Age-Dependent Variation in Metalloproteinase Activity of Isolated Human Bruch’s Membrane and Choroid

Abstract

purpose. To characterize and determine the effect of aging on the matrix metalloproteinase (MMP) component of the extracellular matrix–remodeling mechanism of isolated human Bruch’s–choroid.

methods. Immunohistochemical techniques and western blot analysis were used to detect and localize various members of the MMP family of proteolytic enzymes in the Bruch’s–choroid complex. Gelatin substrate zymography was used to detect and quantify the levels of MMP-2 and -9 in homogenates of Bruch’s–choroid from both macular and peripheral regions of the human fundus. Aging alterations in these enzymes were quantified by densitometric analysis of photographic negatives of the zymography gels.

results. Intact preparations of Bruch’s–choroid showed the presence of inactive forms of two gelatinases (MMP-2, 65 kDa, and MMP-9, 92 kDa), interstitial collagenase (MMP-1, 52 kDa) and stromelysin (MMP-3, 57 kDa). MMP-1 and -3 were localized primarily to Bruch’s membrane. MMP-9 was distributed evenly in Bruch’s membrane with some patchy presence in the choroidal mass. Distribution of MMP-2 was similar to that of MMP-9, but the staining in Bruch’s was much fainter. On gelatin zymography, an active form of MMP-2 (58-kDa species) was frequently observed in peripheral samples but only occasionally in macular regions. The levels of MMP-2 and -9 increased with aging in both the macular and the peripheral regions of the fundus (P < 0.05). MMP-2 levels were lower in macular regions than in the periphery but no such variation was observed with MMP-9. Both these inactive gelatinases could be activated in vitro.

conclusions. A matrix-degrading mechanism essential for extracellular remodeling was shown to be present in Bruch’s membrane. In macular regions, increasing levels of inactive forms of metalloproteinase and scarcity of active forms of MMP-2 suggests possible involvement of impaired extracellular degradation in both aging and macular degeneration.

Bruch’s membrane is a pentalaminated extracellular matrix ¹ (ECM) allowing bidirectional diffusion pathways between the retinal pigment epithelium (RPE) and the choroidal blood supply. Aging is associated with progressive thickening ² due to deposition of matrix components and membranous debris rich in lipids. ^{3 4 5} A consequence of the aging process is an exponential decline in the hydraulic conductivity of Bruch’s membrane. ^{6 7}  

Little is known of the underlying mechanisms of the aging process, but biophysical investigations have implicated a role for the homeostatic turnover of the ECM of Bruch’s membrane. ^{6 7} The observation that most of the decline in transport capacity occurs at an early age, in the absence of gross morphologic alterations, has led to the concept of a continuous remodeling process for Bruch’s membrane. Consequential changes in the structural framework leading to reduced transport may contribute to deposition and stabilization of lipid-rich debris in later life. The effect of such deposits on the normal ECM turnover process may culminate in both the observed aging of Bruch’s membrane and its transition to disease. 

Homeostatic ECM turnover is a delicate balance of coupled biosynthetic and degradative processes. Breakdown of the ECM is mediated by a family of Zn²⁺-dependent enzymes called matrix metalloproteinases (MMPs). These enzymes are released as inactive zymogens and on activation are capable of digesting all components of the ECM. ^{8 9 10 11} Secretion of MMP-1, -2, -3, and -9 has been demonstrated in cultured RPE and choroidal endothelial cells, and thus the machinery for modeling of Bruch’s membrane is present in geographically appropriate compartments. ^{12 13 14 15} Presence of these enzymes has not been demonstrated previously in Bruch’s membrane. The potent proteolytic activity of MMPs is restrained by the presence of tissue inhibitors of metalloproteinases (TIMPs). Both TIMP-2 and -3 have been found in Bruch’s membrane, and TIMP-3 is thought to be a normal component of this ECM. ^{14 16} The degree of ECM breakdown is therefore controlled by the temporal release of MMPs and their inhibition by TIMPs. ^{8 9 10 11}  

Abnormalities in either the biosynthetic or degradative pathways for ECM turnover have the potential to alter morphologic and functional characteristics of Bruch’s membrane. In Sorsby’s fundus dystrophy, a mutation in the TIMP-3 gene ¹⁷ is associated with lipid-rich deposits on the inner aspects of a thickened Bruch’s membrane. This rare disorder demonstrates the importance of tightly regulated ECM turnover for the normal maintenance and function of Bruch’s membrane. The mechanism by which mutant TIMP-3 leads to the observed pathophysiological course or its effects on transportation through Bruch’s membrane remain unknown. 

