Liver Sinusoidal Endothelial Cells Are a Site of Murine Cytomegalovirus Latency and Reactivation^▿

Experimental BMT and establishment of latent mCMV infection.

Sex-mismatched but otherwise syngeneic BMT with male BALB/c mice as BMC donors and female BALB/c mice as BMC recipients was performed as described previously (83). In brief, hematoablative conditioning of 8- to 10-week-old recipients was achieved by total-body gamma irradiation with a single dose of 6.5 Gy. BMT was performed 6 h later by infusion of 5 × 10⁶ femoral and tibial donor BMCs into the tail veins of the recipients. At ca. 2 h after BMT, recipients were infected with 10⁵ PFU of mCMV (strain Smith ATCC VR-194, purchased in 1981; the stock 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, see reference 92) and include the absence of infectivity in key organs of CMV tropism (liver, spleen, lungs, and salivary glands) as well as reduction of viral DNA load in tail vein blood to less than one copy per 10⁴ leukocytes in a longitudinal analysis using M55/gB-specific real-time quantitative PCR (qPCR) (99). Clearance of viral DNA from blood is the criterion that takes longest to be fulfilled in the BMT model, usually ∼8 months (62, 99). Animals were bred and housed under specified-pathogen-free conditions in the Central Laboratory Animal Facility (CLAF) of the Johannes Gutenberg University. Animal experiments were approved by the Landesuntersuchungsamt Rheinland-Pfalz according to German federal law § 8 Abs. 1 TierSchG under permission number 177-07/021-28.

Quantitation of viral and cellular genomes by qPCR.

Quantitation of the viral gene M55/gB, the diploid cellular gene pthrp (71), and the (male) sex-determining region on chromosome Y, gene sry (60), formerly known as the testis-determining gene on chromosome Y, gene tdy (42), was performed by qPCR with primers 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) (GenBank accession 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 primer gB_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 used as probe PTHrP_Taq_Probe1, oligonucleotide 5′-CAAGGGCAAGTCCATCCAAG-3′ (nt 155 to 174) served as forward primer PTHrP_Taq_For1, and oligonucleotide 5′-GGGACACCTCCGAGGTAGCT-3′ (nt 255 to 236) served as reverse 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 as probe Tdy_Taq_Probe1, oligonucleotide 5′-AAGCGCCCCATGAATGC-3′ (nt 113 to 129) served as forward primer Tdy_Taq_For1, and oligonucleotide 5′-CCCAGCTGCTTGCTGATCTC-3′ (nt 216 to 197) served as reverse primer Tdy_Taq_Rev1, yielding an amplification product of 104 bp.

Quantitation was performed on the model 7500 real-time PCR system (Applied Biosystems, Foster City, CA) by using dually labeled probes containing a fluorescent reporter at the 5′ end, a quencher at the 3′ end (6-carboxyfluorescein reporter and 6-carboxytetramethylrhodamine quencher system; Operon Biotechnologies, Cologne, Germany), and the TaqMan universal PCR master mix (Applied Biosystems). All reactions were performed in a total volume of 20 μl, containing 10 μl of 2× TaqMan universal master mix, a 1 μM concentration of each primer, and a 0.25 μM concentration of the probe. Two-step PCR was performed with the following cycler conditions: an initial step for 10 min at 95°C to activate DNA polymerase followed by 50 cycles of 15 s at 95°C and a combined primer annealing/extension step for 1 min at 60°C, during which time data were collected. Standard curves for quantitation were established by using graded numbers of linearized plasmid pDrive_gB_PTHrP_Tdy (101) as the template.

DNA isolation from host tissues and cells. (i) Isolation of DNA from latently infected liver tissue.

To remove circulating intravascular leukocytes, livers from latently infected mice were perfused by standard procedures with Gey's balanced salt solution (GBSS; 137 mM NaCl, 5 mM KCl, 1.6 mM CaCl₂·2H₂O, 0.9 mM MgCl₂·6H₂O, 0.3 mM MgSO₄·7H₂O, 0.2 mM KH₂PO₄, 1.7 mM Na₂HPO₄·2H₂O, 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 milligrams of liver homogenate was used for DNA isolation with the DNeasy blood and tissue kit (catalog no. 69506; Qiagen) as described previously (113).

(ii) Isolation of DNA from MACS-enriched cells.

