Galactosyl carbohydrate residues on hematopoietic stem/progenitor cells are essential for homing and engraftment to the bone marrow

The role of carbohydrate chains in leukocyte migration to inflamed sites during inflammation and trafficking to the lymph nodes under physiological conditions has been extensively characterized. Here, we report that carbohydrate chains also mediate the homing and engraftment of hematopoietic stem/progenitor cells (HSPCs) to the bone marrow (BM). In particular, we found that transplanted BM cells deficient in β-1,4-galactosyltransferase-1 (β4GalT-1) could not support survival in mice exposed to a lethal dose of irradiation. BM cells obtained from mice deficient in β4GalT-1 showed normal colony-forming activity and hematopoietic stem cell numbers. However, colony-forming cells were markedly rare in the BM of recipient mice 24 h after transplantation of β4GalT-1-deficient BM cells, suggesting that β4GalT-1 deficiency severely impairs homing. Similarly, BM cells with a point mutation in the UDP-N-acetylglucosamine 2-epimerase/N-acetylmannosamine kinase gene, encoding a key enzyme in sialic acid biosynthesis, showed mildly impaired homing and engraftment abilities. These results imply that the galactosyl, but not sialyl residues in glycoproteins, are essential for the homing and engraftment of HSPCs to the BM. These findings suggest the possibility of modifying carbohydrate structures on the surface of HSPCs to improve their homing and engraftment to the BM in clinical application.


HSPC population and colony-forming activities of BM cells. Hematopoietic cells, including lym-
phocytes, neutrophils, monocytes, and red blood cells, are present in the peripheral blood of β4GalT-1 −/− mice, although leukocytosis and mild anemia are observed 8 . The T/B cell ratio and CD4/CD8 cell ratio were normal in the spleen and thymus of these mice, respectively (data not shown). To examine the hematopoietic stem cell (HSC) population in BM cells, we performed flow cytometry analysis using various cell surface markers (Fig. 1A). The number of pHSCs 31 defined as lineage (Ter-119, B220, CD3e, CD4, CD8a, Gr-1, and CD11b, IL-7R)-negative, Sca-1 + , c-Kit + , Flk-2 − , CD150 + , CD34 −/low was comparable between β4GalT-1 −/− and β4GalT-1 +/− mice (Fig. 1B). In addition, colony formation by BM cells from β4GalT-1 −/− mice was comparable to that of BM cells from β4GalT-1 +/− mice (Fig. 1C), consistent with our previous reports showing that colony formation in the presence of granulocyte-colony stimulating factor or IL-3 was comparable between these mice 8 . These results imply that the HSC population and colony-forming activity of HSPCs are not affected by β4GalT-1 deficiency.
To further examine HSPC populations in BM cells, as illustrated in Fig. 2A 32 , we performed flow cytometry analysis according to the gating scheme (Fig. S1), as described previously 31 . The numbers of multipotent progenitor (MPP), common myeloid progenitor (CMP), common lymphoid progenitor (CLP), granulocyte-macrophage progenitor (GMP), and megakaryocyte-erythroid progenitor (MEP) are shown in Fig. 2B. The number of MPP increased 3-fold in β4GalT-1 −/− BM cells compared to β4GalT-1 +/− BM cells. Although the numbers of CMP and GMP were equivalent between the genotypes, the numbers of MEP and CLP decreased by approximately half in β4GalT-1 −/− BM cells compared to β4GalT-1 +/− BM cells. These results suggest that HSC differentiation was disturbed to some extent downstream of MPP.

