Direct Intramuscular Injection of Plasmid DNA Encoding Angiopoietin-1 but not Angiopoietin-2 Augments Revascularization in the Rabbit Ischemic Hindlimb
Background—Angiopoietin-1 (Ang1) and angiopoietin-2 (Ang2) have recently been identified as ligands for the endothelial cell–specific Tie2 receptor. Little is known regarding the impact of these Tie2 ligands on postnatal neovascularization. Accordingly, we tested the hypothesis that gene transfer of plasmid DNA encoding Ang1 and Ang2 could modulate collateral vessel development in a rabbit model of hindlimb ischemia.
Methods and Results—pAng1* (n=15), pJFE control (no Ang1* insert) (n=9), pAng2 (n=9), pcDNA3 control (no Ang2 insert) (n=10), or saline (n=5) was injected intramuscularly into the rabbit ischemic hindlimb. Collateral vessel development and limb perfusion were assessed before and 30 days after treatment. Calf blood pressure ratio (ischemic to normal hindlimb) was increased 30 days after Ang1* gene transfer versus controls (Ang1*, 0.90±0.02; pJFE, 0.76±0.05; saline, 0.77±0.03; P<0.05). Angiographic score was higher (P<0.05) in the pAng1* group (0.63±0.02) than in the pJFE (0.51±0.03) or saline (0.52±0.02) group. Maximal (postpapaverine) blood flow in the ischemic limb was higher (P<0.05) after pAng1* (67.8±4.9 mL/min) than pJFE (51.2±4.4 mL/min) or saline (52.9±4.9 mL/min). Capillary density and capillary/muscle fiber ratio (242±12/mm2 and 0.89±0.06, respectively) were higher (P<0.01) with pAng1* than pJFE (172±11/mm2 and 0.64±0.05) or saline (166±10/mm2 and 0.67±0.05). Neovascularization was not enhanced with pAng2.
Conclusions—Ang1 but not Ang2 gene transfer produces anatomic and physiological evidence of enhanced collateral vessel formation. Ang1 may modulate neovascularization in adult animals and thus represents a feasible therapeutic strategy for patients with tissue ischemia. The role of Ang2 in postnatal neovascularization remains to be clarified.
The Tie2 receptor, an endothelium-specific receptor tyrosine kinase, is required for integrity and survival of vascular endothelial cells.1 2 3 Angiopoietin-1 (Ang1), a 70-kDa secreted glycoprotein, has recently been identified as a ligand for the Tie2 receptor.4 Targeted disruption of the gene encoding for Ang1 causes severe vascular abnormalities in the developing mouse embryo, leading to death by embryonic day 12.5.5 These data thus imply a critical role for Ang1 in the development of the normal embryonic vasculature. Angiopoietin-2 (Ang2)6 has also been shown to function as a ligand, with similar affinity to Ang1, for the Tie2 receptor; in contrast to Ang1, however, Ang2 does not phosphorylate the Tie2 receptor but instead appears to represent a naturally occurring antagonist of Ang1.6 Transgenic overexpression of Ang2 disrupts blood vessel formation in the mouse embryo.6 The impact of Ang1 and Ang2 on postnatal neovascularization is unknown. Specifically, the effect of Ang1 and Ang2 on neovascularization in adult animals has not previously been evaluated.
Previous studies7 8 9 10 11 12 13 14 15 16 have established that skeletal striated muscle can take up and express foreign genes transferred as nonviral, covalently closed pDNA. Indeed, ischemic skeletal muscle acts to enhance transgene expression.8 14 When the ischemic muscles of the rabbit hindlimb are transfected with plasmid DNA that encodes for a secreted protein, augmented neovascularization may be achieved. Accordingly, we tested the hypothesis that gene transfer of plasmid DNA encoding Ang1 and Ang2 could modulate collateral vessel development in a rabbit model of hindlimb ischemia.