Cells of defined cell surface marker phenotypes were enriched by positive (immuno)magnetic cell sorting (MACS) as described below, and DNA was then extracted with the DNeasy blood and tissue kit (Qiagen) as outlined previously (113), according to the manufacturer's spin column protocol from step 1c onward (87).

(iii) Isolation of DNA from MACS-enriched LSECs for limiting dilution analysis.

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 each LSEC preparation sample. DNA was eluted in 200 μl elution buffer. The isolated DNA was precipitated with 500 μl of absolute ethanol and 20 μl of 3 M sodium acetate, pH 4.5, overnight at −20°C. Finally, precipitated DNA was dissolved in 5 μl Tris buffer, pH 8.0, and the entire sample was used for M55/gB-specific real-time quantitation.

Virus reactivation in liver tissue explants from latently infected mice.

Livers were perfused as described above and were transferred into minimal essential medium plus GlutaMAX (Invitrogen, Paisley, United Kingdom) supplemented with 10% fetal calf serum (PAA, Pasching, Austria) and 1% penicillin-streptomycin (Invitrogen). They were then immediately cut into pieces with an edge length of 2 to 4 mm, and these were plated separately into 24-well plates (catalog no. 353047; BD Biosciences) with 1 ml medium per well. At defined intervals, 300 μl of cell culture supernatant was transferred onto semiconfluent mouse embryo fibroblasts for viral plaque assay (PFU assay) by the technique of centrifugal enhancement of infectivity as described in greater detail elsewhere (references 64 and 83 and references therein). In all explant cultures, the sampled medium was replaced with 300 μl of fresh medium.

Preparation of hepatocyte suspensions.

Hepatocytes were isolated from mouse livers as described previously (113), except that the collagenase buffer was supplemented with 0.05% NB 4 standard-grade collagenase (Serva, Heidelberg, Germany).

Preparation of nonparenchymal liver cell suspensions.

Nonparenchymal but liver-resident cells (NPLCs), which account for about 35% to 40% of total liver cells (31, 112) and for about 6% of the total liver mass (13), were isolated from mouse livers as described previously (50, 58) with the following modifications. Briefly, mice were euthanized, and the portal vein was dissected and cannulated with a 26-gauge needle. Thereafter, the posterior vena cava was opened to allow outflow 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 5 ml/min. The liver was then excised, and connective tissue and the gallbladder were removed. The organ was mechanically separated using forceps, followed by a 45-min incubation in GBSS buffer supplemented with 0.05% collagenase A and 2,000 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 mesh and thereafter through a sterile 100-μm nylon cell strainer (BD Biosciences) and was washed two times with GBSS at 400 × g for 10 min. For density gradient separation, the pellet containing the NPLCs was resuspended in 3 ml GBSS and added 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) to remove debris and enrich the NPLCs. To avoid exsiccation of the cells, the suspension was overlaid with 2 ml GBSS and centrifuged at 400 × g for 15 min with the brake turned off. The layer of low-density cells at the interface was harvested, and the NPLCs were washed with GBSS at 400 × g for 10 min. After being washed, the pellet was resuspended in 10 ml GBSS for cell counting.

Separation and phenotype analysis of NPLCs.

The following MicroBeads and antibodies were used for MACS and fluorescence-activated cell sorting (FACS): anti-fluorescein MicroBeads (catalog no. 130-048-701; Miltenyi Biotec, Bergisch Gladbach, Germany), rat anti-mouse CD146 (LSEC-antigen) MicroBeads (catalog 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 MicroBeads (catalog no. 130-052-001; Miltenyi Biotec), rat anti-mouse CD45R MicroBeads (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-Fcγ III/II receptor, clone 2.4G2, catalog no. 553142; BD Biosciences), fluorescein isothiocyanate (FITC)-conjugated rat anti-mouse CD31 Ab (clone 390, catalog no. 558738; BD Biosciences), FITC-conjugated rat anti-mouse CD106 Ab (clone 429, catalog no. 553332; BD Biosciences), rat anti-mouse LSEC-antigen CD146 Ab (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 protein R1 (anti-mSIGN-R1) Ab (clone ER-TR9, catalog no. MCA2394A647; AbD Serotec), 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, were a generous gift from A. Hamann, Berlin, Germany.

(i) Immunomagnetic cell sorting.

