VCAM-1 expression in adult hematopoietic and nonhematopoietic cells is controlled by tissue-inductive signals and reflects their developmental origin

The vascular cell adhesion molecule 1 (VCAM-1), an immunoglobulin (Ig)–like transmembrane adhesion molecule, is highly conserved in evolution and participates in a variety of cellular functions in health and disease. Human VCAM-1 has 2 isoforms, the predominant 7 Ig-domain isoform and a minor, alternatively spliced isoform with 6 Ig domains, whereas in mice the second isoform consists of the first 3 domains attached to the cell membrane through a glycosylphosphatidylinositol (GPI) anchor.^1,2  VCAM-1 is minimally expressed on most resting vascular endothelial cells and is inducible in many tissue vascular beds following injury or stress. Because of this activation, VCAM-1 has been implicated in the pathophysiology of certain autoimmune diseases, atherosclerosis, and allograft rejection.^3-8  VCAM-1 is constitutively expressed in bone marrow (BM) stromal/endothelial cells and certain classes of hematopoietic cells (B cells, follicular dendritic cells, and macrophages^9-15 ). Its major ligand is the integrin very late antigen 4 (VLA-4; α4β1) with binding sites located in the first and fourth Ig domains, whereas other ligands bind with less affinity and include α4β7, α9β1, and α_Dβ2.^16-18  Deletion of the VCAM-1 gene produced a phenotype similar to deletion of the α4 gene, with embryonic lethality due to cardiac and placental developmental abnormalities, although few VCAM-1^–/– mice survived.^19,20  Several investigators then adopted approaches to overcome the embryonic lethality observed in VCAM-1–null mice. Cybulsky et al generated mice with a homozygous deletion of domain 4 (VCAM-1^D4D),²¹  resulting in expression of the VCAM-1 isoform containing one binding site. Embryonic survival was significantly reduced and was strain dependent, but surviving mice showed a reduction of VCAM-1 protein to less than 10% of wild type, allowing informative studies in the adult animals.^22-24  In other studies,²⁵  Tie2-cre transgenic mice were bred with floxed VCAM-1 mice to allow VCAM-1 deletion in hematopoietic cells and endothelium through cre-loxP–mediated recombination. These viable mice exhibited mild leukocytosis with elevated immature B cells in the peripheral blood (PB), with reduced immature B cells in the BM, consistent with the role of VCAM-1 in the retention of immature lymphocytes in the marrow. In a similar approach, mice carrying the Mx-cre transgene bred with mice homozygous for the floxed VCAM-1 gene, subsequently deleted neonatally by interferon-mediated cre recombination, led to a reduction of immature B cells in the BM, and an increase of these cells in the circulation.²⁶  Additional in vitro studies have shown that the VLA-4/VCAM-1 interaction is involved in the adhesion of human B cells to follicular dendritic cells of germinal centers in vitro and a link has been reported between VCAM-1 function and the rescue of germinal center B cells or thymocytes from apoptosis.^13,14,27  VCAM-1 may also function in maturation and costimulation of T cells.^28-30  In contrast to results from chimeric α4^–/– mice showing postnatal dependency of B and T cells on α4 integrins,³¹  B and T lymphocytes can develop in the absence of VCAM-1.^25,26  Later studies reported that neonatally induced inactivation of VCAM-1 impaired responses to Toxoplasma encephalitis.³² 

Herein, we explored the pattern of VCAM-1 expression among hematopoietic and nonhematopoietic cells from BM and from selective tissues, before and after the use of mobilizing agents. Several novel insights about VCAM-1 regulation in these cell populations are revealed by studying both control and VCAM-1–ablated animals and these are described in the present work.

VCAM-1 mice used in our experiments were created through interbreeding of VCAM-1 floxed (VCAM-1^f/f) mice with Tie2cre transgenic mice.²⁵  When VCAM-1^f/f males are bred with Tie2cre⁺ females, all progeny, regardless of the presence of the cre transgene, have a constitutional deletion of 1 VCAM-1 allele (VCAM-1^Δ). This scheme seems to afford a more complete deletion of VCAM-1 in hematopoietic tissues of Tie2cre⁺VCAM-1^f/Δ versus Tie2cre⁺VCAM-1^f/f animals following subsequent breeding.²⁵  Consequently, mice used for our experiments had the Tie2cre⁺VCAM-1^f/Δ genotype and controls were of cre^–VCAM-1^f/f genotype (Figure 1). Observations in these mice were carried out from a few weeks of age up to over a year. Mice bred and maintained under specific pathogen-free conditions in University of Washington facilities were provided with irradiated food and autoclaved water ad libitum.

