B cell and chondrocyte isolation
Mice were obtained from Jackson Labs, and were used in the experiments at 8–10 weeks of age. Spleen cells were independently prepared from three female mice of the following 11 inbred strains: 129/SvJ, A/J, AKR/J, BALB/cJ, C3H/HeJ, C57BL/6J, DBA/2J, MRL/MpJ, NZB/BinJ, NZW/LaC, SMJ. Single-cell suspensions of spleen cells were prepared by lysis in ACK buffer (Cambrex, East Rutherford, N.J.). B cells were positively selected using B220 magnetic beads (Miltenyi Biotec, Auburn, CA) on LS cell separation columns (Miltenyi Biotec, Auburn, CA), and then plated at 4 × 106
cells/ml in complete RPMI-1640 medium prior to freezing. Murine costal chondrocytes were isolated from rib cages obtained from 3 day old newborn mice using published methods (Gosset et al., 2008
) from the following strains: A/J, AKR/J, BALB/cJ, BALB/cByJ, C3H/HeJ, C57BL/6J, and DBA/2J. The isolated cells were frozen from 3 independent preparations for each strain, and stored at −80 °C before use.
Microarray expression analysis
RNA was purified and oligonucleotide microarray data was generated using the Affymetrix GeneChip Mouse Genome 430 2.0 Array (~39,000 transcripts) for all samples using previously described methods (Guo et al., 2007
). There were 69 B cell samples, obtained from 11 mouse strains (129/SvJ, A/J, AKR/J, BALB/cJ, C3H/HeJ, C57BL/6J, DBA/2J, MRL/MpJ, NZB/BinJ, NZW/LaC, SMJ) that were exposed to 2 treatment conditions: 1) control and 2) Anti-IgM and CD40 stimulation. There were 3–4 independent samples analyzed for each strain and treatment condition. The same microarrays were used to analyze gene expression in chondrocyte cultures, and 3 independent preparations were analyzed for each strain.
B cell data processing and statistical analysis
The probe intensity data generated from all 69 arrays were read into the R software environment (http://www.R-project.org
) directly from the .CEL files using the R/affy package (Gautier et al., 2004
), which was also used to extract and manipulate probe level data to assess data quality and to create expression summary measures. The array data were also checked for quality using GCOS (Gene Chip Operating Software) from Affymetrix. Normalization was carried out using the robust multiarray average (RMA) method (Irizarry et al., 2003
) to generate one expression measure for each probe set on each array.
The arrays have 45,100 probesets correspond to ~39,000 transcripts. The following analyses were applied to each probeset to identify probesets that are differentially expressed among strains with large fold change. A one-way ANOVA (Analysis of Variance) model was applied to test the whether a gene was differentially expressed among the mouse strains; the basal and stimulated B cell conditions were analyzed separately. The average expression level for each probeset was calculated for each strain. Since RMA signals are on a log2 scale, the fold change was defined as 2 to the power of the maximum average expression level minus the minimum expression level. Probesets with an ANOVA p-value<10−10 and a fold change greater than 10 were identified as genes that were highly significantly differentially expressed. There were 257 and 243 such probesets that corresponded to 183 and 179 genes for the basal and stimulated B cells, respectively. The differentially expressed probesets were selected for further cluster analysis.