In age-related macular degeneration (AMD), the pathophysiological features of Bruch’s membrane are analogous to exaggerated aging with many similarities to Sorsby’s fundus dystrophy. ^{18 19} Although mutations in TIMP-3 have not been observed in AMD, ^{20 21} the large deposition and thickening of Bruch’s membrane suggests abnormal control of ECM turnover. Immunostaining intensity for TIMP-3 increases in the elderly, ¹⁶ and it is therefore possible that overexpression of TIMP-3 affects the degradative capacity for turnover of Bruch’s membrane. However, this apparent increased expression of TIMP-3 in the elderly may simply reflect the thickening of Bruch’s membrane. 

The control and regulation of the degradative arm of ECM remodeling has been shown to be complex, and knowledge of the system in Bruch’s membrane is rudimentary. The present investigation was therefore undertaken to identify the presence in Bruch’s membrane of MMPs known to be released by the RPE and choroid, to assess the potential for activation and to quantify the effect of aging on the gelatinase component (MMP-2 and -9) of the MMP family of enzymes. Quantitative studies on Bruch’s membrane alone are difficult, if not impossible. The membrane cannot be isolated from the underlying choriocapillaris in a consistent form, because of the presence of intercapillary columns on the outermost aspects of Bruch’s membrane. In the present investigation we therefore used the intact Bruch’s–choroid complex to quantify enzymatic activity. Macular samples were used principally for determination of aging changes in gelatinase activity. Other studies of western blot analysis and MMP activation required larger quantities of tissue and were performed on Bruch’s membrane from the peripheral regions of the fundus. 

Methods

Tissue Preparation

Thirty-two pairs of normal human eyes (age range, 17–82 years) were obtained from the Bristol Eye Bank (UK), and one eye from each donor was available for this study. Subsequent to removal of corneas for graft surgery, the eyes reached our laboratory on ice in saline-moistened sterile containers within 48 hours of death. The globes were hemisected by a circumferential incision around the pars plana, and the anterior portion, lens, and vitreous were discarded. In the absence of gross disease, macular and peripheral regions of the fundus were sampled by removing full-thickness trephines. For immunohistochemical analyses, 6-mm diameter macular samples were obtained from two eyes and 4-mm diameter peripheral samples from six eyes. Four 6-mm diameter trephines per eye were removed from the temporal periphery of four eyes for western blot analyses. Two 6-mm diameter trephines were obtained from the macular and nasoperipheral regions of one eye of 22 donors for zymographic quantification of MMP-2 and -9 and construction of an activity age profile. MMP activation studies were performed on three 5-mm diameter trephines from the temporal periphery of each eye of four donors. The remaining eight macular samples were used in other biophysical investigations. 

Immunohistochemistry

Full-thickness macular and peripheral trephines of retina and choroid were fixed in 4% paraformaldehyde, embedded (Tissue-Tek; Miles, Elkhart, IN), and frozen in isopentane, precooled in liquid nitrogen. Cryostat sections (7-μm-thick) were obtained on gelatin-subbed slides and after pretreatment with 0.5% blocking reagent (Boehringer Mannheim, Mannheim, Germany), were incubated overnight at 4°C with primary mouse monoclonal antibodies against MMP-1, -2, -3, and -9 at a concentration of 10 μg/ml (MMP/TIMP Antibody Sampler Kit; Calbiochem, Cambridge, UK). Subsequent procedures were identical with those outlined by Limb et al., ²² except that in this study, a monoclonal secondary antibody was used (rabbit anti-mouse immunoglobulins; Dako, Glostrup, Denmark). Negative control samples were prepared by omitting either the primary antibody or secondary antibody and incubating in 0.5% blocking reagent. 

