INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Colonization of cystic fibrosis (CF) individuals with Pseudomonas aeruginosa represents a significant negative milestone in the progression of this disease. Once colonized, patients are subject to the damaging effects of various secreted virulence factors and to the inflammatory response of the host immune system. A key component of colonization is the adhesion of type IV pili to asialo-GM1 receptors on the surface of epithelial cells (26, 45, 48; for a review, see reference 21). Type IV pili are composed of pilin polymers arranged in a helical structure with five subunits per turn (19, 41). The portion of the pilin protein responsible for cell binding is located near the C terminus (amino acids 129 to 142) in a -turn--turn loop subtended from a disulfide bond (5, 6, 23, 36). A 12- or 17-amino-acid sequence (depending on the specific strain) in this loop interacts with receptors on epithelial cells. For CF individuals, the overproduction of the R domain of the mutant CF transmembrane conductance regulator can lead to an increased level of asialo-GM1 and, accordingly, an increased binding of P. aeruginosa (3, 26, 45). Functional studies of pilin have indicated that only the last pilin subunit (the tip) of a pilus interacts with epithelial cell receptors (31). To interfere with bacterial adhesion, anti-pilin antibodies will need to recognize residues that are normally located at the C-terminal loop of pilin (32). Structural studies have indicated that this loop is dominated by main chain residues; and this may explain why pilins from distinct strains bind the same receptor despite sequence variation and the presence of both 12- and 17-amino-acid loops. Generating antibodies to the C-terminal pilin loop may be useful in reducing or eliminating colonization (15, 22, 47). Table 1 lists the pilin loop sequences from several strains of P. aeruginosa.

Pseudomonas exotoxin A (here called PE), a prominent virulence factor secreted by P. aeruginosa, is cytotoxic for mammalian cells by virtue of its ability to enter cells by receptor-mediated endocytosis and then, after a series of intracellular processing steps, translocate to the cell cytosol and ADP-ribosylate elongation factor 2 (17, 25, 38, 39). This results in the inhibition of protein synthesis and cell death. It is possible to generate a nontoxic mutant toxin that has no ADP-ribosylating activity PE (reference 33 and this study).

PE is composed of three prominent structural domains and one minor subdomain (Fig. 1) (1). The N-terminal domain (Ia) is responsible for receptor binding and the middle domain (II) has translocating activity, while the C-terminal domain (III) is an ADP-ribosyl transferase (24). Subdomain Ib (located between domains II and III in the primary sequence) has no known function and can be deleted without loss of toxin activity.

View larger version (29K): [in this window] [in a new window]   FIG. 1.   Shown in cartoon form is the domain organization of PE (1). PE64 lacks the loop region of domain Ib. PE64pil includes the insertion of the pilin loop (residues 129 to 142) of the PAK strain of P. aeruginosa. The deletion of glutamic acid 553 (indicated by a dot) removes an active site residue (33) and produces proteins PE64553 and PE64553pil with no ADP-ribosylating activity. The Ib loop is shown in light shading and the pilin loop in darker shading.

As a virulence factor, PE can kill polymorphonuclear leukocytes, macrophages, and other elements of the immune system (44). In this way, toxin-mediated destruction of local immune cells may contribute to the maintenance of P. aeruginosa infections (43, 52). The importance of PE as a virulence factor has been confirmed by results showing that toxin-producing strains are more virulent than nontoxogenic ones (53) and by data from murine models of Pseudomonas infection where the presence of anti-PE antibodies reduced pathogenicity and extended life (16, 42, 49).

Here, we report on the development with a wholly recombinant vaccine. The deletion of glutamic acid at position 553 of PE (PE553) produces a protein that exhibits all toxin functions with the exception of ADP-ribosylation (33). PE553, which is noncytotoxic for cells, animals, or humans, is a potential platform for vaccine development. Between domains II and III is the small subdomain termed Ib. It is composed of a seven-amino-acid loop subtended from a disulfide bond. Because deletion of this structure to produce a protein we term PE64 (Fig. 1) causes no loss of toxin activity, it is an attractive location for the insertion of third-party sequences, especially loop sequences. Previously, we reported that the Ib loop could be replaced by sequences from the V3 loop of HIV gp120 (18). Inserts of 14 or 26 amino acids were accommodated without disturbing PE functions (18).

