Jagged1 Instructs Macrophage Differentiation in Leprosy

As circulating monocytes enter the site of disease, the local microenvironment instructs their differentiation into tissue macrophages (MΦ). To identify mechanisms that regulate MΦ differentiation, we studied human leprosy as a model, since M1-type antimicrobial MΦ predominate in lesions in the self-limited form, whereas M2-type phagocytic MΦ are characteristic of the lesions in the progressive form. Using a heterotypic co-culture model, we found that unstimulated endothelial cells (EC) trigger monocytes to become M2 MΦ. However, biochemical screens identified that IFN-γ and two families of small molecules activated EC to induce monocytes to differentiate into M1 MΦ. The gene expression profiles induced in these activated EC, when overlapped with the transcriptomes of human leprosy lesions, identified Jagged1 (JAG1) as a potential regulator of MΦ differentiation. JAG1 protein was preferentially expressed in the lesions from the self-limited form of leprosy, and localized to the vascular endothelium. The ability of activated EC to induce M1 MΦ was JAG1-dependent and the addition of JAG1 to quiescent EC facilitated monocyte differentiation into M1 MΦ with antimicrobial activity against M. leprae. Our findings indicate a potential role for the IFN-γ-JAG1 axis in instructing MΦ differentiation as part of the host defense response at the site of disease in human leprosy.

As circulating monocytes enter the site of disease, the local microenvironment instructs their differentiation into tissue macrophages (MΦ). To identify mechanisms that regulate MΦ differentiation, we studied human leprosy as a model, since M1-type antimicrobial MΦ predominate in lesions in the self-limited form, whereas M2-type phagocytic MΦ are characteristic of the lesions in the progressive form. Using a heterotypic co-culture model, we found that unstimulated endothelial cells (EC) trigger monocytes to become M2 MΦ. However, biochemical screens identified that IFN-γ and two families of small molecules activated EC to induce monocytes to differentiate into M1 MΦ. The gene expression profiles induced in these activated EC, when overlapped with the transcriptomes of human leprosy lesions, identified Jagged1 (JAG1) as a potential regulator of MΦ differentiation. JAG1 protein was preferentially expressed in the lesions from the self-limited form of leprosy, and localized to the vascular endothelium. The ability of activated EC to induce M1 MΦ was JAG1-dependent and the addition of JAG1 to quiescent EC facilitated monocyte differentiation into M1 MΦ with antimicrobial activity against M. leprae. Our findings indicate a potential role for the IFN-γ-JAG1 axis in instructing MΦ differentiation as part of the host defense response at the site of disease in human leprosy.

Author Summary
Mycobacterial diseases, such as leprosy, continue to be serious causes of mortality and morbidity worldwide. They pose a unique treatment challenge due to their ability to modify the immune response in infected individuals. For example, in leprosy there are two distinct manifestations of the disease, each characterized by the immune response of the individual. One results in a more disseminated and severe form of the disease, lepromatous leprosy, and the other is a more limited form with marked antimicrobial activity,

Introduction
When circulating monocytes enter the site of disease, local cues from the tissue microenvironment direct their differentiation into specialized MF equipped for diverse tasks [1][2][3]. While classically activated M1 MF with antimicrobial activity promote host defense against intracellular pathogens, alternatively activated (M2) MF perform homeostatic functions including phagocytosis critical to tissue remodeling [1][2][3][4][5][6]. In leprosy, the divergence of MF functional programs correlate with the clinical disease spectrum [7][8][9]. In the self-limited, tuberculoid (Tlep) form of leprosy, disease lesions contain well-organized granulomas with M1 MF, expressing the MF marker CD209, but negative for the haptoglobin receptor CD163, yet armed with antimicrobial effector function [7]. By contrast, in the progressive, lepromatous (L-lep) form of leprosy, patient lesions are characterized by disorganized granulomas containing MF which co-express CD209 and CD163 but lack antimicrobial activity. Instead, these MF are programmed with phagocytic function, which results in the accumulation of host-derived lipids and favors mycobacterial growth [10,11], and are therefore referred to as M2 MF. These data raise the question regarding the mechanisms by which clues from the microenvironment influence MF programming at the site of infection.
As a gatekeeper to circulating monocytes that enter disease lesions, the microvasculature is poised to deliver key differentiation cues. The very cells which allow monocytes to exit the blood and enter the site of disease, i.e. EC, were shown to trigger monocyte differentiation into MF [12], specifically of the M2 type [13]. Therefore, unstimulated EC have the ability to instruct M2 MF differentiation, yet the conditions that might alter EC to instruct M1 MF differentiation are not known. Here, we explore how the EC-monocyte interface can influence M1 MF differentiation, including upregulation of antimicrobial activity, in the context of leprosy as a human disease model.

