Correspondence Douglas C. Woodhams, Department of Microbiology and Immunology, A-5301 Medical Center North, Vanderbilt University Medical Center, Nashville, TN 37232, USA. Tel: +1 615 343 9449; Fax: +1 615 343 7392 Email: email@example.com
Innate immune mechanisms of defense are especially important to ectothermic vertebrates in which adaptive immune responses may be slow to develop. One innate defense in amphibian skin is the release of abundant quantities of antimicrobial peptides. Chytridiomycosis is an emerging infectious disease of amphibians caused by the skin fungus, Batrachochytrium dendrobatidis. Susceptibility to chytridiomycosis varies among species, and mechanisms of disease resistance are not well understood. Previously, we have shown that Australian and Panamanian amphibian species that possess skin peptides that effectively inhibit the growth of B. dendrobatidis in vitro tend to survive better in the wild or are predicted to survive the first encounter with this lethal pathogen. For most species, it has been difficult to experimentally infect individuals with B. dendrobatidis and directly evaluate both survival and antimicrobial peptide defenses. Here, we demonstrate differences in susceptibility to chytridiomycosis among four Australian species (Litoria caerulea, Litoria chloris, Mixophyes fasciolatus and Limnodynastes tasmaniensis) after experimental infection with B. dendrobatidis, and show that the survival rate increases with the in vitro effectiveness of the skin peptides. We also observed that circulating granulocyte, but not lymphocyte, counts differed between infected and uninfected Lit. chloris. This suggests that innate granulocyte defenses may be activated by pathogen exposure. Taken together, our data suggest that multiple innate defense mechanisms are involved in resistance to chytridiomycosis, and the efficacy of these defenses varies by amphibian species.
It is likely that multiple factors underlie disease resistance. The joint effects of environmental conditions, pathogen biology and host biology may regulate the timing and severity of chytridiomycosis. One aspect of host biology that may contribute to disease resistance is the immune defense system. Amphibians have a well-developed immunity featuring both adaptive and innate responses (Carey, Cohen & Rollins-Smith, 1999; Apponyi et al., 2004). Amphibian skin is a primary defensive barrier against infection by environmental pathogens; it is also the site of infection of B. dendrobatidis in post-metamorphic amphibians. Limited lymphocytic infiltration in the skin of frogs susceptible to severe chytridiomycosis suggests that these animals develop poor adaptive immune responses against B. dendrobatidis (Berger et al., 1998; Pessier et al., 1999; Berger, Speare & Skerratt, 2005b). However, very little is known about the possible role of circulating lymphocyte and granulocyte populations in the response to chytridiomycosis. Here, we investigated differences in populations of circulating lymphocytes, monocytes, neutrophils, eosinophils and basophils in infected and uninfected control Litoria chloris to determine whether changes in any of these populations would reflect an ongoing immune response to B. dendrobatidis. Because the adaptive immune response may be ineffective against B. dendrobatidis, innate defenses such as epithelial barriers, competitive microbiota (Harris et al., 2006), phagocytic cells and antimicrobial peptides may be central to effective protection.
Amphibian skin produces large quantities of host-defensive peptides (Apponyi et al., 2004; Pukala et al., 2006). Although alkaloids and amines can also be present in amphibian skin secretions, many of the secreted peptides have antimicrobial activities. In addition, antimicrobial peptides within naturally secreted mixtures can act together in synergy (Westerhoff et al., 1995; Rollins-Smith et al., 2002; Patrzykat & Douglas, 2005). In vitro experiments show that many isolated skin peptides and natural peptide mixtures have activity against B. dendrobatidis (Rollins-Smith & Conlon, 2005; Woodhams et al., 2006a). In Australian rainforest amphibians, population trends associated with chytridiomycosis were correlated with skin peptide defenses (Woodhams et al., 2006a). In a study of amphibians at an upland rainforest at El Copé, Panama, the resistance of some species to emerging chytridiomycosis was predicted based on skin peptide defenses (Woodhams et al., 2006b). Here, we examine the susceptibility of four Australian species to experimental B. dendrobatidis infection and correlate susceptibility or resistance with antimicrobial peptide defenses in vitro. We suggest that species variation in these defenses partially accounts for the disparity in disease susceptibility among species.