To produce a chimeric protein that displays pilin in a near-native conformation, we replaced the Ib domain of PE by amino acids 129 to 142 of pilin (Fig. 1) including the disulfide bond that links cysteines 129 to 142. This chimeric protein is characterized here as a candidate vaccine designed to produce antibodies that will interfere with Pseudomonas adherence and neutralize PE.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and growth conditions. The bacterial strains, plasmids, and oligonucleotides used in this study are listed in Table 2. Pseudomonas strains used for adherence studies were grown on Luria-Bertani agar and then in M9 minimal medium (KD Medical, Bethesda, Md.) supplemented with 0.4% glucose at 30°C without shaking. Cultures in late log phase were routinely used for adhesion assays.

Oligoduplex formation and plasmid construction. A 54-bp sense oligonucleotide with cohesive ends for PstI and encoding the 12-amino-acid pilin loop of the PAK strain was annealed with a 54-bp antisense oligonucleotide in 10 mM Tris-HCl and 50 mM NaCl (pH 7.4) (oligonucleotide sequences are listed in Table 2). Annealing was accomplished by heating to 94°C for 5 min followed by cooling to 25°C over a period of 40 min. Plasmids pPE64 and pPE64553 (see Table 2), encoding enzymatically active and inactive PE, respectively, were digested with PstI at residue 1470 (see FitzGerald et al. [18]). Ligation with the phosphorylated pilin oligoduplex destroyed the PstI site and introduced a unique SpeI site. A XhoI/SpeI double digest was used to check for the correct orientation of the insert. Final constructs were verified by dideoxy double-strand sequencing.

Antibodies and proteins. The PK99H mouse monoclonal antibody and purified pilin protein were gifts from Randall Irvin, University of Alberta, Alberta, Canada. Horseradish peroxidase-conjugated anti-mouse immunoglobulin G (IgG) and anti-rabbit IgG antibodies (Jackson ImmunoResearch Laboratories, West Grove, Pa.) were used at a 1:2,000 dilution to detect primary antibodies in Western blots and enzyme-linked immunosorbent assays (ELISAs).

Chimera protein expression and purification. Chimeric proteins were expressed in Escherichia coli and recovered from inclusion bodies as previously described (4). Briefly, strain BL21(DE3) was transformed with plasmids harboring a T7 promoter upstream of the initial ATG of the toxin-expressing vectors. Cultures were grown in Superbroth (KD Medical) with ampicillin (50 µg/ml) and then induced for protein expression by the addition of IPTG (isopropyl-D-thiogalactopyranoside) (1 mM). After 2 h of further culture, bacterial cells were harvested by centrifugation. Following cell lysis, expressed proteins were recovered in inclusion bodies. Proteins were solubilized with guanidine HCl (6.0 M) and 2 mM EDTA (pH 8.0) plus dithioerythreitol (65 mM). Solubilized proteins were then refolded by dilution into a redox-shuffling buffer (4). Refolded proteins were dialyzed against 20 mM Tris and 100 mM urea (pH 7.4); adsorbed on Q Sepharose (Amersham Pharmacia Biotech); washed with 150 mM NaCl, 20 mM Tris, and 1 mM EDTA (pH 6.5); and eluted with 280 mM NaCl, 20 mM Tris, and 1 mM EDTA. Eluted proteins were diluted fivefold and then adsorbed onto a MonoQ column (HR 10/10; Amersham Pharmacia Biotech) and further purified by the application of a linear salt gradient (0 to 0.4 M NaCl in Tris-EDTA, pH 7.4). PE proteins eluted between 0.2 and 0.25 M NaCl. Final purification was achieved with a gel filtration column (Superdex 200; Amersham Pharmacia Biotech) in phosphate-buffered saline (PBS), pH 7.4.

Cell cultures. A549 (ATCC CCL-185), and L929 (ATCC CCL-1) cells were maintained in Dulbecco's modified Eagle's medium F12 (DMEM F12) supplemented with 10% fetal bovine serum, 2.5 mM glutamine, a standard concentration of penicillin and streptomycin (100 U of penicillin/ml and 100 µg of streptomycin/ml; Gibco BRL, Grand Island, N.Y.) (further designated complete medium) in 5% CO₂ at 37°C. Cells were fed every 2 to 3 days and passaged every 5 to 7 days. For assays, cells were seeded into 24- or 96-well plates and grown to confluence.