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

(ii) Two-color cytofluorometric analysis.

To determine the purity of isolated LSECs, the cells were incubated with Dil-conjugated AcLDL for 1 h at 37°C. After incubation, the cells were washed and saturated with 1 μg CD16/CD32 monoclonal Ab per million cells to block Fc receptor binding sites. After a washing step, cells were labeled with the FITC-conjugated Ab anti-LSEC antigen CD146. Stainings were carried out in FACS buffer (10 mM EDTA and 20 mM HEPES in PBS containing 0.4% [wt/vol] bovine serum albumin and 0.003% [wt/vol] NaN₃). All labeling procedures were performed on ice to minimize receptor capping or receptor internalization. The analysis was performed with a FACSort (BD Biosciences) using CellQuest 3.3 software for data processing. Fluorescence channel 1 (FL-1) represents fluorescein fluorescence (green), and FL-2 represents Dil fluorescence (red).

(iii) Cytofluorometric cell sorting.

For the cytofluorometric purification of CD31⁺ CD146⁺ cells, density gradient-enriched NPLCs (see above) were saturated with 1 μg CD16/CD32 monoclonal Ab per million cells to block Fc receptor 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 receptor binding 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 binding 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 sorter used was a FACSVantage cell sorter (BD Biosciences). The excitation wavelengths were 488 and 633 nm. Band pass filters for 525, 575, and 660 nm were used 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 scatter plot 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 in polystyrene tubes.

Cultivation of MACS-enriched LSECs.

MACS-purified LSECs were resuspended in Dulbecco's modified Eagle medium plus GlutaMAX (Gibco) supplemented with 10% fetal calf serum (PAA, Pasching, Austria) and 1% penicillin-streptomycin (Gibco), and were incubated for 24 h in collagen-coated six-well plates at a density of 3 × 10⁶ LSECs per six-well culture. For collagen coating of six-well plates, collagen R solution (Serva, Heidelberg, Germany) was diluted 1:10 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-purified LSECs were cultivated for 24 h on fibronectin-coated coverslips in 24-well plates at a density of 8 × 10⁵ cells per well in the presence of 1.5 μg BODIPY FL-conjugated AcLDL (catalog no. L3485; Molecular Probes) per ml. After three washes with PBS, cell nuclei were stained by incubation of the coverslips for 5 min with the DNA-binding, blue-fluorescing dye Hoechst 33342 (catalog no. H3570; Molecular Probes) (1:5,000 in PBS). Finally, cells were washed three times in PBS, and the coverslips were mounted with GelMount 295 aqueous mounting medium (catalog no. G0918; Sigma-Aldrich, Steinheim, Germany).

Construction of plasmids.

Recombinant plasmids were constructed according to established procedures, and enzyme reactions were performed as recommended by the manufacturers. Synthetic transcripts were prepared according to the instructions given in the MEGAscript SP6 kit (Ambion, Austin, TX) technical manual (3).

(i) Plasmid pDrive_M86 for the synthesis of M86 in vitro transcripts.

A DNA fragment of M86/MCP (19, 89) was amplified by PCR from DNA derived from a virus stock of mCMV-BAC (bacterial artificial chromosome-cloned mCMV MW97.01) (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 124,380 (GenBank accession no. U68299; complete genome) (89), served as forward and reverse primers, respectively. The 2,467-bp amplification product containing an M86 sequence fragment from map positions 124,361 to 126,827 was inserted into the pDrive cloning vector (Qiagen, Hilden, Germany) by means of UA-based ligation to generate the 6,320-bp plasmid pDrive_M86. For use as a template in the in vitro transcription, plasmid pDrive_M86 was linearized by digestion 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 generated by subcloning the HindIII/XbaI restriction fragment of plasmid pBlue-ie3 (62) in the vector plasmid pSP64_poly(A) (Promega, Madison, WI). An 831-bp ie3 fragment consisting of a part of exon 2, intron 2, exon 3, and a part of exon 5 was amplified from plasmid pSP64_IE3_poly(A) with primers IE3-SalI_For (5′-AAAGTCGACCGGCCGTCACTTGG-3′), representing map positions 5,876 to 5,889 (GenBank accession no. 330543), and IE3-BamHI_Rev (5′-AAAGGATCCCGTGTGGCCTGTACC-3′), representing map positions 8,529 to 8,543. The SalI and BamHI restriction sites are underlined. After digestion of the amplificate with BamHI and SalI, plasmid pSP64_IE1_IE3_poly(A) was generated by cloning of the restriction fragment into the BamHI- and SalI-digested plasmid pSP64_IE1_poly(A) (62).