K-mean cluster analysis was used to group the differentially expressed probesets into groups with similar expression profiles. Each probeset is associated with an 11 dimensional vector that corresponds to the average expression levels in the 11 mouse strains. The distance between each pair of probesets is measured by the correlation coefficient of the two vectors. When two probesets have a strong positive correlation, they are considered to have similar profiles. The K-mean clustering algorithm is an unsupervised classification algorithm that separates the probesets into a predefined number of groups; and each group contains probesets with similar profiles. Since the number of clusters must be pre-specified (before the analysis is complete), different numbers were tested. The number of genes within a cluster with a representative profile should be large enough to allow identification of genes that are regulated by a common factor, yet not so large that genes with distinct profiles are clustered together. Through empirical testing, we found that specification of 40 clusters each for the B cell basal and stimulated conditions, met this criteria. The cluster analysis was performed using Spotfire DecisionSite 8.2.1 (http://www.spotfire.com/
Haplotype-based computational genetic mapping
The average gene expression profile for genes within cluster 24 was used as the input data. Then, genes with a haplotypic pattern that matched this gene expression were identified using the previously described haplotype mapping method (Wang and Peltz, 2005
). To cover the entire genome, the haplotype blocks were produced by analysis of 8.3 million SNPs among 16 inbred mouse strains that were identified in the NIEHS database (Frazer et al., 2007
). Of note, CAST/EiJ, MOLF/EiJ, PWD/PhJ, WSB/EiJ are wild-derived strains, which we could not productively incorporate into a haplotype map structure that is useful for computational mapping (Wang et al., 2005
). Therefore, the genome-wide haplotype map was constructed using the 3.4M SNPs that are polymorphic among the 12 other strains (Frazer et al., 2007
). Within a haplotypic block, the SNPs display a limited level of variation that can be quantified by measuring the linkage-disequilibrium (LD) among the SNPs. In brief, our previously described methods (Liao et al., 2004
; Wang et al., 2005
) were used to partition a chromosome into a set of haplotype blocks that maximize the within-block LD measure and minimize the between-block LD measure. The average pair-wise LD measure among the component SNPs was used to represent the degree of within-block LD, and
times the average LD measure was used as the score for a candidate block; where n
is the number of SNPs in that block. For partitioning a chromosome into a set of haplotype blocks, the total score for a candidate partition was the sum of the scores of the individual blocks within the partition. The optimal partition was identified through maximizing the score. The haplotype blocks in this optimal partition were those that maximize the within-block LD measure and minimize the between-block LD measure. There were 228,885 haplotype blocks produced by this method. The average number of SNPs and haplotypes per block was 12.02 and 2.86, respectively. According to our previous method (Liao et al., 2004
), only the 141,014 (61.6%) blocks that had more than 3 SNPs were used for genetic mapping. Available phenotypic datasets (Liao et al., 2004
) (MHC, aromatic hydrocarbon response, H2-Ea
gene expression) were used to assess computational mapping results generated using this extended haplotype map. In all cases, the genetic loci that were known to be responsible for the inter-strain differences were identified with the expanded database.
Candidate gene selection
First, we determined which of the 2222 correlated genes were expressed in chondrocytes, which were used as a surrogate for macrophages. For this analysis, we used a gene expression dataset prepared from chondrocytes isolated from 7 different strains. Chondrocytes and macrophages are derived from a common mesenchymal stem cell (Caplan, 1991
); and chondrocytes have been shown to have a common pattern of chemokine production (Borzi et al., 1999
), antigen expression (Summers et al., 1995
) and important functional properties (class II histocompatibility antigen expression, antigen presentation to lymphocytes, induction of mixed and autologous lymphocyte stimulation, production of reactive oxygen intermediates) that are specifically associated with macrophages (Rathakrishnan and Tiku, 1993
