Differential DARC/ACKR1 expression distinguishes venular from non-venular endothelial cells in murine tissues

Mice

C57BL/6 (catalog number 027) and BALB/c mice (RRID: IMSR_JAX:000651, catalog number 000651), 6–12 weeks old, were purchased from Charles River or Jackson Laboratories. DARC^–/– mice [25] were bred at the University of York and used as tissue donors. Apoe ^−/− mice [26] were obtained from Jackson Laboratories (RRID: IMSR_JAX:002052, catalog number 002052). BM chimeras were generated by irradiating C57BL/6 mice (2 × 650 Rad) followed by intravenous (IV) injection of unfractionated DARC^–/– BM mononuclear cells and a rest period of more than 12 weeks before use. Mice were housed under specific pathogen-free conditions in accordance with NIH guidelines. Experimental protocols were approved by the Institutional Animal Care and Use Committee at Harvard Medical School.

Construction of expression plasmids

Monoclonal anti-mouse DARC antibody

The rat cell line PC-12 (ATCC) was grown in F-12 K media (Gibco) with 2.5% fetal bovine serum (FBS) and 15% horse serum. The human embryonic kidney cell line HEK-293 was grown in DMEM (Corning) with 10% FBS. Both cell lines were stably transfected in 10-cm Petri dishes with 2 μg of plasmid using Lipofectamin 2000 (Invitrogen) following the manufacturer’s protocol. PC-12 cells were transfected with DARC-eGFP fusion protein and linker LG (named DARC-eGFP1) and HEK-293 were transfected with DARC-eGFP fusion protein and linker GGGGSAAA (named DARC-eGFP3). The antibiotic G418 (GIBCO) was added at a final concentration of 400 μg/mL 24 h after transfection. The selection medium was renewed every 3–4 days. Stable transfectants were further selected by cell sorting based on GFP expression. PC-12 DARC-eGFP transfectants were conditioned overnight with media supplemented with 10 mM sodium butyrate solution to boost transgene expression before immunization. Adult rats were immunized four times at 2-week intervals with 5 × 10⁶ sodium butyrate-conditioned DARC-eGFP PC-12 cells. For the first immunization, transfectants were suspended in complete Freund’s adjuvant and injected by sub-cutaneous (s.c.) and intra-peritonal (i.p.) routes; the first booster injection of transfectants suspended in incomplete Freund’s adjuvant was performed by i.p. and subsequent injections were performed without adjuvants by the i.p. route. Immunization and fusion of hybridomas were performed by Abpro Biotechnology company under a service contract. Cloning and sub-cloning of the hybridomas was performed in the Dana Farber Monoclonal Antibody Core, Boston. Immune sera were screened by flow cytometry for reactivity with DARC ectodomains using HEK-293 cells expressing DARC-eGFP fusion protein and RBCs from wildtype (WT) or DARC^–/– mice. Following splenocyte fusion, twelve 96-well plates were screened by flow cytometry. Two wells showed reactivity against mouse DARC. One clone producing an anti-mouse DARC MAb was isolated, expanded, and subcloned; the MAb was determined to be a rat IgG2a,k isotype. See Additional file 1: Figure S1.

Antibodies for immunofluorescence staining

Confocal microscopy and image analysis

Two whole mounts of fresh murine omentum were prepared and stained for CD31 and DARC, as described above, and z-stacks of confocal micrographs were acquired (10 or 15 images at 2.9 μm vertical step intervals). Fluorescent images were used for 3D rendering using Imaris software. Supporting data values for all surface areas are included in Additional file 2.

Confocal images were acquired on an Olympus Fluoview BX50WI inverted microscope with × 10/0.4, ×20/0.5, and × 40/1.3 objectives. Image stacks for three-dimensional reconstructions were acquired at 1–4.5 μm Z-intervals. Image analysis was performed using Volocity (Improvision) or Imaris (Bitplane) software.