Isolation of total RNA from LSECs of latently infected mice.

RNA was isolated using the RNeasy minikit (catalog no. 74106; Qiagen) as described in the manufacturer's animal cell protocol (88). Cultivated LSECs were lysed directly in the cell culture well by adding lysis buffer. Lysates were collected with a cell scraper and homogenized by using QIAshredder spin columns (catalog no. 79656; Qiagen). Freshly isolated LSECs were pelletized, resuspended in 350 μl of lysis buffer, and homogenized by using QIAshredder spin columns. RNA was bound to the RNeasy Mini spin column, followed by digestion of contaminating DNA with DNase using the RNase-free DNase set (catalog no. 79254; Qiagen). After two washing steps, bound RNA was eluted twice in 45 μl of RNase-free water. To ensure that all contaminating DNA was removed, a second DNase digestion was performed. The RNA was cleaned by using the RNeasy minikit according to the RNA cleanup protocol (Qiagen) (88). Finally, the RNA was eluted in 50 μl RNase-free water.

Analysis and quantitation of viral transcripts.

Quantitation of viral transcripts from genes m123 (ie1), M122 (ie3), M112-M113 (e1), and M55 (gB) was performed by one-step reverse transcriptase (RT) qPCRs (RT-qPCRs) with primers and probes specified previously (100). For the IE1-IE3 duplex RT-qPCR, the dually labeled probe ie1-taq1 consisted of a 6-carboxyfluorescein reporter at the 5′ end and a Black Hole Quencher (BHQ1a) at the 3′ end. Probe ie3-taq1 consisted 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 codes for 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. Oligonucleotide 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 primer M86_Taq_Rev1, yielding an amplification product of 65 bp.

Quantitations were performed on a model 7500 real-time PCR system (Applied Biosystems) by using dually labeled probes containing a fluorescent reporter at the 5′ end and a quencher at the 3′ end (Operon Biotechnologies). The corresponding in vitro transcripts IE1, IE3, early 1 (E1), gB, and MCP were generated as outlined above and described previously (62, 100) and were used as standards for quantitation. Standard titrations ranged from 10⁶ to 10 in vitro transcripts, with 100 and 10 transcripts being measured in duplicate and triplicate, respectively. IE1 plus IE3 (duplex RT-qPCR), E1, gB, and MCP transcripts in 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 transcripts in 500 ng of total RNA was extrapolated. Reactions for gB and MCP transcripts were performed in a total volume of 25 μl, containing 5 μl of 5× Qiagen OneStep RT-PCR buffer, 1 μl of Qiagen OneStep RT-PCR enzyme mix, 668 μM of each deoxynucleoside triphosphate, 0.76 μM of each primer, 0.26 μM probe, an additional 1.5 mM concentration of MgCl₂, and 0.132 μM ROX (5-carboxy-X-rhodamine) as the passive reference. The reaction mixtures for IE1 plus IE3 transcripts in the duplex RT-qPCR and for E1 transcripts in the singleplex RT-qPCR were modified in that the primer concentrations were 0.4 μM and 1 μM, respectively. Reverse transcription was performed for 30 min at 50°C. The cycle protocol for cDNA amplification started with an activation step for 15 min at 95°C and was followed by 50 cycles of denaturation at 94°C for 15 s and a combined primer annealing/extension step at 60°C for 1 min. The efficacy of 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 exclude amplification from template DNA instead of RNA.

Virus reactivation from latently infected cells.

For identifying cell types capable of virus reactivation, LSEC-enriched (positive MACS) and LSEC-depleted (MACS flowthrough) NPLCs derived from three latently infected XY-XX chimeras were transplanted into naïve female recipients that were immunocompromised (by 6.5 Gy of total-body gamma irradiation 24 h prior to transfer). One aliquot from each donor cell fraction was used to determine viral DNA load and the proportion of donor-derived sry-positive cells. Both fractions were adjusted in PBS to 2 × 10⁷ cells/ml, and 1 ml of the respective cell suspension was administered intravenously to each recipient in two rounds at 500 μl. One day after transplantation, the lungs were excised and cut into pieces with an edge length of around 2 to 4 mm. The lung explants were placed into 24-well culture plates with no feeder cells and were kept in culture for up to 8 weeks. At weekly intervals, 300 μl of cell culture supernatant was replaced with fresh culture medium and was transferred onto semiconfluent mouse embryo fibroblasts for a virus plaque assay (PFU assay) by the technique of centrifugal enhancement of infectivity (see above).