The Ang1* in pJFE plasmid (pAng1*) and Ang2 in pcDNA3 plasmid (pAng2) were generated by Regeneron Pharmaceuticals, Inc. Ang1* is a genetically engineered variant of Ang1 with a modified NH2-terminus and mutated Cys245 that is easier to express and purify but retains properties similar to those of Ang1 in all assays.6 The pJFE plasmid incorporates the SRα promoter, an ampicillin gene for selection in Escherichia coli, as well as the SV40 replication origin for transient expression in COS cells. The Ang1* insert in pAng1* is ≈1.5 kb. The pJFE control does not contain the Ang1* insert. The pAng2 plasmid consists of 1.5 kb of Ang2 cDNA inserted into pcDNA3. The expression of pAng2 and pcDNA3 is transcriptionally regulated by the CMV IE promoter, and they also contain the ampicillin resistance gene. The pcDNA3 control contains no Ang2 insert.
The extent to which angiopoietin may modulate neovascularization of ischemic tissues was investigated in 48 New Zealand White rabbits by use of a previously described model of hindlimb ischemia.17 18 All protocols were approved by St Elizabeth’s Institutional Animal Care and Use Committee.
Intramuscular Gene Transfer
A 25-gauge needle was advanced through a small skin incision and used to perform direct injection of plasmid DNA or to inject saline at 4 different sites in 3 major thigh muscles of the ischemic hindlimb. For each rabbit, 125 μg of pAng1* (n=15), pJFE (n=9), pAng2 (n=9), or pcDNA3 (n=10) per 0.5 mL of normal saline was injected at each of 4 sites (total, 500 μg/2.0 mL) as previously described.14 Additional controls received saline only.
Calf Blood Pressure Ratio
Calf blood pressure was measured in both hindlimbs as previously described17 immediately before (day 10) as well as 1 month after transfection (day 40).
Selective internal iliac angiography was performed as previously described.17 Quantitative angiographic analysis of collateral vessel development was used to derive an angiographic score for each film, defined as the ratio of grid intersections crossed by opacified arteries divided by the total number of grid intersections in the medial thigh. The internal luminal diameter of the main collateral (stem) artery issuing from the internal iliac artery was measured with a previously validated automated edge-detection method.19
In Vivo Doppler Flow Measurement
Blood flow was quantified in vivo before selective internal iliac angiography on days 10 and 40 with a 0.018-in Doppler guidewire (Cardiometrics) as previously described20 at rest and after bolus injection of 2 mg of papaverine (Sigma). Given the similarity in body mass for all rabbits used in this study (3.8 to 4.2 kg), no attempts were made to normalize perfusion to tissue mass.
Capillary Density and Capillary/Muscle Fiber Ratio
Vascular density was evaluated at the microvascular level in light microscopic sections taken from the ischemic hindlimbs. Tissue specimens were embedded in OCT compound (Miles), snap-frozen in liquid nitrogen, cut into 5-μm-thick sections, and stained for alkaline phosphatase with an indoxyl-tetrazolium method to detect capillary endothelial cells as previously described.21
Human Ang1 and Ang2 Gene Expression in Ischemic Skeletal Muscle
Gene expression was evaluated by reverse transcription–polymerase chain reaction (RT-PCR) in 20 additional male New Zealand White rabbits with hindlimb ischemia killed at 3, 7, 14, and 30 days after transfection with pAng1* and pAng2 (n=2 at each time point) or control plasmids (pJFE and pcDNA3, n=2 at 7 days after transfection). In the 8 rabbits killed at 3 and 7 days after transfection, remote tissues (gastrocnemius muscle in the same limb, semimembranous muscle of the contralateral limb, heart, lung, and liver) were also retrieved for analysis of human Ang1 and Ang2 mRNA. To ensure specificity and avoid amplification of endogenous rabbit Ang1, each primer was selected from a region that is not conserved among different species. Sequences of primers used for Ang1 were 5′-GGATTCTAGTTGTGGTTTGTG-3′ (sense) and 5′-GGTGGTTTG ATGCTTGTGG-3′ (antisense). Sequences of primers used for Ang2 were 5′-CCAAGTGAGCAGGACTGTTCTT-3′ (sense) and 5′-TGATGTTCTCCAGCACTTGCAG-3′ (antisense). RT-PCR was performed according to the manufacturer’s protocol (Access RT-PCR System, Promega). The PCR products for Ang1 and Ang2 were 220 and 495 bp, respectively. RT-PCR products were analyzed by 2% agarose gel electrophoresis.
All results were expressed as mean±SEM. Statistical significance was evaluated by unpaired Student’s t test for comparisons between 2 means and ANOVA followed by Scheffé’s procedure for >2 means. A value of P<0.05 was considered to denote statistical significance.
