that were immunocompro-mised (by 6.5 Gy of total-body gamma irradiation 24 h prior to transfer). Onealiquot from each donor cell fraction was used to determine viral DNA load andthe proportion of donor-derived sry-positive cells. Both fractions were adjusted inPBS to 2 � 107 cells/ml, and 1 ml of the respective cell suspension was admin-istered intravenously to each recipient in two rounds at 500 �l. One day aftertransplantation, the lungs were excised and cut into pieces with an edge length ofaround 2 to 4 mm. The lung explants were placed into 24-well culture plates withno feeder cells and were kept in culture for up to 8 weeks. At weekly intervals,300 �l of cell culture supernatant was replaced with fresh culture medium andwas transferred onto semiconfluent mouse embryo fibroblasts for a virus plaqueassay (PFU assay) by the technique of centrifugal enhancement of infectivity (seeabove).

Mathematical calculations and statistical analysis. (i) Calculation of meanvalues and their 95% confidence limits. Mean values and standard deviationswere calculated by using the standard deviation calculator provided on thewebsite http://invsee.asu.edu/srinivas/stdev.html. On the basis of the so-deter-mined mean value, the standard deviation, and the number of samples, the 95%confidence intervals (CI) of the mean were calculated by a program provided onthe website http://glass.ed.asu.edu/stats/analysis/mci.html.

(ii) Limiting dilution analysis. The frequency (reciprocal of the most probablenumber [MPN]) of cells carrying latent viral DNA and the corresponding 95% CIof the MPN were estimated by limiting dilution analysis (66) using the maximum-likelihood method for calculation (35). The calculation is based on the Poissondistribution equation �ln f(0), where is the Poisson distribution parameterlambda and f(0) is the experimentally determined fraction of negative samples/replicates, that is, in this specific case, samples negative for viral DNA in theqPCR for graded numbers of cells seeded. By definition, is 1 for an f(0) of 1/e.Accordingly, in a semilogarithmic plot of cell numbers seeded (abscissa) and �lnf(0) (ordinate), the MPN is revealed as the abscissa coordinate of the point ofintersection between 1/e and the calculated regression line.

(iii) Calculation of positive samples at a given Poisson distribution parameterlambda. If a number of events (n), that is, the number of viruses that reactivatedin this specific case, occurs in a total number of samples (N), that is, in platedlung tissue pieces in this specific case, then the Poisson distribution parameter iscalculated by n/N. Since each positive sample may comprise one event ormore than one event, the number of positive samples is actually less than thenumber of events. The fraction of negative samples [F(0)] can be calculatedaccording to the Poisson distribution equation as F(0) e�. Accordingly, thefraction of positive samples can be calculated by the equation F(�0) 1 � F(0).In turn, the number of events can be higher than the number of positive samples.So, if the fraction of negative samples [F(0)] is known, the Poisson distributionparameter can be calculated as �ln F(0). The fractions of positive samplescontaining n events [F(n)] can be calculated by the formula F(n) /n � F(n � 1)for all n � 0. The total number of events is then the sum of all F(n) multipliedby the total number of samples.

(iv) Significance analysis with two-by-two contingency tables. Two-sided P val-ues from observed two-by-two contingency tables were calculated by using Fish-er’s exact test (method of the sum of small P values) provided for online calcu-lation on the website http://www.quantitativeskills.com/sisa/statistics/fisher.htm,according to the recommendations of SISA (Simple Interactive Statistical Anal-ysis) (1, 110). Groups are regarded as being significantly different for the testparameter, virus reactivation in LSEC-enriched and LSEC-depleted cells in thisspecific case, if P is �0.05.

RESULTS

Establishment of reactivatable viral latency in the livers ofsex-mismatched bone marrow chimeras. BMT leads to a mixedchimerism in transplantation recipients with cells of hemato-poietic lineages being donor derived and repopulating the re-cipients’ bone marrow stroma and lymphoid tissues. Usingmale donors and female recipients, the Y chromosome, rep-resented by the sry/tdy gene (42, 60), which is detectable byqPCR, can thus serve as a molecular marker to identify and



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

quantitate donor-derived cells in organs of the XY-XX chime-ras (Fig. 1A). Viral pathogenesis and immune system recon-stitution in this model have been studied in great detail previ-ously (for a review, see reference 45). Notably, mCMV

infection of BMT recipients was found to inhibit the engraft-ment of hematopoietic donor-derived stem and progenitorcells in the recipients’ bone marrow stroma by nonproductiveinfection of stromal cells associated with a deficient expressionof hemopoietins (72, 105). Whereas this can result in a lethalbone marrow aplasia under conditions of low-dose BMT, ahigh-dose syngeneic or even MHC-I-disparate BMT facilitatestimely immune reconstitution and control of productive infec-tion in host organs by CD8 T cells (2, 46, 84), leading to virallatency (64). Here we have chosen the high-dose BMT proto-col to control acute infection and establish latency in theXY-XX chimeras. As shown in Fig. 1B for three individualchimeras after clearance of productive infection as well asqPCR-controlled clearance of the viral genome from vascularcompartment passenger leukocytes (99; data not shown), res-ident cells in perfused liver tissue harbored latent viral ge-nomes at a load that did not apparently correlate with theproportion of donor-derived sry� cells of hematopoietic origin.Variance in the proportion of sry� cells likely reflects variablelymphoid cell infiltration of the liver, but this was not investi-gated further here. The definition of latency implies that viralgenomes are not replication defective but can be reactivated toproductive infection (96). This was verified by liver tissue ex-plant cultures (Fig. 1C). Whereas absolutely no virus was re-covered from homogenized tissue material derived from thesame livers (negative data not shown), recurrence occurredfrom the explanted tissue pieces with an incidence of �20%.

The latent viral genome localizes to nonparenchymal livercells. Liver tissue can roughly be subdivided into liver paren-chymal cells, namely, the hepatocytes, and NPLCs, which in-clude LSECs, Kupffer-type macrophages, liver-specific naturalkiller (NK) or pit cells, and hepatic stellate cells, formerlyknown as fat-storing and vitamin A-rich perisinusoidal Ito cells(55). Separating liver tissue into hepatocyte and NPLC frac-tions revealed an enrichment of latent mCMV DNA in theNPLC fraction (Fig. 2A, top panel), which accounts for only�6.5% of the liver volume but �40% of the total number ofhepatic cells. This finding excludes the main virus producer inacute infection, the hepatocyte (97), from the candidate list oflatently infected liver cell types. In agreement with the fact thatNPLCs comprise long-lived liver-resident cells and short-livedrepopulating cells of hematopoietic origin, �40 to 50% of thecells in the NPLC fraction were of the sry� genotype (Fig. 2A,bottom panel). The NPLC fraction was then further subdividedby positive immunomagnetic cell sorting specific for a panel ofcell surface marker molecules for phenotype screening (Fig.2B). This localized latent viral DNA to two cell surface marker-defined cell populations, namely, to CD31� NPLCs and toCD146� NPLCs (Fig. 2B, top panel), the two cell populationswith the lowest proportion of cells of donor genotype (Fig. 2B,bottom panel). This excludes a major contribution of donor-derived hematopoietic lineage NPLCs, including Kupffer cells,liver NK cells, and also hepatic stellate cells (6), to the load ofthe latent viral genome in the liver. Recipient origin combinedwith expression of CD146, also known as the LSEC antigenthat is expressed on vascular but not on lymphoid ECs (98),pointed to LSECs as the predominant cellular site of hepaticmCMV latency.

Identification of CD31� CD146� LSECs as a cellular site ofhepatic mCMV latency. Since CD31 is a known marker for

FIG. 1. Reactivation of mCMV from liver explants of latently infectedbone marrow chimeras. (A) Experimental regimen. Sex-mismatched al-logeneic BMT was performed with male BALB/c mice (XY; sry�) asBMC donors and female BALB/c mice (XX; sry�) as BMC recipients thatwere immunocompromised by total-body gamma irradiation with a singledose of 6.5 Gy and infected with wild-type (WT) mCMV, strain Smith.Viral latency and reactivation were usually studied at a minimum of 8months after acute infection, when the viral genome was naturally clearedfrom circulating leukocytes. Note that throughout this paper, individuallytested bone marrow chimeras are numbered consecutively. (B) Quanti-tation of latent viral genomes and of donor-derived cells in the livers ofthree individual bone marrow chimeras (mice 1 to 3) at 9 months afterBMT and infection. M55/gB- and sry-specific qPCRs were performed withDNA isolated from 25-mg pieces of the livers. Circles represent triplicatemeasurements of each sample DNA (see inset legend). Median values aremarked. The viral genome load is shown normalized to 1 � 105 liver cells.Tapered wedges symbolize the proportions of donor (D)- and recipient(R)-derived cells. (C) Reactivation of infectious mCMV in 24 liver explantcultures from each of the same three individual mice for which latent viralDNA load and chimerism are shown in panel B. Bars represent numbersof positive liver explant cultures (left ordinate scale) and incidences ofreactivation (right ordinate scale).