Granulocyte–colony-stimulating factor (G-CSF; Neupogen Filgrastim; Amgen, Thousand Oaks, CA) treatments were given via an osmotic pump (Alzet Model 2001; Durect, Cupertino, CA) for 5 days at 100μg/kg per day (3 experiments) or by intraperitoneal injection twice daily at 200 μg/kg for 4 days (1 experiment). All mice were tested on the fifth day. Flt-3 ligand alone (FL; 50 ng/kg per day; kind gift from S. Lyman, Immunex, Seattle, WA) or the combination of G-CSF (100 μg/kg per day) and FL was administered with osmotic pump for 7 days.

Anti–VCAM-1 (MK/2) antibodies, unconjugated for immunohistochemistry (IHC) and phycoerythrin (PE)–conjugated for FACS, were from Southern Biotech (Birmingham, AL). Fluorochrome-conjugated antibodies to B220, CD3, TER119, Mac-1, Gr-1, and CD117 (c-kit), and fluorescein isothiocyanate (FITC)–conjugated secondary antibodies and streptavidin allophycocyanin (APC) were from BD-Pharmingen (San Diego, CA). Antibodies used for endothelial cells were biotinylated CD105 (endoglin), MECA32 (BD-Pharmingen), biotinylated isolectin B4 (Vector Labs, Burlingame, CA), labeled acetylated low-density-lipoprotein (diI-Ac-LDL; Biomedical Technologies, Stoughton, MA), rabbit anti–human von Willebrand factor (VWF; Abcam, Cambridge, MA), and PE-conjugated anti–mouse CD202 (Tie2; eBioscience, San Diego, CA). Anti–rabbit Alexa 594, streptavidin 488 (Invitrogen/Molecular Probes, Carlsbad, CA), and streptavidin PE (Biomeda, Foster City, CA) were used as secondary reagents.

Antibody-stained cells were analyzed on a FACSCalibur (BD Immunocytometry Systems, San Jose, CA) using the CELLQuest program.

Tissues for immunohistochemistry were fixed in 4% paraformaldehyde, sucrose treated, embedded in TissueTek (Sakura Finetechnical, Tokyo, Japan), and frozen in acetone/dry ice. Femurs/tibiae were similarly fixed, decalcified in 10% EDTA (ethylenediaminetetraacetic acid; 7 to 10 days at 4°C), sucrose treated, then embedded and frozen. Air-dried frozen sections were washed, blocked for endogenous peroxidase activity and nonspecific staining, stained with appropriately diluted antibodies, and processed following the manufacturer's instructions (ABComplex/HRP; Dako, Carpinteria, CA).

Culture–colony-forming unit (CFU-C) assays were performed using a methylcellulose mixture (Methocult GF; Stem Cell Technologies, Vancouver, BC, Canada), as described previously.³³ 

Endothelial cells from bone were derived by treating cleaned and flushed femurs with 0.125% Trypsin/0.2% EDTA for 15 minutes at 37°C; cells were discarded, bones minced, then incubated with 0.1% collagenase (1-3 hours at 37°C). After washing and removing bone fragments, cells were plated in collagen I–coated wells (BD BioCoat; BD Biosciences, Bedford, MA) using endothelial-cell medium as previously described,³⁴  adding 250 ng/mL endothelial mitogen (Biomedical Technologies), 10 ng/mL recombinant murine vascular endothelial cell growth factor (rm-VEGF) and 7.5 ng/mL basic fibroblast growth factor (both from Peprotech, Rocky Hill, NJ). Nonadherent cells were removed after 24 hours and fresh medium was added. Adherent cells were dissociated with either 1:1 cell dissociation solution/phosphate-buffered saline (PBS) (Sigma Aldrich, St Louis, MO), or 1:10 TrypLE Express/PBS (Invitrogen). Bone fibroblasts were derived by plating bone fragments or cells from above in Dulbecco modified Eagle medium (DMEM) with 20% fetal bovine serum (FBS), cultured for 4 to 7 weeks, and dissociated with TrypLE Select. Endothelial cells from heart or lung were isolated by mincing, then washed extensively to remove blood, collagenase treated and plated in 2% gelatin-coated wells/dishes with endothelial cell medium. Endothelial cells or fibroblasts were used for evaluation by fluorescence-activated cell sorting (FACS) or for DNA and RNA studies.