Immunoblots of Western Transfers from Sodium Dodecyl Sulfate–Polyacrylamide Gel Electrophoresis

For western blot analysis, four 6-mm peripheral trephines from each eye were pooled and homogenized with 50 μl Tris-HCl buffer (50 mM Tris-HCl, 10 mM CaCl₂, and 0.25% Triton X-100[ pH 7.4]) followed by centrifugation at 9000 rpm for 30 minutes at 4°C. A 20-μl aliquot of supernatant was activated by incubation with 1 mM aminophenylmercuric acetate (APMA) for 60 minutes. Treated and untreated samples were subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) with the buffer system of Laemmli, ²³ using 8% or 10% gels. Proteins were then transferred electrophoretically to nitrocellulose sheets. The resultant blots were performed with anti-MMP-1, -2, -3, and -9 monoclonal antibodies (MMP/TIMP Antibody Sampler Kit, Calbiochem) at a concentration of 1.5 μg/ml in 1% milk for 2 hours. The blots were then incubated with sheep anti-mouse secondary antibody conjugated to horseradish peroxidase (1:1000 dilution; Amersham, Amersham, UK) for 1 hour followed by exposure to enhanced chemiluminescence detection reagents (Amersham) for 1 minute. Blots were then processed for autoradiography. 

Zymography for Metalloproteinase Activity

Tissue samples consisting of 6-mm diameter full-thickness trephines from macular and peripheral regions of each eye were transferred to phosphate-buffered saline (PBS; Sigma, Poole, UK) and the retina gently peeled away. Exposed RPE cells were removed by gentle brushing with a fine sable hairbrush. Using a dissection microscope, the Bruch’s membrane–choroid complex was carefully separated from the underlying sclera and stored at −40°C until used. 

Each sample was homogenized in 30 μl Tris-HCl buffer and after centrifugation, the supernatant was adjusted to 0.522 mg/ml protein with Tris-HCl buffer. The 10% SDS-PAGE procedure was identical with that described for immunoblotting, except that these gels contained 0.1% gelatin substrate. Each gel also included one lane of reference gelatinase activity ²⁴ and 0.1 ng human MMP-2/MMP-9 (Chemicon International; Temecula, CA). Subsequent incubation and staining procedures were as previously described. ¹⁵  

Specific gelatinases were identified by their respective molecular weights. Gels were then photographed on 5 × 4-cm cut film (Ortho Plus black-and-white copy film; Ilford, Basildon, UK) and developed (Ilfotec; Ilford) for 3 minutes at 20°C. The resultant negatives were scanned on a laser densitometer (LKB Ultrascan; Pharmacia Biotech, St. Albans, UK). Gel-to-gel variation due to degree of destaining was controlled by incorporating an internal gelatinase standard. The integrated sample area of each MMP band was divided by that of the MMP standard running in each gel. This ratio was multiplied by the activity of the MMP standard to obtain activity of specific MMPs in individual tissue samples. Such quantification of activity is often difficult, because the relationship between the densitometric absorbance and the concentration of protein is usually nonlinear. However, this relationship approximated to a linear function over 0% to 0.1% gelatin substrate (Fig. 1) , and a quantitative analysis was therefore possible. Macular and peripheral fundus samples were processed from 22 donors, allowing the construction of an age profile for MMP-2 and -9. 

Activation, Reduction, and Alkylation of Gelatinase Proenzymes

MMPs are secreted as inactive proenzymes bound to their respective TIMPs. ^{8 9 10 11} Removal of the propeptide and TIMPs leads to activation. In SDS-PAGE, binding of SDS leads to partial activation without removal of propeptide. These inhibitory constraints can be chemically removed by APMA activation or by reduction and alkylation of TIMPs. ^{8 15} This ability to activate MMPs was assessed in peripheral samples of Bruch’s–choroid from two young (24 and 29 years) and two old (59 and 77 years) eyes. Three 5-mm Bruch’s–choroid trephines were obtained from each eye and pooled and homogenized in Tris-HCl buffer. After centrifugation at 4°C at 9000 rpm for 30 minutes, 80 μl supernatant was divided into four aliquots. To activate MMPs, one aliquot was adjusted to 1 mM APMA and incubated at 37°C for 60 minutes. To remove TIMPs, one aliquot was adjusted to 2 mM dithiothreitol and incubated at 37°C for 30 minutes. These reduced samples were then adjusted to 5 mM iodoacetamide and incubated at 37°C for another 30 minutes for acetylation. To assess the combined action of MMP activation and TIMP removal, one aliquot was subjected to both activation and reduction-alkylation. The unused aliquot served as a control. All samples were then analyzed by gelatin zymography. 