Quantification of bacterial adherence. To quantify the association of Pseudomonas with A549 cells, we followed the adhesion assay described by Chi et al. (8). Briefly, A549 cells were grown in a 24-well plate (antibiotic-free medium) to a density of approximately 2 × 10⁴ cells per well. Cells were washed three times in Hanks' balanced salt solution without serum and were overlaid with 0.5 ml of DMEM F12 complete medium without fetal bovine serum. A multiplicity of infection of 20 was achieved by adding 10 µl of an appropriate bacterial dilution. Plates were incubated for 1 or 2 h at 37°C and 5% CO₂.

Determination of binding to asialo-GM1 by ELISA. Plates (96-well) were coated with asialo-GM1 or monsialo-GM1 (Sigma Chemical Co., St Louis, Mo.) that had been solubilized in methanol. A 100-µl solution of ganglioside (5 µg/ml) was added to each well and evaporated at 4°C until dry. Wells were washed three times with PBS and blocked with fish gelatin-PBS (BioFX, Randallstown, Md.) for 16 h at 4°C. Test proteins in blocking buffer were added at various concentrations. After incubation for 1 h at 22°C, the supernatant was removed and bound protein was detected with heat-inactivated anti-PE64553pil serum (1:100) as the primary antibody. For competition studies, proteins at 0.2 µg/ml were incubated with 2 µg of asialo-GM1 or monosialo-GM1/ml for 30 min at room temperature. Samples were then added to asialo-GM1-coated plates as above.

Cytotoxicity assay. The inhibition of protein synthesis by PE64 and PE64pil on L929 cells was determined as described previously (38). For assessing toxin neutralization activity, the same proteins (at 1 µg/ml) were incubated for 30 min at 22°C with rabbit sera diluted to 1:100. Samples at the appropriate dilution were added to individual wells containing L929 cells.

Production of polyclonal antibodies. PE64553pil (200 µg per injection) was injected subcutaneously into a total of four rabbits: two rabbits were coinjected with Freund's adjuvant while two rabbits received no adjuvant. Subsequent injections included three biweekly injections at the same dose with or without incomplete Freund's adjuvant. About 12 ml of serum was recovered biweekly from each rabbit. The sera were heat inactivated for 20 min at 56°C, and dilutions thereof were used for assays without further purification.

Synthetic peptides. Peptide 1 (acetyl-KCTSDQDEQFIPKGCSK-NH₂) containing the complete C-terminal loop of PAK pilin protein, peptide 2 (acetyl-DEQFIPK-NH²) containing the core cell-binding sequence, and peptide 3 (acetyl-QIDPEFK-NH₂) with the scrambled amino acids of peptide 2 were custom synthesized by Sigma Genosys. Peptide 1 was oxidized to allow the formation of a disulfide bond. The same peptides were also synthesized with a biotin label.

Inhibition of adhesion. To assess antibody-mediated inhibition of adherence, anti-PE64553pil rabbit sera were incubated at dilutions from 1:20 to 1:100 with 4 × 10⁵ bacteria at 22°C for 30 min. Bacteria were then centrifuged, resuspended in DMEM without supplements, and added to confluent monolayers of A549 cells at a multiplicity of infection of 20 for 1 to 2 h. Adherence was determined as described above. Immune sera taken after the fourth injection were compared to prebleed samples taken from the same rabbits.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Vaccine design. To generate a PE-based pilin vaccine, we synthesized an oligonucleotide duplex that encoded amino acids 129 to 142 of pilin from the PAK strain of P. aeruginosa. The construction of a nontoxic PE vector whereby a unique PstI site was introduced in place of subdomain Ib was previously reported (18). Ligation of the pilin oligoduplex into the PstI-cut vector was followed by several characterization steps to confirm the presence of the pilin insert in the correct orientation. The pilin oligoduplex was ligated into PE vectors to produce the plasmids pPE64pil and pPE64553pil (enzymatically active and inactive, respectively). Final constructs were confirmed by double-stranded DNA sequencing. Vectors were constructed without a bacterial secretion sequence, allowing recombinant proteins to be expressed as inclusion bodies.

