Recombinant human macrophage colony-stimulating factor-induced thrombocytopenia in dogs


Richard A. Nash, MD, Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North, D1-100, PO Box 19024, Seattle, WA 98109–1024, USA. E-mail:


Summary. To characterize recombinant human macrophage-colony stimulating factor (rhM-CSF)-associated thrombocytopenia (TCP), in vivo studies were performed in dogs, including the biodistributions and recoveries of radiolabelled autologous and allogeneic platelets. rhM-CSF induced a reversible, dose-dependent decrease in platelet counts. The number of megakaryocytes in spleen and marrow of rhM-CSF-treated dogs was increased two to threefold. Recoveries of allogeneic platelets transfused from rhM-CSF-treated donors into tolerized recipients (n = 3) were not significantly different from allogeneic baseline studies (93 ± 10% of baseline values at 24 h and 90 ± 1% at 40 h), whereas autologous platelets infused back into rhM-CSF-treated donors had decreased recoveries (45 ± 2% of baseline values at 24 h, P = 0·03 and 20 ± 4% at 40 h, P = 0·001). Platelet biodistribution studies showed increased accumulation of radiolabelled platelets over the spleens and livers of rhM-CSF-treated dogs. Histochemistry showed increased levels of platelet-specific antigen (CD41; glycoprotein IIb) associated with Kupffer cells. The sensitivity of platelets from rhM-CSF-treated dogs to activation from thrombin, as measured by expression of P-selectin (CD62P), was not significantly different when compared with baseline studies (P = 0·18; n = 4). These results support the concept that rhM-CSF induces an activation of the monocyte–macrophage system (MMS), which causes a reversible TCP in a dog model.

In preclinical and clinical studies of recombinant human macrophage colony-stimulating factor (rhM-CSF), thrombocytopenia (TCP) has been a dose-dependent limiting toxicity (Nemunaitis et al, 1991; Sanda et al, 1992; Minasian et al, 1995; Baker & Levin, 1998). The mechanism involved in M-CSF-associated TCP may contribute to the accelerated removal of platelets that has been observed in certain clinical conditions or with the administration of other cytokines. Therefore, the nature of rhM-CSF-associated TCP required further characterization. These studies were performed to determine whether changes in the platelets contributed to their accelerated removal.

Macrophage colony stimulating factor (M-CSF; CSF-1) is a cytokine involved in the proliferation and differentiation of the monocyte/macrophage cell lineage from progenitors to mature macrophages (Koike et al, 1986; Roth & Stanley, 1992). M-CSF, a homodimeric glycoprotein produced by multiple cell types, is predominantly cleared in the liver and spleen by binding to receptors (CD115) of macrophages with subsequent endocytosis of the receptor–ligand complex and intracellular degradation (Bartocci et al, 1987). rhM-CSF induces expression of maturation-linked surface antigens, and the synthesis and secretion of cytokines, including interleukin 6 (IL-6), IL-8, granulocyte–macrophage (GM)-CSF and granulocyte (G)-CSF (Motoyoshi, 1998). In vivo, rhM-CSF induced a dose-dependent increase in monocyte levels (Bukowski et al, 1994) with no consistent changes in the total white blood cell, lymphocyte or neutrophil counts (Hume et al, 1988; Munn et al, 1990; Nemunaitis et al, 1991; Minasian et al, 1995; Baker & Levin, 1998). In mice, it caused an increase in the weights of both the liver and spleen, and increased the number of macrophages within each of these organs. Additionally, these organs exhibited an increased phagocytic function, as measured by their ability to clear carbon particles from the blood (Hume et al, 1988; Baker & Levin, 1998).

In this study, we used a preclinical canine model to characterize the thrombocytopenic effects of rhM-CSF administration. We assessed whether accelerated removal of platelets from the peripheral circulation was mediated by non-specific activation of the monocyte–macrophage system (MMS), and whether cytokine-induced changes to platelets contributed to the development of TCP.

Materials and methods

Experimental animals.  Litters of random-bred dogs were raised at the Fred Hutchinson Cancer Research Center (FHCRC) or obtained from commercial kennels licensed by the US Department of Agriculture. The dogs weighed from 8 to 13 kg (median 10 kg) and were 8–19 months old (median 12 months).

