The elastic fiber plays an important mechanical role in extensible tissues such as muscular blood vessels, lung, and skin (for reviews, see Rosenbloom et al., 1993; Dietz and Mecham, 2000; Starcher, 2000). The proteins elastin, fibrillin 1 and 2, and microfibril-associated glycoproteins (MAGPs) 1 and 2 are known to be components of the elastic fiber. Fibrillin 1 and 2 monomers are assembled in the ECM to form fibers by mechanisms that are not fully understood. MAGP-1 and MAGP-2 have been immunolocalized to fibrillin-containing microfibrils; however, the function of these two proteins is entirely unknown. Fibrillin-containing microfibrils are thought to serve as a framework for the deposition of elastin and thus forming the elastic fiber superstructure (Rosenbloom et al., 1993; Kielty and Shuttleworth, 1995). Fibrillin-containing fibers are also found without elastin in some tissues (Gibson and Cleary, 1987; Wright et al., 1994; Zhang et al., 1995).
Mutations in fibrillin 1 and 2 genes have been shown to produce proteins that cause a dominant negative disruption of ECM microfibril assembly and result in Marfan syndrome (MFS) and congenital contractural arachnodactyly (CCA), respectively (Abraham and Perejda, 1982; Lee et al., 1991). The phenotypic spectrum of both genetic disorders is variable; however, individuals with MFS typically have joint laxity of the hips, elbows, and knees. CCA results in congenital contractures of the hands, feet, elbows, and knees, and this suggests a role for fibrillin 2 in the flexor tendon specifically. Importantly, contractures are not a typical phenotype of MFS and only appear in the most severe presentation of the disease—neonatal MFS. The disruption of these tissues in MFS and CCA provides genetic evidence of an important role for these proteins in connective tissue. The etiology of the malformations that result from fibrillin 1 and 2 mutation is unknown.
Little is known about the localization of elastic fiber proteins or their role within ligaments or tendons. While elastin has been studied in ligaments by chaotropic extraction of whole tissue (Ross and Bornstein, 1969) and electron microscopy (Gross, 1949) for decades, an analysis of elastic fiber distribution in the ligament or tendon has not been reported. Fibrillin 1 was historically known as oxytalin fibrils in early electron microscopy studies, until the protein was identified and named by Sakai et al. (1986). Sakai et al.'s publication is the only report of which we are aware that identifies fibrillin 1 in the tendon by immunohistochemistry, although which tendon was not specified. Due to the congenital contracture phenotype of CCA, it has been assumed that fibrillin 2 was a component of tendons and ligaments; however, to our knowledge, the protein has not yet been formally identified immunohistochemically in these tissues. Both MAGP-1 and MAGP-2 have been identified in the bovine nuchal ligament and Achilles tendon (Gibson et al., 1986, 1998). While it is thought that MAGP-1 and -2 primarily associate with the fibrillins, it is not known if they are also incorporated into the ECM independent of fibrillin-containing microfibrils.
Each tendon and ligament has a unique mechanical job to accomplish, and therefore the anatomy can vary greatly. We have chosen to examine the role of elastic fiber proteins in the canine flexor digitorum profundus (FDP) tendon because it contains three functional regions that are also found in many other tendons and ligaments: 1) a tendon insertion (IS)-to-bone region, 2) an avascular/tensional (AV/T) region, and 3) a fibrocartilagenous (FC) region (Fig. 1) (Evans and Christensen, 1979; Vogel and Koob, 1989). The AV/T region primarily experiences axial loading, while the FC region is formed in response to compressive force (Ploetz, 1938; Vogel and Koob, 1989; Benjamin and Ralphs, 1998).
Herein we have determined the locations of elastin and the elastic fiber-associated proteins fibrillin 1, fibrillin 2, MAGP-1, and MAGP-2 within the distinct regions of a digital flexor tendon in an effort to gain information about the functional roles of these proteins in the tendon and ligament. These data provide new knowledge of the varied mechanical microenvironments of the tendon. Additionally, differences in the distribution of fibrillins 1 and 2 and MAGP-1 and -2 provide intriguing insights into functional variation.