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

ECs, although not expressed exclusively by ECs (32, 78), it wasat hand to suggest that the viral-genome-carrying, recipient-derived CD31� NPLCs and CD146� NPLCs defined abovemay actually be one and the same cell population, specifically,the LSECs. We therefore performed two-color cytofluoromet-ric staining of NPLCs for the expression of CD31 and CD146,followed by cell sorting of a coexpressing CD31� CD146�

population (Fig. 3). Besides this LSEC population, presortanalysis also revealed NPLC subpopulations characterized bythe phenotypes CD31� CD146� and CD31� CD146�, whichinclude CD31� stellate cells (116) and most leukocytes, as wellas a minor subpopulation with the phenotype CD31�

CD146low, which includes a subpopulation of CD49b� NKcells (reference 98 and our own data not shown). The latentviral genome clearly copurified with the sorted CD31�

CD146� population predominantly of the sry-negative geno-type, whereas the different quantities of cells located outsidethe sorting gate and originating primarily from the donor col-lectively contained only trace amounts of viral genomes de-tected in some but not all replicate qPCRs of a sample. Sinceno physical separation method gives an absolute purity, the lownumbers of viral genomes detected in the LSEC-depleted cellfraction may rather represent a contaminant of LSECs, whichcontain �100-fold-higher numbers of viral genomes. Thesefindings were reproduced in five independent cell sorts per-formed with NPLCs derived individually from five latentlyinfected mice (Fig. 3), and they confirmed a preceding sortexperiment performed with NPLCs pooled from livers of threelatently infected mice (data not shown). Finally, immunomag-netically purified CD146� cells, which did not contain numer-able numbers of CD11c� DCs or CD11b� macrophages (neg-ative data not shown), were classified as LSECs functionally bytheir high endocytotic capacity to take up AcLDL (10, 12, 77,82) ex vivo (Fig. 4A) as well as after a 24-h period in cellculture (Fig. 4B).

Donor-derived MHC-II� cells expressing the LSEC antigenCD146 do not harbor the latent viral genome. The cell surfacemarker phenotype of LSECs is debated in the literature (for areview, see reference 33). Whereas LSECs were reported toconstitutively express MHC-II as well as CD11c, sharing thesemarkers with DCs (reviewed in references 57 and 67), Katz andcolleagues presented conflicting results showing low or absentexpression of both MHC-II and CD11c by LSECs (50). Tomake confusion perfect, Onoe and colleagues (79) concludedthat murine LSECs express MHC-II but not CD11c. Impor-tantly, as shown above (Fig. 2), in our model, CD11c� cellswere primarily donor derived and did not carry latent viralDNA, which excludes DCs as well as putative CD11c� LSECsfrom the list of candidates. In our hands, and in agreementwith Katz and colleagues (50), immunomagnetically purifiedFIG. 2. Phenotyping of liver cells that carry latent viral DNA.

(A) The latent viral genome localizes to nonparenchymal liver cells. At9.5 months after sex-mismatched BMT and infection, DNA was pre-pared from 25 mg of unseparated liver tissue from three individualbone marrow chimeras (mice 4 to 6), from 5 � 106 isolated hepatocytes(chimeras 7 to 9), and from NPLCs (chimeras 10 to 12). Note thattechniques for optimizing the purity of hepatocytes and NPLCs aremutually exclusive. Latent viral DNA loads and chimerism were de-termined by qPCRs specific for the viral gene M55/gB and the alloso-mal cellular gene sry. (B) DNA load and chimerism in subsets ofNPLCs. NPLCs were isolated from a pool of five livers, and CD4�,CD31�, CD106�, CD11b�, CD11c�, CD45R�, and CD146� subsetsthereof were purified by immunomagnetic cell sorting (MACS). DNA

was prepared from 5 � 106 cells of each sorted subset, and genesM55/gB and sry were quantitated by qPCR for determining viral DNAload and chimerism. Throughout, dot symbols represent triplicatemeasurements from each sample and the short horizontal bars markthe median values. Negative data are indicated below the dotted line.Tapered wedges symbolize the proportions of donor (D)- and recipient(R)-derived cells.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

CD146� cells all coexpressed high levels of CD31 but failed tocoexpress CD4, CD45R, and CD11c (negative data notshown); yet, a small subset of the cells indeed expressedMHC-II (Fig. 5A), in accordance with the work of Onoe et al.(79). Since various MHC-II-negative cell types conditionallyexpress MHC-II upon activation by cytokines, by gamma in-terferon in particular (106), we surmised that the MHC-II�

CD146� cells may not represent an independent cell popula-tion but rather LSECs in a state of activity, which might in partexplain the different findings in the literature. This assumption,however, proved to be wrong. Cytofluorometric cell sorting ofimmunomagnetically enriched CD146� cells into MHC-II�

and MHC-II� subsets (Fig. 5A) revealed a clear-cut distinctionin that the MHC-II� cells were in the majority sry negative andthus recipient derived, whereas the MHC-II� cells were for themost part of the sry� genotype and thus donor derived (Fig.5B). Importantly, the latent viral DNA localized primarily tothe recipient-derived MHC-II� population of LSECs (Fig.5C). Since our interest was focused on the cellular site ofmCMV latency in the liver, we did not here further investigate

the precise nature of the donor-derived MHC-II� CD146�

population.The latent viral genome localizes to CD146� cells coexpress-

ing the murine homolog of L-SIGN. Whereas LSECs quanti-tatively dominate the hepatic EC population, most cell surfacemarkers are shared between microvascular LSECs and con-ventional ECs of the liver macrovasculature. As discussed inthe critical review article by Elvevold and colleagues on thecontroversial identity of LSECs (33), the human C-type lectinL-SIGN, a homolog of DC-SIGN also known as DC-SIGN-related molecule DC-SIGNR, is strongly and constitutivelyexpressed on LSECs and lymph node endothelium but not onDCs, other hepatic cells, or other endothelia. As shown byGeijtenbeek and colleagues (38), the murine DC-SIGN ho-molog mSIGN-R1 is more closely related to L-SIGN than toDC-SIGN itself because it is specifically expressed by LSECsand not by DCs and may thus be referred to as murine L-SIGN(mL-SIGN).

It was therefore of interest to use ml-SIGN (mSIGN-R1) incombination with the LSEC antigen CD146 to distinguish be-

FIG. 3. Purification of CD31� CD146� LSECs by two-color cytofluorometric cell sorting. (A) At 10 months after sex-mismatched BMT andinfection, density gradient-enriched NPLCs derived from livers of five individual mice were separated by cytofluorometric cell sorting into CD146�

CD31� LSECs and all remaining non-LSEC NPLCs. Shown are two-dimensional dot plots of CD31-stained and CD146-stained NPLCs of onerepresentative mouse out of five mice tested, before (Pre-sort) and after (Post-sort) cell sorting. The presort dot plot represents a total of 6,676cells, with 2,345 cells (35%) contained within the indicated sort gate and 4,331 cells outside of the sort gate. Postsort dot plots reveal the enrichment(upper plot) and the depletion (lower plot) of CD31� CD146� LSECs, respectively. Percentages indicate the proportions of cells within theindicated electronic gates. (B) For the five individually tested mice (mice 13 to 17), latent viral DNA loads and chimerism were determined byqPCRs specific for M55/gB and sry, respectively, in all NPLCs before cell sorting (NPLC fraction), in CD31� CD146� LSECs (LSEC-enrichedfraction within the electronic gate), and in non-LSEC NPLCs (LSEC-depleted fraction outside the electronic gate). For the meaning of symbols,see the legend for Fig. 2.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

tween LSECs and conventional ECs. It is instructive to recallthat, unlike mL-SIGN, CD146 is not expressed on lymph nodeendothelium (98), so that coexpression of these two markerssimultaneously excludes lymphoid ECs and conventional ECs.As shown in Fig. 6, by M55/gB-specific qPCR after two-colorcytofluorometric cell sorting, the latent viral genomes localizeto mL-SIGN� CD146� cells, which further supports the con-clusion that LSECs harbor the latent viral DNA. It should benoted that these cells were genotyped as sry�, that is, as recip-ient derived (data not shown).

In summary of the series of phenotyping studies, mCMVlatency is established primarily in the recipient-derived popu-lation of MHC-II� CD11b� CD11c� CD31� mL-SIGN�

CD146� LSECs.Viral gene expression in latently infected LSECs is limited

to IE genes. By definition, viral latency implies the absence ofvirus production (96) but does not exclude limited transcrip-tional activity by temporary desilencing of certain viral genes.

As shown previously for mCMV latency in the lungs, the majorIE (MIE) locus is sporadically active during latency withoutfurther progression in the productive transcriptional program(40, 62; reviewed in reference 93). Since such a transcriptionalactivity is relevant to immune control of latency by peptide-specific CD8 T cells (100), we tested transcription in purifiedCD146� LSECs derived from five latently infected XY-XXchimeras individually either directly after the ex vivo purifica-tion or after a 24-h period in cell culture (Fig. 7). In the ex vivosamples, IE1 and IE3 transcripts, which are generated from acommon precursor transcript by differential splicing (51, 52,74), were detected in various quantities in LSEC preparationsfrom all five and from four out of five mice, respectively.Despite the presence of IE3 transcripts encoding the essentialtransactivator of early (E)-phase gene expression (4, 74), thetranscriptional program did not proceed to the E1/M112-M113transcript (17, 24), and accordingly, the E-phase M55/gB andthe late-phase M86 transcripts were undetectable.