Preparation of aorta

To induce atherosclerosis, Apoe^–/– mice were fed a Western diet (Harlan Teklad) for 12 weeks prior to analysis [28]. Aortic root tissue was frozen in OCT and 5-μm thick sections were cut with a cryostat. Sections were fixed in isopropanol for 10 min at 4 °C and blocked with PBS + 10% goat serum + 0.5% BSA. Staining was performed overnight at 4 °C with AF546-anti-DARC antibody + APC-anti-CD31 (MEC 13.3). Samples were fixed in 4% PFA for 10 min, nuclei were stained with Yoyo-1, and sections were mounted using Prolong Gold reagent. Images of the vasa vasorum of aorta were acquired on a Leica SP5 microscope.

Tissue digestion

Lymph node and brain tissues were digested with 2.5 mg/mL Collagenase D (Roche), 50 μg/mL DNAse I (Roche), and 0.4× protease inhibitor (Roche) in digestion buffer at 37 °C. Digestion of LNs was adapted from Fletcher et al. [29], such that tissues were digested for 20 min without protease inhibitor followed by three 10-min incubations with protease inhibitor. Brain tissues were digested for 15–20 min at 37 °C. Enrichment of ECs was performed by centrifugation at 5000 g for 30 min at 4 °C in a dextran solution (17% dextran (Sigma, catalog number 31392)/20 mM HEPES). Skin, colon, and small intestine tissues were digested with 2.5 mg/mL Collagenase D (Roche), 50 μg/mL DNAse I (Roche), and 1× protease inhibitor (Roche) in digestion buffer for 30 min at 37 °C on a rotisserie wheel. Colon and small intestine were washed with 5% FBS and 25 mM HEPES, followed by 2 mM EDTA and 25 mM HEPES, and finally 10% FBS, 5 mM EDTA, and 15 mM HEPES prior to enzymatic digestion to remove epithelial cells. Pancreatic and adipose tissues were digested with 1.25 mg/mL Collagenase D (Roche), 50 μg/mL DNAse I (Roche), and 1× protease inhibitor (Roche) in digestion buffer for 30 min at 37 °C on a rotisserie wheel. Cells were resuspended in FACS buffer (PBS (Lonza), 5% FBS (Invitrogen-Gibco), 5 mM EDTA (Boston BioProducts)) for analysis.

Flow cytometry

Single-cell suspensions were immunostained, washed, and resuspended in FACS buffer containing 7-AAD viability stain (BioLegend) for immediate acquisition using a BD FACS CANTO, BD LSRII (BD Biosciences), or CytoFlex (Beckman Coulter) and analyzed using FlowJo software (Treestar).

RT-qPCR

EC subsets were FACS-sorted using an Aria I cell sorter (BD Biosciences) and collected in Trizol (Ambion). RNA extraction was performed prior to cDNA synthesis using the iScript kit (BioRad). qPCR was performed on a LightCycler 480 II (Roche) using SybGreen Quantifast kit (Qiagen). Relative gene expression was calculated using the ∆CT method. See list of primers in Table 1.

Intravital microscopy (IVM)

IVM of subiliac lymph node, skin, BM, and cremaster muscle was performed as previously described [5, 30–32] using an IV-500 intravital microscope (Mikron Instruments), equipped with a Rapp OptoElectronic SP-20 xenon flash lamp system and QImaging Rolera-MGi EMCCD camera.

Western blot analysis

Protein extracts were loaded on a 4–12% Tris-Glycine precast gel (Invitrogen). Precision Plus protein Kaleidoscope standards (Bio-Rad, catalog number 161-0375) were used as controls for molecular weight. Heat-denaturation of protein samples was performed. Membranes were incubated with anti-mouse DARC MAb and revealed using HRP-linked anti-rat antibody (Cell Signaling, RRID AB_10694715, catalog number 7077S). Anti-β-actin antibody (Sigma, RRID AB_476743, catalog number A5316) was used for loading control. Immunoreactive proteins were detected with Pierce ECL Western Blotting Substrate (Thermo Scientific).

Chemokine binding assay

Chemokine binding to RBCs was assessed as previously described [33], with slight modifications. RBCs were incubated with increasing chemokine concentrations (10^–10–10^–5 M or 10^–11–10^–7 M of CXCL8 or mCXCL1, respectively), with either anti-human DARC MAb Fy6 (gift from Dr. M. Uchikawa, Japanese Red Cross) or anti-mouse DARC MAb. Mean fluorescence of MAb staining was analyzed by FACS. IC₅₀ was calculated using GraphPad Prism software.