Dogs were observed for indications of disease for at least 60 d before entry into the study. All dogs were immunized against leptospirosis, distemper, hepatitis, parvovirus and papillomavirus. The research was performed according to the principles outlined in the Guide for the Care and Use of Laboratory Animals, Institute of Laboratory Animal Resources, National Research Council. All dogs were housed in kennels certified by the American Association for Accreditation of Laboratory Animal Care and were examined at least twice daily. The Institutional Animal Care and Use Committee of the FHCRC approved the research protocols.

rhM-CSF administration.  The recombinant human rhM-CSF was produced and provided by Genetics Institute (Cambridge, MA, USA). For administration, rhM-CSF was diluted in normal saline (Abbott Laboratories, Chicago, IL, USA), containing 0·5% normal heat-inactivated dog serum. rhM-CSF was administered twice daily, subcutaneously at a dose of 300 µg/kg/d in three dogs, 500 µg/kg/d in four dogs and 600 µg/kg/d in two dogs for 7 d, and at 300 µg/kg/d in two dogs and 600 µg/kg/d in two dogs for 14 d, with the last two dogs being used to study biodistribution of re-infused 111In-labelled autologous platelets.

In order to test whether repeated administration of the cytokine could reproduce similar changes in blood cell counts, two of the dogs were given a second course of rhM-CSF. Dogs were initially given 300 µg/kg/d for 14 d and then later they received rhM-CSF at 300 µg/kg/d for a further 7 d.

Biodistribution of platelets by gamma camera imaging. Platelets were procured from treated dogs before, during and after 14 d of rhM-CSF administration. Platelets labelled with 111In were then infused to study their distribution via gamma camera imaging (General Electric 400 A/T, Fairfield, CT, USA). The techniques employed have been described previously (Heyns et al, 1982; Snyder et al, 1986; Nash et al, 1995). Dynamic acquisition by the camera/computer was begun immediately upon injection of the platelets. This first data set consisted of a series of dynamic images, each for 30 s and continuing for 90 min. Immediately after the dynamic acquisition, individual 5-min long static images were taken of the anterior chest/abdomen, anterior pelvis and hind legs, anterior head and fore legs, and posterior chest/abdomen. On d 2 after platelet injection, additional 5-min long static images of the anterior abdomen, posterior abdomen, anterior head and anterior pelvis were acquired. To keep the dogs immobile during the gamma camera imaging, general anaesthesia was administered, as described previously (Nash et al, 1995).

Quantification of platelet recoveries in peripheral blood.  Quantification of the recoveries of re-infused radiolabelled autologous platelets at all doses administered was studied by methods described previously (Weiden et al, 1976), and radioactivity in serial blood samples drawn post infusion was detected with a Minaxi 5000 γ Counter (Packard, Downers Grove, IL, USA). A weighted mean model to determine platelet recoveries and survivals was used to analyse the data, with blood samples taken at 0·5, 18, 24, 40 and 64 h.

Recoveries of transfused allogeneic platelets from rhM-CSF-treated dogs [at 500 µg/kg/d for 7 d (n = 3)] into tolerized recipients were studied in three healthy dog pairs. Suitable donor–recipient pairs for allogeneic clearance studies were first identified by platelet crossmatch by enzyme-linked immunosorbent assay, and selected recipients were then tolerized to donor platelets by techniques previously described (Slichter & Bean, 1991). Platelets were procured from a treated animal, radiolabelled and divided equally into two syringes. The radiolabelled platelets in one syringe were then transfused into a matched, tolerized recipient dog, and the other syringe was re-infused into the rhM-CSF-treated donor. The weekly platelet studies were continued for at least 2 weeks after rhM-CSF was discontinued, in order to ensure that the development of alloantibodies was not causing increased allogeneic clearance.

All of the recovery study results were expressed as a percentage of the platelets in circulation at 24 and 40 h post infusion after the start of rhM-CSF treatment as compared with percentages of circulating radiolabelled platelets at the same time-points in the baseline studies.

Immunohistological studies of the liver and spleen in rhM-CSF-treated dogs.  Four dogs were euthanized at the time of the platelet nadir (d 7 of rhM-CSF treatment) and underwent necropsy to assess the changes induced by rhM-CSF in the liver and spleen. Samples were embedded in optimal cutting temperature (OCT) compound (Miles, Elkhart, IN, USA), snap frozen, sectioned at 6 µm and stained by a streptavidin–horseradish peroxidase technique, according to the manufacturer's instructions (Zymed, South San Francisco, CA, USA) and previously described methods (Danilenko et al, 1995). Normal controls consisted of liver and spleen from four dogs not given rhM-CSF.