Lower-Limb Calf Blood Pressure Ratio
At 10 days after induction of ischemia (immediately before transfection), calf blood pressure ratio (BPR) was similar among all groups (pAng1*, 0.43±0.02; pJFE, 0.41±0.03; pAng2, 0.42±0.02; pcDNA3, 0.43±0.03; and vehicle, 0.44±0.03; P=NS) (Figure 1⇓). By day 40 (30 days after transfection), BPR had improved among all 5 groups. BPR for pAng1* (0.90±0.02), however, was significantly higher (P<0.05) than that for pJFE (0.76±0.05) or vehicle (0.77±0.03) (Figure 1A⇓). BPR on day 40 was similar for pAng2 (0.74±0.04), pcDNA3 (0.77±0.02), and vehicle (0.77±0.03) (Figure 1B⇓).
Angiographic Assessment of Collateral Circulation
At baseline, there was no significant difference in angiographic score among the pAng1*, pJFE, pAng-2, pcDNA3, and vehicle (0.41±0.03, 0.42±0.03, 0.43±0.02, 0.42±0.02, and 0.44±0.04, respectively; P=NS) (Figure 2⇓). By day 40, however, angiographic score for pAng1* (0.63±0.02) was higher (P<0.05) than for pJFE (0.51±0.03) or vehicle (0.52±0.02) (Figure 2A⇓). Day 40 angiographic score was similar for pAng2 (0.54±0.02), pcDNA3 (0.51±0.02), and vehicle (0.52±0.02; P=NS) (Figure 2B⇓). Representative angiograms recorded from pAng1*-, pJFE-, pAng2-, and pcDNA3-treated animals at day 40 are shown in Figure 3⇓. The principal angiographic finding accounting for the increased angiographic score in Ang1*-treated animals was an increase in so-called midzone collateral vessels.
Quantitative angiography also documented increased luminal diameter of the principal stem collateral artery issuing from the internal iliac artery between days 10 and 40 in all animals. At day 10, luminal diameter of the principal stem artery was similar among all groups. By day 40, however, mean diameter of the stem artery for pAng1* (1.48±0.05 mm) exceeded that of pJFE (1.26±0.05 mm) and vehicle (1.28±0.06 mm, P<0.01) (Figure 4A⇓). There was no significant difference in the diameter of main stem artery among pAng2, pcDNA3, and saline at day 40 (1.31±0.06, 1.29±0.06, and 1.28±0.06 mm, respectively; P=NS) (Figure 4B⇓).
Intravascular Doppler Measurements of Blood Flow
Resting and maximal blood flow were similar for 5 groups at day 10 (Table⇓). Resting blood flow in the ischemic limb was also similar for these same 5 groups at day 40. Maximal blood flow for pAng1* (67.8±4.9 mL/min), however, was higher (P<0.05) than that recorded for either pJFE (51.2±4.4 mL/min), vehicle (52.9±4.9 mL/min), pAng2 (50.1±2.8 mL/min), or pcDNA3 at day 40 (51.3±3.8 mL/min). Resting and maximal blood flows in the nonischemic limb were similar among these 4 groups at both days 10 and 40.
Capillary Density and Capillary/Muscle Fiber Ratio
Capillary density in the pAng1*-treated group (242±12/mm2) was significantly higher (P<0.01) than that in the pJFE (172±11/mm2) and vehicle (166±10/mm2) groups (Figure 5⇓). Capillary/muscle fiber ratio was similarly greater (P<0.01) for the pAng1* rabbits (0.89±0.06) versus pJFE (0.64±0.05) and vehicle (0.67±0.05). There were no significant differences in capillary density or capillary/muscle fiber ratio among pAng2 (180±15/mm2 and 0.69±0.06, respectively), pcDNA3 (175±14/mm2 and 0.69±0.05, respectively), and vehicle.
Light microscopic signs of frank myonecrosis were not observed in any group.