; Tiku et al., 1985
). Moreover, chondrocytes have been shown to respond to IFNβ, and their response is of importance to inflammatory arthritis (Corr et al., 2011
; Palmer et al., 2004
). To do this, the MAS5 calls for the chondrocyte gene expression data were analyzed using the R/affy package (Gautier et al., 2004
). If a probe set was determined to be “present” for all 3 replicates of that strain, the probe set was labeled as expressed in this strain. For this analysis, a probe set was labeled as expressed if it was present in at least one of the 7 strains analyzed. Secondly, we identified the genes with a unique pattern of expression in C3H/HeJ chondrocytes. To do this, we used an ANOVA model to identify genes whose absolute expression level was at least 2-fold different in C3H/HeJ chondrocytes relative to the average level of expression in the other 6 strains. This ANOVA model had a nested feature that was written as: gene expression ~ isC3H + strain(isC3H), where the primary variable of interest is a variable to indicate whether the strain is C3H and different mouse strains are nested within the primary variable. To identify the genes that were commonly found in the different gene lists, all gene symbols were converted to Entrez Gene ID using the IDconverter (Alibes et al., 2007
) program. Then, the genes that were expressed in both lists with common gene identifiers were selected using R (www.r-project.org
B cell analysis
B cells were purified from splenic cells using a Miltenyi Biotec B Cell Isolation Kit (Cat. # 130-090-862). The cell density was adjusted using complete media (RPMI-1640 with 10% heat-inactivated FBS, P/S, non-essential amino acids, pyruvate, and L-Gln). Cells were incubated at 37°C, 5% CO2. After purification, cells were adjusted to 0.8 × 106/ml. 1 hour later, cells were stimulated with indicated concentrations of recombinant mouse interferon-β (R&D, Cat. # 12400-1). At indicated time points, cells were harvested for either real-time PCR analyses or Western Blot. For real-time PCR analysis, total RNA was extracted using the Qiagen RNeasy Mini Kit (Valencia, CA, Cat. # 74104) and QIAshredder (Cat. # 79654). 2 µg total RNA from each sample was used for 1st-strand cDNA synthesis that was performed using Invitrogen SuperScript III (Carlsbad, CA Cat. # 18080-051). 0.5 µl cDNA each sample was used for realtime PCR (Qiagen SYBR GreenER, Cat. # 56465). The data were collected and analyzed using an Applied Biosystems 7900HT, and all values were normalized relative to the expression level of β-Actin. The following primer sequences were used: Isg20 5’-TCC CTG AGG CTG CTG TGT AAG-3’, 5’-TGG GGG AGT GTT CTT GGT TTT-3’; Zbp1 5’-GTA GCC CCC AGA CCA CAG AAC-3’, 5’-GCA-AGG-TCG-GTT-CCA-CTT-CTT-3’; Mx1 5’-GCC AGG ACC AGG TTT ACA AGG-3’, 5’-TCC AGG AAC CAG CTG CAC TTA-3’; Irf7 5’-CAC CCC CAT CTT CGA CTT CAG-3’, 5’-GAC CCA GGT CCA TGA GGA AGT-3’; Bst2 5’-GCT GGA GAA TCT GAG GAT CCA A-3’, 5’-AAG CAG AAC TCC CTC CCC ACT-3’; β-Actin 5’-TGA CGT TGA CAT CCG TAA AGA CC-3’, 5’-AAG GGT GTA AAA CGC AGC TCA-3’. For western blotting, the cells were lysed in 1× SDS loading buffer, boiled, and the viscosity was reduced by incubation with Benzonase (Novagen, Gibbstown N.J., Cat. # 70664) at 4°C for 1 hour before further analyses.
Antibodies for Western Blot
IRF7 (Santa Cruz Biotechnology, Santa Cruz CA, clone H-246), pSTAT1 (Cell Signaling, clone Tyr701), β-Actin (Cell Signaling, Boston MA, clone 13E5), CD14 (BD, clone rmC5-3). Alkaline-phosphatase-conjugated secondary antibodies were Promega or Santa-Cruz Biotechnology.
Macrophage stimulation with polyI:C or Toll-like receptor ligands
CD11b+ peritoneal macrophages were purified from C3H/HeJ or C57BL/6 mice using Miltenyi Biotec CD11b MicroBeads (Auburn CA, Cat #. 130-049-601)., and the cells were adjusted to a final density of 3.75 × 105 cells per ml in a 96-well plate. Each well contained 60,000 cells that were stimulated in the presence or absence of recombinant soluble CD14 (Cell Sciences, Canton MA, Cat. # CRCC03) and the following TLR ligands for 24 hr: Pam3CSK4 for TLR1/2; HKLM (heat killed Listeria monocytogenes) for TLR2; polyI:C for TLR3; LPS-EK for TLR4; ST-FLA (flagellin from Salmonella typhimurium) for TLR5; FSL1 (Pam2CGDPKHPKSF) for TLR6/2; ssRNA40 for TLR7; CpG ODN1826 for TLR9. After 24 hours, media were collected for either Western blot or EIA analyses. Poly I:C was obtained from Amersham Biosciences (Pittsburg PA) and the other Toll-like receptor ligands were from Invivogen (Cat. # tlrl-kit1m).