Results

Immunohistochemical Localization of MMPs in Human Bruch’s–Choroid

The results of immunohistochemical staining for MMPs were consistent among all six samples and are presented in Figures 2A 2B 2C 2D 2E 2F . Samples from both the macular and the peripheral regions of the fundus showed immunoreactivity against antibodies to MMP-2 and -9. MMP-2 was visualized as an uneven band over Bruch’s membrane but, occasionally, small localized regions of heavy staining were observed (Figs. 2A , macular, and 2C, peripheral regions). Intercapillary columns showed more intense staining, and a patchy choroidal presence was also observed. MMP-9 was distributed evenly along the entirety of Bruch’s membrane, and some patchy staining was also observed in the large choroidal mass (Figs. 2B , macular, and 2D, peripheral regions). Reactivity toward antibodies to MMP-1 and -3 (Figs. 2E 2F , respectively) was assessed only in peripheral samples and was shown to be present predominantly in Bruch’s membrane, with a punctate distribution and stronger staining of the columns between capillaries. 

Western Blot Analysis

The anti-MMP antibodies used in this study were capable of recognizing both active and inactive forms of their respective MMPs. Proteins extracted from Bruch’s–choroid from the peripheral fundus showed the presence of MMP-1, -2, -3, and -9 (Fig. 3) . All these enzymes were present in their inactive forms, determined by their molecular weights. APMA-mediated activation resulted in partial conversion to the lower molecular weight activated forms for MMP-2 and -9 (Fig. 3) . These were represented by the 58- and 84-kDa bands, respectively (Fig. 3) . By contrast, MMP-1 and -3 were not activated by APMA. 

Characterization of Gelatinases in Human Bruch’s–Choroid Samples

All gelatinolytic activities were inhibited by replacing Ca²⁺ in the zymography reaction buffer with 10 mM EDTA, confirming that the measured enzymic activities were those of the metalloproteinases. 

Gelatin zymography of human Bruch’s–choroid consistently demonstrated the presence of two major MMP enzymes of molecular weights 65 kDa (MMP-2) and 92 kDa (MMP-9) in both macular and peripheral regions of the fundus (Fig. 4) . A trace of a 58-kDa gelatinase species, an activated form of MMP-2 proenzyme, was regularly observed in the peripheral regions but only occasionally in the macular areas (Fig. 4B) . Active forms of MMP-9 were never observed in this study. The activities of both the MMP-2 and -9 proenzymes increased with increasing age of the donor (P < 0.05; Fig. 5 ). Activities were expressed per unit protein, but, because the total protein content of the 6-mm trephine remained invariant with age (P > 0.2), the results could also be expressed in terms of unit area. Generally, activity of MMP-2 gelatinase was lower in the macular regions than in the periphery. No such regional differences could be detected for the MMP-9 gelatinase. 

A further two bands of gelatinolytic activity were detected at 225 kDa and 130 kDa (Fig. 4) . These may have been higher molecular weight forms (homo- and heterodimers) of MMP-9; both disappeared on activation. APMA activation resulted in a reduction in the molecular weights of both enzymes. The 92- and 65-kDa proteolytic activities were shifted to 84 and 58 kDa, respectively (Fig. 6) . An additional 61-kDa species was also observed. 

The presence of MMP-2 and -9 by western blot analysis was in agreement with the results of zymography, except that the 58-kDa active form of MMP-2, clearly identified on zymography, was not detected by western blot analysis. 

Discussion

This study has shown that interstitial collagenase (MMP-1), stromelysin (MMP-3), and two gelatinases (MMP-2 and -9) were present in human Bruch’s membrane, and that the level of the two inactive gelatinases increased with the age of the donor. Regional differences were apparent in the levels of the two gelatinases. The level of MMP-9 remained invariant, but that of MMP-2 was lower in the macular region than in the periphery. Given that the thickness of Bruch’s membrane increases with age and that of choroid decreases, it is likely that the observed increase in MMP levels occurs mainly in Bruch’s. Studies are in progress to subfractionate the Bruch’s–choroid complex using an excimer laser–based technique ²⁵ and to quantify the age-dependent alteration in both MMP activity and levels of TIMP-3. 