Human Ang1 and Ang2 Gene Expression in Ischemic Skeletal Muscle
Human Ang1 mRNA was detected in sections of semimembranous muscles retrieved from days 3 to 14 after gene transfer. No human Ang1 mRNA was detected in tissue samples retrieved from the distal ischemic hindlimb (gastrocnemius muscle) or sites remote from ischemic hindlimb, including the semimembranous muscle of contralateral limb, as well as the heart, lung, and liver (Figure 6A⇓). Rabbit semimembranous muscles injected with pJFE plasmid were consistently negative for human Ang1 mRNA. The presence or absence of human Ang2 mRNA in tissue sections from similar sites in rabbits transfected with pAng2 followed a pattern similar to that described for pAng1* (Figure 6B⇓). Rabbit semimembranous muscles injected with pcDNA3 plasmid were consistently negative for human Ang2 mRNA.
Blood vessel development and remodeling are regulated by receptor tyrosine kinases and their ligands.22 23 Targeted disruption of the endothelial cell–selective receptor tyrosine kinases vascular endothelial growth factor (VEGF)-R1,24 VEGF-R2,25 Tie1,3 and Tie226 in each case leads to embryonic lethality with severely defective vasculature.27 Ang1 was first identified to be a secreted ligand for the Tie2 receptor in 1996 by Davis et al.4 Subsequently, Ang2 was identified as a relative of Ang1 by homology screening and was shown to bind but not phosphorylate the Tie2 receptor.6 When constitutively expressed in transgenic mice, Ang2 was demonstrated to antagonize the effects of Ang1, leading to disruption of angiogenesis in vivo.6 Tie2 and the angiopoietins are expressed later in development than the VEGF receptors and their respective ligands,28 and the expression of Ang1 appears to become more widespread as development proceeds.4 These observations have suggested a role for Ang1 in angiogenesis that is both later than and distinct from that of VEGF.
VEGF and Ang1 also have been shown to differ in at least 2 other important respects. VEGF is clearly mitogenic for endothelial cells and is also upregulated by hypoxia,29 30 2 features consistent with its role as an endogenous mediator of angiogenesis. In contrast, Ang1 does not induce growth response in cultured endothelial cells4 and is downregulated by hypoxia.31 A growth-promoting effect on endothelial cells has conventionally been viewed as the sine qua non for an angiogenic growth factor.32 Ang1 is thus an intriguing example of an endothelial cell–specific growth factor that is not mitogenic for endothelial cells yet promotes angiogenesis in vivo.
There are at least 2 possible mechanisms by which Ang1 may have contributed to augmented neovascularity in rabbit ischemic hindlimb. First, recent studies from our laboratory have established that Ang1 but not Ang2 is chemotactic for endothelial cells.32a Excess soluble Tie2 but not Tie1 receptor was shown to abolish the chemotactic response of endothelial cells toward Ang1. The fact that the in vivo results of the present experiments involving Ang1 better parallel in vitro analyses of migration as opposed to proliferation is consistent with notions regarding the relative contributions of these activities to angiogenesis. In a classic experiment performed in the rat cornea, Sholley et al33 showed that vascular sprouting could be induced and could continue for >2 days despite irradiation treatment sufficient to suppress DNA synthesis. Angiogenic activity in this model was thus interpreted to reflect the dominant impact of endothelial cell migration. The absence of endothelial cell chemotaxis in response to Ang2 is further consistent with the absence of any effect of Ang2 on in vivo angiogenesis in the rabbit ischemic hindlimb model and supports the notion of endothelial cell migration as a critical element for in vivo angiogenesis.
A second mechanism potentially contributing to the augmented angiogenesis observed after Ang1 gene transfer in the present study is the possibility of a synergistic effect between exogenous Ang1 and endogenous VEGF expression. Shortly after the onset of hindlimb ischemia, VEGF expression is naturally upregulated and remains so for up to 14 days,34 presumably in response to tissue hypoxia. The precedent for a synergistic effect of 2 angiogenic cytokines administered exogenously has been previously established in vitro35 36 and in vivo.19 More recent in vitro37 and in vivo38 experiments have suggested that such synergism may also result from exogenous growth factor administration together with native expression of a second angiogenic cytokine.