Latently infected LSECs proceed to E-phase gene expres-sion upon cultivation. A 24-h period in cell culture enhancedIE transcription and induced a progression to E1 transcriptionin LSECs from four out of five mice tested, and thresholdamounts of the M55/gB transcripts were detected in the cellcultures occasionally (Fig. 7). In contrast, true late M86 tran-scripts encoding the essential MCP (107) remained absentthroughout, so that virus capsid assembly could not take place.Consequently, complete reactivation to infectious virions didnot occur in cell cultures of latently infected LSECs (negativedata not shown).

Frequency of latently infected LSECs and copy number ofthe latent viral genome. A limiting dilution approach was usedto determine the frequency (minimum estimate) of viral-ge-nome-carrying, latently infected cells present in a populationof purified CD146� LSECs. Graded numbers of cells in 12replicates were subjected to M55/gB-specific qPCR using plas-mid pDrive_gB_PTHrP_Tdy as the standard (101). The cutoffcycle threshold value for distinguishing between viral-DNA-negative and -positive samples was defined by extrapolation ofthe standard curve to one template and was found to be 41(Fig. 8A). Accordingly, samples were classified as negativesamples when no signal was obtained after 41 amplificationcycles. The frequency estimate (MPN for a of 1) was thenobtained from the proportions of negative samples accordingto the Poisson distribution equation and was found to be 1latently infected cell in 14,600 (95% CI, 9,300 to 23,000)LSECs (Fig. 8B).

With this information and based on the latent viral genomeload determined by qPCR for the very same LSEC preparation(mean, 593; 95% CI, 427 to 760 genomes per 106 cells; six testsamples), it became possible to calculate the average numberof viral genomes per latently infected LSEC as 9 (range, 4to 17).

Virus can reactivate in lung explant cultures after LSECtransfer. As shown in Fig. 1C, latent mCMV can reactivate inliver explant cultures, and in general, tissue explants have longbeen known to provide a supporting microenvironment forefficient virus reactivation (20, 49, 114). Although we have sofar demonstrated the presence of latent viral DNA in LSECs,there remained the question of whether these genomes werefully intact and accounted for the reactivation observed in

FIG. 4. Functional analysis of LSECs. (A) Immunomagneticallyenriched ex vivo LSECs, with no prior cultivation, were incubated for30 min with Dil-conjugated AcLDL and counterstained with FITC-conjugated anti-LSEC (anti-CD146) Ab. Cells were analyzed by two-color flow cytometry for endocytotic uptake of AcLDL (FL-2, carbo-cyanine dye Dil) and expression of CD146 (FL-1, FITC). (Left panel)Physical properties of LSECs as defined by size (forward scatter [FSC])and granularity (sideward scatter [SSC]). The live gate for excludingdead cells and debris is indicated. (Right panel) Two-dimensional dotplot of AcLDL and CD146 staining of 30,000 cells analyzed, with 4,600cells displayed as dots. Percentages of cells located in the four quad-rants are indicated and reveal the purity of the LSEC preparation inthe upper right quadrant. (B) Immunomagnetically enriched CD146�

LSECs were cultured for 24 h in the presence of 1.5 �g BODIPYFL-labeled AcLDL. The endocytotic activity of LSECs was determinedby the uptake of AcLDL visualized by fluorescence microscopy withgreen cytoplasmic staining for AcLDL and blue staining of cell nucleiby Hoechst dye 33342. The bar marker represents 20 �m.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

unseparated liver explants, as only this fulfills the definition oflatency in LSECs.

We approached this problem by intravenous adoptive trans-fer of purified CD146� LSECs, which were derived from la-tently infected BALB/c mice after sex-mismatched BMT, intoimmunocompromised and uninfected female BALB/c recipi-ents (Fig. 9). Since transferred LSECs are mostly trapped inthe vascular bed of the lungs at 24 h after transfer, as revealedwith normal, male LSEC donors by sry-specific qPCR in femalerecipients (data not shown), virus reactivation was studied inlung explant cultures with the hope that the heterologous en-vironment of lung tissue would also support virus reactivationin LSECs. Specifically, two indicator recipients each received2 � 107 purified LSECs, the highest feasible cell dose fortransfer, whereas four control recipients each received 2 � 107

NPLCs depleted of LSECs from the very same latently in-fected donors. At 24 h after the cell transfer, lungs of therecipients were each subdivided into 48 tissue pieces andplated into two 24-culture plates (Fig. 9A). It was not beforeweek 5 after explantation that infectious virus became detect-able in a total of 3 out of 96 cultures in the LSEC group,

whereas during the observation period of 8 weeks, no reacti-vation was observed in the total of 192 explant cultures of theLSEC-depleted NPLC control group. Importantly, whereas re-activation was observed in explants from both recipients ofLSEC-enriched NPLCs, no reactivation was observed in ex-plants from four recipients of LSEC-depleted NPLCs (Fig.9B). According to Fisher’s exact probability test (Fig. 9C1), thedifference between the two groups is significant, with a P valueof 0.036 (two-sided). This means that reactivation occurredmore frequently from purified LSECs than from all the re-maining NPLCs in an LSEC-depleted population.

The LSEC preparations regularly contained �5% non-LSEC NPLCs. So, if one argues as a counterhypothesis that thethree observed reactivation-positive lung pieces in the LSECtransfer group, which according to the Poisson distributionequals three reactivation events, rather came from 5% con-taminating non-LSEC NPLCs, which is a worst-case assump-tion, an extrapolated number of 60 reactivations should haveoccurred in 96 explants from an LSEC-depleted NPLC prep-aration. Since, according to the Poisson distribution, more thanone reactivation could occur from any of the plated lung

FIG. 5. The latent viral genome localizes to CD146� MHC-II� cells of recipient origin. (A) Immunomagnetically enriched CD146� cellsderived from livers of latently infected bone marrow chimeras at 11 months after sex-mismatched BMT and infection were separated into CD146�

MHC-II� and CD146� MHC-II� subsets by two-color cytofluorometric cell sorting. Shown are two-dimensional dot plots of MHC-II-stained andCD146-stained LSECs of one representative mouse out of three mice tested, before (Pre-sort) and after (Post-sort) cell sorting. The presort dotplot represents a total of 7,524 cells, with 6,089 cells (81%) within gate 1 (G1) representing CD146� MHC-II� cells and 752 cells (10%) withingate 2 (G2) representing CD146� MHC-II� cells. Postsort dot plots reveal the purity of the sorted subsets. Percentages reveal the proportions ofcells contained within the indicated electronic gates. (B and C) Chimerism (B) and latent viral DNA loads (C) determined for three individuallytested mice (mice 18 to 20) by qPCRs specific for sry and M55/gB, respectively, in all immunomagnetically enriched cells before cytofluorometriccell sorting (CD146� fraction) as well as in sorted cells of G1 (CD146� MHC-II� fraction) and G2 (CD146� MHC-II� fraction). For the meaningof symbols, see the legend for Fig. 2.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

pieces, 60 postulated reactivations should have led to 45 virus-positive lung explant cultures out of 96 cultures tested, insteadof none in 192 cultures, as was actually observed. The proba-bility for observing such a difference by chance (null hypothe-sis) is zero (Fisher’s test, P � 10�10, two-sided) (Fig. 9C2).These findings and statistical considerations strongly supportthe conclusion that reactivation originated from LSECs.

These data also allow us for the first time to give a minimalestimate of the frequency of reactivations. According to themeasured frequency of viral-genome-carrying LSECs (Fig. 8),a total of 4 � 107 transferred LSECs contained 2,740 (95% CI,1,740 to 4,300) latently infected cells, which led to three reac-tivations. Thus, the frequency of latently infected LSECs inwhich virus reactivated was in the region of 1 in 1,000, that isf is approximately 10�3, under the specific conditions of theexperimental model studied here.

Altogether, these data have identified LSECs as the maincellular site of reactivatable mCMV latency in the liver.

DISCUSSION

LSECs are a highly interesting cell population constitutingthe interface between the microvascular sinusoidal compart-ment of the liver and the liver parenchyma. LSECs sense

FIG. 6. Detection of latent viral genomes in LSECs coexpressingCD146 and the murine homolog of L-SIGN. CD146� SIGN-R1�

LSECs were purified from density gradient-enriched NPLCs, followedby two-color cytofluorometric cell sorting. NPLCs were derived frommouse livers at 12 months after sex-mismatched BMT, and cell sortingswere performed individually for four latently infected chimeras (mice21 to 24). Viral genomes were quantitated by M55/gB-specific qPCR inNPLCs prior to cell sorting (left panel) as well as in sorted CD146�

SIGN-R1� LSECs derived thereof (right panel). Note that sorted cellswere mostly of the sry� genotype and thus recipient derived (data notshown). Filled circles represent triplicate qPCR data for each sampleDNA preparation normalized to 1 � 105 cells. Median values areindicated by short horizontal bars.