Results
Given the critical role the microvasculature plays in the transmigration of circulating leukocytes, it is poised to deliver instructive cues to monocytes entering the site of disease [12,14,15]. To investigate how the microvasculature, specifically EC, influence MF differentiation, we chose leprosy as a model, focusing on M1 and M2 MF that expressed CD209, and the relative expression of CD163, either low and high, reflecting the major MF phenotypes at the site of disease and endowed with distinct functional programs. We hypothesized that a resting microenvironment leads EC to instruct monocyte differentiation into M2 MF with phagocytic function; whereas, perturbations in the local microenvironment may direct monocytes to differentiate into M1 MF with antimicrobial activity (Fig 1A).
When human monocytes were cultured in the presence of several types of EC, but not vascular smooth muscle cells, we observed differentiation of monocytes into MF expressing CD209 (S1 Fig). As a first step to explore how the endothelial microenvironment may influence M1 vs. M2 MF differentiation, EC were treated with various regulatory cytokines before adding human peripheral blood mononuclear cells. After co-culture for 48h, monocyte differentiation was assessed by flow cytometry. For most of the cytokines used in the pre-treatment, EC triggered differentiation into a comparable percentage of M2 MF co-expressing CD209 and CD163 ( Fig 1B). However, only IFN-γ-treated EC facilitated monocyte differentiation into the CD209+CD163 neg M1 MF phenotype associated with host defense [7,16]. These M1 MF also expressed higher levels of CD40 [17,18] (Fig 1C and 1D), but not the dendritic cell marker CD1a (S2 Fig). To assess whether this effect was specific to the Type II IFN, IFN-γ, we also tested the Type I IFN, IFN-α, which failed to mediate similar EC-triggered MF differentiation ( Fig 1E). Monocyte differentiation into CD209+ MF was dependent on direct contact with EC and did not occur when monocytes were treated with IFN-γ in the absence of EC (S3A and Next, we compared the MF derived from activated vs. resting EC for antimicrobial vs. phagocytic characteristics, given that our previous studies showed these programs to be divergent [7]. The M1 MF (CD209 + CD163 neg ) induced by IFN-γ-treated EC, when compared to the M2 MF induced by resting EC (CD209 + CD163 + ), were found to: i) take up less oxidized low density lipoprotein (oxLDL) (p <0.01) (Fig 1F), ii) produce greater amounts of pro-inflammatory cytokines in response to stimulation with a mycobacterial TLR2/1 ligand (p <0.05) (Fig 1G), and iii) express greater levels of the vitamin D antimycobacterial pathway genes Cyp27b1, VDR and cathelicidin [7] (Fig 1H). Therefore, IFN-γ, which is known to play a critical role in host defense against M. leprae and other mycobacterial pathogens [4,8,19,20], licensed the EC to instruct monocyte differentiation into M1 MF programmed with upregulation of the vitamin D antimicrobial pathway.

Diverse compounds can facilitate endothelial cell-driven antimicrobial macrophages
IFN-γ is a potent inflammatory mediator that regulates an extensive gene program in EC [19][20][21]. We envisioned that small molecules that would facilitate EC-driven M1 MF differentiation may do so through convergence upon shared regulatory mechanisms (Fig 2). From a small molecule library generated by diversity-oriented synthesis (n = 642) [22], 24 compounds (3.7%) were identified which when used to treat EC, promoted M1 MF differentiation as measured by cell surface phenotype (Fig 3A). Two structurally distinct families, naphthyridines and tetrahydro-pyrrolo-triazolo-pyridazindiones (tptp) accounted for 13 (54%) of the "hits", and subsequent experiments with compounds from each of these families confirmed that upon treatment of EC, they triggered differentiation of M1 MF ( Fig 3B). As with IFN-γ, this effect was EC-dependent, since the compounds failed to directly trigger monocytes to become MF that express CD209 (S4 Fig). Among 81 naphthyridine analogs [22], 34 (42%) prompted EC to instruct monocyte differentiation into M1 MF (S5 Fig), indicating some specificity among naphthyridines. As with IFN-γ treated EC cultures, M1 MF derived from the compound-activated EC cultures were significantly less phagocytic than the M2 MF derived from the resting EC cultures (p < 0.0001) (Fig 3C and 3D). We chose the most effective compound (naphthyridine 105A10) for further analysis and found that 105A10-treated EC triggered MF that were also more responsive to TLR2/1 activation in terms of induction of pro-inflammatory cytokines ( Fig 3E).