Statistical analysis

Results are expressed as means ± SEM. All statistical analyses were performed in Prism (GraphPad Software) or Excel Software. Means between two groups were compared by two-tailed t-test. Means between three or more groups were compared by one-way or two-way ANOVA with Tukey’s or Dunnett’s Multiple Comparison Test. The numbers of replicates or animals are indicated in figure legends.

Validation of anti-mouse DARC mAb specificity

After immunization of a rat with DARC-eGFP transfected PC-12 cells, hybridoma lines were generated and supernatants screened by FACS. One hybridoma supernatant stained HEK-293 cells transfected with DARC or DARC-eGFP (Fig. 1a), as well as RBCs from C57BL/6 WT mice (Fig. 1b), but not untransfected HEK-293 cells or DARC^–/– RBCs. This hybridoma was subcloned and determined to produce a single MAb of IgG2a,k isotype. Both supernatant and purified IgG recognized DARC protein in lysates of DARC or DARC-eGFP transfected HEK-293 cells by Western blot (Fig. 1c). No signal was detected with the parental untransfected HEK-293 line. Robust immunoreactivity was also observed with TER-119⁺ RBCs from Balb/c mice. The MAb did not cross-react with RBCs from other species, including rat and human (Additional file 3: Figure S2). Confocal microscopy of peripheral lymph node (PLN) sections revealed that anti-DARC MAb colocalized with MAb MECA-79, which specifically delineates peripheral node addressin (PNAd), a marker of HEVs [34]. No microvascular reactivity was detected in DARC^–/– PLNs (Fig. 1d).

To assess whether the anti-DARC MAb is sensitive to DARC occupancy with chemokines, we assessed staining of WT RBCs in the presence of CXCL1 or CXCL8. Neither chemokine affected binding of the anti-mouse MAb, whereas both blocked staining of human RBCs with anti-human DARC MAb Fy6 (Additional file 3: Figure S2C), consistent with previous studies [35, 36]. Thus, although the exact epitope recognized by our anti-mouse DARC MAb was not identified, it is unlikely to overlap with the chemokine binding region(s) of DARC.

DARC expression is restricted to V-ECs

Confocal micrographs of whole mounts of murine adipose tissue and skin showed that DARC immunoreactivity was exclusively biased toward venules (Fig. 2a). No signal was detected in either arterioles or the vast majority of capillaries. On V-ECs, DARC appeared to be enriched at intercellular junctions between adjacent ECs.

To quantify the frequency of DARC⁺ and DARC^– microvessels, we categorized CD31⁺ microvascular segments based on their morphology and caliber in whole mounts of omentum. Virtually all PCVs and CVs were DARC⁺, whereas arterioles and downstream capillaries were consistently DARC^–. The transition from DARC^– to DARC⁺ ECs occurred either at the point where capillaries merged to form a PCV or slightly upstream in pre-venular capillaries (Fig. 2b and Additional file 4: Figure S3). DARC staining in venules was generally uniform and without obvious gaps in immunoreactivity, suggesting that essentially all V-ECs express DARC. Analogous distribution patterns were detected in multiple tissues, including cremaster muscle, bladder, conjunctiva, brain, Peyer’s patches, mesenteric LNs (MLNs), thymus, pancreas, small intestine, and colon (Fig. 3). DARC was undetectable in large veins or arteries, including the lumen of the aorta (Additional file 5: Figure S4 and Additional file 6: Figure S5A). However, DARC was readily detected in vasa vasorum around the aorta of WT mice, and expression in these vessels appeared to be further enhanced in atherosclerotic aortas in Apoe ^–/– mice (Additional file 6: Figure S5B).

Flow cytometric analysis of DARC expression on microvascular EC subsets

To more precisely define the frequency and distribution of DARC⁺ and DARC^– ECs, we stained single-cell suspensions from normal tissues with fluorophore-conjugated anti-mouse DARC MAb combined with markers of leukocytes (CD45), myeloid phagocytes (CD11b), RBCs (TER-119), the pan-endothelial marker CD31, and gp38, which detects lymphatic ECs (LECs). This staining strategy combined with gating on characteristic forward and side light scatter profiles allowed simultaneous discrimination of V-EC (CD45^–TER-119^–CD11b^–gp38^–CD31⁺DARC⁺), NV-ECs (CD45^–TER-119^–CD11b^–gp38^–CD31⁺ DARC^–) and LECs (CD45^–TER-119^–CD11b^–gp38⁺CD31⁺) in multiple tissues (Fig. 4a).