Materials and methods
We bred four species of wild-caught adult frogs at the Amphibian Research Centre in Melbourne, Victoria, Australia. From a single egg clutch of each species, we raised full siblings through metamorphosis under conditions restricting the potential for exposure to B. dendrobatidis. We randomly allocated 40 juvenile frogs of each species to exposure and control treatments. These species overlap in range in south-eastern Australia, and were chosen for their variability in size under the hypothesis that development of lethal chytridiomycosis would take longer in species with a larger skin surface area. Species included (mean snout–vent length at 2 months of age±sd in brackets): orange-eyed treefrogs, Litoria chloris (17.40±1.18 mm), green tree frogs, Lit. caerulea (40.24±3.80 mm), spotted marsh frogs, Limnodynastes tasmaniensis (23.21±2.15 mm) and great barred frogs, Mixophyes fasciolatus (40.08±2.57 mm). We housed frogs individually in 1.6 L transparent plastic enclosures (18 × 11 × 12 cm) with ventilated lids (Critter Keeper™, Rolf Hagen & Co. Inc: Mansfield, MA, USA). The bottoms of containers were covered by a 2–3 cm layer of aquarium gravel, with a depression at one end that allowed access to standing water. Drain holes 2 cm from the bottom of containers maintained water levels. Water was renewed through low-volume inlets for 15 min daily. We maintained all enclosures on a 12L:12D photoperiod and placed them on shelving in 20 blocks, each block containing all treatments. We fed frogs every other evening with crickets dusted in a vitamin supplement and occasionally with houseflies. Enclosures had separate drain pipes, minimizing the opportunity for disease transmission.
Survival following exposure to B. dendrobatidis
In order to compare patterns of survival among juvenile frogs of four species following standardized exposure to a pathogen, we exposed 20 individuals of each species to B. dendrobatidis zoospores and 20 to culture broth lacking zoospores (sham-infected controls) in May 2001. Batrachochytriu dendrobatidis (isolate GibboRiver-Llesueuri-00-LB-1) culture was obtained from CSIRO in Geelong, Victoria. To expose the frogs, we placed them in 100 mL specimen jars containing 2 mL of water and either 150 μL of B. dendrobatidis culture broth, diluted to contain c. 5000 zoospores, or 150 μL of sterile broth. We then placed the jars horizontally and rolled them every 15 min for 15 h to thoroughly expose frogs to the zoospores, and then placed the frogs into their enclosures. Because the tree frogs were prone to climbing out of the solution, this method insured exposure to the pathogen. This level of exposure is likely more severe than occurs in nature but was used to ensure infection. We monitored the survival of all frogs for 108 days after B. dendrobatidis exposure. Once the animals became fully symptomatic for chytridiomycosis, they were euthanized in 0.2% tricaine methane sulfonate (TMS). Clinical signs of severe disease include erythema of ventral skin, lethargy and delayed righting reflex. These signs are indicative of mortality within hours (Berger, Speare & Hyatt, 1999; Berger et al., 2004, 2005b). Surviving frogs including controls were maintained at the Amphibian Research Centre following the experiment. We analyzed survival data with a logrank test on censored survival data in StatXact 4.0 (CYTEL Software Corporation).
Infection status of frogs exposed to B. dendrobatidis
To examine the status and intensity of infection, we removed one toe from each frog before exposure, 10 days after exposure and at death or 108 days following experimental exposure for surviving frogs. Following the protocol of Berger, Speare & Kent (1999), we diagnosed infection status from histological sections without reference to treatment. We assessed the intensity of infection as follows: we determined the density of B. dendrobatidis sporangia on the toes of deceased frogs by dividing the average number of sporangia per cross section by the average perimeter of sections from that species. We estimated perimeters from a sample of five sections per species measured by superimposing a section on a hemocytometer with a 0.0025 mm2 grid. Using a Kruskal–Wallis test in SPSS v.12 (SPSS Inc.), we compared the infection intensities among species.