Identification of candidate regulatory genes leading to the induction of antimicrobial M1 MΦ in leprosy
Having identified structurally diverse compounds that mimicked IFN-γ, we sought to use these compounds to explore the mechanisms by which EC trigger this differentiation. Since IFN-γ signaling is primarily through STAT-1, we sought to determine if active compounds from both families increased the phosphorylation of STAT-1 in treated EC. Active compounds from both tptp and naphthyridine families of compounds failed to induce phosphorylated STAT-1 ( Fig  4A). To determine whether the various stimuli induce a common gene signature in EC, we measured the gene expression profiles in EC treated with either IFN-γ, IFN-α or one of four active small molecules (two naphthyridines: 105A9, 105A10 and two tptp family members: 104B11, 104C2). IFN-γ induced a broad profile (n = 3675 probes >1.25-fold induction), by comparison, the four compounds induced a more restricted profile, (range n = 1248-1935 probes >1.25-fold induction). A high proportion of the genes induced by the four compounds (24-29%) overlapped with the IFN-γ signature (hypergeometric p values for enrichment: 5.35 x 10 −32 to 2.80 x 10 −103 , Fig 4B).  To identify the genes triggered in activated EC with relevance to leprosy, we overlapped three profiles: i) induced by IFN-γ in EC, ii) induced by at least one of the four small molecules in EC; and, iii) preferentially expressed at the site of disease in the self-limited T-lep vs. the progressive L-lep form of leprosy (Figs 2 and 5A). This analysis identified 166 candidate regulatory genes, of which 50 were induced by at least two of the four compounds (S1 Table).
We next tested the role of the 50 common genes in facilitating EC-directed M1 MF differentiation. EC were transfected with siRNA against each of these candidate genes, and then treated with IFN-γ to induce the M1 polarizing microenvironment, followed by co-culture with primary human PBMC. In this context, monocyte differentiation into CD163+ MF would reflect that the M1 MF-polarizing effect of IFN-γ treated EC was being inhibited by the siRNA. Across five separate experiments, eight genes significantly inhibited the effect that IFNγ exerts on the EC-driven M1 MF phenotype (    After confirming that JAG1 is induced on EC following stimulation with IFN-γ (S7 Fig), we then assessed JAG1 expression at the site of disease. JAG1 expression in leprosy lesions was validated by immunohistochemistry, which demonstrated that JAG1 was expressed within the dermis and the granulomas in T-lep, but not L-lep lesions (Fig 5E, S8 Fig). We also noted perivascular labeling of JAG1 in proximity to CD209 + MF (Fig 5F), as well JAG1 expression in the microanatomic locations in which M1 MF (CD209 + CD163 neg ) were found (Fig 5F). In addition, there appeared to be JAG1 staining in the epidermis of both the T-lep and L-lep lesions which is consistent with the known role of JAG1 in keratinocyte differentiation and maturation [23,24]. Blinded analysis of JAG1 immunohistochemical staining determined a significant (p = 0.0063) increased positive staining in T-lep sections, scores ranged from 0 (absent) to 4 (highly positive). Together, these data indicated that JAG1 expression correlated with M1 MF accumulation at the site of disease in leprosy.