MATERIALS AND METHODS
Sample Collection and Histological Preparation
The FDP tendon was harvested from 20- to 30-kg adult mongrel canines according to a protocol approved by the animal studies committee of Washington University. Tendons were excised at the tendon-bone junction and proximal to the first annular pulley. Samples were fixed in 10% buffered formalin for 24–36 hr at 4°C and embedded in paraffin blocks, and 5-micron sections were cut. Slides were baked overnight at 55°C, de-paraffinized in xylene, and rehydrated into phosphate-buffered saline (PBS). For general structure, hematoxylin and eosin (H&E) staining was performed by standard techniques. Resorcin fuchsin stain (Poly Scientific, Bay Shore, NJ) was used for staining elastic fibers.
Fibrillin 1 antibodies.
FBN1-Pro is a rabbit polyclonal antibody made against the unique proline-rich domain of human fibrillin 1 and has been previously described (Trask et al., 1999). Antibodies to FBN1, FBN2, MAGP-1, and elastin were generous gifts from Dr. Robert Mecham, Washington University, St. Louis, MO. FBN1-CT is a rabbit polyclonal antibody directed toward the carboxyterminal domain of human fibrillin 1 and has been previously described (Ritty et al., 1999).
Fibrillin 2 antibodies.
FBN2-Gly is a rabbit polyclonal antibody made against the unique glycine-rich region of human fibrillin 2.
NLR3 is a rabbit polyclonal made against amino acids Met-1 to Thr-1114 of human fibrillin 2 (Trask et al., 1999). Neither antibody reacts against analogous regions of recombinant fibrillin 1 proteins by Western blot. MAGP-1 is a rabbit polyclonal antibody made against the human sequence and has been previously described (Robb et al., 1999). The MAGP-2 antibody is a rabbit polyclonal antibody made against an amino terminal region that shares little similarity with MAGP-1 and was a generous gift from Dr. Michael Shipley, Washington University, St. Louis, MO. Western blotting has demonstrated the specificity of the two MAGP antibodies (data not shown). The elastin antibody was made against a bacterially expressed human elastin sequence. A rabbit anti-human antibody was used to detect fibronectin (Sigma, St. Louis, MO).
De-pariffinized sections were blocked with bovine serum albumin and nonimmune serum and then stained with the primary antibodies described above for 45 min at room temperature. After washing, the primary antibodies were detected with either chicken anti-rabbit Alexa 488 or 594 (Molecular Probes, Eugene, OR). Parallel controls without primary antibody were carried out with chicken anti-goat and chicken anti-rabbit secondary antibodies and did not show reactivity with canine epitopes. Some slides were also stained with Hoescht 33258 to visualize nuclei.
A Zeiss Axiophot microscope mounted with a Zeiss AxioCam HR digital camera was used to capture 14-bit 1300 × 1030 pixel TIF images.
Desmosine and Hydroxyproline Content
Digital flexor tendons were collected from four adult canine mongrels, as described above. For each tendon, one sample was collected from areas 1–6, as indicated in Figure 1. All samples were frozen and then freeze-dried. The samples (3–10 mg) were placed in 1.5-ml microfuge tubes (secure-lock, Fisher Scientific, Pittsburgh, PA) with 500 μl of 6 N HCl and hydrolyzed for 24 hr at 100°C. The samples were evaporated to dryness and dissolved in 500 μl of water. Desmosine analysis was performed as described previously (Starcher and Conrad, 1995) using 25 μl of sample hydrolysate. Hydroxyproline was quantitated by amino acid analysis on 12.5 μl of sample, and protein was determined with 0.25 μl of sample using a ninhydrin assay procedure (Starcher, 2001).
To compare desmosine and hydroxyproline content variation by anatomical zone, one-way analysis of variance was performed between the sets of four samples taken from each area of measurement.
The locations of elastic fiber-associated proteins within various functional areas of the FDP tendon are described below, and only representative images are shown. As the fibrillins, MAGPs, and elastin frequently share similar tissue distribution and macromolecular assemblies, immunohistochemistry often yielded staining that was similar in appearance for these five proteins. The images were chosen to limit redundancy and yet to highlight both similarities and differences from area to area and between the proteins themselves. Figure 1 is a cartoon depicting the three functionally distinct regions of the FDP tendon and the six areas that were evaluated immunohistochemically and biochemically. Table 1 summarizes the immunohistochemical findings, and Table 2 summarizes the biochemical data.