Femurs from control and VCAM-1–ablated mice were flushed, and recovered cells were washed, incubated for 2 hours in Iscoves modified Dulbecco medium (IMDM) plus 10% FBS to remove adherent cells, counted and plated in IMDM with 10% FBS, 100 ng/mL murine stem-cell factor (Peprotech), 10 μL/mL of murine interleukin 3 (IL-3; kind gift from K. Kaushansky, San Diego, CA), 100 ng/mL G-CSF, 50 ng/mL granulocyte macrophage–colony-stimulating factor (GM-CSF; Peprotech), and 3 U/mL erythropoietin (Epoetin; Amgen, Thousand Oaks, CA).

Mononuclear cells from blood of mobilized control or ablated animals were isolated by density gradient separation (Histopaque 1083; Sigma Chemical), washed, and cultured in the same media as BM cells. Spleens from control and VCAM-1–ablated mice were teased, triturated, washed, and plated in 10% FBS, 100 ng/mL murine IL-2, 50 ng/mL murine IL-7 (both from Peprotech), and 10^–4 M 2-mercaptoethanol (Sigma Chemical), or in the same cytokine cocktail as the PB cells. Samples were collected for analysis at day 0 and after 3 to 7 days in culture.

Genomic DNA and total RNA were isolated from mouse tails or cell cultures using DNA or RNA isolation kits (Gentra Systems, Minneapolis, MN) according to the manufacturer's instructions. Isolated RNA was reverse-transcribed using oligo-dT (Invitrogen) or random nonamer (Sigma Chemical) primer and SuperScript II enzyme (Invitrogen). The following primers were used in polymerase chain reactions (PCRs): for genomic DNA to distinguish between wild type, floxed, and Δ alleles of the VCAM-1 gene (5′-ATCGATCCCTGGATATGTCG-3′,5′-AGGCTATCAGTGAGACCCACA-3′,5′-TGGGCTGTCTATCTGGGTTC-3′); mVCAM-1 7 Ig domain-specific primers (sense, 5′-AAGTCTGTGGATGGCTCGT-3′; antisense, 5′-AGATGCGCAGTAGAGTGCAA-3′); VCAM-1/GPI-linked–specific primers (sense, 5′-GATGGGAGAATGAAGTCTCAGATTA; antisense, 5′-TTCAGCTTCCTCACAGAGCA); Tie2-specific primers (sense, 5′-TGTCAATCAGGCCTGGAAATAC-3′; antisense, 5′-GAGGAGGGAGAATGTCACTAAGC-3′)²¹  or (sense, 5′-CCTTCCTACCTGCTA-3′; antisense 5′-CCACTACACCTTTCTTTACA-3′)³⁵ ;discoidindomainreceptor2(DDR2)–specific primers (sense, 5′-CTGTCGGATGAGCAGGTTAT-3′; antisense, 5′-CTCGGCTCCTTGCTGAAGAA-3′)³⁶ ; and murine hypoxanthine phosphoribosyltransferase (HPRT)–specific primers (sense 5′-GCTGGTGAAAAGGACCTCT-3′; antisense 5′-CACAGGACTAGAACACCTGC-3′)²²  or murine β-actin–specific primers (sense, 5′-GCTGTATTCCCCTCCATCGTG-3′, antisense, 5′-CACGGTTGGCCTTAGGGTTCAG-3′).

Levels of stromal cell–derived factor 1 (SDF-1) and soluble VCAM-1 (sVCAM-1) were assayed in the cytokine analysis core facility of Fred Hutchinson Cancer Research Center, as previously described.³⁷  SDF-1 and sVCAM-1 kits were purchased from R&D Systems (Minneapolis, MN).