To obtain single-cell suspensions, a number of mouse tissues, including skin, LNs, adipose tissue, colon, small intestine, brain, and pancreas, were mechanically dissociated and incubated with a mixture of proteases. Pilot experiments revealed that the epitope recognized by the anti-mouse DARC MAb on both ECs and RBCs was susceptible to degradation by neutral proteases (Additional file 7: Figure S6). We developed, for each tissue, specific digestion protocols that could reliably preserve DARC staining on both RBCs and V-ECs using distinct combinations of protease cocktails and protease inhibitors (Fig. 4b).

Using this approach, we assessed the frequency of V-ECs among total blood ECs (BECs) in each tissue sample by flow cytometry (Fig. 4c). The recovery of DARC⁺ V-ECs relative to the total yield of BECs was similar in skin and LNs while adipose tissue, colon, small intestine, brain, and pancreas yielded a lower frequency of DARC⁺ V-ECs among BECs. Several factors could potentially contribute to the observed differences between tissues in apparent V-EC frequency, as detected by flow cytometry. (1) V-ECs in some tissues may contain a variably sized population that does not express DARC. However, given the uniformity of venular DARC staining in whole mount preparations, we consider this unlikely. (2) Proteases used to prepare single-cell suspensions for FACS analysis may have destroyed DARC epitopes on a fraction of V-ECs in some tissues, which could potentially lead to an underestimation of the DARC⁺ EC fraction in those tissues. However, our tissue digestion protocols were devised to mitigate this effect as far as possible, and we monitored Ter-119⁺ RBCs in tissue preparations to ensure conservation of DARC epitopes. (3) The relative rate of recovery of V-EC versus NV-EC may differ between tissues. (4) The observed measurements may reflect true differences in V-EC abundance between tissues.

To independently address the latter two (non-exclusive) points, we conducted a digital 3D analysis of a whole mount staining of murine omentum by quantifying the percentage of DARC⁺ venular surface area in the total vascular (CD31⁺) surface area. In the chosen sample, the DARC⁺ fraction was approximately 14% of the total CD31⁺ surface area (Additional file 8: Figure S7). By comparison, the average frequency of DARC⁺ V-ECs among total ECs in adipose tissue was approximately 4%, as detected by flow cytometry (Fig. 4c). This apparent difference must be interpreted with caution for several reasons. Firstly, in order to obtain a sufficient number of ECs from adipose tissue for FACS analysis, we pooled subcutaneous and epididymal fat pads, which might have a lower venular density. Secondly, in order to obtain sufficient resolution of vascular structures of interest within whole-mount preparations, we acquired 3D scans from carefully selected regions that contained, within the same field of view, all microvascular structures of interest, i.e., arterioles, capillary network, and at least one complete venular tree. This limits the choice of suitable areas and may have introduced a selection bias that could result in over-representation of venules. Finally, it is also possible that ECs in different microvascular segments may differ in size. For example, if the average luminal surface area covered per V-EC is greater than that covered by capillary ECs, the venular surface area would be fractionally larger than the percentage of V-ECs. Our current technology does not allow us to accurately resolve individual ECs in intact tissues, so we cannot address this issue with certainty.

Interestingly, a topological analysis of the microvascular anatomy in murine MLNs [37] estimated that HEVs composed approximately 12% of the overall vascular length, which is in good agreement with our flow cytometry data showing 14% DARC⁺ V-ECs among BECs in MLN.

Consistent with reports of DARC expression in human skin [38], murine DARC was detectable by FACS on subsets of LECs in LNs, adipose tissue, and skin (Additional file 9: Figure S8); however, DARC expression on LECs was too low to be detectable in these tissues by confocal microscopy (Figs. 1 and 2).