Our data further confirm the concept that Ang1/Tie2 mediates vessel maturation or remodeling.3 5 The resting blood flow measured 1 month after gene transfer was no different in hindlimbs transfected with pAng1* versus control plasmid; the luminal diameter of the stem artery (internal iliac artery) in the hindlimb, however, was significantly larger in the pAng1*-transfected group. The fact that the increase in luminal diameter was not due to increased flow suggests that Ang1 may have had a direct effect on vessel maturation or remodeling that may be conferred by recruitment and interaction of endothelial cells with periendothelial support cells, as previously suggested.4 27
In adult animals, as in the developing embryo, Ang1 is expressed widely, whereas Ang2 is selectively expressed in ovary, uterus, and placenta.6 This intriguing observation has been interpreted27 as further evidence for Ang2 in physiological disruption of established vasculature in reproductive organs characterized by the need for cyclic vessel development and regression. In this study, we were unable to show any effect, most notably any inhibitory effect, of Ang2 alone on tissue neovascularization. Previous data have suggested that antagonism of Ang1 by Ang2 requires an 8-fold dose differential of Ang2:Ang1.6 In the setting of hindlimb ischemia, it is possible that such a differential was not achieved with the present dose of pAng2, allowing endogenous angiogenic cytokines to predominate and obviate any potentially disruptive influences of the pAng2 transgene.
The results of the present study demonstrate that intramuscular administration of Ang1 is sufficient to enhance revascularization in a rabbit model of lower-extremity ischemia, although no attempt was made in this study to determine the corresponding extent of functional improvement. The anatomic evidence of revascularization in response to Ang1 administration was observed at 2 levels. Necropsy examination documented an increase in vascularity at the capillary level, consistent with the classic definition of angiogenesis formulated by Klagsbrun and Folkman.39 In addition, systematic quantification of angiographic recordings established that the diameter and number of angiographically visible collateral arteries in the Ang1-treated animals exceeded that in the controls. Successful augmentation of neovascularization after intramuscular gene transfer of Ang1 is similar to that previously achieved with intramuscular gene transfer of VEGF in the same animal model14 and later in patients with critical limb ischemia.40 41 The data involving both VEGF and Ang1 imply that ischemic skeletal muscle represents a promising target for gene therapy with naked plasmid DNA.
Like VEGF, Ang1 and Ang2 contain hydrophobic sequences at the N-terminal region typical of a secretory signal sequence.4 Successful intramuscular transfection of naked DNA in the present study is in large part the result of the secreted nature of the gene product. Our previous studies have shown that low-efficiency transfection with genes encoding for secreted proteins may achieve biologically meaningful effects not realized by transfection with genes encoding for proteins that remain intracellular.42 43 Expression of human Ang1 and Ang2 transgenes was detected by RT-PCR for as long as 14 days after gene transfer in the present study. No gene expression was detected in remote tissues, including contralateral muscles, lung, heart, and liver. Intramuscular administration of human Ang1 and Ang2 thus results in site-specific gene expression.
In summary, these findings provide the first demonstration that angiopoietin can augment angiogenesis in vivo, despite the absence of a growth-promoting effect on endothelial cells in culture. The mechanisms responsible for this effect remain to be fully elucidated. It is conceivable, however, that intramuscular administration of Ang1* plasmid may constitute a feasible adjunct or alternative therapeutic agent for revascularization of ischemic tissues.
Dr Shyu is supported in part by a research fellowship from Shin Kong Wu Ho-Su Memorial Hospital, Taipei, Taiwan.
- Received March 13, 1998.
- Revision received June 11, 1998.
- Accepted June 13, 1998.
- Copyright © 1998 by American Heart Association
Sato TN, Qin Y, Kozak CA, Audus KL. Tie-1 and tie-2 define another class of putative receptor tyrosine kinase genes expressed in early embryonic vascular system. Proc Natl Acad Sci U S A. 1993;90:9355–9358.
Maisonpierre P, Suri C, Jones P, Bartunkova S, Wiegand S, Radziejewski C, Compton D, McClain J, Aldrich T, Papadopoulos N, Daly T, Davis S, Sato T, Yancopoulos G. Angiopoietin-2, a natural antagonist for Tie2 that disrupts in vivo angiogenesis. Science. 1997;277:55–60.
Wolff JA, Malone RW, Williams P, Chong W, Acsadi G, Jani A, Felgner PL. Direct gene transfer into mouse muscle in vivo. Science. 1990;247:1465–1468.
Stratford-Perricaudet LD, Makeh I, Perricaudet M, Briand P. Widespread long-term gene transfer to mouse skeletal muscles and heart. J Clin Invest. 1992;90:626–630.