FIG. 7. Viral gene expression in latently infected LSECs ex vivo and after cultivation. Immunomagnetically enriched CD146� LSECs wereseparately prepared from livers of five latently infected mice (mice 25 to 29) at 11 months after sex-mismatched BMT and infection and were testedfor viral transcripts either directly ex vivo (Ø cultivation) or after a 24-h period of cultivation (24h cultivation). Highly purified, DNA-free totalRNA was prepared from 4 � 106 LSECs, and triplicate 1/25 aliquots of each sample were used for the absolute quantitation of IE1, IE3, E1,M55/gB, and M86/MCP transcripts by the respective gene-specific RT-qPCRs using corresponding synthetic transcripts as standards. Theexperimentally determined numbers of transcripts were normalized to 500 ng of total RNA. Dot symbols represent triplicate measurements foreach sample, with the median values marked by short horizontal bars. Negative data are indicated below the dotted line. Since genes M55/gB andM86/MCP have no exon-intron structure, PCRs were performed also with no RT (open circles) to control for amplification from contaminatingDNA and were found to be negative throughout.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

pathogens through Toll-like receptors and interact with pas-senger lymphocytes. Notably, they can act as antigen-present-ing cells but induce tolerance of T cells instead of effectorfunctions to avoid immunopathological inflammation thatwould interfere with their physiological scavenger function (re-viewed in reference 57).

The fact that mCMV targets LSECs during an acute infec-tion in vivo was shown recently by cell-type-specific recombi-nation of a fluorescence (enhanced-green fluorescence protein[EGFP])-tagged reporter virus expressing EGFP upon removalof a loxP-flanked transcriptional stop sequence in Cre-trans-genic mice expressing Cre under the control of a cell-type-selective promoter (97). Specifically, recombination in Tie2-creor Tek-cre mice, which express Cre selectively in vascular butnot in lymphoid ECs (26), led to the coexpression of EGFPand intranuclear viral IE1 protein in cells of EC morphology,identified as LSECs by their expression of CD31 and LYVE-1

combined with their localization in sinusoids distant from ma-crovascular vessels. Although virus recombined in LSECs ofTie2-cre mice was in principle able to spread to neighboringhepatocytes, the amount of virus released by LSECs was verylow and contributed little to the overall virus production in theliver. In contrast, reporter virus that recombined selectively inhepatocytes of Alb-Cre mice accounted for almost all of thevirus production in the liver (97). Importantly, LSECs have noeffective barrier function against mCMV infection of the liver,and a replication passage through LSECs is not required (97)since the virus can reach the hepatocytes directly, probably viathe fenestrae in the discontinuous sinusoidal endothelium (15).Transcytosis of LSECs (109), as suggested for other hepato-tropic viruses (16, 28), does not appear to be a relevant mech-anism for mCMV since, in most organs, floxed reporter virusfailed to traverse endothelia of Tie2-cre mice without beingrecombined (97).

As we have shown here, LSECs are a site of mCMV latency,whereas hepatocytes are not. In agreement with previous con-clusions (8, 11), this shows clearly that establishment of latencyis not positively correlated with virus productivity in a certaincell type, although the overall virus production in the host islikely of importance for efficient virus dissemination to thecellular sites of latency and thus determines the gross load oflatent viral genomes in host tissues (25, 91). That latency is notestablished in hepatocytes may relate to the fact that the in-fection is highly cytopathogenic in this cell type, leading to celldestruction and large plaque-like lesions in the liver paren-chyma (45). LSECs, in contrast, are ideal candidates for la-tency due to their very limited virus productivity (97). Thisreduces cytopathogenicity but, unlike with nonproductive in-fection of nonpermissive cell types, maintains the potential forreactivation to a low-level productive infection, in accordancewith the classical definition of herpesviral latency given byRoizman and Sears (96).

Another feature of ECs that favors them for latency is theirvery low turnover, although to our knowledge there exist nodata specifically for the microvascular LSECs. As far as mac-rovascular ECs are concerned, they are among those cellsexhibiting the lowest proliferation level in the body, with only0.01% of cells engaged in cell division at any time (80). Usinghematopoietic chimeras, Crosby and colleagues (30) have de-termined that only 0.2% to 1.4% of the vascular ECs werederived from donor hematopoietic progenitors within 4months after irradiation and hematopoietic recovery. This mayexplain also why we observed that MHC-II� CD146� LSECsin the latently infected XY-XX chimeras at 12 months afterBMT were primarily recipient derived, although the literatureprovides reasonable evidence to suggest that ECs, includingLSECs, are descendants of bone marrow-derived hemangio-blastic endothelial progenitor cells involved in maintenanceangiogenesis and liver regeneration (5, 22, 36, 37, 43, 108). Asadditional evidence for a hematopoietic origin of LSECs, themurine hematopoietic stem cell marker Sca-1 (Ly-6A/E) wasfound to be constitutively expressed on murine LSECs (70). Inaccordance with a low rate of LSEC replacement after BMT,our data also show a low but nonnegligible proportion (up to10%) of donor-derived sry� LSECs, with some variance be-tween individual mice (Fig. 3B and 5B). Alternatively, predom-inance of the recipient genotype in LSECs of BMT chimeras

FIG. 8. Frequency of LSECs carrying latent viral genomes. Limit-ing dilution analysis of immunomagnetically enriched CD146� LSECsby M55/gB-specific qPCR was used to estimate the frequency of LSECsthat carry latent viral DNA. (A) Standard curve for the quantitation ofM55/gB by qPCR using log10-graded numbers of linearized plasmidpDrive_gB_PTHrP_Tdy as templates. The y-axis interception (oneDNA molecule) of the extrapolated regression line reveals a cutoffcycle threshold (Ct) value of 41 cycles. Accordingly, experimental sam-ples are classified as samples negative for viral DNA when no signalwas obtained after 41 amplification cycles in the qPCR. (B) Gradednumbers of LSECs derived from a pool of livers of three latentlyinfected BALB/c mice at 8 months after sex-mismatched BMT andinfection were tested in 12 replicates for the presence of viral DNA.The frequency estimate is based on the experimentally determinedfractions of negative replicates. The log-linear plot shows the Poissondistribution graph and its 95% CI (shaded area) calculated by themaximum-likelihood method. The MPN, which is the reciprocal of thefrequency, is revealed as the abscissa coordinate (dashed arrow) ofthe point of intersection between 1/e and the calculated regression line.It gives us the number of cells containing, on average, one latentlyinfected cell. P, probability value indicating the goodness of fit.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

may reflect endogenous regeneration from a liver-resident en-dothelial progenitor cell, an argument that is supported by thefact that the liver has a hematopoietic function in ontogenesisand establishes microchimerism after transplantation (dis-cussed in reference 57). Whatever mechanism applies, our datahave shown that more than 90% of MHC-II� CD146� LSECsare recipient derived during mCMV latency many months afterBMT, so that replacement with the progeny of donor-derived

endothelial progenitor cells must be a very slow process. Sincethe frequency of latently infected LSECs is rather low, how-ever, we cannot formally exclude the possibility that the latentviral genome resides in the minor fraction of donor-derivedLSECs, although we currently see no rationale for the specu-lation that only donor-derived and not recipient-derivedLSECs support latent infection, in particular since liver-resi-dent LSECs are rapidly targeted by mCMV during acute in-

FIG. 9. Virus reactivation from latently infected LSECs by cell transfer and tissue explantation. (A) Experimental regimen. Immunomagneti-cally LSEC (CD146�)-enriched and LSEC (CD146�)-depleted NPLCs derived from pooled livers of three latently infected BALB/c donor miceat 8 months after sex-mismatched BMT and infection were transferred intravenously (i.v.) into CMV-naïve, immunocompromised (6.5 Gy ofgamma irradiation 24 h before cell transfer) female BALB/c recipients. Of each population, 2 � 107 cells were transferred per recipient mouse.At 24 h after cell transfer, recipients’ lungs were each cut into 48 pieces and the explants were kept in culture for 8 weeks, with weekly monitoringfor the occurrence of infectious virus in the culture supernatants. (B) Reactivation incidences in 96 and 192 explant cultures from two and fourrecipients of LSEC-enriched and LSEC-depleted NPLCs, respectively. (C1) Two-by-two contingency tables (observed-value table and table ofvalues expected according to the null hypothesis of random distribution) for significance analysis using Fisher’s exact test. (C2) Two-by-twocontingency tables (observed and expected values) under the rebuttable presumption that reactivation in the LSEC-enriched fraction resultedexclusively from contaminating 5% (worst-case assumption) of non-LSEC NPLCs. Differences between observed- and expected-value tables areconsidered significant for a P of �0.05 (two-sided).



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

fection before hematopoietic reconstitution can replace recip-ient-derived LSECs with donor-derived LSECs.