Bone marrow cells, permeabilized for 15 minutes in PBS with 10 mM EDTA and 0.1% Triton X-100, were mixed with propidium iodide at 2 μg/mL and analyzed by FACS using CELLQuest and ModFit LT (Verity, Topsham, ME) software.

Tests of significance were performed using the Student t test (Microsoft Excel 2000 software; Microsoft, Redmond, WA).

VCAM-1 expression in hematopoietic cells was assessed by flow cytometry using well-characterized antibodies. In control mice of young or old age (α4^f/f/cre^–; Figure 1), nearly half of all BM cells were VCAM-1⁺ (Figure 2A). The proportion of VCAM-1⁺ cells was lower in the spleen. Cells from all hematopoietic lineages were represented in the BM VCAM-1⁺ population of VCAM-1^f/f mice (Gr-1⁺, Mac-1⁺, CD3⁺, B220⁺, or TER119⁺ cells) and 59.6% of kit⁺ cells were VCAM-1⁺ (Figure 2A). We also examined the expression of VCAM-1 at the RNA level in BM sorted cells (Gr1⁺/VCAM-1⁺ and Gr1⁺/VCAM-1^–) and by immunohistochemistry. In Figure 3 (top 2 panels), a broad VCAM-1 positivity in BM, consistent with scattered positive cells, is seen. In the spleen there is positivity throughout the red pulp, reflecting the presence of VCAM-1⁺ myeloid cells. Within the largely negative germinal centers, small groups of cells highly positive for VCAM-1 are seen, likely representing follicular dendritic cells.¹⁴  Importantly, BM and spleen from VCAM-1–deficient animals showed no staining with VCAM-1 (Figure 3, right panels) and proportions of VCAM-1⁺ cells by FACS were also very low (Figure 2).

In contrast to high VCAM-1 positivity in the hematopoietic tissues of normal mice, circulating cells were virtually devoid of VCAM-1 expression and as such, they were no different from VCAM-1–ablated mice (Figure 2B).

VCAM-1 expression in nonhematopoietic cells was initially assayed by immunohistochemical analyses of tissues. In control mice, endothelial cells in the BM, spleen, and lymphoid tissues (lymphatic and blood vessels) were clearly positive for VCAM-1, as were some stromal cells of the thymus (Figure 3). There were no VCAM-1⁺ endothelial cells in tissues from ablated mice, although rare, scattered positive cells could be seen. We next initiated endothelial cell cultures from bone, lung, or heart, and tested them after in vitro expansion. Expression of endothelial cell markers (Tie2, VWF, diI-Ac-LDL, and isolectin B4 uptake³⁸ ) was seen in all cultures by immunofluorescence labeling, and the cells formed tubular networks in Matrigel (Becton Dickinson, Bedford, MA; see also Supplemental Figure S1B, available at the Blood website [see the Supplemental Materials link at the top of the online article]). VCAM-1 mRNA was detected in endothelial cell cultures from bone, lung, and heart (Figure 4 and data not shown) from control animals, and at much lower levels in endothelial cell cultures from ablated animals, in contrast to the virtual absence of VCAM-1 mRNA from hematopoietic cells residing in bone marrow from the same animals (Figure 4).

In addition to endothelial cell cultures, the level of VCAM-1 ablation was tested in stromal cell cultures (positive for DDR-2 and procollagen 1α1; Figure 4 and data not shown) from crushed bones. Unexpectedly, we found a significant level of VCAM-1 gene excision in fibroblast cultures from ablated animals (Figure 4), similar to that seen in some endothelial cell cultures. Parallel cultures of early passage fibroblasts from control animals showed VCAM-1 expression by virtually all CD45^– cells (Supplemental Figure S2B). Also, Tie2 mRNA was detected in cultured fibroblasts (Supplemental Figure S2A), but at lower levels than in cultured endothelial cells.