Table 1. Summary of immunohistological findings
This table summarizes the immunohistological findings for five proteins in transverse and longitudinal sections from six areas (Fig. 1) in the FDP tendon. (++++) indicates the strongest staining observed, (+) indicates slight staining, and (–) no staining observed. To distinguish between internal and external layers and between different staining appearances the following notes are added: 1. Fibrillar, thick and oriented along axis of tendon. 2. Pericellular around linear cell arrays of the interior. 3. Outer cells layers of dorsal and palmar aspects. 4. Distributed throughout. 5. Fibrous, thin, wispy and with random orientation.
Within the FC zone
Palmar to the FC zone
Proximal to the FC region
Table 2. Desmosine and hydroxyproline content of FDP tendon regions
Source of samples
pM Des/mg prot
μg OHP/mg prot
pM Des/μM OHP
Samples from four tendons were analyzed in parallel for hydroxyproline and desmosine concentration per milligram of total protein and the results expressed as mean +/− one standard deviation. The anatomical areas from which the samples were taken correspond to Figure 1. One way analysis of variance was performed to compare the groups as follows: The vincular membrane contains approximately 11X more elastin as compared to all other samples, (*P < 0.001). The tendon near the insertion site had 2X more elastin than the rest of the tendon, excluding the elastic anchor area 3, (**P = 0.005). The elastic anchor area contained 4-5X more elastin than any other area of the tendon body, (***P < 0.01). Hydroxyproline concentration did not differ significantly among the areas measured (P = 0.26).
873 ± 607 (*)
3.18 ± 0.28
283 ± 207 (*)
62 ± 13 (**)
3.13 ± 0.13
19.8 ± 3.6 (**)
199 ± 137 (***)
3.70 ± 0.73
56.8 ± 42.1 (***)
38 ± 10
3.23 ± 0.10
11.5 ± 3.1
33 ± 0.6
3.33 ± 0.26
10.0 ± 0.8
Proximal to FC region
44 ± 8.0
3.33 ± 0.05
12.8 ± 2.2
In the canine FDP tendon, and in several other quadrupeds (Vogel and Koob, 1989), a FC zone is located dorsally beneath the metacarpel-phalangeal joint and occupies approximately half of the tendon diameter (Figs. 1 and 2a). For this reason, it is useful to distinguish the FC zone from the region of the tendon within which it is contained. We evaluated the FC zone of the FDP tendon as well as the areas immediately proximal, palmar, and distal. Similar to the AV/T region (Fig. 1), the FC region also does not have a vascular supply. Proximal to the FC zone in area 6, staining for each of the five elastic fiber proteins revealed a similar peritendonous circumferential distribution that was most prominent superficially and lessened moving into the tendon body. By longitudinal section, large fibers were visible. Transverse sections revealed highly defined areas of strong staining for the fibrillins, MAGPs, and elastin. We interpret this pattern as the transection of elastic fibers running longitudinally in the direction of force. These fibers were largely confined to the outermost cell layers (not shown).
Elastin was not detectable within the FC zone; however, it was quite prominent, as were the other microfibrillar proteins, within the first few layers on the palmar side of the tendon opposite (but not within) the fibrocartilage zone (Fig. 2b). Transverse sections of the palmar aspect of the FC region demonstrated the same patchy appearance seen in area 6 that probably represents the transection of longitudinally running elastic fibers (Fig. 2c). MAGP-1 and -2 and fibrillins 1 and 2 all demonstrated the same pattern in the superficial layer palmar to the FC zone.
Interestingly, within the FC zone, MAGP-1 displayed a staining pattern that was different from MAGP-2 or the fibrillins (Fig. 2d–f). MAGP-1 demonstrated prominent staining that had a diffuse filamentous appearance by transverse and longitudinal sections. MAGP-2 and fibrillins 1 and 2 stained only pericellularly within the fibrocartilage, and no filamentous or fibrous pattern was seen. For these three proteins, pericellular staining was evident by both longitudinal and transverse sections, and not evenly distributed throughout, but restricted to certain cell populations. Similar pericellular staining was observed in several other areas along the length of the tendon and appeared to be specific for the immediate pericellular ECM, but not the cells themselves. Pericellular staining was most often observed around internal tendon fibroblasts in linear arrays within the tendon body.