JAG1 triggers antimicrobial macrophages in an endothelial celldependent manner
We next investigated whether JAG1 could instruct the differentiation of monocytes into M1 MF with antimicrobial function. We found that soluble JAG1 (sJAG1) facilitated EC-driven M1 MF differentiation (Fig 6A and 6B). Furthermore, overexpression of JAG1 in EC, as well as addition of a JAG1 agonist peptide to the co-cultures, induced the differentiation of monocytes into the M1 MF phenotype (S9 Fig). In contrast, addition of sJAG1 to monocytes alone did not induce MF differentiation (S10 Fig). Given that JAG1 is known to activate Notch 1 signaling, we determined whether Notch-downstream genes were upregulated by the addition of JAG1 to the EC/monocyte co-cultures. In comparison to untreated EC, the addition of JAG1 led to the mRNA upregulation of three prototypic Notch-downstream genes in MF, HES1, SOCS3 and RBPJ (S11 Fig).
Differentiation of monocytes in the presence of sJAG1 and EC yielded M1 MF with decreased phagocytosis (Fig 6C) and heightened induction of vitamin D-dependent antimicrobial pathway genes (Fig 6D). To determine whether EC treated by either IFN-γ or JAG1 induced differentiation of monocytes into MF with antimicrobial activity, MF differentiated in the presence of treated EC were infected with live M. leprae, and the antimicrobial response measured according to the ratio of M. leprae RNA to DNA [25,26] (Fig 6E). As compared to MF differentiated in the presence of resting EC (i.e. treated with media), the MF induced by culture with EC treated with either IFN-γ or sJAG1 showed significant antimicrobial activity. Therefore, when monocytes encounter JAG1 in the context of EC, a differentiation program is triggered, resulting in M1 MF, defined by a CD209 + CD163 neg phenotype and antimicrobial function. The presence of IFN-γ, JAG1-expressing EC and CD209 + CD163 neg MF in the selflimited form of leprosy suggests that the IFN-γ-JAG1-antimicrobial MF differentiation pathway contributes to host defense at the site of disease in leprosy.

Discussion
Our understanding of MF immunobiology has been significantly advanced through understanding of the pathways by which microbial ligands and/or cytokines program monocytes to differentiate into M1 and M2 MF [27,28]. However, it is not clear how local tissue signals can differentially program the MF response. Signals from endothelium are involved; this default pathway triggers M2 MF differentiation [13]. However, the mechanisms by which monocytes, upon entering the site of disease via the endothelium, are instructed to differentiate into M1 MF remain elusive [1][2][3]. Here, we hypothesized that if EC were to encounter the proper signals, the EC microenvironment would instruct monocytes to differentiate into M1 MF, equipped for host defense against intracellular pathogens at the site of disease. By studying leprosy as a model, we provide evidence that upregulation of JAG1 on endothelium instructs monocytes to differentiate into M1 MF with antimicrobial activity.
Our data indicate that the induction of JAG1 is involved in EC instruction of M1 MF differentiation. In addition, the concomitant induction of Notch 1-downstream genes including HES1, SOCS3 and RBP-J mRNAs was detected in the differentiated M1 MF. These findings are consistent with the known ability of JAG1 to signal via Notch 1 receptors [29], and with reports that Notch 1 signaling, via SOCS3 and RBP-J [30][31][32] through reprogramming of mitochondrial metabolism [33], contributes to M1 MF differentiation. Nevertheless, since JAG1 is known to signal via several distinct receptors [34,35], further work is necessary to identify the physiologically relevant interactions responsible for EC-driven M1 MF differentiation. Not only does IFN-γ induce JAG1 on EC which can influence monocyte differentiation, IFN-γ also augments TLR-induced regulation of JAG1 expression in differentiated MF [36]. Further studies will be required to elucidate how JAG1 can contribute to MF differentiation, plasticity, function and proliferation at the site of disease [37].
In addition to the role of JAG1 in regulating innate immune responses via MF differentiation, evidence suggests a role for JAG1 in regulating adaptive T cell responses. Patients with Alagille syndrome, in which JAG1 mutations result in a multisystem disorder [34,38], can exhibit altered Th1 responses [35], implicating JAG1 induced signaling in T cell differentiation. In vitro studies have also shown that JAG1 expression on keratinocytes promotes dendritic cell maturation, which could also influence T cell responses [39]. Therefore, the expression of JAG1 by resident cells in tissue can influence both innate and adaptive immune responses.
Under resting conditions, EC instruct monocytes to differentiate in M2 MF [13]. M2 MF are highly phagocytic, and are involved in clearing various biomolecules relevant for tissue repair, removal of excess metabolic products as well as clearance of debris. However, in the context of M. leprae infection, M2 MF can phagocytize the bacteria, but are unable to mount an antimicrobial response. Furthermore, these M2 MF take up host-derived lipids, providing necessary nutrients for mycobacterial growth [40]. Therefore, the induction of M1 MF is required for host defense against this intracellular pathogen, as these MF are weakly phagocytic but exhibit a strong antimicrobial response. One direct signal at the site of infection is production of IL-15, which directly triggers M1 MF differentiation. In addition, our data demonstrates that IFN-γ induces JAG1 expression on EC, which also facilitates differentiation of monocytes into M1 MF. In the self-limited form of leprosy, JAG1 expression is restricted to microanatomical regions of the granuloma enriched for M1 MF. Therefore, our findings support the concept that the IFN-γ-JAG1 axis is involved in the EC instruction of the antimicrobial MF response against M. leprae at the site of infection. The ability to model how the microenvironment influences the immune response at the site of disease has become feasible because of advances in analyzing increasingly complex systems. We used a cell co-culture system in which we integrated small molecule screening with gene expression profiles to look for recurrent motifs in gene activation patterns associated with ECtriggering M1 MF. Since none of the molecular signals we identified recapitulate the antimicrobial MF phenotype on their own, our findings indicate that the emergent properties inherent to more complex heterotypic systems allowed for their discovery [41]. As such, this approach provides a strategy to identify potential drugs or biologic agents that would otherwise not be identified in experiments exploring direct effects on monocyte differentiation into antimicrobial MF. The identification of JAG1 and other small molecules that can harness the local microenvironment to augment innate immune responses at the site of disease may hold promise for combating intracellular pathogens.