Bone marrow cellularity and progenitor content were tested in VCAM-1–ablated young animals (9-14 weeks old) as well as in older animals (> 40 weeks old) to uncover age-related changes. As seen in Table 1, BM cellularity of young adult VCAM-1–ablated animals was lower than controls. This difference, likely attributable partially to their smaller (about 10%) weight, was maintained when older animals from both groups were compared. There was a moderate decrease in BM progenitor content in young ablated animals, but in older animals of either genotype, no significant differences were seen, although progenitors were significantly increased in both, consistent with previously published data.³⁹  The proportion of lineage-affiliated cells, Gr-1⁺, Mac1⁺, and TER119⁺, in BM was not significantly different between the 2 groups (Supplemental Table S1). However, there were lower levels of T cells (P < .05) and B220⁺ cells.

A modest but significant increase was seen in spleen cellularity of ablated animals, which persisted with age, when the total cellularity increased in both groups (Table 1). There were also more splenic progenitors in ablated animals, both young and old, but no significant differences in the proportion of B cells, T cells, myeloid, or TER119⁺ erythroid cells in the spleen.

Significant and consistent differences in circulating cells were found between ablated and control groups. Circulating white blood cells (WBCs) were significantly higher in the ablated group (Table 1), with a higher proportion of B220⁺ cells and a relative decrease in Mac-1⁺ cells (Supplemental Table S1). More pronounced and sustained differences were seen in the number of circulating hematopoietic progenitors (CFU-Cs). It is notable that the progenitor differences in peripheral blood between VCAM-1–ablated animals and controls were maintained with age (Table 1).

Since cellularity and progenitor content in the spleen were higher in VCAM-1–ablated mice, we suspected that the spleen continuously sequesters progenitor cells from circulation, so we studied splenectomized animals. Both WBCs and circulating progenitors increased (Figure 5) in splenectomized mice (5.3-fold for CFU-GMs, 22.4-fold for BFU-Es, and 6.2-fold for CFU-GEMMs). In the BM of splenectomized animals, there was an increased CFU-C content per femur (149 376 ± 15 131/femur in controls compared with 112 065 ± 17 447 in the ablated group), but no significant differences in the proportion of different lineages within this population. It is speculated that the increase in progenitor content in the marrow after splenectomy is attributable at least in part to some rehoming of circulating progenitor cells.⁴⁰ 

To test the ability of VCAM-1^Δ/Δ mice to mobilize hematopoietic progenitors beyond their increased levels at baseline, we treated them with several mobilizing agents. First, they were treated with anti–VLA-4 antibody. In contrast to the expected and observed responses in control animals, both in the WBCs (10 761 ± 1230 vs 5040 ± 480/μL at baseline) and in the CFU-Cs (736 ± 172/mL vs 47 ± 6.3 at baseline; n = 5; P < .05), there was no significant response from VCAM-1–ablated mice (WBCs, 9965 ± 1148/μL vs 9970 ± 1120; CFU-Cs, 972 ± 181 vs 683.6 ± 165/mL at baseline; n = 5; P = .28). These data provided assurance that the previously described anti–VLA-4–induced mobilization⁴¹  was due to the disruption of VLA-4 interaction with VCAM-1 rather than other α4 integrin ligands, and unlikely to be the result of extraneous antibody effects.

We then examined the response to G-CSF (G-CSF-5d; Figure 6). WBCs increased similarly in VCAM-1^f/f and VCAM-1^Δ/Δ mice (P > .5), but mobilized CFU-Cs in blood were significantly higher in VCAM-1–deficient animals (P < .05). At peak mobilization, BM cellularity was reduced both in VCAM-1^f/f mice and VCAM-^+Δ/Δ animals (P < .05), but progenitor content in BM was not significantly different between the two groups. Spleen cellularity increased in both groups, but was higher in controls than in VCAM-1–deficient spleens (P<.01). Spleen progenitor content was also significantly higher in controls (Figure 6) compared with ablated animals (P < .05). Consistent with the increased proliferative activity in spleen, the proportion of kit⁺ cells was higher in control spleens compared with VCAM-1–deficient spleens (29.3% ± 1.0% vs 16.76% ± 0.5%, respectively; P < .001). Furthermore, following G-CSF treatment, VCAM-1⁺ cells in the spleens of VCAM-1^f/f animals were significantly increased compared with untreated animals (38.85% vs 22.26%). This increased expression was confirmed by immunohistochemistry after mobilization (data not shown). In contrast to spleen, control mice had a decrease in VCAM-1 protein expression in BM at the end of cytokine treatment (Figure 2B).