Overall, the AV/T region of the tendon had the lowest staining for the elastic fiber proteins, compared to the rest of the tendon. Staining was the most prominent within a thin area just at the outermost cell layers, and antibodies to the five elastic fiber proteins produced generally similar patterns in this area (Fig. 3a). Fibrous MAGP-1 and -2 staining was visible at the tendon surface and was most prominent on the dorsal and palmar aspects. Here, as with the FC zone, there was pericellular staining within certain populations of the tendon interior for fibrillin 1 and especially fibrillin 2, but not elastin (Fig. 3b). The fibrillin 2 staining was fairly restricted to these interior cells and stained little at the outer layers of the tendon, where fibrillin 1 staining was prominent.
By transverse section, MAGP-1 and -2 staining of interior cell populations revealed elaborate interfascicular ECM that was contiguous between distant cells (Fig. 3c), although we did not determine if the cell membrane extends through these spaces. The connections were only apparent within restricted areas that were generally deeper layers in the lateral regions of the tendon body.
Insertion Site Region
We have focused on two areas within this region. The first is a vincular membrane and its interface with the tendon that runs along the dorsal aspect from the insertion site 2 cm proximally (Fig. 1, areas 1 and 3). The second area is the most distal centimeter of the tendon body nearest the insertion to bone (Fig. 1, area 2).
The dorsal vincular membrane is a conduit for small blood vessels that supply the tendon at the insertion site. Generally, the vessels remain peritendonous, although a few may penetrate the tendon body (unpublished observations). The membrane is thin and flat and forms a triangular shape as it runs longitudinally along the dorsal surface of the tendon. At the proximal end of the interface between the tendon and the membrane (Fig. 1, area 3) we have identified a high concentration of elastic fiber proteins within the tendon body. This appears to be a specialized structure—an elastic anchor—designed to deliver a large amount of elastic recoil and dampen the stresses at that point during articulation. Resorcin fuchsin staining for elastin of a continuous section that spans the tendon and continues into part of the vincular membrane reveals an abundance of large elastic fibers by light microscopy (Fig. 4a–c). Not surprisingly, each antibody to elastic fiber proteins stained the elastic anchor strongly (Fig. 5a).
The abundance of elastic fiber proteins seen in the elastic anchor continues throughout the vincular membrane. Resorcin fuchsin staining as well as immunofluorescent staining for MAGP-1 and -2, fibrillin1 and 2, and elastin demonstrated the presence of relatively thick fibers running in register (Fig. 5b–f). It appeared that the direction of the fibers was essentially parallel to the outer edge of the triangular membrane.
The last centimeter of the tendon near the insertion site was also very rich in elastic fiber proteins. This region was higher than all other areas of the tendon body with the exception of the elastic anchor (Table 1, Fig. 6). Overall, staining for the fibrillins and MAGPs was greatest on the dorsal aspect within 1–2 cm of the insertion site, which is the interface with the vincular membrane, but was also plentiful along the palmar aspect for 2–3 cm from the insertion. Throughout the interior of the tendon, staining for fibrillins 1 and 2 was prominent and fibrous in appearance around the linear cell arrays and was even more abundant within the dense cell layers that surround the central collagen core. Moving proximally, the pericellular distribution of the two fibrillins diverged as fibrillin 1 was localized near the tendon surface while fibrillin 2 was more prominent in the interior, although still present in the fibers on the surface. Elastin was not detected in the interior pericellular regions that were positive for the fibrillins and MAGPs.
The MAGP-1 and -2 staining pattern was also widespread and intense and had a distribution similar to the fibrillins. Interestingly, throughout the rest of the tendon, the staining intensity of MAGP-1 and MAGP-2 was approximately equal; however, near the insertion site MAGP-2 had greater staining intensity than MAGP-1 and was the highest degree of MAGP-2 staining anywhere in the tendon. MAGP-2 was quite prominent along the linear cell arrays between the collagen fascicles of the tendon interior (Fig. 6d).