Effectiveness of natural mixtures of skin peptides against B. dendrobatidis
In order to test for a correlation between the survival rate of infected frogs and skin peptide defenses against B. dendrobatidis, we shipped five Lit. chloris [mean weight (g)±sd: 2.43±0.32], six Lit. caerulea (21.17±3.85), five Lim. tasmaniensis (3.56±0.34) and five M. fasciolatus (29.88±2.58) of the control frogs of the four species from the Amphibian Research Centre, Melbourne, Victoria, to James Cook University (JCU), Townsville, Queensland, in October 2002. Because sampling of skin peptides may interfere with measurement of survival time, we only sampled peptides from the control animals after completion of the infection experiment. At JCU, we housed frogs individually at room temperature (c. 24°C) with a 12L:12D photoperiod and regular feeding schedule with vitamin-dusted crickets. After a 3-week acclimation, we collected skin peptide secretions by bathing frogs in collection buffer containing 100 μM norepinephrine–bitartrate salt (Sigma–Aldrich, St. Louis, MO, USA) for 15 min (Woodhams et al., 2006a). We passed the skin secretions over C-18 Sep-Pak cartridges (Waters Corporation, Milford, MA, USA) to enrich for cationic and hydrophobic peptides and shipped the Sep-Paks to Vanderbilt University in Nashville, Tennessee. We eluted the peptides bound to the Sep-Paks with 70% acetonitrile, 29.9% water, 0.1% trifluoroacetic acid (v/v/v) and concentrated them to dryness by centrifugation under vacuum. To determine total skin peptides recovered after Sep-Pak purification, we followed the manufacturer's instructions for Micro BCA Assay (Pierce, Rockford, IL, USA), except that bradykinin (RPPGFSPFR) (Sigma) was used to establish a standard curve (Woodhams et al., 2006a).
We diluted peptide mixtures to concentrations between 1.6 and 500 μg mL−1 and incubated them with B. dendrobatidis zoospores to determine whether the peptides could inhibit growth after 7 days of incubation at 23°C. We determined the minimal inhibitory concentration (MIC in μg mL−1) for each species (Rollins-Smith et al., 2002). MIC is defined as the lowest concentration of added peptide at which no significant growth of B. dendrobatidis was observed. We calculated the mean MIC equivalents for each species (Woodhams et al., 2006b). An MIC equivalent is the total amount of peptides (μg) recovered from each frog per 1 cm2 surface area, divided by the experimentally determined MIC (μg mL−1) for each species. Surface area was calculated using the standard equation from McClanahan & Baldwin (1969): surface area (cm2)=9.90 (weight in grams)0.56.
We used a Jonckheere–Terpstra trend test carried out in StatXact 4.0 to test for differences among species in the distribution of MIC equivalents along the ordered gradient formed by per cent survival. We compared the MIC levels of the three groups of species with differing survival rates (Lit. caerulea and M. fasciolatus, 5%, Lit. chloris, 35%, and Lim. tasmaniensis, 100%).
Response of leukocyte populations to infection in Lit. chloris
This experiment was designed to determine how leukocyte populations change in response to environmental temperature or infection with B. dendrobatidis. Litoria chloris were chosen because they showed a non-constant survival curve and intermediate susceptibility to chytridiomycosis compared with the other species tested here (Fig. 1) and were therefore most likely to display an immune response such as changes in leukocyte populations after exposure to B. dendrobatidis. At the Amphibian Research Centre, 72 captive-bred frogs were raised for 3 months post-metamorphosis to c. 20 mm snout–vent length. We then randomly allocated them to treatments. The experiment had a factorial design, with two levels of infection (yes or no) crossed with three levels of a thermal regime (constant 20°C, naturally fluctuating temperatures ranging from 14 to 23°C+/−1.8°C sd, and naturally fluctuating temperatures with two nights of cooling down to 8°C 12 days after exposure to B. dendrobatidis). Because temperature can affect amphibian immunity and mortality due to chytridiomycosis (Woodhams, Alford & Marantelli, 2003; Berger et al., 2004; Pounds et al., 2006; Raffel et al., 2006), three thermal regimes were chosen to represent constant laboratory conditions, natural temperature variation and a cold-shock scenario. There were 12 frogs in each treatment, housed in two enclosures each containing six individuals. Samples were collected from four individuals in each group, and the data for individual frogs were used as replicates in analyses. We inoculated frogs as described above with 15 000 zoospores or sterile broth.
On day 27 post-infection, four frogs from each group enclosure were anesthetized with 0.01% TMS. We collected blood for differential leukocyte counts in heparinized microhematocrit tubes from a cut in the abdominal vein. We then euthanized the frogs in 0.02% TMS, and verified their infection status by histology of the skin. Insufficient volume was available from these juvenile frogs for total white and red blood cell counts, but relative counts were made from blood smears. We stained blood smears with Diff-Quik (Dade Behring Inc: Newark, DE, USA) for identification of lymphocytes, monocytes, neutrophilic, eosinophilic and basophilic leukocytes. We calculated the ratio of leukocytes to erythrocytes by counting the number of leukocytes among at least 300 erythrocytes.