Isolation of LSECs gave us for the first time an opportunityto estimate the frequency of latently infected tissue-residenthost cells and to determine the latent viral genome copy num-ber per latently infected cell. The frequency of latently infectedLSECs among total LSECs was found to be in the region of10�4 (0.01%) under the conditions of the model studied here.It is important to emphasize that the precise number is not afixed value but certainly depends on the degree of virus dis-semination during acute infection, which determines the over-all load of latent viral DNA in host tissues (91). Nevertheless,the order of magnitude was found to be the same for hCMVlatency in hemopoietin-mobilized mononuclear cells fromhealthy individuals undergoing natural latent infection,namely, 0.004 to 0.01% (102). Based on the latent viral DNAload determined here for the liver, the average copy number oflatent viral genomes per latently infected LSEC was in theregion of 10 copies, specifically 9 and ranging from 4 to 17. Thisappears to be a more or less fixed value for latent CMVs,because similar copy numbers of latent viral episomes (14)were estimated for hCMV latency in hemopoietin-mobilizedmononuclear cells ex vivo and in cultured granulocyte-mac-rophage progenitors, namely, 2 to 13 genomes and 1 to 8genomes, respectively (102). Similarly, Pollock and colleaguesfound that peritoneal-exudate macrophages can harbor latentmCMV with a frequency of �1 in 50,000 cells (0.002%) and acopy number of 1 to 10 per cell (85). Thus, strikingly, both thefrequency of latently infected cells and the copy numbers oflatent viral genomes were in comparable ranges in such differ-ent cell types and across CMV and host species.

These copy numbers in latency are actually astoundinglyhigh, since at least during acute infection in vivo, the proba-bility for �10 virus hits per latently infected cell is very low, inparticular if we consider the low prevalence of latently infectedcells. Why should a latent viral genome accumulate in so fewcells of a certain cell type by random hits? Although we do notknow how many viral genomes reach the cell nucleus duringacute infection of a cell in vivo, there is evidence to suggest thatproductive infection usually initiates monoclonally from a sin-gle successful viral genome. Specifically, two-color in situ viralDNA hybridization of liver tissue sections after intravenouscoinfection of mice with equal doses of different mCMV vari-ants, for instance, wild-type or revertant virus and a replica-tion-competent mutant virus, almost exclusively revealed dis-tinct foci of infection of either color (41). In hepatocytes, lackof cytokinesis after mitosis can lead to cells with two nuclei.Most impressively, cases in which wild-type virus replicated inone nucleus and mutant virus in the second nucleus of suchbinuclear hepatocytes were occasionally observed (90). Onlyafter coinfection of mice with wild-type virus and a replication-or dissemination-deficient mutant virus was an increased inci-dence of cellular coinfection observed due to the selectiveforce on the mutant (23). Coinfection indicates that oligoclonalvirus replication can indeed occur in individual cell nuclei, butapparently this is the exception rather than the rule.

We can discuss two possibilities. One is that viral genomesenter the cell nucleus by random hits and remain silencedduring acute infection. The problem is that we currently cannotexplain the relatively high average copy numbers. Alterna-

tively, and actually more likely, latent viral genomes might bederived from a single hit or a few hits by amplification throughDNA replication, as proposed already by Roizman and Searsfor herpes simplex virus latency in neurons (96). The questionremains, however, why this postulated amplification leads toonly about 10 and not many more copies, as would be the casein productive infection.

As we have documented previously for the model of mCMVlatency in the lungs (reviewed in reference 93), not all latentmCMV genomes are transcriptionally silent at any time. In-stead, a few viral genomes are desilenced at the MIE locus andgenerate spliced MIE transcripts IE1 and IE2, but not usuallyIE3 (40, 62). Notably, whereas recent work has shown that theIE1 protein is dispensable for the reactivation of latent mCMV(18), the IE3 transcript encodes a transactivator protein that isessential for downstream E-phase gene expression and thus forviral replication (4, 74). Upon stimulation of the MIE en-hancer by tumor necrosis factor alpha, however, alternativesplicing of the IE1-IE3 precursor transcript to IE3 mRNA wasinduced; nevertheless, viral gene expression did not advance tothe expression of the essential M55/gB transcript, and conse-quently, latency was maintained (101). From this, it was pro-posed that multiple molecular checkpoints control latency.This situation is precisely what was found here for ex vivo-purified LSECs. Although IE1 and IE3 transcripts werepresent, viral gene expression did not proceed to the genera-tion of M112-M113 (E1), M55 (gB), and M86 (MCP) mRNA,so that infectious virions could not be produced. Interestingly,although a 24-h cultivation of LSECs advanced reactivatedviral gene expression to E1 and gB mRNA, the replicativecycle could not be completed because MCP was not expressed,so that viral nucleocapsids could not assemble. It should benoted that attempts to enforce virus reactivation in culturedLSECs by treatment with gene-desilencing drugs such as thehistone deacetylase inhibitors trichostatin A (115), sodium bu-tyrate (61), and valproic acid (39) all failed (data not shown).The identification of a further, so-far-unknown, checkpointbetween gB and MCP provides additional evidence in supportof the multiple-checkpoint model of latency and reactivation(63; reviewed in reference 93).

There remained the question of whether full reactivation tovirus recurrence can occur at all from LSECs, a question thatis crucial for defining LSECs as a cellular site of latency andreactivation rather than as nonproductive, viral-DNA-harbor-ing cells. In fact, to our knowledge, previous work documentingthe presence of viral DNA in ECs (see the introduction) hasnot documented productive virus reactivation specifically fromECs. Indeed, this is not at all trivial to prove, since the usualregimes for inducing virus recurrence, namely, in vivo immu-nodepletion or tissue explantation (for reviews, see references47 and 92) can, obviously, not identify the cell type. Further-more, although infectious virus could be recovered from la-tently infected host tissues by these reactivation protocols, theactual frequency of cells in which virus reactivated necessarilyremained unknown. As suggested by the model of mCMVreactivation in the lungs following total-body gamma irradia-tion, the strongest in vivo inducer known so far, most tran-scriptional reactivations still stop at downstream checkpoints,and only few reactivation events progress to the release ofinfectious virus (63).



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

We approached this problem by intravenous adoptive trans-fer of purified LSECs, which were derived from latently in-fected donor mice, into CMV-naïve recipient mice, so thatreactivation could originate only from the transferred cells. Asevery physical purification of cells includes contaminatingother cell types, it was decisive to transfer LSEC-depletedNPLCs as a control. While no reactivation was observed fromnon-LSEC NPLCs, the frequency of reactivation in the trans-ferred LSEC population was found to be in the range of 10�7

relative to all LSECs and 10�3 relative to the fraction ofLSECs harboring latent viral DNA. This is a minimum esti-mate, since not all transferred LSECs are trapped in the lungsand since no one can know if explanted lung tissue provides theoptimal microenvironment for virus reactivation from trappedLSECs. In any case, these data have shown with statisticalsignificance that virus reactivates more frequently from anLSEC-enriched NPLC population than from an LSEC-de-pleted NPLC population. These functional reactivation data,combined with the localization of the viral DNA, identifiedLSECs as a cellular site of reactivatable mCMV latency in theliver.

Conclusion. Like in the cell-type-specific Cre-inducible re-porter virus recombination (97) discussed above, small amountsof LSEC-derived reactivated virus associated with little-to-nocytopathogenicity might spread to more productive neighbor-ing cell types, which then account for bulk virus recurrence.Thus, in liver transplantation, reactivation may be initiated inthe sinusoidal endothelium, from which the virus can spread tohepatocytes, leading to cytopathogenic parenchymal infectionand ultimately to graft failure.

ACKNOWLEDGMENTS

We thank Steffen Schmitt and Julia Altmaier (FACS Core Facility)for performing cytofluorometric cell sorting and the CLAF team foranimal care. In the beginning of the project, the LSEC-specific AbME-9F1 was generously supplied by Alf Hamann, Charite UniversityMedicine, Berlin, Germany.

Extramural support was provided by the Deutsche Forschungsge-meinschaft, SFB 490, individual project E2. Special thanks go to theDr. Hans-Joachim and Ilse Brede Stiftung for a generous donation.

REFERENCES

1. Agresti, A. 1992. A survey of exact inference for contingency tables. Stat.Sci. 7:131–177.

2. Alterio de Goss, M., R. Holtappels, H.-P. Steffens, J. Podlech, P. Angele, L.Dreher, D. Thomas, and M. J. Reddehase. 1998. Control of cytomegalovirusin bone marrow transplantation chimeras lacking the prevailing antigen-presenting molecule in recipient tissues rests primarily on recipient-derivedCD8 T cells. J. Virol. 72:7733–7744.

3. Ambion. 2009. MEGAscript™ high yield transcription kit instruction man-ual (manual version 0209). Ambion, Austin, TX.

4. Angulo, A., P. Ghazal, and M. Messerle. 2000. The major immediate-earlygene ie3 of mouse cytomegalovirus is essential for viral growth. J. Virol.74:11129–11136.

5. Asahara, T., H. Masuda, T. Takahashi, C. Kalka, C. Pastore, M. Silver, M.Kearne, M. Magner, and J. M. Isner. 1999. Bone marrow origin of endo-thelial progenitor cells responsible for postnatal vasculogenesis in physio-logical and pathological neovascularization. Circ. Res. 85:221–228.