Since there was considerably less CFU-C accumulation in the spleens of VCAM-1^Δ/Δ mice after G-CSF mobilization (Figure 6), the total progenitor level from all hematopoietic organs (total BM, total blood volume, and spleen) was significantly reduced between ablated animals and controls (1.82 ± 0.2 × 10⁶ CFU-Cs in f/f vs 1.28 ± 0.1 × 10⁶ in Δ/Δ; P = .04). To test whether the increased numbers in blood were simply due to decreased progenitor return to tissues, or decreased adhering to tissue vascular beds, rather than to increased progenitor release from the BM, we followed the kinetics of progenitor clearance from the blood after stopping G-CSF administration. As seen in Figure 6B, the rate of reduction in circulating progenitor cells after cessation of G-CSF treatment was similar in control and VCAM-1–ablated animals, suggesting that the higher CFU-C levels in blood are not the result of decreased migration to tissues and/or increased survival in circulation.

To further pursue the issue of decreased splenic response in ablated mice, we used more intense mobilization with combinations of G-CSF and FL.^42-44  Following a 7-day mobilization with G-CSF plus FL, a strong response was elicited by both control and VCAM-1–ablated animals well above levels achieved with either cytokine alone (Figure 6). However, the total progenitor content in BM, spleen, and blood was again higher in control mice (7.64 ± 2.1 × 10⁶ CFU-Cs in f/f vs 4.43 ± 1.35 × 10⁶ in Δ/Δ; P = .21), largely as a result of differences in splenic progenitor content, and consistent with data after only G-CSF mobilization (Figure 6). Cell-cycle analyses in BM and spleen also suggested decreased cycling in spleens of ablated mice after mobilization (Supplemental Table S2). Moreover, in G-CSF plus FL–mobilized normal animals, there was a significant decrease in VCAM-1 protein (from 49.63% ± 2.87% to 14.8% ± 0.4% VCAM-1⁺ cells) and mRNA expression (Figure 2C) in BM compared with baseline levels.

There was very little VCAM-1 expression in circulating cells after mobilization in both control and deficient animals (Figure 2B). To explore whether VCAM-1 expression could be induced in mobilized cells, the latter were cultured in a cytokine cocktail. After 3 days in culture and at a time when there was no appreciable increase in cell number, there were significant increases in VCAM-1 mRNA (Figure 7) and protein expression (from < 1% to 10.1%). There was an additional increase (to 28.6%) in VCAM-1⁺ cells at day 6 following cell proliferation. To test whether all proliferating cells express or activate VCAM-1 expression, we also cultured splenocytes under IL-7 and IL-2 stimulation. Despite significant proliferation, the latter failed to show VCAM-1 expression (Figure 7). Further, since VCAM-1 can be up-regulated in nonproliferating endothelial cells by inflammatory stimuli, we also tested induction of VCAM-1 expression in mature hematopoietic cells from control mice following an inflammatory stimulus (thioglycolate-induced peritonitis). Cells accumulated in high numbers in the inflamed peritoneum, but no expression of VCAM-1 was seen at either 4 or 16 hours after thioglycolate injection (data not shown).

Because of the previously emphasized role of SDF-1 in mobilization, we explored changes in SDF-1 levels in BM and PB after mobilization. Significant decreases in BM SDF-1 levels at the peak of mobilization were seen in wild-type, VCAM-1^f/f, and VCAM-1^Δ/Δ mice (Table 2). This is consistent with several previous studies⁴⁵  and references therein. Moreover, sVCAM-1 was also decreased in BM of all G-CSF–treated animals (Table 2). Variable levels of SDF-1 were present in the PB, but increases in sVCAM-1 were seen in control animals (P < .001), as previously reported.³⁶  VCAM-1–ablated animals showed lower levels of sVCAM-1 in BM and circulation, consistent with incomplete ablation of VCAM-1 in certain populations of VCAM-1–expressing cells.