Co-culture experiments
Co-culture experiments were carried out as previously described [22]. In short, Primary human endothelial cells (EC) were plated to confluence in a 96 well plate. After adherence, endothelial cells were activated by indicated treatments for a period of 5 hours and subsequently washed 2-3 times to ensure removal of activation treatment. We then added human peripheral blood mononuclear cells (PBMC) at a ratio of 3 PBMC to 1 EC. Cultures were incubated at 37°C and 7% CO 2 for a period of 48hrs. Human Umbilical Vein Endothelial Cells (HUVEC) were purchased from Lonza and used from passages 4-8. Peripheral blood mononuclear cells were isolated from healthy donors (UCLA Institutional Review Board # 92-10-591-31) using Hypaque Ficoll gradients (GE Healthcare).

MΦ TLR activation cytokine response
PBMC/EC Co-cultures were harvested after 48 hours of incubation and CD14 + MF were subsequently purified using a CD14 positive selection bead assay (Miltenyi Biotec) (purity > 95%). CD14 + MF from each condition (DMSO, IFN-γ and 105A10) were plated in equal number in 96 well flat bottom plates and stimulated with 10μg/ml TLR2/1 ligand (EMC Microcollections). After 24 hours of stimulation supernatants were harvested and characterized by CBA for production of MIP1-β, TNF-α and IL-6.

Real time PCR
cDNA was generated using iScript cDNA synthesis reagent (Biorad) following manufacturers guidelines. Primers (IDT) were used for determining mRNA expression of CYP27b, CAMP, VDR, and JAG1. SYBR Green PCR Master Mix (BioRad) was used for Real Time PCR reactions and data was normalized to h36B4 gene expression (IDT). Expression values were calculated as previously described [7].

Vitamin D pathway regulation
CD14+MF from co-cultures were harvested and purified as previously mentioned. After purification, MF were plated in 10% FCS with 25-D 3 (10 −8 M) (Biomol) and incubated for 24 hrs. Cells were then harvested and analyzed for CAMP, VDR and Cyp27b1 gene expression by qPCR.

Antimicrobial assays
Viable bacteria stocks of M. leprae were obtained from Dr. James L. Krahenbuhl of the National Hansen's Disease Programs, Health Resources Service Administration, Baton Rouge, LA. For antimicrobial assays, Endothelial/PBMC co-cultures were set up as previously mentioned. After 48 hours of incubation CD14 + MF (>90% CD209 + ) were isolated from co-cultures for infection with M. leprae. Co-culture conditioned (Media, IFN-γ and sJAG1) CD14 + MF were cultured in RPMI with 10% FCS (Omega Scientific) in the presence of live M. leprae (MOI 10:1). Infected cells were subsequently stimulated with IFN-γ (10ng/ml) in the presence of 25-D 3 (10 −8 M) after 24 hours of infection. To measure antimicrobial activity in M. lepraeinfected MF (5 days post infection) we followed the protocol as previously described [40,42]. In short, qPCR was performed to determine levels of bacterial 16S rRNA and genomic element DNA (RLEP). Expression levels of h36B4 were also evaluated to determine infectivity between all the conditions. The M. leprae 16S rRNA and RLEP primers used were as previously described [25,26].