Impaired hematopoietic reconstitution by β4GalT-1 −/− BM cells. We previously reported that
β4GalT-1 is involved in the biosynthesis of selectin ligands, indicated by the impairment of inflammatory responses and wound healing in β4GalT-1 −/− mice 8,9 . As the carbohydrate chains in HSPCs may also regulate homing and engraftment to the BM, we first examined the extent of hematopoietic reconstitution in lethally irradiated mice (Table 1), which generally die within 10 days unless transplanted with functional BM cells. Remarkably, the lethally irradiated mice intravenously transplanted with β4GalT-1 −/− BM cells also died within 10 days, whereas those transplanted with β4GalT-1 +/− cells survived for more than 60 days. Conversely, lethally irradiated β4GalT-1 −/− mice transplanted with BM cells from wild type C57BL/6 mice also survived for more than 60 days, while lethally irradiated β4GalT-1 −/− mice transplanted with β4GalT-1 −/− BM cells died within 10 days. Collectively, these results indicated that β4GalT-1 activity in transplanted BM cells, but not in recipient mice, is critical for hematopoietic reconstitution. It is known that BM cells engraft to the BM more efficiently when they are directly transplanted into the BM cavity using the intra-BM transplantation (IBM-BMT) method 33 . Similar to the intravenous BM transplantation (IV-BMT), lethally irradiated wild type mice transplanted with BM cells from β4GalT-1 −/− mice by IBM-BMT died within 10 days, indicating that β4GalT-1 −/− BM cells could not reconstitute hematopoiesis even when the IBM-BMT method was used.
www.nature.com/scientificreports www.nature.com/scientificreports/ We then transplanted a mixture of β4GalT-1 −/− and β4GalT-1 +/− BM cells to ensure survival (Table 2), and the former cells were labeled with green fluorescent protein (GFP) to enable tracking. Strikingly, GFP-positive cells were rarely detected in the spleen, thymus, and peripheral blood of recipient mice 9 weeks after transplantation even when 90% of the transplanted cells were β4GalT-1 −/− (Table 2, Exp. 2), whereas more than 90% of the  www.nature.com/scientificreports www.nature.com/scientificreports/ cells in these tissues were GFP-positive when only GFP-positive wild type cells were transplanted (Table 2, Exp. 4). These results suggest that β4GalT-1 −/− BM cells are impaired in homing and engraftment after transplantation.
Colony formation by transplanted β4GalT-1 −/− BM cells. To examine the homing ability of the transplanted HSPCs, splenocytes and BM cells were collected from the recipient mice at 3 h and 24 h after IV-BMT, and analyzed using a colony formation assay. Colony-forming transplanted β4GalT-1 −/− cells were approximately 0.5-and 0.3-fold as abundant in recipient splenocytes as in the transplanted β4GalT-1 +/− cells at 3 h and 24 h after IV-BMT, respectively. Similarly, the colony-forming ratio of transplanted β4GalT-1 +/− cells increased from 1.5% to 6% between 3 h and 24 h, but that of transplanted β4GalT-1 −/− cells was less than 0.5% (Fig. 3A,B). These results indicated that β4GalT-1 deficiency severely impaired the homing ability of transplanted HSPCs to the BM.
In these experiments, C57BL/6 mice were used as the recipient, although the donor cells were harvested from β4GalT-1 −/− or β4GalT-1 +/− mice on a mixed 129/Sv and C57BL/6 genetic background 34   www.nature.com/scientificreports www.nature.com/scientificreports/ the transplantation must be immunologically compatible, since both 129/Sv and C57BL/6 mice have an H-2 b haplotype. To completely exclude potential immunological effects, we also used immunodeficient NOD/SCID mice as the recipient for comparison. Lethally irradiated NOD/SCID mice transplanted with β4GalT-1 −/− BM cells also died within 10 days (Table 1), and the number of colony-forming transplanted cells in recipient BM cells was severely reduced compared with that of transplanted β4GalT-1 +/− cells (Fig. 3C). These results indicate that the homing deficiency of β4GalT-1 −/− BM cells was not due to immunological rejection.