The present results clearly demonstrate the existence of MMP degradative mechanisms in Bruch’s membrane and strengthen the remodeling hypothesis for continuous turnover of this ECM. MMPs-1, -3, and -9 were present only as inactive forms. This was in marked contrast to MMP-2, of which a small amount of the 58-kDa active form was frequently detected in the peripheral regions but only occasionally in the macula. The active form of MMP-2 was observed on zymography but was undetectable on western blot analysis. This discrepancy may be intrinsic to the technique, in that zymography is dependent on enzymatic hydrolysis allowing activity amplification of the small amount of enzyme present, whereas the detection limit by immunoblotting is dependent on the quantity of protein. The presence of active MMP-2 in the periphery implies active remodeling and may explain why the decline with aging in hydraulic conductivity of Bruch’s membrane is less marked in peripheral regions than in the macula. ^{6 7}  

The origin of the various MMPs found in Bruch’s–choroid remains unknown. The three potential sources are RPE cells, choroidal cells, and plasma in the choroidal vessels. Cultured RPE cells have been reported to synthesize and secrete MMP-1, -2, -3, and -9 and TIMPs, ^{12 13 14 26 27} and these enzymes and their inhibitors have been shown to be incorporated into the interphotoreceptor matrix. ²⁸ Furthermore, cultured choroidal microcapillary endothelial cells and pericytes also have shown the ability to synthesize and secrete TIMPs, ¹⁴ and a number of studies have shown the presence of MMPs in plasma. ²⁹ There are two pathways whereby these enzymes may be incorporated into Bruch’s membrane. First, the enzymes may be released from plasma, RPE, and/or choroidal cells and then diffuse into Bruch’s membrane. This is certainly a possibility for the smaller molecular weight forms such as MMP-1 (52 kDa), MMP-2 (65 kDa), and MMP-3 (57 kDa), because the molecular weight exclusion limit for Bruch’s membrane is approximately 65 to 75 kDa. ³⁰ Second, release of MMPs may be coincident with the synthesis of structural components of Bruch’s membrane and therefore may be incorporated passively into the ECM of Bruch’s. Such a pathway would allow incorporation of higher molecular weight enzymes such as MMP-9. An observation in support of this hypothesis is the finding that levels of TIMP-3 correlate with the amount of ECM and in particular with excessive deposition such as drusen. ¹⁶ Thus the age-related increase in gelatinase activity of normal human Bruch’s–choroid may be caused by the increased deposition of various types of collagen and other ECM components. In other systems, MMP and TIMP expression is regulated by signaling from ECM receptors. ³¹ One study suggests that the integrity of Bruch’s membrane may serve to regulate RPE functions in MMP and TIMP secretion. ²⁸  

The mediators responsible for activation of MMPs in Bruch’s membrane remain unknown. Chemical modification by APMA and/or reduction and alkylation showed that MMPs of Bruch’s membrane retained the potential for activation. MMP-2 activation was associated with the formation of two active products, one of 58 kDa and a small amount of a 61-kDa species; these findings support those in a previous study. ³² Despite this potential for activation, endogenously activated enzymes were rarely observed in macular regions. It is likely that aging may limit access of mediators of activation to their progelatinase targets, because the observed aging decline in hydraulic conductivity ^{6 7} implies decreased porosity of the membrane. It is also possible that the substrate of these enzymes undergoes considerable age-related structural modification and may not be susceptible to proteolytic action. A number of studies have shown that human collagen susceptibility to collagenase is reduced with aging because of the increase of intermolecular cross-links within collagen fibrils. ^{33 34} It has been reported that there is a 50% decrease in collagen solubility of Bruch’s membrane between birth and 90 years of age in both the macular and the peripheral regions. ³⁵ These aging changes, together with deposition of lipids, proteins, and abnormal proteoglycans, ³⁶ are likely to limit access of both activators to their respective MMPs and MMPs to their substrates, leading to inefficient degradation and further accumulation of extracellular components. 

In conclusion, this study has demonstrated an age-related increase in the level of inactive gelatinases in Bruch’s–choroid and regular occurrence of active forms of MMP-2 in peripheral regions and their noted scarcity in macular regions. The former is likely to be associated with age-related thickening and deposition within Bruch’s membrane, whereas the latter may suggest diminished remodeling in macular regions with consequences for accumulation of basal laminar deposits associated with pathophysiological features in AMD. 