To test for overall differences among temperature treatments and between infected and uninfected frogs, we used multivariate oneway randomization tests [R. A. Alford, RAMAN v 1.73, based on Manly (1991)]. When these were significant, we then used Mann–Whitney U-tests to determine which blood parameters differed significantly.
Survival following experimental exposure to B. dendrobatidis
Survival following exposure to B. dendrobatidis zoospores varied among species (logrank test on censored survival data, P<0.001, Fig. 1). All (18/18) of the exposed Lim. tasmaniensis survived to the termination of the experiment at day 108, whereas only 5% of exposed Lit. caerulea (1/20) and M. fasciolatus (1/20) survived. Survival of chytrid-exposed Lit. chloris was intermediate, with 35% of the animals (7/20) surviving to the end of the experiment (Fig. 1). The survival times of individual frogs of each species also reflected their comparative sensitivity to lethal infection. The median survival time of Lit. caerulea was only 8.9 days (mean±se: 28.8±2.1), whereas the median survival time of M. fasciolatus and Lit. chloris was 21.5 days (56.6±3.9) and 26.8 days (65.1±7.4), respectively. All of the Lim. tasmaniensis survived longer than the 108-day experimental period (Fig. 1).
Infection status of exposed frogs
No exposed animals were positive for infection by B. dendrobatidis on the toe removed before chytrid exposure. The level of infection detectable on the toes at 10 days post infection varied with species. Sixty per cent (12/20) of the Lit. caerulea and 20% (4/20) of M. fasciolatus were positive for B. dendrobatidis, whereas none of the Lit. chloris (0/20) or Lim. tasmaniensis (0/18) were positive at 10 days post infection (Table 1). All of the dead animals including those euthanized with severe clinical chytridiomycosis were heavily infected as determined by the histology of the toe clips, but infection was not detected in surviving frogs 108 days after experimental exposure (Table 1).
Table 1. Infection status and intensity of infection of four species of Australian anurans diagnosed by histology of skin sections 10 days post-infection, after death or euthanasia, or after surviving 108 days post-exposure to Batrachochytrium dendrobatidis.
The intensity of infection at the time of death differed significantly among Lit. caerulea, Lit. chloris and M. fasciolatus (Kruskal–Wallis test, χ22=20.248, P<0.001; Table 1). Batrachochytriu dendrobatidis sporangia were the most dense on the toes of M. fasciolatus. Histological sections of the toes of Lim. tasmaniensis toes often revealed missing or sloughing stratum corneum and were all negative for infection. Therefore, infection intensity was not determined for this species.
Effectiveness of natural mixtures of skin peptides against B. dendrobatidis
Natural mixtures of skin peptides from all four species were capable of inhibiting the growth of B. dendrobatidis in vitro. Significant dose-dependent inhibition was detected for skin peptides from individuals of Lit. caerulea (mean MIC=271 μg mL−1), Lit. chloris (mean MIC=267 μg mL−1) and Lim. tasmaniensis (mean MIC=133 μg mL−1). Individual M. fasciolatus did not provide sufficient peptides to test. However, peptides pooled from groups of three individuals were tested. These mixtures of peptides inhibited B. dendrobatidis growth, and the mean MIC was estimated at 272 μg mL−1. Representative experiments showing growth inhibition curves and MICs of individual frogs are shown (Fig. 2). MIC equivalents cm−2 differed among species (Kruskal–Wallis test, χ32=10.475, P=0.0149). The pattern of MIC equivalents mirrors the pattern of survival in these species; species differ in their effectiveness of skin peptide defenses per surface area of skin along the ordered gradient of per cent survival after B. dendrobatidis exposure. A significant trend is indicated (Jonckheere–Terpstra statistic=178, exact two-tailed P=0.001).
Response of leukocyte populations to infection in Lit. chloris
Leukocyte populations in the blood of Lit. chloris did not differ significantly among the three temperature treatments (multivariate randomization test, P=0.595). They did differ between infected and uninfected frogs across the temperature treatments (multivariate randomization test, P=0.011, Table 2). Neutrophilic and eosinophilic granulocyte counts were lower, and basophilic granulocyte counts were higher in the blood of infected frogs (Mann–Whitney U-tests, P=0.006, 0.046 and 0.026, respectively).