Primary monoclonal antibodies (mAb) were diluted 1:10 for optimum reactivity as established by titration. Negative controls consisted of an isotype-matched irrelevant mAb or omission of the primary antibody. The number, distribution and immunophenotype of platelets in connection to immune cells in liver and spleen was defined by canine-specific mAbs to major histocompatibility complex (MHC)-class II antigen (CA2.1C12), CD11b (CA16.3e10) (Cobbold & Metcalfe, 1994), CD18 (CA1.4E9) (Moore et al, 1990) and mAb 2F9 [specific for gpIIb (CD41); 2F9] (Burstein et al, 1991).

Flow studies on expression of cell surface adhesion markers of platelets and leucocytes.  Blood samples from four rhM-CSF-treated dogs [500 µg/kg/d for 7 d (n = 2); 600 µg/kg/d for 7 d (n = 2)] were drawn into syringes containing the anticoagulant acid citrate dextrose (ACD). To exclude the influence of non-specific platelet activation, blood was drawn with and without the platelet activation inhibitor prostaglandin E1 (PGE1). Thrombin activation tests were conducted to determine whether rhM-CSF administration resulted in the preactivation of platelets (Peng et al, 1996). Platelets from rhM-CSF-treated dogs obtained before and at d 5 of cytokine administration were thrombin activated, probed with antibodies to P-selectin (CD62P) and glycoprotein (gp)IIb (CD41) 2F9 (Burstein et al, 1991), then fixed and assayed as previously described (Peng et al, 1996). Two thousand events were captured via FACScan and analysed with lysysii software (Becton Dickinson, San Jose, CA, USA).

To study the possible changes in platelet–leucocyte interactions induced by rhM-CSF, samples of peripheral blood from two rhM-CSF-treated dogs (at 600 µg/kg/d for 7 d) were obtained in EDTA vacutainers and stained with antibodies to CD18 and CD41. Staining was performed by techniques previously described (Nash et al, 1995). Lymphocyte, monocyte and neutrophil populations were gated according to their characteristic forward (FSC) versus side scatter (SSC) profiles. Changes to white-blood-cell-associated integrins were detected by CD18-PE. Platelet binding to each of these cell types was detected with CD41-FITC. Fluorescence was measured on 10 000 events with a FACScan flow cytometer (Becton Dickinson). Dead cells and debris were excluded by live gating using conventional forward and side scatter light characteristics of canine leucocytes. Data were analysed using lysysii software.

Statistical analysis.  The significances of rhM-CSF-induced dose-dependent changes in blood cell counts, platelet P-selectin (CD62P) expression and platelet recovery studies were evaluated using paired t-tests.


rhM-CSF induced a transient dose-dependent decrease in platelet counts in the dog

Administration of rhM-CSF to healthy dogs in all groups studied resulted in decreased platelet counts within 24 h of administration that reached the nadir by d 5 ± 1. The decrease ranged from 65 ± 10% at 300 µg/kg/d of rhM-CSF (baseline counts 307·4 × 109/l ± 161·8, counts at nadir 98·8 × 109/l ± 47·4, P = 0·0001) to 92 ± 4% at 600 µg/kg/d of rhM-CSF (baseline counts 325·3 × 109/l ± 93·5, counts at nadir 24 × 109/l ± 11·7, P = 0·0001) (Fig 1). The observed decrease in platelet counts in rhM-CSF-treated dogs was dose dependent, being 1·4-fold greater at the highest dose used (600 µg/kg/d) as compared with the lowest (300 µg/kg/d) (P = 0·0009, n = 13). There was a two to threefold increase in megakaryocyte numbers in the red pulp of spleen and bone marrow at the time of the platelet nadir. In spite of continued administration of rhM-CSF, platelet counts began to rise and approached baseline levels by d 11 in all dogs studied.

Figure 1.

Platelet counts in representative dogs receiving rhM-CSF for 14 d at 300 µ/kg/d [two dogs: BC9326 (●), D881 (▪)] and 600 µ/kg/d [two dogs: D973 (▴), E009 (◆)]. There was a dose-dependent decrease in platelet counts with a nadir by d 5 ± 1 with a subsequent rise in counts in spite of continued cytokine administration. Platelet counts recovered to the baseline levels during week 2 of rhM-CSF administration.