Although VCAM-1⁺ cells of hematopoietic origin have been previously noted, no systematic studies of VCAM-1 expression patterns among hematopoietic cells at baseline and following stress have been reported.

VCAM-1 expression in normal BM is present among cells of all lineages, both immature and mature. The overall expression of VCAM-1 in almost half of BM cells is consistent with VCAM-1 expression in a subset of progenitor cells or kit⁺ cells, and likely a subset of primitive stem cells, although further studies are needed to document this notion. Thus, VCAM-1 expression in at least a common myeloid progenitor (CMP), may question a prior conclusion²⁶  about the expression and functional influence of VCAM-1 only in late cells, beyond the stage of common progenitors.

In contrast to VCAM-1 positivity of myeloid cells in BM and spleen, it is of interest that PB cells do not display VCAM-1 (Figure 2B). This is true not only for cells during steady-state hematopoiesis, but also after cytokine-induced mobilization, when significant increases in the numbers of progenitor and mature cells are seen in periphery. To explore the reasons for the absence of VCAM-1 expression by PB cells, we considered the following possibilities: whether VCAM-1^– cells were preferentially mobilized, both at steady state and after G-CSF; whether VCAM-1 positivity was masked or VCAM-1 expression was silenced once the cells were in the PB environment; or whether another common characteristic of cells in the PB was responsible for the VCAM-1^– phenotype. If preferential mobilization of VCAM-1^– cells occurred, one would expect enrichment for VCAM-1⁺ cells in the BM of controls at the peak of mobilization. Instead, a decrease in VCAM-1 expression was found in BM at the peak of mobilization (Figure 2) in G-CSF and a further decrease in G-CSF plus FL (from 49.6% ± 2.8% to 14.8% ± 0.4%). The possibility of a masking epitope was a viable one; therefore, we examined transcription of VCAM-1 in PB cells. We found that PB cells, from animals both mobilized and at steady state, were not only negative for VCAM-1 protein, but had no detectable levels of VCAM-1 mRNA (Figure 7). Next, we explored whether the known differences in cell cycling between PB and BM^46,47  might be responsible for differences in VCAM-1 expression. Indeed, after cytokine-induced proliferation in vitro, mobilized cells displayed a significant expression of VCAM-1 mRNA and protein. However, in vitro culture could have generated new progeny with a VCAM-1–expressing phenotype from VCAM-1^– cells, but this was unlikely since VCAM-1 levels were increased prior to increases in cell numbers, suggesting activation of VCAM-1 before cell division. Although these data cannot entirely dissociate activation alone versus cells in cycle under a strong cytokine influence, the fact that proliferating lymphoid cells in vitro do not express VCAM-1 suggests some specificity for myeloid cells, or the activation of molecular pathways leading to VCAM-1 expression only in myeloid cells. Whatever the mechanism, differences in VCAM-1 expression between PB and BM or spleen suggest that VCAM-1 expression is actively maintained by tissue-derived signals, but is down-regulated or silenced in the periphery. This view is consistent with absence of VCAM-1 expression in free cells (macrophages) in peritoneal fluid in thioglycolate-induced aseptic peritonitis.

Studies of basal hematopoiesis in VCAM-1–ablated mice and their age-matched controls reveal certain notable differences. Among them, the most conspicuous is a modest increase in circulating WBCs, but a significant and sustained increase in circulating hematopoietic progenitors. Although an increased progenitor content is seen in the spleen, its enhancement following splenectomy suggests release from the BM with subsequent sequestration in the spleen. Given the extravascular location of hematopoietic cells in BM, how is the enhanced release of progenitor cells explained? Is it secondary to VCAM-1 absence from hematopoietic cells themselves or to VCAM-1 ablation from endothelial cells? We have already argued that a preferential release of VCAM-1^– hematopoietic cells present in BM is an unlikely explanation for this phenomenon. Are VCAM-1^– endothelial cells responsible? We believe that absence of VCAM-1 not only from endothelial cells, but especially from BM stromal cells, tested for the first time in the present studies, is likely responsible for this process, suggesting a key role for these cells in the retention of (α4 positive) hematopoietic cells within the BM extravascular space. This ongoing, active retention is disrupted in the VCAM-1^Δ/Δ mice and a further disruption of this process can be effected through a proteolytic mechanism in G-CSF–induced mobilization. Indeed, histochemical evaluation of VCAM-1 in BM in prior studies^48,49  showed marked reduction in VCAM-1 following G-CSF, and this was attributed to a decrease in nonhematopoietic cells.^10,11  However, this decrease, in light of our new data showing lower proportions of VCAM-1⁺ hematopoietic cells in BM (Figure 2) following G-CSF, needs to be reinterpreted, including both hematopoietic cells, and stroma.^48,49 