Mathematical calculations and statistical analysis. (i) Calculation of mean values and their 95% confidence limits.

Mean values and standard deviations were calculated by using the standard deviation calculator provided on the website http://invsee.asu.edu/srinivas/stdev.html. On the basis of the so-determined mean value, the standard deviation, and the number of samples, the 95% confidence intervals (CI) of the mean were calculated by a program provided on the website http://glass.ed.asu.edu/stats/analysis/mci.html.

(ii) Limiting dilution analysis.

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

(iii) Calculation of positive samples at a given Poisson distribution parameter lambda.

If a number of events (n), that is, the number of viruses that reactivated in this specific case, occurs in a total number of samples (N), that is, in plated lung tissue pieces in this specific case, then the Poisson distribution parameter is calculated by λ = n/N. Since each positive sample may comprise one event or more than one event, the number of positive samples is actually less than the number of events. The fraction of negative samples [F(0)] can be calculated according to the Poisson distribution equation as F(0) = e^−λ. Accordingly, the fraction 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 distribution parameter can be calculated as λ = −ln F(0). The fractions of positive samples containing n events [F(n)] can be calculated by the formula F(n) = λ/n × F(n − 1) for all n > 0. The total number of events is then the sum of all F(n) multiplied by the total number of samples.

(iv) Significance analysis with two-by-two contingency tables.

Two-sided P values from observed two-by-two contingency tables were calculated by using Fisher's exact test (method of the sum of small P values) provided for online calculation on the website http://www.quantitativeskills.com/sisa/statistics/fisher.htm, according to the recommendations of SISA (Simple Interactive Statistical Analysis) (1, 110). Groups are regarded as being significantly different for the test parameter, virus reactivation in LSEC-enriched and LSEC-depleted cells in this specific case, if P is <0.05.

Establishment of reactivatable viral latency in the livers of sex-mismatched bone marrow chimeras.

BMT leads to a mixed chimerism in transplantation recipients with cells of hematopoietic lineages being donor derived and repopulating the recipients' bone marrow stroma and lymphoid tissues. Using male donors and female recipients, the Y chromosome, represented by the sry/tdy gene (42, 60), which is detectable by qPCR, can thus serve as a molecular marker to identify and quantitate donor-derived cells in organs of the XY-XX chimeras (Fig. 1A). Viral pathogenesis and immune system reconstitution in this model have been studied in great detail previously (for a review, see reference 45). Notably, mCMV infection of BMT recipients was found to inhibit the engraftment of hematopoietic donor-derived stem and progenitor cells in the recipients' bone marrow stroma by nonproductive infection of stromal cells associated with a deficient expression of hemopoietins (72, 105). Whereas this can result in a lethal bone marrow aplasia under conditions of low-dose BMT, a high-dose syngeneic or even MHC-I-disparate BMT facilitates timely immune reconstitution and control of productive infection in host organs by CD8 T cells (2, 46, 84), leading to viral latency (64). Here we have chosen the high-dose BMT protocol to control acute infection and establish latency in the XY-XX chimeras. As shown in Fig. 1B for three individual chimeras after clearance of productive infection as well as qPCR-controlled clearance of the viral genome from vascular compartment passenger leukocytes (99; data not shown), resident cells in perfused liver tissue harbored latent viral genomes at a load that did not apparently correlate with the proportion of donor-derived sry⁺ cells of hematopoietic origin. Variance in the proportion of sry⁺ cells likely reflects variable lymphoid cell infiltration of the liver, but this was not investigated further here. The definition of latency implies that viral genomes are not replication defective but can be reactivated to productive infection (96). This was verified by liver tissue explant cultures (Fig. 1C). Whereas absolutely no virus was recovered from homogenized tissue material derived from the same livers (negative data not shown), recurrence occurred from the explanted tissue pieces with an incidence of ∼20%.

The latent viral genome localizes to nonparenchymal liver cells.