Elastin and Hydroxyproline Content
Desmosine cross-links are unique to elastin; thus desmosine content can be used as a measure of mature elastin content. Biochemical analysis of the desmosine content in transverse sections from six areas (Fig. 1) of the FDP tendon confirms and extends the immunofluorescent findings (Table 2). The vincular membrane had the highest elastin concentration of any of the areas evaluated (P < .001). Minor vascularity was present in the membrane; however, the fraction of vessel contribution to total elastin content could not be determined. Within the tendon body, the elastic anchor (vincular membrane anchor, Fig. 1, area 3) was four to five times higher in elastin concentration than the four other areas of the tendon body (P < .01). This finding is impressive in that the desmosine content was determined on the total cross section, and histology shows that most of the elastin is in one small area relative to the total area of the section; therefore, the desmosine per milligram of protein measurement would be diluted by the total protein content of the rest of the section. The high variation in the vincular membrane and elastic anchor samples was the result of difficulties in the collection of analogous regions between tendons due to the small dimensions of these areas. As seen histologically, the insertion site (Fig. 1, area 2) contained more elastin than the avascular region or the FC region (Fig. 1, areas 4–6; P = .005). The hydroxyproline content demonstrated no significant differences throughout the tendon (P = .26). Because of the consistency of collagen content, the ratio of desmosine to hydroxyproline mirrored the overall elastin content characteristics of each region.
We have determined the locations of elastin and elastic fiber-associated proteins within functionally distinct regions of the canine FDP tendon. We show that the FDP tendon is a tissue that contains fibrillins 1 and 2 and MAGP-1 and -2 (Table 1). Biochemical analysis of transverse sections from a vincular membrane and five distinct areas of the FDP tendon has demonstrated the presence of elastin in each area (Table 2).
Throughout the tendon, elastic fibers could be distinguished by resorcin fuchsin stain and by immunodetection in both transverse and longitudinal sections. The specific locations of elastic fiber proteins detailed in the “Results” section revealed consistent patterns. The elastic fibers were not dispersed uniformly, but were confined to restricted locations. Generally, elastic fibers were most prominent within the first few peritendonous cell layers of the dorsal and palmar aspects. The depth of the fibers varied between the different functional zones. Specifically, elastic fibers were confined to the outermost cell layer within the AV/T region, but were more abundant and present in deeper cell layers opposite (palmar to) the fibrocartilage zone, proximal to the FC region, and near the insertion site.
Wherever elastin was found, immunohistological detection of fibrillins 1 and 2 and MAGP-1 and -2 resulted in similar staining patterns. This is not surprising, as each protein is thought to be a potential component of the elastic fiber superstructure. However, the fibrillins are also known to be deposited without elastin in some tissues, e.g., kidney (Zhang et al., 1995) and ciliary zonules (Wright et al., 1994). Indeed, we also observed the fibrillins and MAGPs in areas of the digital flexor tendon where elastin was not detected, i.e., the interior of the tendon body.
It has been experimentally demonstrated that fibrillins 1 and 2 share overlapping expression in some tissues (Sakai et al., 1986; Zhang et al., 1994, 1995; Mariencheck et al., 1995), and the overlapping phenotypes of MFS and CCA provide similar genetic evidence (Milewicz et al., 2000). While co-localization of fibrillin 1 with fibrillin 2 often occurs, there are tissues in which expression does not overlap. For example, fibrillin 2 is thought to be exclusively expressed in auricular cartilage (Zhang et al., 1994), while fibrillin 1 may be the only fibrillin in the suspensory ligament of the lens (ciliary zonules) of the eye (Wright et al., 1994).
The data presented here demonstrate areas of overlapping fibrillin 1 and 2 distribution, as well as areas where one form is more prominent than the other. Within the tensional regions, such as the AV/T region, fibrillin 2 stained interior cells more prominently than those near the surface of the tendon, where fibrillin 1 was more evident (Fig. 3b and c). This pattern was also apparent within 2 cm of the insertion site (Fig. 6e and f). Fibrillin 1 seemed to be the predominant form in areas of elastic fiber, while fibrillin 2 was the predominant fibrillin in the pericellular matrix of the tendon internal fibroblasts. A phenotypic manifestation of CCA, which results from mutations in fibrillin 2, is flexion contractures of the hands, feet, elbows, and knees (Ramos-Arroyo et al., 1985). Similar contractures are not a common manifestation of MFS, although they may be present at the most severe end of the phenotypic spectrum—neonatal MFS. It may be that fibrillin 2 plays an important role in the proper organization and growth of flexor tendons during development. Its location in the pericellular ECM of tendon internal fibroblasts lends support to this hypothesis.