Protein expression and purification. Using the T7 expression system described by Studier et al. (51), four PE-related proteins were expressed in E. coli. These were PE64, PE64553, PE64pil, and PE64553pil (Fig. 1). Each protein was expressed separately and purified to near homogeneity. Expression was induced by the addition of IPTG for 2 h, followed by harvesting of bacterial pellets. Inclusion bodies were recovered from lysed bacteria. Proteins were then denatured and renatured from inclusion bodies as outlined in Materials and Methods. Briefly, proteins were solubilized in guanidine HCl and a reducing agent and then renatured with a redox-shuffling buffer (4). Refolded proteins were dialyzed against Tris-urea, loaded onto Q Sepharose, and then recovered with a step gradient (0.15 and 0.28 M NaCl). The proteins eluted at 0.28 M NaCl were diluted and applied to a MonoQ column which was then developed with a linear salt gradient. Gel filtration was used as the final purification step for PE64553pil.

Characterization of PE64pil proteins. Proteins were initially analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Fig. 2A and C). Substantially pure proteins were isolated by using the purification scheme outlined above. By Western blot analysis, PE64pil and PE64553pil proteins reacted with PK99H, a monoclonal antibody to the C-terminal loop of pilin (Fig. 2B). The same antibody also reacted with soluble preparations of these proteins, indicating that the pilin insert was exposed on the surface of the chimeric protein (data not shown). PE proteins without inserts did not react with the PK99H antibody (Fig. 2B).

View larger version (65K): [in this window] [in a new window]   FIG. 2.   Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (A and C) and Western blot analysis (B) of PE proteins and pilin. (A) Lanes 1 to 4 show substantially pure PE proteins (4 to 5 µg of protein was loaded per lane) after MonoQ chromatography. A small amount of dimer is noted at 130 kDa. From left to right, the proteins loaded were PE64, PE64pil, PE64553, and PE64553pil. Purified PAK pilin was added to lane 5. (B) Lanes 6 to 10 contain the same proteins as shown in panel A but were probed with a monoclonal antibody to the pilin loop. (C) Lane 11 is PE64553pil after gel filtration chromatography. Standard proteins and their molecular masses in kilodaltons are indicated.

View larger version (11K): [in this window] [in a new window]   FIG. 3.   Toxicity of PE64pil compared to PE64. To assess the effect of introducing a third-party loop into PE, we compared the toxicity of PE64 () with PE64pil (). Increasing concentrations of each protein were added to L929 cells and, after an overnight incubation, inhibition of protein synthesis was determined. Results are expressed as percent control compared to cells receiving no toxin. Error bars represent 1 standard deviation (SD) of the mean from triplicate wells.

View larger version (99K): [in this window] [in a new window]   FIG. 4.   PE64pil and PE64553pil interact with immobilized asialo-GM1. (A) Various concentrations of PE64pil or PE64 were added to plates coated with asialo-GM1, and binding was determined by reactivity with rabbit anti-PE followed by a peroxidase-labeled goat anti-rabbit IgG antibody. Absorbance at 450 nm was used to monitor binding. (B and C) To investigate ganglioside specificity, a competition assay was devised whereby soluble asialo-GM1 (aGM1) or monosialo-GM1 (GM1) at 2 µg/ml was preincubated with PE64pil (B) or PE64553pil (C), and the percent residual binding determined as described in the legend to panel A. For panels B and C, graphs show the mean of a representative triplicate experiment. Error bars represent 1 SD. N.A., no addition of competitor.

Rabbit immune response to PE64553pil. To test the ability of the toxin-pilin protein to generate relevant antibody responses, four rabbits were injected with the PE64553pil protein. Two rabbits (numbered 87 and 88) received the protein plus adjuvant (complete Freunds for the first injection followed by incomplete Freunds for subsequent injections), and two rabbits (numbered 89 and 90) received the protein alone. Two hundred micrograms of protein per injection was given subcutaneously for a total of four cycles spaced approximately 2 weeks apart (Fig. 5). Anti-pilin titers were determined using an ELISA assay where biotinylated pilin peptides were immobilized on strepavidin-coated plates. Over the period of immunization, anti-pilin titers increased in all four animals (Fig. 5). However, the speed and extent of the response were greater in the two rabbits that received antigen plus adjuvant. To avoid complement-mediated bacterial killing (see below), immune sera were heat inactivated. This treatment did not significantly alter antibody titers in the ELISA assay (data not shown).