Liver tissue can roughly be subdivided into liver parenchymal cells, namely, the hepatocytes, and NPLCs, which include LSECs, Kupffer-type macrophages, liver-specific natural killer (NK) or pit cells, and hepatic stellate cells, formerly known as fat-storing and vitamin A-rich perisinusoidal Ito cells (55). Separating liver tissue into hepatocyte and NPLC fractions revealed an enrichment of latent mCMV DNA in the NPLC fraction (Fig. 2A, top panel), which accounts for only ∼6.5% of the liver volume but ∼40% of the total number of hepatic cells. This finding excludes the main virus producer in acute infection, the hepatocyte (97), from the candidate list of latently infected liver cell types. In agreement with the fact that NPLCs comprise long-lived liver-resident cells and short-lived repopulating cells of hematopoietic origin, ∼40 to 50% of the cells in the NPLC fraction were of the sry⁺ genotype (Fig. 2A, bottom panel). The NPLC fraction was then further subdivided by positive immunomagnetic cell sorting specific for a panel of cell 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 to CD146⁺ NPLCs (Fig. 2B, top panel), the two cell populations with 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 of the latent viral genome in the liver. Recipient origin combined with expression of CD146, also known as the LSEC antigen that is expressed on vascular but not on lymphoid ECs (98), pointed to LSECs as the predominant cellular site of hepatic mCMV latency.

Identification of CD31⁺ CD146⁺ LSECs as a cellular site of hepatic mCMV latency.

Since CD31 is a known marker for ECs, although not expressed exclusively by ECs (32, 78), it was at hand to suggest that the viral-genome-carrying, recipient-derived CD31⁺ NPLCs and CD146⁺ NPLCs defined above may actually be one and the same cell population, specifically, the LSECs. We therefore performed two-color cytofluorometric 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, presort analysis also revealed NPLC subpopulations characterized by the phenotypes CD31⁺ CD146⁻ and CD31⁻ CD146⁻, which include CD31⁻ stellate cells (116) and most leukocytes, as well as a minor subpopulation with the phenotype CD31⁻ CD146^low, which includes a subpopulation of CD49b⁺ NK cells (reference 98 and our own data not shown). The latent viral genome clearly copurified with the sorted CD31⁺ CD146⁺ population predominantly of the sry-negative genotype, whereas the different quantities of cells located outside the sorting gate and originating primarily from the donor collectively contained only trace amounts of viral genomes detected in some but not all replicate qPCRs of a sample. Since no physical separation method gives an absolute purity, the low numbers of viral genomes detected in the LSEC-depleted cell fraction may rather represent a contaminant of LSECs, which contain ∼100-fold-higher numbers of viral genomes. These findings were reproduced in five independent cell sorts performed with NPLCs derived individually from five latently infected mice (Fig. 3), and they confirmed a preceding sort experiment performed with NPLCs pooled from livers of three latently infected mice (data not shown). Finally, immunomagnetically purified CD146⁺ cells, which did not contain numerable numbers of CD11c⁺ DCs or CD11b⁺ macrophages (negative data not shown), were classified as LSECs functionally by their high endocytotic capacity to take up AcLDL (10, 12, 77, 82) ex vivo (Fig. 4A) as well as after a 24-h period in cell culture (Fig. 4B).

Donor-derived MHC-II⁺ cells expressing the LSEC antigen CD146 do not harbor the latent viral genome.