IHC and confocal microscopy
Immunoperoxidase and immunofluorescence labeling were carried out on frozen patient tissue sections. For immunoperoxidase staining, samples were initially blocked with normal horse or goat serum prior to labeling with monoclonal antibodies (JAG1 (Abcam), vWF (AbD Serotec), CD163 (AbD Serotec) and appropriate isotype controls). Sections were then labeled with biotinylated horse anti-mouse IgG or biotinylated goat anti-rabbit IgG. After labeling, sections were counterstained with hematoxylin and visualized using the ABC Elite system (Vector Laboratories). In order to determine protein co-localization in tissue sections, two-color immunofluorescence and confocal microscopy were performed. For Immunofluorescence, sections were labeled with rabbit anti-human JAG1, anti-CD163 (IgG1), anti-CD209 (IgG2b), anti-vWF (IgG1) and appropriate isotype controls. Subsequently samples were labeled with isotype-specific, fluorochrome (A488 or A568)-labeled goat anti-mouse/rabbit immunoglobulin antibodies (Molecular Probes). Nuclei were stained with DAPI (4',6'-diamidino-2-phenylindole). Double immunofluorescence of skin sections was examined using a Leica-TCS-SP MP inverted single confocal laser scanning and a two-photon laser microscope (Leica, Heidelberg, Germany) at the Advanced Microscopy/Spectroscopy Laboratory Macro-Scale Imaging Laboratory, California NanoSystems Institute, University of California at Los Angeles [26]. Blinded review of IHC samples was carried out and positive staining was scored on the scale of 0 (absent) to 4 (highest staining) relative to isotype controls. Fishers exact test was used to determine significance.

Transfections
siRNA transfections were carried out on 2x10 4 HUVEC in 96 well plates and 7x10 3 HUVEC in 384 well plates. siRNA for candidate genes, siControl and siGlow were obtained from Dharmacon as was the transfection reagent Dharmafect 4. siRNA transfections were performed according to manufacturer's recommendations using 100nM concentration of siRNA. Decrease in message in transfected cells was confirmed by qPCR and protein expression. Ectopic expression cassettes for JAG1, GFP and M11-empty vector were obtained from Genecopoeia. Plasmid transfections were carried out on HUVEC that were grown to 80-90% confluence. HUVEC were harvested and transfected with 1μg DNA using the AMAXA transfection device and HUVEC Nucleofect kit (Lonza). To determine transfection efficiency, control cells were characterized for GFP production. In addition, surface expression of transfected JAG1 was confirmed by flow cytometry.

Microarrays
For microarrays performed on compound and cytokine treated HUVEC, ECs were seeded in 6 well plates at 1X10 6 cells/well. Single wells were stimulated for five hours with DMSO, IFN-γ, IFN-α and compounds 104B11, 104C2, 105A9 and 105A10 at concentrations noted earlier.
After incubation, mRNA was harvested using Trizol (Invitrogen), followed by RNeasy Minelute Cleanup Kit (Qiagen). mRNA samples for all arrays were processed using the Affymetrix Human U133 plus 2 platform and analyzed as previously described [22].

Statistical analysis
Statistical significance (<.05) of experimental values was calculated using a paired two-tailed Student's t-test. Hypergeometric p values were calculated using the online resource (http:// systems.crump.ucla.edu/hypergeometric) (Tom Graeber laboratory, UCLA).

Ethics statement
Patient samples were obtained with approval from the IRB of the University of California Los Angeles, the Institutional Ethics Committee of Oswald Cruz Foundation and the University of Southern California School of Medicine. All subjects were legal adults and provided written informed consent before participating in the study [26].  Table. Candidate genes used in siRNA screen. Top fifty candidate genes used in siRNA screen. Fold change of gene expression in treated over media control HUVEC samples from microarray assays. In addition, the fold change of gene expression in T lep vs L lep tissue biopsy microarrays(BT/LL) and P-values (represent significance of differences in gene expression in T lep and L lep tissue biopsy microarrays) are displayed in the right two columns. (TIF)