Thrombocytopenic effects of rhM-CSF were reproducible in two dogs upon the second administration of rhM-CSF (300 µg/kg/d for 7 d) 1 month after the initial course (300 µg/kg/d for 14 d). However, a less pronounced decrease in platelet counts was observed in one these dogs.

No significant changes were observed in red blood cell, total white blood cell, neutrophil, lymphocyte or monocyte counts. There were no signs of adverse clinical effects other than the development of thrombocytopenia.

Platelets from rhM-CSF-treated dogs have normal survivals when transfused into allogeneic recipients

In the initial dosing studies, administration of rhM-CSF resulted in a decrease of autologous radiolabelled platelet recoveries (n = 9) to 63 ± 6% of baseline levels at 24 h and to 35 ± 8% of baseline levels at 40 h after infusion. This decrease in recoveries was not dose dependent and persisted throughout the cytokine course in spite of an observed rise in platelet counts after the initial nadir at d 5 ± 1 (Table I). Studies performed upon the discontinuation of cytokine revealed that both platelet counts and recoveries had returned to their baseline levels in all dogs studied.

Table I.  Autologous platelet recovery in response to rhM-CSF administration.
Dog IDDose of rhM-CSF
Days of rhM-CSF
Platelet recovery during
rhM-CSF treatment
Platelet recovery post rhM-CSF treatment
1st week2nd week2nd week or 3rd week
24 h40 h24 h40 h24 h40 h
  1. All platelet recovery data are given as the percentage of platelets in circulation compared with baseline (pretreatment) autologous platelet recovery values.

E05450074915  9572
E10750074119  127109
E16950074627  105113
Total  63 ± 635 ± 8  102 ± 7 91 ± 8

Recoveries (n = 3) of allogeneic platelets transfused from rhM-CSF-treated donors into tolerized non-rhM-CSF-treated recipients were comparable to baseline (pretreatment) recoveries of allogeneic transfused platelets from the same donor (93 ± 10% of baseline value at 24 h). Decreased autologous recoveries were observed (45 ± 2% of baseline values at 24 h) when a portion of these radiolabelled autologous platelets was infused into the rhM-CSF-treated platelet donors (P = 0·03) (Table II). Circulating autologous and allogeneic platelet recoveries (n = 3) were 20 ± 4% and 90 ± 1% (P = 0·001) of baseline levels at 40 h respectively.

Table II.  Allogeneic rhM-CSF-treated platelet recoveries in transfusion-tolerized recipient dogs.
dog ID
Dose of rhM-CSF
Days of rhM-CSF
Autologous platelet recovery*
1st week on rhM-CSF treatment
dog ID (donor)
Allogeneic platelet recovery
24 h40 h24 h40 h
  • *Autologous platelet recoveries for E054, E107 and E169 were also presented in Table I. Radiolabelled platelets from each of these dogs were split into two equal fractions. One fraction of the radiolabelled platelets was infused into a paired allogeneic recipient not treated with rhM-CSF and the other fraction was re-infused back into the rhM-CSF-treated autologous recipient of the platelets. Recoveries of radiolabelled platelets in the rhM-CSF-treated autologous recipients were significantly less than in the normal allogeneic recipients. All recovery data is given as the percentage of platelets in the circulation compared with baseline (pretreatment) autologous or platelet recovery values.

  • All recovery data is given as the percentage of platelets in the circulation compared with baseline (pretreatment) allogeneic platelet recovery values.

E05450074915E245 (E054)9588
E10750074119E246 (E107)7490
E16950074627E243 (E169)10992
Total  45 ± 220 ± 4  93 ± 1090 ± 1

Increased uptake of radiolabelled autologous platelets in the liver and spleen of rhM-CSF-treated dogs

Biodistributions of radiolabelled platelets studied by gamma camera imaging (n = 2) showed platelet accumulation in the livers and spleens of rhM-CSF-treated dogs. The intensity of the organ-to-heart radioactive signal ratio in the two dogs was increased by 19% and 103% over the liver, and 34% and 35%, respectively, over the spleen as compared with baseline organ-to-heart ratios (Fig 2).

Figure 2.