The cell surface marker phenotype of LSECs is debated in the literature (for a review, see reference 33). Whereas LSECs were reported to constitutively express MHC-II as well as CD11c, sharing these markers with DCs (reviewed in references 57 and 67), Katz and colleagues presented conflicting results showing low or absent expression of both MHC-II and CD11c by LSECs (50). To make confusion perfect, Onoe and colleagues (79) concluded that murine LSECs express MHC-II but not CD11c. Importantly, as shown above (Fig. 2), in our model, CD11c⁺ cells were primarily donor derived and did not carry latent viral DNA, which excludes DCs as well as putative CD11c⁺ LSECs from the list of candidates. In our hands, and in agreement with Katz and colleagues (50), immunomagnetically purified CD146⁺ cells all coexpressed high levels of CD31 but failed to coexpress CD4, CD45R, and CD11c (negative data not shown); yet, a small subset of the cells indeed expressed MHC-II (Fig. 5A), in accordance with the work of Onoe et al. (79). Since various MHC-II-negative cell types conditionally express MHC-II upon activation by cytokines, by gamma interferon in particular (106), we surmised that the MHC-II⁺ CD146⁺ cells may not represent an independent cell population but rather LSECs in a state of activity, which might in part explain the different findings in the literature. This assumption, however, proved to be wrong. Cytofluorometric cell sorting of immunomagnetically enriched CD146⁺ cells into MHC-II⁺ and MHC-II⁻ subsets (Fig. 5A) revealed a clear-cut distinction in that the MHC-II⁻ cells were in the majority sry negative and thus recipient derived, whereas the MHC-II⁺ cells were for the most part of the sry⁺ genotype and thus donor derived (Fig. 5B). Importantly, the latent viral DNA localized primarily to the recipient-derived MHC-II⁻ population of LSECs (Fig. 5C). Since our interest was focused on the cellular site of mCMV 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 coexpressing the murine homolog of L-SIGN.

Whereas LSECs quantitatively dominate the hepatic EC population, most cell surface markers are shared between microvascular LSECs and conventional ECs of the liver macrovasculature. As discussed in the critical review article by Elvevold and colleagues on the controversial identity of LSECs (33), the human C-type lectin L-SIGN, a homolog of DC-SIGN also known as DC-SIGN-related molecule DC-SIGNR, is strongly and constitutively expressed on LSECs and lymph node endothelium but not on DCs, other hepatic cells, or other endothelia. As shown by Geijtenbeek and colleagues (38), the murine DC-SIGN homolog mSIGN-R1 is more closely related to L-SIGN than to DC-SIGN itself because it is specifically expressed by LSECs and 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) in combination with the LSEC antigen CD146 to distinguish between LSECs and conventional ECs. It is instructive to recall that, unlike mL-SIGN, CD146 is not expressed on lymph node endothelium (98), so that coexpression of these two markers simultaneously excludes lymphoid ECs and conventional ECs. As shown in Fig. 6, by M55/gB-specific qPCR after two-color cytofluorometric cell sorting, the latent viral genomes localize to mL-SIGN⁺ CD146⁺ cells, which further supports the conclusion that LSECs harbor the latent viral DNA. It should be noted that these cells were genotyped as sry⁻, that is, as recipient derived (data not shown).

In summary of the series of phenotyping studies, mCMV latency is established primarily in the recipient-derived population 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 of virus production (96) but does not exclude limited transcriptional activity by temporary desilencing of certain viral genes. As shown previously for mCMV latency in the lungs, the major IE (MIE) locus is sporadically active during latency without further progression in the productive transcriptional program (40, 62; reviewed in reference 93). Since such a transcriptional activity is relevant to immune control of latency by peptide-specific CD8 T cells (100), we tested transcription in purified CD146⁺ LSECs derived from five latently infected XY-XX chimeras individually either directly after the ex vivo purification or after a 24-h period in cell culture (Fig. 7). In the ex vivo samples, IE1 and IE3 transcripts, which are generated from a common precursor transcript by differential splicing (51, 52, 74), were detected in various quantities in LSEC preparations from all five and from four out of five mice, respectively. Despite the presence of IE3 transcripts encoding the essential transactivator of early (E)-phase gene expression (4, 74), the transcriptional program did not proceed to the E1/M112-M113 transcript (17, 24), and accordingly, the E-phase M55/gB and the late-phase M86 transcripts were undetectable.

Latently infected LSECs proceed to E-phase gene expression upon cultivation.

A 24-h period in cell culture enhanced IE transcription and induced a progression to E1 transcription in LSECs from four out of five mice tested, and threshold amounts of the M55/gB transcripts were detected in the cell cultures occasionally (Fig. 7). In contrast, true late M86 transcripts encoding the essential MCP (107) remained absent throughout, so that virus capsid assembly could not take place. Consequently, complete reactivation to infectious virions did not occur in cell cultures of latently infected LSECs (negative data not shown).

Frequency of latently infected LSECs and copy number of the latent viral genome.