6. Baba, S., H. Fujii, T. Hirose, K. Yasuchika, H. Azuma, T. Hoppo, M. Naito,T. Machimoto, and I. Ikai. 2004. Commitment of bone marrow cells tohepatic stellate cells in mouse. J. Hepatol. 40:255–260.

7. Bain, M., M. Reeves, and J. Sinclair. 2006. Regulation of human cytomeg-alovirus gene expression by chromatin remodeling, p. 167–183. In M. J.Reddehase (ed.), Cytomegaloviruses: molecular biology and immunology.Caister Academic Press, Wymondham, Norfolk, United Kingdom.

8. Balthesen, M., L. Dreher, P. Lucin, and M. J. Reddehase. 1994. Theestablishment of cytomegalovirus latency in organs is not linked to localvirus production during primary infection. J. Gen. Virol. 75:2329–2336.

9. Balthesen, M., M. Messerle, and M. J. Reddehase. 1993. Lungs are a majororgan site of cytomegalovirus latency and recurrence. J. Virol. 67:5360–5366.

10. Benten, D., A. Follenzi, K. K. Bhargava, V. Kumaran, C. J. Palestro, and S.Gupta. 2005. Hepatic targeting of transplanted liver sinusoidal endothelialcells in intact mice. Hepatology 42:140–148.

11. Bevan, I. S., C. C. Sammons, and C. Sweet. 1996. Investigation of murinecytomegalovirus latency and reactivation in mice using viral mutants andthe polymerase chain reaction. J. Med. Virol. 48:308–320.

12. Blomhoff, R., C. A. Drevon, W. Eskild, P. Helgerud, K. R. Norum, and T.Berg. 1984. Clearance of acetyl low density lipoprotein by rat liver endo-thelial cells. J. Biol. Chem. 259:8898–8903.

13. Blouin, A., R. P. Bolender, and E. R. Weibel. 1977. Distribution of or-ganelles and membranes between hepatocytes and nonhepatocytes in therat liver parenchyma. A stereological study. J. Cell Biol. 72:441–455.

14. Bolovan-Fritts, C. A., E. S. Mocarski, and J. A. Wiedeman. 1999. Peripheralblood CD14� cells from healthy subjects carry a circular conformation oflatent cytomegalovirus genome. Blood 93:394–398.

15. Braet, F., and E. Wisse. 2002. Structural and functional aspects of liversinusoidal endothelial cell fenestrae: a review. Comp. Hepatol. 1:1–17.

16. Breiner, K. M., H. Schaller, and P. A. Knolle. 2001. Endothelial cell-mediated uptake of a hepatitis B virus: a new concept of liver targeting ofhepatotropic microorganisms. Hepatology 34:803–808.

17. Buhler, B., G. M. Keil, F. Weiland, and U. H. Koszinowski. 1990. Charac-terization of the murine cytomegalovirus early transcription unit e1 that isinduced by immediate-early proteins. J. Virol. 64:1907–1919.

18. Busche, A., A. Marquardt, A. Bleich, P. Ghazal, A. Angulo, and M. Mes-serle. 2009. The mouse cytomegalovirus immediate-early 1 gene is notrequired for establishment of latency or for reactivation in the lungs. J. Vi-rol. 83:4030–4038.

19. Chee, M., S. A. Rudolph, B. Plachter, B. Barrell, and G. Jahn. 1989.Identification of the major capsid protein gene of human cytomegalovirus.J. Virol. 63:1345–1353.

20. Cheung, K.-S., and D. J. Lang. 1977. Detection of latent cytomegalovirus inmurine salivary and prostate explant cultures and cells. Infect. Immun.15:568–575.

21. Chi, J.-T., H. Y. Chang, G. Haraldsen, F. L. Jahnsen, O. G. Troyanskaya,D. S. Chang, Z. Wang, S. G. Rockson, M. van de Rijn, D. Botstein, and P. O.Brown. 2003. Endothelial cell diversity revealed by global expression pro-filing. Proc. Natl. Acad. Sci. USA 100:10623–10628.

22. Choi, K., M. Kennedy, A. Kazarov, J. C. Papadimitriou, and G. Keller.1998. A common precursor for hematopoietic and endothelial cells. Devel-opment 125:725–732.

23. Cicin-Sain, L., J. Podlech, M. Messerle, M. J. Reddehase, and U. H. Ko-szinowski. 2005. Frequent coinfection of cells explains functional in vivocomplementation between cytomegalovirus variants in the multiply infectedhost. J. Virol. 79:9492–9502.

24. Ciocco-Schmitt, G. M., Z. Karabekian, E. W. Godfrey, R. M. Stenberg, A. E.Campbell, and J. A. Kerry. 2002. Identification and characterization ofnovel murine cytomegalovirus M112-113 (e1) gene products. Virology 294:199–208.

25. Collins, T., C. Pomeroy, and M. C. Jordan. 1993. Detection of latentcytomegalovirus DNA in diverse organs of mice. J. Infect. Dis. 168:725–729.

26. Constien, R., A. Forde, B. Liliensiek, H. J. Grone, P. Nawroth, G. Ham-merling, and B. Arnold. 2001. Characterization of a novel EGFP reportermouse to monitor Cre recombination as demonstrated by a Tie2 Cre mouseline. Genesis 30:36–44.

27. Conway, E. M., and P. Carmeliet. 2004. The diversity of endothelial cells: achallenge for therapeutic angiogenesis. Genome Biol. 5:207.

28. Cormier, E. G., R. J. Durso, F. Tsamis, L. Boussemart, C. Manix, W. C.Olson, J. P. Gardner, and T. Dragic. 2004. L-SIGN (CD209L) and DC-SIGN (CD209) mediate transinfection of liver cells by hepatitis C virus.Proc. Natl. Acad. Sci. USA 101:14067–14072.

29. Crispe, I. N. 2003. Hepatic T cells and liver tolerance. Nat. Rev. Immunol.3:51–62.

30. Crosby, J. R., W. E. Kaminski, G. Schatteman, P. J. Martin, E. W. Raines,R. A. Seifert, and D. F. Bowen-Pope. 2000. Endothelial cells of hematopoi-etic origin make a significant contribution to adult blood vessel formation.Circ. Res. 87:728–730.

31. Daoust, R., and A. Cantero. 1959. The numerical proportions of cell typesin rat liver during carcinogenesis by 4-dimethylaminoazobenzene (DAB).Cancer Res. 19:757–762.

32. DeLisser, H. M., P. J. Newman, and S. M. Albelda. 1994. Molecular andfunctional aspects of PECAM-1/CD31. Immunol. Today 15:490–495.

33. Elvevold, K., B. Smedsrod, and I. Martinez. 2008. The liver sinusoidalendothelial cell: a cell type of controversial and confusing identity. Am. J.Physiol. Gastrointest. Liver Physiol. 294:G391–G400.

34. Emery, V. C. 1998. Relative importance of cytomegalovirus load as a riskfactor for cytomegalovirus disease in the immunocompromised host.Monogr. Virol. 21:288–301.

35. Fazekas de St. Groth, S. 1982. The evaluation of limiting dilution assays.J. Immunol. Methods 49:R11–R23.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

36. Fujii, H., T. Hirose, S. Oe, K. Yasuchika, H. Azuma, T. Fujikawa, M.Nagao, and Y. Yamaoka. 2002. Contribution of bone marrow cells to liverregeneration after partial hepatectomy in mice. J. Hepatol. 36:653–659.

37. Gao, Z., V. C. McAlister, and G. M. Williams. 2001. Repopulation of liverendothelium by bone-marrow-derived cells. Lancet 357:932–933.

38. Geijtenbeek, T. B. H., P. C. Groot, M. A. Nolte, S. J. van Vliet, S. T.Gangaram-Panday, G. C. F. van Duijnhoven, G. Kraal, A. J. M. van Oost-erhout, and Y. van Kooyk. 2002. Marginal zone macrophages express amurine homologue of DC-SIGN that captures blood-borne antigens in vivo.Blood 100:2908–2916.

39. Gottlicher, M., S. Minucci, P. Zhu, O. H. Kramer, A. Schimpf, S. Giavara,J. P. Sleeman, F. L. Coco, C. Nervi, P. G. Pelicci, and T. Heinzel. 2001.Valproic acid defines a novel class of HDAC inhibitors inducing differen-tiation of transformed cells. EMBO J. 20:6969–6978.

40. Grzimek, N. K. A., D. Dreis, S. Schmalz, and M. J. Reddehase. 2001.Random, asynchronous, and asymmetric transcriptional activity of enhancer-flanking major immediate-early genes ie1/3 and ie2 during murine cytomeg-alovirus latency in the lungs. J. Virol. 75:2692–2705.

41. Grzimek, N. K. A., J. Podlech, H.-P. Steffens, R. Holtappels, S. Schmalz,and M. J. Reddehase. 1999. In vivo replication of recombinant murinecytomegalovirus driven by the paralogous major immediate-early promoter-enhancer of human cytomegalovirus. J. Virol. 73:5043–5055.

42. Gubbay, J., J. Collignon, P. Koopman, B. Capel, A. Economou, A. Mun-sterberg, N. Vivian, P. Goodfellow, and R. Lovell-Badge. 1990. A genemapping to the sex-determining region of the mouse Y chromosome is amember of a novel family of embryonically expressed genes. Nature 346:245–250.