Static anterior gamma camera images taken 90 min after infusion of 111In-radiolabelled platelets by gamma camera imaging at baseline (A) and at d 8 post rhM-CSF administration (B). There was a 103% increase in the accumulation of platelets over the liver (L) and a 35% increase over spleen (S) relative to heart blood pool (H) compared with the baseline in one of the two dogs evaluated.

rhM-CSF induced increased uptake of platelets by Kupffer cells in liver and macrophages in spleen

Immunohistology revealed an increased intensity of the stain for platelet-specific antigen CD41 associated with Kupffer cells from dogs treated with rhM-CSF for 7 d as compared with untreated control dogs. It was not possible to determine whether the CD41 was expressed on the surface of Kupffer cells or was present within the cells. Based on MHC II and CD11b expression in serial sections, there were no differences in the number or distribution of Kupffer cells between livers of control and rhM-CSF-treated dogs. Staining for CD41 in the red pulp of rhM-CSF-treated spleens was markedly diminished, probably as a result of decreased platelet counts in peripheral blood and the smaller storage pool of platelets in the spleen. In contrast, marginal zones of the spleens were two to three times wider than normal and strongly positive for CD41.

rhM-CSF administration expanded a minor leucocyte population and increased the expression of adhesion markers on leucocytes

There was a minor population of cells (FSCLOSSCHI) that increased from 2% to 13% of all leucocytes (Fig 3). The intensity of fluorescence detected from phycoerthrin-conjugated CD18 (CD18-PE) bound to this and other leucocyte subpopulations (monocytes, neutrophils and lymphocytes) increased by d 5 of rhM-CSF administration (Fig 4).

Figure 3.

Expansion of FSCLOSSCHI CD18+ cell population with concurrent contraction of the lymphocyte population. The percentages reported represent the percentage of each leucocyte subset versus the total leucocytes gated. R1 and R5 = FSCLOSSCHI, R2 and R6 = neutrophils, R3 and R7 = monocytes, and R4 and R8 = lymphocytes. (FSC, forward scatter; SSC, side scatter).

Figure 4.

Platelets binding to peripheral blood leucocytes by flow cytometry. Y-axis, CD18-PE (integrin); X-axis, CD41-fluorescein isothiocyanate (FITC) (gpIIb). Lower left box in d 1 panels indicates the negative population; the horizontal line distinguishes between the high and low levels of CD18. Increased expression of CD18 on FSCLOSSCHI cells, neutrophils, monocytes and lymphocytes was observed by d 5 after rhM-CSF treatment. Platelets bound to FSCLOSSCHI cells, neutrophils and monocytes, but not lymphocytes, and this did not change qualitatively with M-CSF treatment.

Baseline flow cytometry studies showed platelets binding to the FSCLOSSCHI cells, monocytes and neutrophils, but negligible binding to lymphocytes. The binding of platelets to these peripheral blood leucocyte subsets was not qualitatively changed after rhM-CSF administration (Fig 4).

P-selectin expression was detected on 49 ± 5% of baseline platelets upon ex vivo thrombin stimulation. Platelets from rhM-CSF-treated dogs activated ex vivo by thrombin at the same dose did not show a significant increase in P-selectin expression (62 ± 1%, P = 0·18, n = 4).


Thrombocytopenia has been associated with the administration of a number of cytokines, including GM-CSF (Nash et al, 1995), M-CSF (Koike et al, 1986; Sanda et al, 1992; Minasian et al, 1995), G-CSF (Lindemann et al, 1989), IL-2 (Paciucci et al, 1990), stem cell factor (SCF) (Schuening et al, 1993) and Flt3 ligand (unpublished observations). There may be a common mechanism responsible for cytokine-related decreases in platelet counts. Increased blood levels of M-CSF associated with thrombocytopenia have been demonstrated in such diverse clinical settings as pregnancy (Yong et al, 1992), malaria (Lee et al, 1997) and sepsis (Francois et al, 1997), supporting the concept that the mechanism may be M-CSF mediated.