View larger version (17K): [in this window] [in a new window]   FIG. 5.   Anti-pilin antibody titers (1:100) postimmunization with PE64553pil with and without adjuvant. Sera were collected from each of four rabbits (numbered 87 to 90) at various times, diluted 1:100, and then added to strepavidin-coated plates that had been loaded with biotinylated pilin peptides. Rabbit IgG was detected by the addition of a peroxidase-conjugated goat anti-rabbit antibody. Rabbits 87 () and 88 () received adjuvant while rabbits 89 () and 90 () did not.

Inhibition of P. aeruginosa (PAK strain) adhesion by post immunization sera. Sera taken 2 weeks after the last injection were assayed for blocking activity by the bacterial adherence assay. Compared to prebleeds, immune sera at various dilutions blocked adherence of the PAK strain of P. aeruginosa (Fig. 6A). Reduction of adherence ranged from 60% at a dilution of 1:100 to 90% at a dilution of 1:20. At a dilution of 1:20, blocking activity was comparable without regard to the presence of adjuvant in the antigen preparation (Fig. 6B).

View larger version (41K): [in this window] [in a new window]   FIG. 6.   Antibody-mediated interference with Pseudomonas adhesion to A549 cells. (A) The PAK strain of P. aeruginosa was incubated with 1:20 to 1:100 dilutions of prebleed or immune (taken after the fourth injection of antigen) sera from rabbit 87. Bacteria were then added to cells, and the percent adhesion was determined by comparison with bacteria that had been incubated in media alone. (B) A 1:20 dilution of sera from each rabbit, prebleed and immune, was tested for antibody mediated interference. (C) Various strains of P. aeruginosa were incubated with immune sera (1:20) from one of the rabbits that received antigen alone (rabbit 90) and one that received antigen plus adjuvant (rabbit 87). For each panel, bars represent the number of bacteria per cell determined by examining 100 A549 cells. The error bars represent 1 SD from the mean of three independent experiments.

Inhibition of P. aeruginosa (various strains) Adhesion by postimmunization sera. Inhibition of PAK strain adhesion confirmed that rabbits responded to the specific pilin sequence that was administered in the vaccine. However, because the C-terminal loop of pilin exhibits considerable sequence variation, it was important to determine the reactivity of the immune sera for other strains of P. aeruginosa. Strains PAO1, 1071, SBI-N, 82935, 82932, 90063, 1244, and M2 were tested for adherence to A549 cells under conditions similar to those used for the PAK strain. The specific cell binding of all strains was reduced in adhesion when heat-inactivated immune rabbit sera were mixed with bacteria at a 1:20 dilution (Fig. 6C). The reduction in adhesion among the different strains was more or less in the range of that for the PAK strain (about 90% reduction).

Toxin-neutralizing response. We also evaluated rabbit antisera for toxin-neutralizing activity. All four of the immunized rabbits receiving a 1:20 dilution of sera neutralized 1.0 µg of toxin/ml completely (Fig. 7). From these results we concluded that the PE-pilin vaccine can generate antibodies of two reactivities: one that blocks adhesion and one that neutralizes the exotoxin.

View larger version (13K): [in this window] [in a new window]   FIG. 7.   Antibody-mediated neutralization of PE toxicity. Immune sera () or prebleed sera () were diluted 1:20 and mixed with PE64 at 1.0 µg/ml. Samples were then diluted to the concentration indicated and added to L929 cells for an overnight incubation. Results are expressed as percent control of protein synthesis compared to cells receiving no toxin. Error bars represent 1 SD of the mean from triplicate wells.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Within the first year of life, 25% of CF individuals are colonized with P. aeruginosa; by age 15, this percentage climbs toward 100% (2). Clearly, strategies to prevent the initial colonization event are needed (2). Here, we report on the development of a chimeric subunit vaccine for generating antibodies that interfere with two important components of Pseudomonas virulence, namely pilin-mediated adherence and the tissue destructive activity of PE. Twelve amino acids from the C-terminal loop of pilin (PAK strain) were inserted at a location in nontoxic PE where they could fold into a near-native conformation and cause little or no disruption of toxin structure. Pilin functionality was confirmed by showing that the chimeric protein acquired the ability to bind asialo-GM1. Previously, it was reported that the V3 loop from gp120 of HIV1 could be accommodated in the same location, while retaining antibody reactivity for conformational-dependent epitopes (18). The result with the pilin insert confirms the broad utility of this toxin-based system for insertion of third party sequences, especially loop structures.