Lethally irradiated wild-type recipient mice were transplanted with GFP-labeled BM cells from β4GalT-1 −/− and β4GalT-1 +/− mice and the femur of each recipient mouse was prepared 24 h after BMT. We observed GFP-labeled BM cells in the recipient femur using fluorescence microscopy ( Fig. 4A). Abundant GFP-positive β4GalT-1 +/− cells were observed in the BM of recipient mice, whereas β4GalT-1 −/− cells were rarely observed. Quantitative analysis showed that approximately 0.2-fold β4GalT-1 −/− cells adhered to the BM of recipient mice compared with β4GalT-1 +/− cells (Fig. 4B). These observations also support the notion that the homing ability of β4GalT-1 −/− HSPCs was severely impaired. Impaired hematopoietic reconstitution by β4GalT-1 −/− fetal liver cells. During embryonic development, hematopoiesis mainly proceeds in the fetal liver 35,36 . We also examined homing and engraftment of immature HSPCs in the fetal liver of embryonic stage 14.5 (E14.5). Lethally irradiated wild-type mice transplanted with β4GalT-1 −/− fetal liver cells also died within 10 days, whereas those transplanted with β4GalT-1 +/− cells survived for more than 15 days (Table 3). Furthermore, the colony-forming ratio of transplanted β4GalT-1 −/− fetal liver cells were approximately 0.22-fold as abundant in recipient BM as they were in transplanted β4GalT-1 +/− fetal liver cells (Fig. 5A). On the other hand, direct colony-forming activity of fetal liver cells was slightly (1.2-fold), but significantly higher in β4GalT-1 −/− mice than it was in β4GalT-1 +/− mice (Fig. 5B). As the difference in colony-forming activity of the fetal liver among the genotypes was small, it may not have affected the homing and engraftment activity. Therefore, homing and engraftment of immature HSPCs in the fetal liver of β4GalT-1 −/− mice was also impaired similar to their mature HSPCs in the BM.
Lectin blot of lineage-negative β4GalT-1 −/− BM cells. To examine the galactosyl carbohydrate residues in HSPCs, lineage-negative BM cells were analyzed by a lectin blot using RCA120 and ECA, which specifically recognize Galβ1-4GlcNAc (Fig. 6). Since most proteins are sialylated at the non-reducing carbohydrate terminus, RCA120-and ECA-reactive bands were rarely detected in both β4GalT-1 +/− and β4GalT-1 −/− lineage-negative BM cells. However, digestion of cells with sialidase generated strong and smeary RCA120-and ECA-reactive bands between 100 and 200 kDa in β4GalT-1 +/− lineage-negative cells, but not in β4GalT-1 −/− lineage-negative cells (Fig. 6). This binding specificity was confirmed by the addition of lactose to specifically block RCA120 binding to Galβ1-4GlcNAc. Considering that the galactosyl residues in high-molecular-weight glycoproteins were lost in lineage-negative BM cells from β4GalT-1 −/− mice, these results suggested that they significantly promoted homing and engraftment.  www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ Colony formation by transplanted Gne (V572L) BM cells. Sialic acids are well known to modify non-reducing terminal carbohydrates, including the galactosyl residues in glycoproteins and glycolipids. Notably, the lack of GNE, a key enzyme in sialic acid biosynthesis, is an embryonic lethal mutation 37 . However, mice with a V572L point mutation in GNE survive for several months, but suffer from a nephrotic-like syndrome because of severe hyposialylation of podocyte glycoproteins 38 . Colony formation by BM cells obtained from such mice was comparable with that of wild-type cells, suggesting that BM cell proliferation and differentiation were normal (Fig. 7A). However, colony-forming transplanted Gne (V572L) cells were about 60% as abundant in recipient BM cells as transplanted wild type cells (Fig. 7B).
The survival of lethally irradiated wild type mice clearly depended on the number of Gne (V572L) BM cells transplanted (Fig. 7C). For example, survival was lower in mice transplanted with 1 × 10 6 Gne (V572L) BM cells (60%) than that in mice transplanted with the same number of wild-type cells (90%). The difference in survival was even larger (0% vs 80%) when 5 × 10 5 cells were transplanted. However, all mice died within 10 days www.nature.com/scientificreports www.nature.com/scientificreports/ when 2 × 10 5 cells were transplanted. Collectively, these data indicated that homing was mildly impaired in Gne (V572L) BM cells, suggesting that sialyl residues ligated to Galβ1-4GlcNAc also played a role in homing and engraftment, which was diminished by the Gne (V572L) point mutation.