43. Gunsilius, E., H. C. Duba, A. L. Petzer, C. M. Kahler, K. Grunewald, G.Stockhammer, C. Gabl, S. Dirnhofer, J. Clausen, and G. Gastl. 2000.Evidence from a leukaemia model for maintenance of vascular endothe-lium by bone-marrow-derived endothelial cells. Lancet 355:1688–1691.

44. Hamilton, J. D., and B. J. Seaworth. 1985. Transmission of latent cytomeg-alovirus in a murine kidney tissue transplantation model. Transplantation39:290–296.

45. Holtappels, R., M. W. Munks, J. Podlech, and M. J. Reddehase. 2006. CD8T-cell-based immunotherapy of cytomegalovirus disease in the mousemodel of the immunocompromised bone marrow transplantation recipient,p. 383–418. In M. J. Reddehase (ed.), Cytomegaloviruses: molecular biol-ogy and immunology. Caister Academic Press, Wymondham, Norfolk,United Kingdom.

46. Holtappels, R., J. Podlech, G. Geginat, H.-P. Steffens, D. Thomas, and M. J.Reddehase. 1998. Control of murine cytomegalovirus in the lungs: relativebut not absolute immunodominance of the immediate-early 1 nonapeptideduring antiviral cytolytic T-lymphocyte response in pulmonary infiltrates.J. Virol. 72:7201–7212.

47. Hummel, M., and M. M. Abecassis. 2002. A model for reactivation of CMVfrom latency. J. Clin. Virol. 25:S123–S136.

48. Jarvis, M. A., and J. A. Nelson. 2007. Human cytomegalovirus tropism forendothelial cells: not all endothelial cells are created equal. J. Virol. 81:2095–2101.

49. Jordan, M. C., and V. L. Mar. 1982. Spontaneous activation of latentcytomegalovirus from murine spleen explants: role of lymphocytes andmacrophages in release and replication of virus. J. Clin. Investig. 70:762–768.

50. Katz, S. C., V. G. Pillarisetty, J. I. Bleier, A. B. Shah, and R. P. DeMatteo.2004. Liver sinusoidal endothelial cells are insufficient to activate T cells.J. Immunol. 173:230–235.

51. Keil, G. M., A. Ebeling-Keil, and U. H. Koszinowski. 1987. Immediate-earlygenes of murine cytomegalovirus: location, transcripts, and translationproducts. J. Virol. 61:526–533.

52. Keil, G. M., A. Ebeling-Keil, and U. H. Koszinowski. 1987. Sequence andstructural organization of murine cytomegalovirus immediate-early gene 1.J. Virol. 61:1901–1908.

53. Klotman, M. E., S. C. Henry, R. C. Greene, P. C. Brazy, P. E. Klotman, andJ. D. Hamilton. 1990. Detection of mouse cytomegalovirus nucleic acid inlatently infected mice by in vitro enzymatic amplification. J. Infect. Dis.161:220–225.

54. Klotman, M. E., D. Starnes, and J. D. Hamilton. 1985. The source ofmurine cytomegalovirus in mice receiving kidney allografts. J. Infect. Dis.152:1192–1196.

55. Kmiec, Z. 2001. Cooperation of liver cells in health and disease. Adv. Anat.Embryol. Cell Biol. 161:1–151.

56. Knolle, P. A., and G. Gerken. 2000. Local control of the immune responsein the liver. Immunol. Rev. 174:21–34.

57. Knolle, P. A., and A. Limmer. 2003. Control of immune responses byscavenger liver endothelial cells. Swiss Med. Wkly. 133:501–506.

58. Knook, D. L., and E. C. Sleyster. 1976. Separation of Kupffer and endo-thelial cells of the rat liver by centrifugal elutriation. Exp. Cell Res. 99:444–449.

59. Koffron, A. J., M. Hummel, B. K. Patterson, S. Yan, D. B. Kaufmann, J. P.Fryer, F. P. Stuart, and M. I. Abecassis. 1998. Cellular localization of latentmurine cytomegalovirus. J. Virol. 72:95–103.

60. Koopman, P., A. Munsterberg, B. Capel, N. Vivian, and R. Lovell-Badge.1990. Expression of a candidate sex-determining gene during mouse testisdifferentiation. Nature 348:450–452.

61. Kruh, J. 1982. Effects of sodium butyrate, a new pharmacological agent, oncells in culture. Mol. Cell. Biochem. 42:65–82.

62. Kurz, S. K., M. Rapp, H.-P. Steffens, N. K. A. Grzimek, S. Schmalz, andM. J. Reddehase. 1999. Focal transcriptional activity of murine cytomega-lovirus during latency in the lungs. J. Virol. 73:482–494.

63. Kurz, S. K., and M. J. Reddehase. 1999. Patchwork pattern of transcrip-tional reactivation in the lungs indicates sequential checkpoints in thetransition from murine cytomegalovirus latency to recurrence. J. Virol.73:8612–8622.

64. Kurz, S. K., H.-P. Steffens, A. Mayer, J. R. Harris, and M. J. Reddehase.1997. Latency versus persistence or intermittent recurrences: evidence for alatent state of murine cytomegalovirus in the lungs. J. Virol. 71:2980–2987.

65. Lautenschlager, I., L. Halme, K. Hockerstedt, L. Krogerus, and E. Taski-nen. 2006. Cytomegalovirus infection of the liver transplant: virological,histological, immunological, and clinical observations. Transpl. Infect. Dis.8:21–30.

66. Lefkovits, I., and H. Waldman. 1979. Limiting dilution analysis of cells inthe immune system, p. 38–82. Cambridge University Press, Cambridge,England.

67. Limmer, A., and P. A. Knolle. 2001. Liver sinusoidal endothelial cells: a newtype of organ-resident antigen-presenting cell. Arch. Immunol. Ther. Exp.49:S7–S11.

68. Limmer, A., J. Ohl, C. Kurts, H.-G. Ljunggren, Y. Reiss, M. Groettrup, F.Momburg, B. Arnold, and P. A. Knolle. 2000. Efficient presentation ofexogenous antigen by liver endothelial cells to CD8� T cells results inantigen-specific T-cell tolerance. Nat. Med. 6:1348–1354.

69. Lohse, A. W., P. A. Knolle, K. Bilo, A. Uhrig, C. Waldmann, M. Ibe, E.Schmitt, G. Gerken, and K.-H. Meyer Zum Buschenfelde. 1996. Antigen-presenting function and B7 expression of murine sinusoidal endothelialcells and Kupffer cells. Gastroenterology 110:1175–1181.

70. Luna, G., J. Paez, and J. E. Cardier. 2004. Expression of the hematopoieticstem cell antigen Sca-1 (LY-6A/E) in liver sinusoidal endothelial cells:possible function of Sca-1 in endothelial cells. Stem Cells Dev. 13:528–535.

71. Mangin, M., K. Ikeda, and A. E. Broadus. 1990. Structure of the mousegene encoding parathyroid hormone-related peptide. Gene 95:195–202.

72. Mayer, A., J. Podlech, S. Kurz, H.-P. Steffens, S. Maiberger, K. Thalmeier,P. Angele, L. Dreher, and M. J. Reddehase. 1997. Bone marrow failure bycytomegalovirus is associated with an in vivo deficiency in the expression ofessential stromal hemopoietin genes. J. Virol. 71:4589–4598.

73. Mercer, J. A., C. A. Wiley, and D. H. Spector. 1988. Pathogenesis of murinecytomegalovirus infection: identification of infected cells in the spleen dur-ing acute and latent infections. J. Virol. 62:987–997.

74. Messerle, M., B. Buhler, G. M. Keil, and U. H. Koszinowski. 1992. Struc-tural organization, expression, and functional characterization of the mu-rine cytomegalovirus immediate-early gene 3. J. Virol. 66:27–36.

75. Mocarski, E. S., Jr., G. Hahn, K. L. White, J. Xu, B. Slobedman, L. Hertel,S. A. Aguirre, and S. Noda. 2006. Myeloid cell recruitment and function inpathogenesis and latency, p. 465–481. In M. J. Reddehase (ed.), Cytomega-loviruses: molecular biology and immunology. Caister Academic Press,Wymondham, Norfolk, United Kingdom.

76. Mutimer, D. 1996. CMV infection of transplant recipients. J. Hepatol.25:259–269.

77. Nagelkerke, J. F., K. P. Barto, and T. J. C. van Berkel. 1983. In vivo and invitro uptake and degradation of acetyl low density lipoprotein by rat liverendothelial, Kupffer, and parenchymal cells. J. Biol. Chem. 258:12221–12227.

78. Newman, P. J. 1994. The role of PECAM-1 in vascular cell biology. Ann.N. Y. Acad. Sci. 714:165–174.

79. Onoe, T., H. Ohdan, D. Tokita, M. Shishida, Y. Tanaka, H. Hara, W. Zhou,K. Ishiyama, H. Mitsuta, K. Ide, and T. Asahara. 2005. Liver sinusoidalendothelial cells tolerize T cells across MHC barriers in mice. J. Immunol.175:139–146.