A corollary to the decrease of VCAM-1 in normal BM after G-CSF mobilization is the increase in sVCAM-1 in circulation.⁴⁸  If VCAM-1 cleavage was critical for G-CSF mobilization, one would have expected that VCAM-1–ablated mice would respond less to such a treatment. Contrary to expectations, our mice responded as well as, or, at suboptimal stimulation, better than controls (Figure 6) by displaying enhanced levels of circulating progenitors. Thus, the proteolytic modulation of VCAM-1 in the bone marrow is not the driving force in G-CSF mobilization, a view consistent with results obtained in protease-deficient mice.⁴⁹  Furthermore, our data suggest that absence of VCAM-1 may facilitate exit from the bone marrow in response to G-CSF–induced signals. This conclusion is consistent with a sustained hematopoietic progenitor egress from the bone marrow in these mice at baseline, similar to one seen in mice with conditional deletion of α4 integrin,³³  the principal ligand of VCAM-1, thereby emphasizing the role of the VLA-4/VCAM-1 pathway in the active retention of progenitor cells and their egress following disruption of this pathway.

In contrast to BM showing decreased VCAM-1 levels, there was an increase in VCAM-1 expression by myeloid cells residing in the spleens of normal animals following G-CSF (Supplemental Figure S3). Therefore, proteolytic conditions for VCAM-1 cleavage were only present in BM, in agreement with prior data.^48-50  However, since progenitor cells were also sequestered in the spleens of VCAM-1–ablated mice either before or after G-CSF, VCAM-1 in the spleen may not be responsible for their sequestration or entry into the spleen, in contrast to a previous speculation.⁴⁸  Nevertheless, since mobilized normal mice accumulated more progenitor cells in their spleens compared with mobilized VCAM-1–ablated mice, one has to postulate that in the absence of VCAM-1, either there is a ceiling in splenic sequestration, or that the proliferative expansion of progenitors within the spleen is limited in the ablated animal. Consistent with this notion is the increased proportion of kit⁺ cells and increased cycling in the spleens of control animals.

In mobilized mice, levels of soluble VCAM-1 in BM were reduced after mobilization, in contrast to PB. Presence of measurable levels of soluble VCAM-1 in VCAM-1–deficient animals (Table 2) provided confirmatory evidence that VCAM-1 is not completely ablated in the vascular beds of all tissues in the deficient mice, as was independently concluded by our RNA studies (Figure 7). Furthermore, SDF-1 levels in BM were reduced in all mice at the time of peak mobilization (Table 2). Whether these observed SDF-1 changes represent the main cause or just another facilitating pathway in mobilization is difficult to state for our or other murine models.

In summary, we have uncovered new information about the biology of VCAM-1 expression in hematopoietic cells by demonstrating that it is constitutively expressed in myeloid cells confined to tissues. Although an association with their proliferative status was seen in vitro and in vivo, the functional significance of VCAM-1 expressed in hematopoietic cells has yet to be explored. Ablation of VCAM-1, in at least a subset of BM fibroblasts, suggests their developmental origin from a Tie2-expressing (mesenchymal?) progenitor, and further studies to pursue this issue are in progress. Increased egress of BM hematopoietic progenitors in the VCAM-1–ablated mouse validates the role of the VCAM-1/VLA-4 pathway in intramarrow retention of hematopoietic progenitors. Further, our postmobilization data provide further clarification about the role of VCAM-1 in G-CSF–induced mobilization.

We are grateful to Devra Batdorf for her assistance with mouse care, husbandry, and mouse splenectomy.