Table 2. Leukocyte populations of juvenile Litoria chloris
Mean count out of 100 leukocytes for five cell types and the ratio of leukocytes to 1000 erythrocytes. Numbers in bold indicate significant differences, P<0.05.
75.36 ± 2.25
71.2 ± 5.41
6.5 ± 1.14
6.22 ± 0.70
Ratio of leukocytes to 1000 erythrocytes
2.7 ± 0.31
2.15 ± 0.52
Varied survival following experimental infection
The emergence of amphibian chytridiomycosis has devastated some species while leaving others almost untouched. Here, we show that when environmental conditions and pathogen strain are held constant, the outcome of infection by B. dendrobatidis varies among species. Lit. caerulea and M. fasciolatus were highly susceptible to chytridiomycosis and experienced 95% mortality. Lit. chloris was less susceptible with 65% mortality, and Lim. tasmaniensis exhibited resistance to infection. Only Lit. chloris showed a non-constant survival curve, reaching a temporary plateau between 50 and 70 days post-exposure (Fig. 1), indicating a potential adaptive immune response. Apponyi et al. (2004) found that Limnodynastes spp. produced large volumes of secretions containing small inactive peptides. Here, we show that Lim. tasmaniensis secretions are active against B. dendrobatidis and this species survives infection better than three other species. Future studies should investigate the active components of Lim. tasmaniensis skin secretions. Survival, compared among these four species, tended to be higher in species with more effective skin peptides. None of these species are known to be in population decline and are of the least conservation concern (Global Amphibian Assessment, 2006, http://www.globalamphibians.org). Although Lit. caerulea and M. fasciolatus are highly susceptible to chytridiomycosis in the laboratory, perhaps environmental or behavioral characteristics provide some protection in the field. Experimental infection experiments in field enclosures may be ethically problematic.
At least some of the variation among species in patterns of survival under controlled environmental conditions appears to be caused by differences in skin peptides. Figure 3 shows that species reaching a level (roughly 0.02 MIC equivalents cm−2) of peptide effectiveness may be protected from chytridiomycosis. This is consistent with a previous study of Australian rainforest species demonstrating populations trends associated with peptide effectiveness against B. dendrobatidis in which species of conservation concern tended to have MIC equivalents cm−2 <0.02 (Woodhams et al., 2006b). It should be pointed out that the x-axis of Fig. 4 of that study should be labeled 0–0.08, and was erroneously labeled in publication.
In our experiments, species with better skin peptide defenses had lower infection intensity and higher survival rates (Table 1, Fig. 3). In Lit. caerulea that died of chytridiomycosis, we observed an infection intensity within the range previously reported by others (mean=94.3 sporangia mm−1, Berger et al., 2005b). There may be a species-specific threshold of infection intensity that leads to mortality and thus determines the length of time survived after exposure to B. dendrobatidis (Carey et al., 2006). Lim. tasmaniensis may also reduce infection loads by shedding infected skin layers more efficiently than other species; this may explain the lack of stratum corneum on many histological sections of the toes of this species. This result is not unexpected, given that intermolt period varies among amphibian species (Duellman & Trueb, 1986; Weldon & du Preez, 2006), and infection intensity can be reduced after molt (Berger et al., 2005b). Excessive skin shedding is a common clinical sign of chytridiomycosis (Nichols et al., 2001; Parker et al., 2002; Berger et al., 2005a,b) and may be a host mechanism to reduce infection loads or a manipulation of the host to increase keratinized substrate for the pathogen. In future studies, a comparison of the rates of shedding and growth rate between infected and uninfected individuals and across species could help determine the effects of shedding on the outcome of infection.