A limiting dilution approach was used to determine the frequency (minimum estimate) of viral-genome-carrying, latently infected cells present in a population of purified CD146⁺ LSECs. Graded numbers of cells in 12 replicates were subjected to M55/gB-specific qPCR using plasmid pDrive_gB_PTHrP_Tdy as the standard (101). The cutoff cycle threshold value for distinguishing between viral-DNA-negative and -positive samples was defined by extrapolation of the standard curve to one template and was found to be 41 (Fig. 8A). Accordingly, samples were classified as negative samples when no signal was obtained after 41 amplification cycles. The frequency estimate (MPN for a λ of 1) was then obtained from the proportions of negative samples according to the Poisson distribution equation and was found to be 1 latently infected cell in 14,600 (95% CI, 9,300 to 23,000) LSECs (Fig. 8B).

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

Virus can reactivate in lung explant cultures after LSEC transfer.

As shown in Fig. 1C, latent mCMV can reactivate in liver explant cultures, and in general, tissue explants have long been known to provide a supporting microenvironment for efficient virus reactivation (20, 49, 114). Although we have so far demonstrated the presence of latent viral DNA in LSECs, there remained the question of whether these genomes were fully intact and accounted for the reactivation observed in unseparated liver explants, as only this fulfills the definition of latency in LSECs.

We approached this problem by intravenous adoptive transfer of purified CD146⁺ LSECs, which were derived from latently infected BALB/c mice after sex-mismatched BMT, into immunocompromised and uninfected female BALB/c recipients (Fig. 9). Since transferred LSECs are mostly trapped in the vascular bed of the lungs at 24 h after transfer, as revealed with normal, male LSEC donors by sry-specific qPCR in female recipients (data not shown), virus reactivation was studied in lung explant cultures with the hope that the heterologous environment of lung tissue would also support virus reactivation in LSECs. Specifically, two indicator recipients each received 2 × 10⁷ purified LSECs, the highest feasible cell dose for transfer, whereas four control recipients each received 2 × 10⁷ NPLCs depleted of LSECs from the very same latently infected donors. At 24 h after the cell transfer, lungs of the recipients were each subdivided into 48 tissue pieces and plated into two 24-culture plates (Fig. 9A). It was not before week 5 after explantation that infectious virus became detectable in a total of 3 out of 96 cultures in the LSEC group, whereas during the observation period of 8 weeks, no reactivation was observed in the total of 192 explant cultures of the LSEC-depleted NPLC control group. Importantly, whereas reactivation was observed in explants from both recipients of LSEC-enriched NPLCs, no reactivation was observed in explants from four recipients of LSEC-depleted NPLCs (Fig. 9B). According to Fisher's exact probability test (Fig. 9C1), the difference between the two groups is significant, with a P value of 0.036 (two-sided). This means that reactivation occurred more frequently from purified LSECs than from all the remaining NPLCs in an LSEC-depleted population.

The LSEC preparations regularly contained <5% non-LSEC NPLCs. So, if one argues as a counterhypothesis that the three observed reactivation-positive lung pieces in the LSEC transfer group, which according to the Poisson distribution equals three reactivation events, rather came from 5% contaminating non-LSEC NPLCs, which is a worst-case assumption, an extrapolated number of 60 reactivations should have occurred in 96 explants from an LSEC-depleted NPLC preparation. Since, according to the Poisson distribution, more than one reactivation could occur from any of the plated lung pieces, 60 postulated reactivations should have led to 45 virus-positive lung explant cultures out of 96 cultures tested, instead of none in 192 cultures, as was actually observed. The probability for observing such a difference by chance (null hypothesis) is zero (Fisher's test, P < 10⁻¹⁰, two-sided) (Fig. 9C2). These findings and statistical considerations strongly support the conclusion that reactivation originated from LSECs.

These data also allow us for the first time to give a minimal estimate of the frequency of reactivations. According to the measured 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 reactivations. Thus, the frequency of latently infected LSECs in which virus reactivated was in the region of 1 in 1,000, that is f is approximately 10⁻³, under the specific conditions of the experimental model studied here.

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

Conclusion.

Like in the cell-type-specific Cre-inducible reporter virus recombination (97) discussed above, small amounts of LSEC-derived reactivated virus associated with little-to-no cytopathogenicity might spread to more productive neighboring cell types, which then account for bulk virus recurrence. Thus, in liver transplantation, reactivation may be initiated in the sinusoidal endothelium, from which the virus can spread to hepatocytes, leading to cytopathogenic parenchymal infection and ultimately to graft failure.