Mouse interferon-β was measured using the PBL VeriKine kit (Piscataway, NJ, Cat. #, 42400-1), and IL-6 was quantified using the R&D Systems kit (Cat. # M6000B). For analysis of sCD14 analysis, 3 ml of serum from the following strains were purchased from Jackson Laboratory: A/J (12 weeks); AKR/J (9 weeks); C3H/HeJ (12 weeks); and C57BB6 (12 weeks). The sera were analyzed using a CD14 EIA kit (Cell Sciences cat. #: CKM034).
Rapid amplification of cDNA ends (RACE) for murine CD14
Cd14 RACE was performed using the Clontech (Mountain View, CA) SMARTer RACE cDNA Amplification Kit (Cat. # 634923). Primer sequences were as following: 5’-outer 5’-CGC ACC GTA AGC CGC TTT AAG GAC-3’, 5’-inner 5’-CTT CCG TGT CCA CAC GCT TTA-3’; 3’-outer 5’-AGC CAG ATT GGT CCA GCG CTT TC-3’, 3-inner 5’-GGC AGA TGT GGA ATT GTA CGG-3’.
In vitro transcription-translation
Cd14 RNA mutants were prepared using T7 polymerase (Thermo Scientific, Cat. # 88856) acting on 200 ng of agarose-fractionated DNA templates, which were PCR-amplified from strain-specific cDNA clones using purified forward primers (sequences below) and a common reverse primer (5’-TTA AAC AAA GAG GCG ATC TCC-3’): C57BL/6 (5’-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA G AGA GAA CAC CAC CGC TGT AAA G-3’ with a C57BL/6 clone); C3H-L (5’-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA GAG ACG CAA TTA GAA TTC ACA GAG with a C3H clone); C3H-S (5’-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA GAA CAA GCC CGT GGA ACC TG-3’ with a C3H clone); and 5’-UU-3’ (5’-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA GAG AGA ACA CCA TCG CTGTAA AG-3’ with a C3H clone). The resulting RNA was precipitated with ammonium acetate, and re-suspended with nuclease-free water; 2 µg RNA was used for in vitro translation (Thermo Scientific, Cat. # 88856); and at the indicated time points, a fixed portion of each reaction was analyzed by immunoblotting with anti-CD14 antibodies. Reactions from 1 µg of pCFE-GFP (Thermo Scientific, Cat. # 88856) were used as a positive control for in vitro transcription, and as a negative control for immunoblotting after in vitro translation.
Nucleic acid extraction from human blood cells
Blood was obtained from anonymous donors at the Stanford University Blood Bank after informed consent was obtained. 100 ul of whole blood was placed in 1 ml PBS/2mM EDTA, and the solution was centrifuged and re-suspended in 1 ml 1× RBC lysis buffer (Biolegend, San Diego, CA, catalogue #420301). The solution was divided in half, and used for either total RNA or genomic DNA extraction. Total RNA was prepared using the QIAGEN RNeasy Mini kit (cat. # 74104). Genomic DNA was obtained by lysing the cells in 500ul SDS buffer (100mM Tris pH 7.4, 5mM EDTA, 200mM NaCl, 0.2% SDS, 100 ug/ml Proteinase K) at 55C for 10 min. The DNA was precipitated with 500 ul isopropanol, washed with 70% Ethanol, and genomic DNA was re-suspended in 100 ul TE1/10 buffer and stored at 4°C before use.