80. Ortega, N., H. Hutchings, and J. Plouet. 1999. Signal relays in the VEGFsystem. Front. Biosci. 4:D141–D152.

81. Pampou, S. Y., S. N. Gnedoy, V. B. Bystevskaya, V. N. Smirnov, E. I.Chazov, J. L. Melnick, and M. E. DeBakey. 2000. Cytomegalovirus genomeand the immediate-early antigen in cells of different layers of human aorta.Virchows Arch. 436:539–552.

82. Pitas, R. E., J. Boyles, R. W. Mahley, and D. M. Bissell. 1985. Uptake ofchemically modified low density lipoproteins in vivo is mediated by specificendothelial cells. J. Cell Biol. 100:103–117.

83. Podlech, J., R. Holtappels, N. K. A. Grzimek, and M. J. Reddehase. 2002.Animal models: murine cytomegalovirus. Methods Microbiol. 32:493–525.

84. Podlech, J., R. Holtappels, N. Wirtz, H.-P. Steffens, and M. J. Reddehase.1998. Reconstitution of CD8 T cells is essential for the prevention ofmultiple-organ cytomegalovirus histopathology after bone marrow trans-plantation. J. Gen. Virol. 79:2099–2104.

85. Pollock, J. L., R. M. Presti, S. Paetzold, and H. W. Virgin IV. 1997. Latentmurine cytomegalovirus infection in macrophages. Virology 227:168–179.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

86. Pomeroy, C., P. J. Hilleren, and M. C. Jordan. 1991. Latent murine cyto-megalovirus DNA in splenic stromal cells of mice. J. Virol. 65:3330–3334.

87. Qiagen. 2006. DNeasy blood and tissue handbook (manual no. 07/2006).Qiagen, Hilden, Germany.

88. Qiagen. 2006. RNeasy mini handbook (manual no. 04/2006). Qiagen,Hilden, Germany.

89. Rawlinson, W. D., H. E. Farrell, and B. G. Barrell. 1996. Analysis of thecomplete DNA sequence of murine cytomegalovirus. J. Virol. 70:8833–8849.

90. Reddehase, M. J. 2006. Preface, p. XXIII-XXVII. In M. J. Reddehase (ed.),Cytomegaloviruses: molecular biology and immunology. Caister AcademicPress, Wymondham, Norfolk, United Kingdom.

91. Reddehase, M. J., M. Balthesen, M. Rapp, S. Jonjic, I. Pavic, and U. H.Koszinowski. 1994. The conditions of primary infection define the load oflatent viral genome in organs and the risk of recurrent cytomegalovirusdisease. J. Exp. Med. 179:185–193.

92. Reddehase, M. J., J. Podlech, and N. K. A. Grzimek. 2002. Mouse modelsof cytomegalovirus latency: overview. J. Clin. Virol. 25:S23–S36.

93. Reddehase, M. J., C. O. Simon, C. K. Seckert, N. Lemmermann, andN. K. A. Grzimek. 2008. Murine model of cytomegalovirus latency andreactivation. Curr. Top. Microbiol. Immunol. 325:315–332.

94. Reeves, M., and J. Sinclair. 2008. Aspects of human cytomegalovirus la-tency and reactivation. Curr. Top. Microbiol. Immunol. 325:297–313.

95. Reeves, M. B., H. Coleman, J. Chadderton, M. Goddart, J. G. P. Sissons,and J. H. Sinclair. 2004. Vascular endothelial and smooth muscle cells areunlikely to be major sites of latency of human cytomegalovirus in vivo.J. Gen. Virol. 85:3337–3341.

96. Roizman, B., and A. E. Sears. 1987. An inquiry into the mechanisms ofherpes simplex virus latency. Annu. Rev. Microbiol. 41:543–571.

97. Sacher, T., J. Podlech, C. A. Mohr, S. Jordan, Z. Ruzsics, M. J. Reddehase,and U. H. Koszinowski. 2008. The major virus-producing cell type duringmurine cytomegalovirus infection, the hepatocyte, is not the source of virusdissemination in the host. Cell Host Microbe 3:263–272.

98. Schrage, A., C. Loddenkemper, U. Erben, U. Lauer, G. Hausdorf, P. R.Jungblut, J. Johnson, P. A. Knolle, M. Zeitz, A. Hamman, and K. Kluge-witz. 2008. Murine CD146 is widely expressed on endothelial cells and isrecognized by the monoclonal antibody ME-9F1. Histochem. Cell Biol.129:441–451.

99. Seckert, C. K., A. Renzaho, M. J. Reddehase, and N. K. Grzimek. 2008.Hematopoietic stem cell transplantation with latently infected donors doesnot transmit virus to immunocompromised recipients in the murine modelof cytomegalovirus infection. Med. Microbiol. Immunol. 197:251–259.

100. Simon, C. O., R. Holtappels, H.-M. Tervo, V. Bohm, T. Daubner, S. A.Oehrlein-Karpi, B. Kuhnapfel, A. Renzaho, D. Strand, J. Podlech, M. J.Reddehase, and N. K. A. Grzimek. 2006. CD8 T cells control cytomegalo-virus latency by epitope-specific sensing of transcriptional reactivation.J. Virol. 80:10436–10456.

101. Simon, C. O., C. K. Seckert, D. Dreis, M. J. Reddehase, and N. K. Grzimek.

2005. Role for tumor necrosis factor alpha in murine cytomegalovirustranscriptional reactivation in latently infected lungs. J. Virol. 79:326–340.

102. Slobedman, B., and E. S. Mocarski. 1999. Quantitative analysis of latenthuman cytomegalovirus. J. Virol. 73:4806–4812.

103. Smedsrød, B. 2004. Clearance function of scavenger endothelial cells.Comp. Hepatol. 3(Suppl. 1):S22.

104. Starzl, T. E. 2001. The “privileged” liver and hepatic tolerogenicity. LiverTranspl. 7:918–920.

105. Steffens, H.-P., J. Podlech, S. Kurz, P. Angele, D. Dreis, and M. J. Redde-hase. 1998. Cytomegalovirus inhibits the engraftment of donor bone mar-row cells by downregulation of hemopoietin gene expression in recipientstroma. J. Virol. 72:5006–5015.

106. Steimle, V., C. A. Siegrist, A. Mottet, B. Lisowska-Grospierre, and B. Mach.1994. Regulation of MHC class II expression by interferon-gamma medi-ated by the transactivator gene CIITA. Science 265:106–109.

107. Streblow, D. N., S. M. Varnum, R. D. Smith, and J. A. Nelson. 2006. Aproteomics analysis of human cytomegalovirus particles, p. 91–110. In M. J.Reddehase (ed.), Cytomegaloviruses: molecular biology and immunology.Caister Academic Press, Wymondham, Norfolk, United Kingdom.

108. Takakura, N., T. Watanabe, S. Suenobu, Y. Yamada, T. Noda, Y. Ito, M.Satake, and T. Suda. 2000. A role for hematopoietic stem cells in promot-ing angiogenesis. Cell 102:199–209.

109. Tuma, P. L., and A. L. Hubbard. 2003. Transcytosis: crossing cellularbarriers. Physiol. Rev. 83:871–932.

110. Uitenbroek, D. G. 1997. Binominal. SISA. http://www.quantitativeskills.com/sisa/. Accessed 30 June 2009.

111. Wagner, M., S. Jonjic, U. H. Koszinowski, and M. Messerle. 1999. System-atic excision of vector sequences from the BAC-cloned herpesvirus genomeduring virus reconstitution. J. Virol. 73:7056–7060.

112. Weibel, E. R., W. Staubli, H. R. Gnagi, and F. A. Hess. 1969. Correlatedmorphometric and biochemical studies on the liver cell. I. Morphometricmodel, stereologic methods, and normal morphometric data for rat liver.J. Cell Biol. 42:68–91.

113. Wilhelmi, V., C. O. Simon, V. Bohm, T. Daubner, S. Emde, D. Strand, A.Renzaho, N. A. Lemmermann, C. K. Seckert, M. J. Reddehase, and N. K.Grzimek. 2008. Transactivation of cellular genes involved in nucleotidemetabolism by the regulatory IE1 protein of murine cytomegalovirus is notcritical for viral replicative fitness in quiescent cells and host tissues. J. Vi-rol. 82:9900–9916.

114. Wilson, E. J., D. N. Medearis, L. V. Barrett, and R. H. Rubin. 1985.Activation of latent murine cytomegalovirus in cardiac explant and cellcultures. J. Infect. Dis. 152:625–626.

115. Yoshida, M., K. Masako, M. Aktia, and T. Beppu. 1990. Potent and specificinhibitor of mammalian histone deacetylase both in vivo and in vitro bytrichostatin A. J. Biol. Chem. 28:17174–17179.

116. Zhou, W., M. Inada, T.-P. Lee, D. Benten S. Lyubsky, E. E. Bouhassira, S.Gupta, and H.-M. Tsai. 2005. ADAMTS13 is expressed in hepatic stellatecells. Lab. Investig. 85:780–788.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

liver sinusoidal endothelial cells are a site of murine

Documents