 Supported by grants from the National Eye Research Centre, the IRIS Fund, and the British Council.

 Submitted for publication January 13, 1999; revised May 25, 1999; accepted June 30, 1999.

 Commercial relationships policy: N.

 

Figure 1.

 

Quantification of MMP activity by zymography. A discreet gradient gel (1.0-mm thick) was prepared incorporating gelatin at concentrations of 0 to 0.5 g/100 ml. It was stained with Coomassie blue and destained in the manner described in the Methods section. After photography, the negatives were scanned and the absorbance of each gradient band plotted against its respective gelatin concentration producing the nonlinear curve shown. The normal working concentration of gelatin for zymography is 0.1 g/100 ml with enzymatic hydrolysis decreasing the concentration. A linear approximation was obtained between absorbance and gelatin concentration in the range 0 to 0.1 g/100 ml, and the integrated band area for specific MMPs could therefore be related directly to the amount of gelatin hydrolyzed. An MMP standard was also incorporated to correct for gel-to-gel variation in the degree of destaining.

Figure 2.

 

Immunohistochemical localization of MMPs in human Bruch’s–choroid. Cryosections of samples from the macular (A, B) and peripheral (C, D, E, F) fundus were immunostained with antibodies against various MMPs and visualized by the alkaline phosphatase–anti-alkaline phosphatase (APAAP) detection system. The illustrations are macular MMP-2 (A) and MMP-9 (B), peripheral MMP-2 (C) and MMP-9 (D), and peripheral MMP-1 (E) and MMP-3 (F). The reddish purple staining indicates a positive reaction. Negative control specimens with primary or secondary antibodies omitted did not stain (not included). Scale bars, 30 μm.

Figure 3.

 

Western blot analysis of MMPs in human Bruch’s–choroid complex. Peripheral sample extracts were electrophoresed in 10% polyacrylamide gels for MMP-1, -2, and -3 and 8% gels for MMP-9. The immunotransfer membrane was probed with monoclonal anti-MMP-1, -2, -3, and -9 antibodies, followed by enhanced chemiluminescence detection. Note that all four MMPs in untreated tissue samples were present in their inactive forms. APMA treatment produced lower molecular weight–activated forms of MMP-2 and -9 but had no effect on MMP-1 and -3. APMA−, untreated extracts of Bruch’s–choroid; APMA+, extracts partially activated by APMA before electrophoresis.

Figure 4.

 

Zymogram analysis of gelatinases in human Bruch’s–choroid from the macular (A) and peripheral regions (B). Lane S: MMP-2 and MMP-9 zymography standards, 0.1 ng. Lanes M1 through M7, extracts from the macula (A); lanes P1 through P7, from periphery (B). Total protein of 5.22μ g was loaded in each well. Age range of donors in lanes 1 through 7 was 17 to 82 years—respectively, 17, 24, 46, 58, 68, 71, and 82 years. Note that band intensities generally increased with age of donor. The level of 65-kDa gelatinase was lower in the macular region (A) than in the periphery (B). A 58-kDa active form of MMP-2 could be seen regularly in the peripheral region (B), but only occasionally in the macular region (A). Other gelatinases were present at positions corresponding to molecular weights of 225 and 130 kDa.

Figure 5.

 

The effect of donor age on the levels of 65-kDa gelatinase (A) and 92-kDa gelatinase (B) in macular and peripheral regions of 22 donor eyes. Both MMP-2 and MMP-9 increased with the age of the donor (P < 0.05) in both the macular and the peripheral regions. Note that the level of 65-kDa gelatinase was lower in the macular region than in the periphery (A), and no regional differences were apparent in the activity of the 92-kDa gelatinase (B).

Figure 6.

 

The effect of APMA activation, reduction, and alkylation on gelatinase activity of human Bruch’s–choroid. Sample extracts were activated by APMA (+APMA), and/or reduced and alkylated (+R/A) before zymography. APMA activation converted latent 92-kDa gelatinase (MMP-9) to an active 84-kDa form and 65 kDa (MMP-2) to active 61- and 58-kDa forms (lane 2). Reduction and alkylation alone did not affect activity of the latent gelatinases (lane 3) but enhanced APMA activation (lane 4).

The authors thank Ann Patmore, Roy Whiston, and Phillip Eaton for technical assistance.