Human CD14 genotyping at SNP rs2569190
The genomic region spanning SNP rs2569190 was amplified (primers: 5'-TCC TGG GGA GAG AGC AGA GGT-3' and 5'-TTT GGT GGC AGG AGA TCA ACA-3') using the following PCR conditions: 95°C 5 min, 35 cycles of 95°C 30 sec, 60°C 30 sec, 72°C 1 min, plus a final 72°C 10 min cycle. The amplified products were subject to Ava II digestion; which digests amplicons containing the A (but not the G) allele. The allelic determinations were confirmed by A- or G-allele specific PCR performed using the following primers: 5'-TCC TGG GGA GAG AGC AGA GGT-3' and 5'-CAG AAT CCT TCC TGT TAC GGT-3' for A allele; 5'-TCC TGG GGA GAG AGC AGA GGT-3' and 5'-CAG AAT CCT TCC TGT TAC GGC-3' for G allele. The PCR amplifications were performed as follows: 95°C 5 min, 5 cycles of 95°C 30 sec, 68°C 30 sec, 72°C 30 sec, another 5 cycles of 95°C 30 sec, 65°C 30 sec, 72°C 30 sec, and 25 cycles of 95°C 30 sec, 63°C 30 sec, 72°C 30 sec, plus the final 72°C 10 min.
Cloning and sequencing of the 5’ UTR of human CD14 mRNA
First strand cDNA from total RNA was synthesized using the Clontech SMARTer RACE Kit (cat. # 634923) and human CD14-specific primers. 5'-AAG GTT CTG GCG TGG TCG CAG AG-3' and 5'-CGG GTG CCG CTG TGT AGG AAA G-3' were used for 5'-RACE and 3'-RACE, respectively, according to the manufacturer's recommended PCR conditions. The 5'-RACE product was subject to a second round of PCR amplification (primer pair: 5'-ACT GAT GAG CTC AAG CAG TGG TAT CAA CGC AGA GT-3'/5'-AGC AGC AGC AGC AAC AAG CAG-3') under the following PCR parameters: 95°C for 5 min, 35 cycles of 95°C for 30 sec, 65°C for 30 sec, 72°C for 2 min, and a final 72°C for 10 min. Purified DNA was subject to Sac I digestion and ligated onto pcDNA 3.1+ (Invitrogen) for sequencing. At least 8 independent clones were analyzed for each blood sample.
In vitro transcription and translation of human CD14
A full length human CD14 construct was produced by ligating the longest 5'-RACE fragment from each of the above donors onto a 3'-RACE human CD14 clone at the Sac I site (located 25 bp 5’ of the initiator ATG) in pcDNA 3.1+ (Invitrogen) vector. The structure of all resulting constructs was confirmed by full-length sequencing. The DNA templates for T7-driven human CD14 transcripts were PCR-amplified from the full-length clone using a common reverse primer (5'-ACT GAT GTT TAA ACT GGG GCA AAG GGT TGA ATT GGT C-3' in one experiment and 5'-TTA CTT GTC GTC ATC GTC TTT GTA GTC GGC AAA GCC CCG GGC CCC TTG G-3' in the other experiment) and the following forward primers: 5'-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA GAG GAT TAC ATA AAC TGT CAG AGG CAG-3' for −142 from ATG, 5'-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA GCA GCC GAA GAG TTC ACA AGT GTG AAG-3' for −119 from ATG, 5'-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA GTC ACA AGT GTG AAG CCT GGA AGC CGG-3' for −107 from ATG, and 5'-GGA AGG AAG GAA GAG ATA ATA CGA CTC ACT ATA GAC AAG TGT GAA GCC TGG AAG CCG GCG-3' for −105 from ATG. The in vitro transcription and translation experiments were performed as described for the murine CD14 experiments. The amount of in vitro translated human CD14 was measured by enzyme immunoassay (R&D, cat. # DC140) according to the manufacturer's instructions. The capture step was performed with an overnight incubation at 4°C, and the color was developed for 1 hour. Two independent experiments produced similar results.