We injected PE64553pil subcutaneously into rabbits as proof of the principle that antibodies with the desired specificities could be produced in an animal. In the future, other routes of administration will be pursued, especially mucosal delivery to airway epithelia. Previously, we compared the subcutaneous route with mucosal delivery of toxin-V3 loop proteins (37). Results of mucosal vaccination indicated that a robust anti-V3 loop response could be achieved with high titer responses of both serum IgG and secretory IgA antibodies. Because the toxin-pilin chimeric protein is a candidate vaccine to prevent Pseudomonas colonization in CF, it will be important to optimize vaccine delivery for mucosal antibody responses at airway epithelia.

Type IV pili, which are composed of pilin homopolymers, are thought to be responsible for the initial binding event that mediates adherence of several gram-negative pathogens to mammalian cells. For Pseudomonas pili, this interaction involves the binding of the C-terminal loop of the last pilin subunit to asialo-GM1 on the surface of epithelia (21). Pilin is a 144-amino-acid protein with its cell-binding loop located between amino acids 129 to 142. Apparently, only antibodies to this loop interfere with adhesion. And while the middle portion of pilin is immunogenic, the C-terminal loop usually fails to generate a strong antibody response (22). To overcome poor immunogenicity, strategies to include strong adjuvants along with pilin sequences have been proposed (22). Here, we used an active site deletion mutant of PE as a combination protein carrier and protein adjuvant. This strategy has resulted in a dual neutralizing response to both pilin and PE. In our vaccine protein, we retain the toxin's binding domain (Fig. 1) and thus promote delivery of the pilin loop to cells expressing the toxin receptor, the low-density lipoprotein receptor-related protein designated LRP (also known as CD91) (29). Because LRP is widely distributed on cells and tissues, including macrophages and other antigen presenting cells (30), we speculate that the PE-carrier system has certain attractive features. It was reported recently that the administration of PE to tracheal epithelia resulted in efficient toxin delivery to submucosal lymph nodes and spleen (13). This bodes well for the mucosal delivery of a PE-based vaccine to CF airways.

The potential value of a Pseudomonas vaccine relates in part to its ability to protect individuals broadly from the strains that are present in the environment. Based on the length of the pilin loop insert, there are two groupings for P. aeruginosa: one group with a 12-amino-acid sequence and one with a 17-amino-acid insert. Both loops apparently bind asialo-GM1 and are thought to exhibit similar structures. Reflecting this, we note that our vaccine protein, containing a 12-amino-acid loop from the PAK strain, was able to generate antibodies that were reactive not only for strains with the shorter loop but also for the SBI-N strain, which displayed the longer loop. Our studies have also provided additional sequence data for pilin and pilin loop sequences. We report here two pilin loop sequences that have not previously been entered in databases (Table 1). Details of complete pilin sequences will be presented in greater detail elsewhere.

Chronic pulmonary colonization by P. aeruginosa is associated with a decline in the clinical course of CF patients. Frequently, antibiotic therapy, even via pulmonary delivery, fails to eradicate P. aeruginosa infections in these patients (50). Controlling P. aeruginosa infections, or better yet, preventing them, has thus become a critical unmet medical need in the care of CF patients (2). To address this, a number of vaccine approaches have been explored, many focused on outer membrane constituents (35, 40, 46), some focused on toxins (7, 14, 20, 34) and some focused on a combination approach (7, 9-12, 28).

Here, we characterized a recombinant fusion protein as a candidate vaccine for generating anti-pilin and anti-toxin responses that interfere with bacterial adhesion and neutralize exotoxin A activity. Results obtained to date support further development and evaluation of this approach.