Transverse sections of the AV/T region revealed a continuous pericellular ECM when stained with antibodies for MAGP-1 and -2 that was interfascicular (Fig. 3c). Membrane dyes have been previously employed to reveal a similar pattern of cell-to-cell connections that extends through interfascicular ECM (McNeilly et al., 1996). Given the fibrillin and MAGP staining presented here, it is likely that the pericellular ECM that these cellular connections pass through contains fibrillins and MAGPs. This suggests a mechanism by which defective fibrillin 2 proteins, which are known to have a dominant negative effect on ECM assembly, might disrupt tendon pericellular ECM organization and could conceivably disrupt normal cell-to-cell signaling during tendon development. Thus, the congenital flexion contractures typical of CCA would occur not through a dominant negative disruption of elastic fiber assembly, but through a disruption of cell-to-cell communication during development.
The function of both MAGP-1 and MAGP-2 is unknown. Previous studies have shown that MAGP-1 and MAGP-2 associate with the fibrillins in the ECM (Henderson et al., 1996; Gibson et al., 1998); however, it is not yet clear if they appear in the ECM independent of the fibrillins. There is recent evidence that MAGP-2 cannot associate with the ECM using the same mechanism employed by MAGP-1 (Segade et al., 2002). Unexpectedly, transverse and longitudinal sections of the FC region stained for MAGP-1 revealed a heavy, fibrous staining pattern within the fibrocartilage zone (Fig. 2d and e). Staining of serial sections with antibodies to MAGP-2 and fibrillins 1 and 2 resulted in a different pattern in this region, as only a light pericellular stain was observed (Fig. 2f). These patterns suggest that within the FC zone only, MAGP-1 has an ECM distribution that may be independent of the fibrillins. Although this pattern was not present in any other area, care must be taken in the interpretation of these results, as differences could be due to variation in antibody sensitivities.
Biochemical analysis of desmosine content confirms and extends the immunohistological findings (Table 2). The vincular membrane had the highest desmosine content, and this was reflected histologically with elastin staining. The membrane appears to act as a conduit for capillaries serving the distal tendon, and the high elastin content may aid its integrity through the repeated cycles of stretch and recoil experienced in vivo. The intense resorcin fuchsin staining (Fig. 4) and immunofluorescent staining of the elastin anchor (Fig. 5a) were confirmed biochemically, and the high concentration of elastic fiber suggests a significant mechanical contribution to the interface between the tendon body and vincular membrane. As the distal phalanx articulates, the membrane is drawn taut along its edge (unpublished observations). This elastin-rich structure probably serves to buffer the force at this interface.
Identification of the specific elastic fiber locations within subregions of the tendon provides insight into the mechanical characteristics of these small zones. Throughout the tendon, elastic fibers were most visible along the dorsal and palmar surfaces. The only exception was the dorsal surface within the FC zone where elastin was not detected. This distribution parallels the directions in which the tendon deforms during normal use, i.e., palmar-dorsal and not lateral. Further, although elastin is present throughout the tendon, the two areas that have the highest concentration of elastin and elastic fiber proteins are also the areas that likely undergo the greatest flexion. The tendon at the insertion site must respond to changes in the angle of the force as the digit rotates in response to muscular contraction. Areas of axial tension such as the AV/T region do not have similar mechanical requirements and have lower elastin content. Similar to the insertion site, the surface of the tendon opposite (palmar to) the FC zone must be capable of larger strain deformations as the tendon wraps around the metacarpal-phalangeal joint. Here, too, elastic fiber staining is locally high (Fig. 2b). Thus it appears that the areas of the tendon that undergo the greatest flex also have the highest elastic fiber content.
It is clear that elastic fibers and associated proteins have an important role in tendon physiology. The presence of elastic fibers within specific subregions of the tendon provides insight into the mechanical characteristics of these areas. Many other tendons and ligaments also contain functional regions similar to those evaluated in this study, and the findings presented here may be broadly applicable. Differential distributions of MAGP-1 within the fibrocartilage and fibrillin 2 within the internal pericellular ECM suggest roles outside of the elastic fiber for these ECM proteins. The relative abundance of MAGP-2 near the tendon insertion to bone implies a greater functional role in this region and merits further investigation. Investigation of these tendon regions in fibrillin 2 and MAGP-1 and -2 null animals is now possible and should yield additional information about the function of these proteins.
We thank Jeremy Herzog for valuable technical assistance and Megan Burns for statistical support.