In this experiment, rhM-CSF administration to dogs induced a dose-dependent decrease in platelet counts, but an increase in circulating monocyte counts was not observed. Although Bukowski et al (1994) demonstrated that rhM-CSF-induced monocytosis was dose dependent, the highest dose used in our study (600 µg/kg/d) may still be below a threshold level for inducing monocytosis in the dog model. However, at this dosage, a minor leucocyte population in the peripheral blood (FSCLOSSCHI) was increased by d 5 of rhM-CSF administration. These cells may represent the previously described expansion of a circulating CD16+ monocyte subset, noted when rhM-CSF was administered to both human (Saleh et al, 1995) and non-human primates (Munn et al, 1990). Platelet counts began to rise in all dogs studied after the nadir around d 5, in spite of continued administration of rhM-CSF. This trend of platelet recovery despite continued rhM-CSF administration was similar to findings previously reported both in humans and in animal models with a number of other cytokines (Lindemann et al, 1989; Schuening et al, 1993; Nash et al, 1995). Even though platelet counts were rising during week 2 of rhM-CSF administration, consumption of platelets continued, as evidenced by persistently decreased platelet recoveries and increased numbers of megakaryocytes in the spleen and marrow. This indicated that the physiological response induced by rhM-CSF was maintained, and excluded the development of neutralizing antibodies against rhM-CSF as a possible cause for the increase in platelet counts during its administration. The expansion of the megakaryocyte pool indicated that accelerated platelet production was the most probable cause of the increase in platelet counts after the initial nadir induced by rhM-CSF.

Activation of the MMS has been proposed as a general mechanism for the development of rhM-CSF-induced TCP even though splenectomy did not decrease the severity of the TCP (Baker & Levin, 1998). Other cytokines (GM-CSF), as previously mentioned, may also induce an increase in platelet consumption through non-specific activation of the MMS (Nash et al, 1995). In our experiments, there was an increased detection of radiolabelled autologous platelets by gamma camera imaging over the spleen and liver during rhM-CSF administration. Increased platelet-specific antigen gpIIb was detected in association with Kupffer cells in the liver and macrophages in the splenic marginal zone. Altogether, this provided the supporting evidence that rhM-CSF-induced MMS activation resulted in the accelerated removal of platelets from the circulation and the development of TCP. Whether platelets found in association with Kupffer cells were bound to the cell membranes or were located within the cells themselves needs to be further elucidated.

Using flow studies, we further investigated whether rhM-CSF induced any changes to the peripheral blood leucocytes and/or platelets, and if those changes contributed to an accelerated removal of platelets from the circulation. Administration of rhM-CSF increased the expression of CD18 on peripheral blood leucocytes, providing evidence for an augmented adhesiveness mediated by members of the CD18 integrin family, lymphocyte-function-associated antigen 1 (LFA1, CD18/CD11a), Mac-1 (CD18/CD11b) and CR4 (CD18/CD11c). However, no change in the binding of platelets to peripheral blood leucocytes (Rinder et al, 1991) after rhM-CSF administration was observed. The contribution of peripheral blood leucocytes to the development of rhM-CSF-associated TCP appears to be minimal, if any at all.

Flow cytometry assays were performed to compare thrombin-mediated P-selectin expression on the membrane surface of platelets at baseline and during treatment with rhM-CSF. As rhM-CSF has been shown to increase both the platelet production and consumption rates by d 5 of administration, changes in P-selectin expression upon stimulation with a submaximal dose of thrombin may reflect the development of a preactivation state after rhM-CSF administration (Peng et al, 1994, 1996). Thrombin-mediated expression of the adhesion marker P-selectin (CD62P) was not significantly different as compared with baseline studies and, therefore, in addition to the normal recoveries of platelets from rhM-CSF-treated dogs in allogeneic recipients, it can be concluded that thrombocytopenia did not result from preactivation or changes in platelets induced by rhM-CSF.

The results from this study support the concept that rhM-CSF induces an activation of the MMS which causes reversible TCP in a dog model. Platelets do not undergo changes that contribute to their premature removal from the circulation. More studies are required to further define the mechanisms involved in the development of TCP that may help in designing treatments for the prevention or correction of the premature removal of circulating platelets found in some diseases that are associated with increased levels of cytokines.


We thank Robert Schaub, PhD, Wyeth, Cambridge, MA, USA, for the gift of the rhM-CSF. We thank Barbara Johnston, DVM, for providing veterinary support for the dog colony, and Ted Graham and Doug Jones for their contributions to the care of the dogs. We are grateful to the technicians of the Shared Canine Resource and Hematology Laboratory for their technical assistance. We thank Laura Bolles for her contribution to the flow cytometry studies, and Chris Davis for generating graphs. We thank Helen Crawford, Bonnie Larson, Lori Ausburn, Diana Tepp, Sue Carbonneau, Karen Carbonneau and Connie Chan for their excellent support in preparing this manuscript. This work was supported by grants DK42716, HL63457, HL47227 and CA15704 from the National Institutes of Health, Department of Health and Human Services, Bethesda, MD, USA.