Human DC preparation and stimulation
After informed consent was provided, peripheral blood was obtained from healthy donors the Stanford Blood Center and monocytes were enriched using RosetteSep® Human Monocyte Enrichment Cocktail (STEMCELL Technologies, cat# 15028) followed by ficoll-hypaque density gradient centrifugation. Monocytes were further purified using MACS® CD14 microbeads (Miltenyi Biotec, cat# 130-050-201), and cultured at 1 × 106 cells/ml in DC media (Iscove’s Modified Dulbecco’s Medium (Gibco) containing 10% human AB serum, 100ug/ml Penicillin-Streptomycin, 2mM L-glutamine, and 50uM 2-ME) supplemented with recombinant human cytokines GM-CSF (100ng/ml, Berlex Laboratories Inc) and IL-4 (20ng/ml, Peprotech) for 6 days at 37°C/10% CO2. These cytokines were replenished on days 2 and 5 of culture. On day 6, the monocyte-derived DCs were assessed for conventional morphology by light microscopy and harvested via gentle scraping with cold 5mM EDTA in PBS. DCs or freshly isolated monocytes were stained with DAPI (Invitrogen), fluorescently conjugated mAbs against CD14, HLA-DR (Biolegend), CD209 (BD Biosciences), or appropriate isotype control mAbs. Their cell surface phenotype was assessed using a BD LSRII (BD Biosciences) insturment, and the data was analyzed using FlowJo software (TreeStar, Inc).
Viable DCs were re-suspended in DC stimulation media in the presence or absence of 1–5 ug/ml sCD14 (R&D Systems, cat# 383-CD-050-CF) at 1 × 106 cells/ml overnight at 37°C/10%CO2. The DC stimulation media contained different TLR agonists (Invitrogen), including LPS (TLR4), PolyI:C (TLR3), Flagellin (TLR5), or Pam3Csk4-(TLR1/2) at a range of concentrations. Supernatants and media only controls were collected 24 hr post stimulation, centrifuged to remove cellular debris, and stored in 96-well plates at −80°C for EIA. DCs were prepared from at least 4 healthy donors, and were used in two independent experiments that produced similar results. Human TNF-α, IL-10, IL-12, and IL-6 were analyzed using R&D Systems EIA kits (Cat. # DTA00C, D1000B, D1200, D6050). Human IFN-α and IFN-β were measured using PBL VeriKine kits (Cat. # 41100, 41410). For statistical analysis, the background-corrected IL-10 measurements were log-transformed, and an ANOVA model was used to assess the effect of sCD14. A fixed effect variable representing the presence or absence of sCD14 in the media as well as two random effect variables representing the donors and dosage of LPS, respectively, were included in the model. The model was evaluated and the significance of the effect of sCD14 was calculated using SAS 9.1 (Cary, NC).
Mixed Lymphocyte Reaction
DC were isolated as described above. On day 6 of culture, DC were harvested and stimulated with LPS in the absence or presence of 1ug/ml sCD14 for 24h and washed 4× with PBS to remove residual stimulants. Bulk CD4 T cells were enriched from peripheral blood from an allogeneic healthy donor using Rosette Sep cocktail (StemCell Technologies), and naïve CD4 T cells were further purified using MACs negative selection kit (Miltenyi Biotec). Naïve CD4 T cells were labeled with CFSE or CellTrace Violet tracking dyes (Invitrogen) prior to addition to washed DC for 7 days (2T and 4T to 1DC ratios). The T cells were re-suspended in fresh media and re-stimulated with PMA (200ng/ml) and Ionomycin (1ug/ml) in the presence of Brefeldin A for 15h and stained with Live Dead Aqua Fixable Dead cell stain for viability (Invitrogen), anti-human mAbs CD3 FITC, TNFalpha PE-CY7, IL-10 APC (BD Biosciences), CD3 APC-CY7, IL-17A PerCP Cy5.5, IFNgamma AF700, IL-13 PE, FOXP3 Pacific Blue, T-bet PerCP Cy5.5 (Biolegend), RORg(t) APC (eBiosciences) according to manufacturer’s protocols. For data analysis: the p-values were calculated using a standard ANOVA model. When the effect of adding sCD14 with different LPS concentrations was assessed, a variable representing the donor and another variable representing whether sCD14 was added and their interaction was included in the ANOVA model. When data for all concentrations of LPS were combined, another variable representing the LPS concentration as well as its interaction with the other variables was added to the model. In both cases, the p values for the main effect of adding sCD14 were reported.