Tsurumi Y, Takeshita S, Chen D, Kearney M, Rossow ST, Passeri J, Horowitz JR, Symes JF. Direct intramuscular gene transfer of naked DNA encoding vascular endothelial growth factor augments collateral development and tissue perfusion. Circulation. 1996;94:3281–3290.
Tripathy SK, Svensson EC, Black HB, Goldwasser E, Margalith M, Hobart PM, Leiden JM. Long-term expression of erythropoietin in the systemic circulation of mice after intramuscular injection of a plasmid DNA vector. Proc Natl Acad Sci U S A. 1996;93:10876–10880.
Takeshita S, Zheng LP, Brogi E, Kearney M, Pu LQ, Bunting S, Ferrara N, Symes JF, Isner JM. Therapeutic angiogenesis: a single intra-arterial bolus of vascular endothelial growth factor augments revascularization in a rabbit ischemic hindlimb model. J Clin Invest. 1994;93:662–670.
Pu LQ, Sniderman AD, Brassard R, Lachapelle KJ, Graham AM, Lisbona R, Symes JF. Enhanced revascularization of the ischemic limb by means of angiogenic therapy. Circulation. 1993;88:208–215.
Asahara T, Bauters C, Zheng LP, Takeshita S, Bunting S, Ferrara N, Symes JF, Isner JM. Synergistic effect of vascular endothelial growth factor and basic fibroblast growth factor on angiogenesis in vivo. Circulation. 1995;92(suppl II):II-365-II-371.
Ku DD, Zaleski JK, Liu S, Brock TA. Vascular endothelial growth factor induces EDRF-dependent relaxation in coronary arteries. Am J Physiol. 1993;265:H586–H592.
Mustonen T, Alitalo K. Endothelial receptor tyrosine kinases involved in angiogenesis. J Cell Biol. 1995;129:895–898.
Dumont DJ, Gradwohl G, Guo-Hua F, Puri MC, Gertsenstein M, Auerbach A, Breitman ML. Dominant-negative and targeted null mutations in the endothelial receptor tyrosine kinase, tek, reveal a critical role in vasculogenesis of the embryo. Genes Dev. 1994;8:1897–1909.
Hanahan D. Signaling vascular morphogenesis and maintenance. Science. 1997;277:48–50.
Thomas KA. Vascular endothelial growth factor, a potent and selective angiogenic agent. J Biol Chem. 1996;271:603–606.
Namiki A, Brogi E, Kearney M, Wu T, Couffinhal T, Varticovski L, Isner JM. Hypoxia induces vascular endothelial growth factor in cultured human endothelial cells. J Biol Chem. 1995;270:31189–31195.
Folkman J, Klagsbrun M. Angiogenic factors. Science. 1987;235:442–447.
Witzenbichler B, Maisonpierre PC, Jones P, Yancopoulous GD, Isner JM. Chemotactic properties of angiotensin-1 and -2, ligands for the endothelial-specific receptor tyrosine kinase Tie2. J Biol Chem. 1998;273:18514–18521.
Mandriota SJ, Pepper MS. Vascular endothelial growth factor-induced in vitro angiogenesis and plasminogen activator expression are dependent on endogenous basic fibroblast growth factor. J Cell Sci. 1997;110:2293–2302.
Van Belle E, Witzenbichler B, Chen D, Silver M, Chang L, Schwall R, Isner JM. Potentiated angiogenic effect of scatter factor/hepatocyte growth factor via induction of vascular endothelial growth factor: the case for paracrine amplification of angiogenesis. Circulation. 1998;97:381–390.
Klagsbrun M, Folkman J. Angiogenesis. In: Sporn MB, Roberts AB, eds. Peptide Growth Factors and Their Receptors II. New York, NY: Springer-Verlag; 1990:459–586.
Baumgartner I, Pieczek A, Manor O, Blair R, Kearney M, Walsh K, Isner JM. Constitutive expression of phVEGF165 following intramuscular gene transfer promotes collateral vessel development in patients with critical limb ischemia. Circulation. 1998;97:1114–1123.
Losordo DW, Pickering JG, Takeshita S, Leclerc G, Gal D, Weir L, Kearney M, Jekanowski J, Isner JM. Use of the rabbit ear artery to serially assess foreign protein secretion after site specific arterial gene transfer in vivo: evidence that anatomic identification of successful gene transfer may underestimate the potential magnitude of transgene expression. Circulation. 1994;89:785–792.