Other potential sources of variation among species include the distribution of granular glands. Some species have granular glands evenly distributed on all skin surfaces while others may have areas of concentrated skin glands (Duellman & Trueb, 1986; Bevins & Zasloff, 1990; Lacombe et al., 2000; Warburg et al., 2000; Apponyi et al., 2004). Antimicrobial peptides may spread across the skin surface within the mucous layer. However, the protective mucous layer may be thin or missing from some skin surfaces. The ventral surfaces of the toes of Lit. caerulea lack granular glands (Berger et al., 2005). Both the toes and inguinal region of anurans may be particularly vulnerable to water-borne infectious zoospores of B. dendrobatidis that can be found on moist substrates (Berger et al., 2005; Lips et al., 2006; Puschendorf & Bolaños, 2006; Weldon & du Preez, 2006). Amphibian species vary in other aspects of skin physiology including structure (Schwinger, Zanger & Greven, 2001) and the presence of lipid glands (Warburg et al., 2000). It is unknown whether species vary in their ability to release defensive peptides onto the skin surface in response to stressors including norepinephrine or infectious agents.
Changes in circulating leukocytes following infection with B. dendrobatidis
Amphibians have well-developed immune systems including both adaptive and innate immune responses (Carey et al., 1999; Rollins-Smith, 2001). Adaptive immunity to B. dendrobatidis has not been reported in the literature. Innate immune function in amphibians is not thoroughly understood (Boman, 1995). For example, little is known about leukocyte population responses to infection in amphibians (Wright & Whitaker, 2001). The typical amphibian response to stressors includes increased neutrophil counts and decreased lymphocyte counts (Bennet & Daigle, 1983; Cooper et al., 1992; Carey, Maniero & Stinn, 1996; Maniero & Carey, 1997). However, Lit. chloris did not show this response in this study and the response did not vary with the thermal regime including a 2 days cold shock. Frogs infected with the amphibian chytrid exhibited relatively fewer circulating neutrophils and eosinophils, and more basophils, than uninfected frogs. This response is similar to the one found in frogs with red leg disease attributed to Pseudomonas hemophila (Kaplan, 1952), which showed decreases in neutrophils and eosinophils and an increase in lymphocytes. Insufficient blood volume precluded the total cell counts, but one explanation of these proportional changes in circulating granulocytes may be that cells were moving toward infected areas. Weldon & du Preez (2006) suggest that the tubercles on the feet and toes of some frogs that are heavily keratinized and likely to contact infectious zoospores are often subjected to abrasion. Recruitment of leukocytes to these areas and inflammatory responses around microscopic ulcerations and lesions common with chytridiomycosis may be important for an adaptive immune defense (Nichols et al., 2001; Parker et al., 2002; Berger et al., 2005). Some innate defense peptides including amphibian temporin may be involved in chemoattraction of leukocytes (Chen et al., 2004; Kurosaka et al., 2005). The bradykinin-like peptides secreted from amphibian skin (Pukala et al., 2006; Suzuki et al., 2007) may act to increase vascular permeability, but this has not been tested in amphibians. These peptide defense mechanisms may be particularly important for skin wounds, although further study is needed to uncover whether amphibian skin peptides can signal an effective adaptive immune response against B. dendrobatidis at areas of skin abrasion.
It is clear from our study that innate immune defenses are an important component of resistance to chytridiomycosis and may partially account for variation in susceptibility among species. Skin peptides may be more essential in immune defense than was previously considered, although little is known about their mode of operation in vivo or their role in the microbial environment of the skin. We suggest that species-specific innate immune defenses are important determinants of infection intensity and subsequent morbidity or resistance to chytridiomycosis.
The authors are grateful for the assistance of T. Langkilde, S. Bell, B. Mott, S. Reilly, L. Reilly, B. Cullen, J. Whittier and D. Cochrane. Thanks are due to volunteers and friends of the Amphibian Research Centre, including C. Steele, R. Hobbs, S. Barber, D. Taylor, T. Dornum, N. Nolas, M. Velik-Lord, A. Crigo, N. Evangelou, P. Oliver, A. Marantelli and P. Marantelli. The authors thank the CSIRO amphibian disease research team, including L. Berger, A. Hyatt, V. Olsen and D. Boyle. Thanks are due to R. Potter and K. Hoek for helpful comments on a previous version of this manuscript. Ethical approval for all experiments was granted from James Cook University (A451 and A699_01) and the Amphibian Research Centre. Subcontracts to R.A.A. and L.R.S. from Integrated Research Challenges in Environmental Biology (IRCEB) grant IBN-9977063 (J.P. Collins, P.I.) and grant IOB-0520847 (L.R.S., P.I.) from the US National Science Foundation funded this research. D.C.W. was partially supported by an NHBLI Immunology of Blood and Vascular Systems Training Grant 5T32 HL069765-05 (J. Hawiger, P.I.).