Discussion
Colony-forming activity of HSPCs and the number of pHSC (short-term [ST]-HSC) were not different between β4GalT-1 −/− BM cells and β4GalT-1 +/− BM cells. Although the number of functional HSC can be quantitatively examined using in vivo limiting dilution assay 39 by BMT of serial diluted BM cells, it was difficult to estimate its number in β4GalT-1 −/− BM cells because homing and engraftment to the BM was severely impaired. In the differentiation pathway of HSC, the number of MPP was higher and those of MEP and CLP were lower in β4GalT-1 −/− BM cells than in β4GalT-1 +/− BM cells. These results suggest that HSC differentiation was disturbed to some extent downstream of MPP. Although the number of MPP increased in β4GalT-1 −/− BM cells, colony-forming activity of HSPCs was comparable between β4GalT-1 −/− and β4GalT-1 +/− BM cells. Therefore, the disturbed differentiation downstream of MPP did not seem to have an influence on our findings that homing and engraftment of β4GalT-1 −/− HSPCs after BMT were severely impaired. Colony-forming activity of HSPCs from β4GalT-1 −/− fetal liver cells slightly increased. These results indicated that the number, proliferation activity, or both of fetal liver HSPCs in β4GalT-1 −/− mice were slightly enhanced, not reduced, which also did not seem to affect the homing and engraftment of transferred β4GalT-1 −/− fetal liver HSPCs.
Several cell adhesion systems such as VLA-4/VCAM-1 and CXCR4/SDF-1 play a major role in HSPC homing and engraftment to the BM, while other systems such as LFA-1/ICAM-1 and sLe x /endothelial selectins play a supportive role. However, these systems interact to ensure efficient homing and engraftment. As β4GalT-1 −/− mice also have compromised biosynthesis of selectin ligands and reduced inflammatory reactions 8 , the contribution of sLe x /endothelial selectins to HSPC homing and engraftment must also be impaired. However, lethally irradiated mice deficient in both P-and E-selectin survive when transplanted with at least 5 × 10 5 BM cells 29 . Furthermore, homing in these mice deficient in both P-and E-selectin is reduced by approximately 50% compared with that in wild-type recipient mice 29 . These results strongly suggest that the defect in the sLe x /endothelial selectins system alone cannot explain the loss of homing and engraftment from β4GalT-1 −/− BM cells. Furthermore, the present results imply that a novel cell adhesion system based on galactosyl carbohydrates promotes HSPC homing and engraftment following transplantation. On the other hand, the abundance of colony-forming transplanted Gne (V572L) cells in recipient tissues, as well as the survival of lethally irradiated recipient mice transplanted with such cells, was highly similar to that previously observed in P/E-selectin double-knockout mice 29 . These results suggest that the impaired homing and engraftment of Gne (V572L) cells is due to the defect in the sLe x /endothelial selectins system. Alternatively, the mild phenotype of Gne (V572L) mice may be attributed to reduced, but not abolished, GNE activity.
During the latter half of mouse embryonic development until birth, hematopoiesis mainly occurs in the fetal liver 35,36 . Neonatal migration of HSPCs from the fetal liver to the adult BM seems to be normal in β4GalT-1 −/− mice, because colony-forming activity of HSPCs and ST-HSC populations in the adult BM were comparable between β4GalT-1 +/− and β4GalT-1 −/− mice. However, homing and engraftment of immature fetal liver HSPCs of β4GalT-1 −/− mice to the adult recipient BM was impaired similar to that in the BM HSPCs of β4GalT-1 −/− mice. These results suggest that fetal liver HSPCs and adult BM HSPCs use the similar galactosyl residues in homing and engraftment to the BM.
Lectin blots show that galactosyl residues in high-molecular-weight glycoproteins were lost in lineage-negative BM cells from β4GalT-1 −/− mice. Accordingly, these glycoproteins are good candidates as critical regulators of HSPC homing and engraftment. We noted that integrins such as integrin α4, α5, αL, and β1 are larger than 100 kDa and contain many possible glycosylation sites, some of which are actually N-glycosylated (Glycoprotein Database, http://jcggdb.jp/rcmg/gpdb/index.action). Therefore, it is possible that the function of VLA-4 (α4β1 integrin), VLA-5 (α5β1 integrin), or LFA-1 (αLβ2 integrin) in HSPC homing and engraftment is compromised by β4GalT-1 and GNE deficiency. Another possibility is that β4GalT-1 and GNE deficiency may disrupt the function of unknown carbohydrate ligands that regulate homing and engraftment. Thus, further studies are necessary to fully elucidate the role of carbohydrate residues in HSPC homing and engraftment.