DARC as a pan-venular marker

Next, we asked whether DARC expression correlates with other known venule-associated markers. For example, P- and E-selectin are constitutively expressed (although challenging to detect by immunostaining) in resting skin venules [30, 39, 40]. Indeed, when mRNA levels of Darc, Selp, and Sele were measured by qPCR in sorted dermal DARC⁺ V-ECs, DARC^– NV-ECs, and LECs, all three mRNA species were detected exclusively in V-ECs (Fig. 5a).

Similarly, HEVs (unlike other microvessels) in PLN express carbohydrate sulfotransferase 4 (Chst4) [41] and fucosyltransferase VII (Fut7) [42], enzymes involved in the biosynthesis of 6-sulfo-SLeX on core 1 O-glycans that comprise the PNAd marker [40, 43]. PNAd is also synthesized by most HEVs in MLN, which additionally expresses MAdCAM-1 [44, 45]. Accordingly, Darc mRNA expression in PLN and MLN was restricted to V-ECs and correlated closely with Chst4 and Fut7 in both tissues and with Madcam1 in MLNs (Fig. 5b, c). These observations were confirmed at the protein level by FACS analysis of PLN and MLN BECs. Of note, while almost all PNAd⁺ and MAdCAM-1⁺ BECs expressed DARC, up to approximately 15% of DARC⁺ ECs were PNAd^– and MAdCAM-1^– (Fig. 5d, e). These ECs likely derived from medullary segments of the venular tree, which are lined by flat ECs and do not express vascular addressins [46].

Since PNAd, MAdCAM-1, and the endothelial selectins are not only restricted to V-ECs, but are also subject to a strong tissue bias at steady-state [47], we also stained BECs for ICAM-1, which is constitutively expressed on ECs throughout the body [48]. Indeed, BECs expressed ICAM-1 in both skin and LNs, whereas PNAd and MAdCAM-1 were absent on dermal BECs [49] (Fig. 5d–f). Interestingly, although ICAM-1 expression was more pronounced and uniform in venules than on other ECs, it was also expressed, albeit at a lower level, on most DARC^– NV-ECs. Thus, when compared to other known venular surface moieties, DARC stands out as a truly pan-venular marker that shows exquisite specificity for V-ECs and is readily detectable by IHS and flow cytometry in every murine tissue analyzed.

Detection of venular DARC expression by IVM

Having determined that our anti-DARC MAb recognizes surface-expressed DARC in frozen sections and whole mount preparations of murine tissues ex vivo and on intact transfectants and isolated ECs in vitro, we asked whether the MAb could pinpoint microvascular DARC expression in vivo. We transplanted irradiated C57BL/6 mice with DARC^–/– BM (Fig. 6a); at 8 weeks after BM transplantation, this resulted in absent DARC expression on RBCs. Thus, upon intra-arterial injection of Alexa Fluor-488 labeled anti-DARC, the MAb remained free to bind endothelial DARC without being captured by DARC⁺ RBCs.

Luminal DARC expression was readily detectable by epifluorescence IVM [5] in HEVs and large collecting venules in subiliac LNs (Fig. 6b). Epifluorescence IVM of intact skull BM [31] identified prominent DARC expression in intermediate venules, collecting venules, post-sinusoidal venules, and sinusoids (Fig. 6c). Segmental microvascular immunoreactivity, particularly at inter-cellular junctions was also readily detected by IVM in intact ear skin and in surgically exposed cremaster muscle (Fig. 6d–f). By contrast, as expected, none of these IVM models showed detectable DARC expression in arterioles or capillaries.

DARC⁺ microvessels uniquely support physiologic leukocyte adhesion

To independently verify that the DARC⁺ microvessels detected by IVM were indeed venules, we made use of the fact that leukocytes undergo constitutive selectin-mediated rolling interactions in venules, but not in capillaries or arterioles in murine ear skin [30]. Thus, we injected fluorescent anti-DARC MAb into mice and then injected fluorescently tagged leukocytes into the carotid artery to quantify leukocyte rolling fractions in dermal DARC⁺ and DARC^– microvessels by IVM [30]. All DARC⁺ microvessels supported robust rolling interactions, whereas rolling was absent in all DARC^– microvessels, indicating that DARC delineates specifically pro-adhesive venules, but not non-adhesive non-venular microvessels (Fig.  6 g, Additional file 10: movie 1).