In clinical applications of BMT, especially in cord blood transplantation, it is essential to enhance the efficiency of HSPC homing and engraftment. The present study suggests the possibility of modifying carbohydrate structures on the surface of HSPCs to improve their homing and engraftment to the BM in clinical application. Indeed, a recent study demonstrated that the ex vivo fucosylation of cord blood cells improved their homing abilities, leading to faster neutrophil and platelet engraftments 40 . Accordingly, it might be possible that enforced galactosylation and sialylation of HSPCs would also improve their homing and engraftment to the BM.
In conclusion, we have demonstrated that β4GalT-1 activity in donor BM cells, but not recipient mice, is critical for hematopoietic reconstitution and homing/engraftment to the BM after transplantation. However, BM cells from Gne (V572L) mice only showed relatively mild impairment. The deficiency of BM cells in sLe x /endothelial selectins system might explain the defect of BM cells from Gne (V572L) mice, but cannot explain the defect of β4GalT-1 −/− BM cells. Collectively, these data suggest that a novel cell adhesion system containing galactosyl or sialyl residues or both may promote homing and engrafting of HSPCs to the BM.

Materials and Methods
Mice. β4GalT-1 −/− mice on a mixed 129/Sv and C57BL/6 genetic background, and mice with a V572L point mutation in GNE [Gne (V572L) mice] on a C57BL/6 background were described previously 34,38 . To produce GFP-labeled BM cells, these mice were crossed with CAGGFP mice on a C57BL/6 background, which were kindly provided by Dr. Okabe at Osaka University 41 . NOD/SCID mice and pseudo-pregnant ICR mice were purchased from Charles River Japan and CLEA Japan, Inc., respectively. Animal experiments were conducted in Preparation of BM cells. BM cells were harvested by aseptically flushing the femur and tibia using a 22-gauge needle with Dulbecco's modified Eagle's medium (DMEM, Life Technologies, Grand Island, NY, USA) containing 5% fetal calf serum (FCS). The obtained cell suspension was filtered through a 70-μm mesh, treated with 140 mM NH 4 Cl in 17 mM Tris-HCl (pH 7.2) buffer for 5 min to lyse red blood cells, washed, and suspended in DMEM with 5% FCS for BMT and the colony formation assay.

Preparation of fetal liver cells.
Oocytes from β4GalT-1 +/− mice with the CAGGFP gene in homozygotes were fertilized in vitro by sperms from β4GalT-1 −/− mice and fertilized two-cell stage eggs were transferred to the oviduct of pseudo-pregnant ICR females. Embryos were collected at E14.5 and fetal livers were prepared by crushing using a plunger on 70-μm strainer in Ca 2+ -and Mg 2+ -free phosphate-buffered saline (PBS) containing 3% FCS. Fetal liver cells were collected after centrifugation, suspended in Hanks' solution, and passed through a 40-μm strainer for transplantation and colony formation assay 42 .
Colony formation assay. Specimens of the femur and spleen were prepared from recipient mice 3 h and 24 h after transplantation. BM cells and splenocytes collected from these specimens were filtered through a 70-μm mesh, suspended in RPMI-1640 medium, and mixed with MethoCult TM GF-M3434 medium (STEMCELL Technologies, Vancouver, Canada) according to the manufacturer's procedure. After incubation at 37 °C and 5% CO 2 for 7-11 days, GFP-positive donor-derived colonies were counted. GFP-positive colonies obtained from recipient tissues were normalized to the total BM cells in the recipient, considering that BM cells in the femur specimen constitute approximately 6.7% of all BM cells in a mouse 43 . The colony-forming ratio of the BM was calculated as the number of GFP-positive colonies per recipient mouse relative to the number of transplanted donor cells (2-5 × 10 6 ). Similarly, the colony-forming ratio in the spleen was calculated as the number of GFP-positive colonies per total spleen relative to the number of transplanted donor cells. Colonies formed by untransplanted donor BM cells were also quantified. When fetal liver cells were transplanted, GFP-positive colonies obtained from recipient femurs were normalized to 3 × 10 7 total BM cells in the femur of recipient mice. Colonies formed by untransplanted donor fetal liver cells were also quantified.