The linkage results were presented at the ECHO Meeting, Paris, France, in December 2006 and will be presented at the Annual Meeting of the German Gastroenterological Association (DGVS), Bochum, Germany, in September 2007.
Potential conflict of interest: Nothing to report.
Genomewide scans of inbred strains of mice have linked the genes encoding the hepatocanalicular cholesterol transporter ABCG5/G8 to gallstone formation. Five nonsynonymous coding single-nucleotide polymorphisms (SNPs) in the orthologous human genes are associated with differences in serum cholesterol and plant sterol levels. We now tested these ABCG5/G8 SNPs for linkage and association with gallstone susceptibility in humans. Prospectively, we collected data from 178 white individuals with gallbladder stones or history of cholecystectomy in 84 families and from 70 stone-free controls, as confirmed by abdominal ultrasound. We performed nonparametric linkage (NPL) analysis of affected sib pairs (ASPs) and association tests of cases and controls. In ASPs, gallstones were strongly linked to the D19H variant of the ABCG8 gene (NPL score = 7.1; P = 4.6 × 10−13). The risk of gallstones in carriers of the 19H allele was significantly increased in randomly selected cases from the ASP cohort compared to the stone-free controls (OR = 3.018; P = 0.017). Consistent with the mouse model, heterozygosity for the lithogenic ABCG8 allele was associated with gallstones in humans; 21.4% of gallstone patients carried the heterozygous D19H genotype, compared with 8.6% of controls (OR = 2.954; P = 0.026). Conclusion: The linkage and association studies identified the cholesterol transporter ABCG5/G8 as a genetic determinant of gallstone formation, or LITH gene, in humans. The function of this transporter and the results of the genetic study taken together indicate that in gallstone-susceptible carriers of the ABCG8 19H allele, cholesterol cholelithiasis is secondary to increased hepatobiliary cholesterol secretion. (HEPATOLOGY 2007.)
Gallstones are common, with a prevalence of up to 20% in Europe and of more than 50% in Amerindians.1–4 Although most gallstones are silent, 25% of gallstone carriers develop symptoms during life and have to undergo cholecystectomy, causing a substantial economic burden for the health care systems.4, 5 At present, the mechanisms of gallbladder stone formation are understood, but knowledge of genes that cause susceptibility to gallstones in humans is lacking.6 Thus, the NIH Action Plan for Liver Disease Research5 specified a major long-term research goal of identifying genetic factors that define populations at risk and might lead to new preventive and therapeutic strategies. In this article, we report the finding of a human “lithogenic” gene associated with increased risk of gallstones, which was identified from an affected sib pair (ASP) study and from the homology with the orthologous murine Lith genes.
As reviewed recently,7, 8 the results of epidemiological studies9 and familial clustering of gallstones indicate that genetic factors influence the risk of gallstone formation. In most patients gallstones are likely to develop as a result of lithogenic polymorphisms in several genes and their interactions with multiple environmental factors, rendering gallstones a complex genetic disease. As estimated in our previous large twin study10 and a recent family study,11 environmental factors account for up to 75% of the phenotypic variation in gallstone disease, but 25% to 29% is determined by genes. Notwithstanding, except for apparently rare gene variants,8, 12, 13 genetic risk factors for most patients with cholesterol gallbladder stones have yet to be identified. Studies of common polymorphisms in the genes encoding apolipoproteins A, B, and E and cholesterol 7α-hydroxylase, as well as cholesteryl ester transfer protein, did not find any associations that have been replicated.2
Unlike this situation in the search for polymorphisms in humans, in inbred mouse strains mapping of quantitative trait loci for gallstone formation has led to the identification of more than 20 susceptibility (Lith) loci.14, 15 Interestingly, Wittenburg et al. established Abcg5/g8 (ATP-binding cassette [ABC] transporters, subfamily G, members 5 and 8; also known as sterolins 1 and 2) as candidate genes in the mouse Lith9 locus.16 ABC transporters use the energy of ATP-binding/hydrolysis to transport substrates across cell membranes. The ABCG5/G8 genes encode protein half-transporters that heterodimerize to form the functional transporter localized in the canalicular membrane of hepatocytes and facilitating cholesterol secretion into bile,17 consistent with the candidacy of this gene pair in cholesterol gallstone susceptibility in mice.15 These experimental observations are in line with clinical data demonstrating that phenotypic variation in cholesterol metabolism and transport is also relevant for the formation of cholesterol gallstones, the predominant stone type in humans.18 Interestingly, a cholesterol-rich diet significantly induces lithogenic bile only in gallstone carriers but not in stone-free controls,19 indicating that intestinal cholesterol absorption and biliary cholesterol secretion are, in part, genetically determined.14, 20
Because the ABCG5/G8 system also transports plant sterols, loss-of-function mutations cause sitosterolemia, a rare autosomal-recessive disease characterized by retention of plant sterols and cholesterol with high serum sterol levels and accelerated atherosclerosis but the absence of gallstones.17, 21 On the other hand, we and others have shown that non-sitosterolemia-associated ABCG5/G8 polymorphisms are associated with decreased serum plant sterol and cholesterol levels.22–25 Hence, the ABCG5/G8 transporter appears to be a key determinant of serum cholesterol level and cholesterol efflux into bile and intestine. We therefore hypothesized that the previously described22 functionally relevant ABCG5/G8 polymorphisms confer an increased risk of gallstone formation and investigated their effects on gallstone susceptibility in affected sib pairs and stone-free controls, employing a combination of genetic linkage and association approaches.
Patients and Methods
Recruitment of Affected Sib Pairs.
For this study, we collected data from affected sib pairs (ASPs) both affected with gallstones (plus the parents, if possible; Table 1). An affected sib pair linkage study is based on the idea of sharing alleles identical by descent: relatives who are phenotypically alike (that is, both are gallstone carriers) have inherited the same alleles from their common ancestors at the disease locus and linked marker loci.26 Thus, in an ASP study, the proportion of alleles shared among the siblings is compared to the proportion expected with no linkage to disease.
Table 1. Summary of Gallstone-Affected Sib Pairs (ASPs)
Affected Siblings (n) per Pedigree
Independent ASPs (n) per Pedigree
Total ASPs (n)
ASPs and normal controls were recruited prospectively at the Department of Medicine III, University Hospital of Cluj-Napoca, Romania, between November 2002 and April 2006. The inclusion criterion for patients and their affected siblings was: (1) previous cholecystectomy for symptomatic gallbladder stones, or (2) cholecystolithiasis, as confirmed by abdominal B-mode ultrasound. Only patients of white ethnicity based on places of birth of parents and grandparents were included. Age, sex, and body mass index (BMI) were recorded on inclusion in the study. The median age of the affected siblings was 55 years (range 24–80 years), and 156 patients (87.6%) were female. The average age and difference in BMI of the siblings were 5.8 years and 4.6 kg/m2, respectively, indicating that age and BMI were similar within pedigrees.
The controls were collected prospectively according to the epidemiological criteria of Grimes and Schulz27: free of the outcome of interest (gallbladder disease), representative of the population at risk of the outcome, and selected independent of exposure of interest, in this case, risk factors for gallstones. In particular, controls were comparable to cases in (1) ethnicity, (2) geographical residence, (3) hospitalization, and (4) time of recruitment. Controls were included if they had no personal or family history of gallstone disease or cholecystectomy. Controls were not related to cases, and in all controls, the presence of gallstones was excluded by abdominal ultrasound performed on inclusion in the study. All patients gave informed consent, and the local ethics committee approved the study protocol.
Genomic DNA was isolated from EDTA anticoagulated blood using the membrane-based QIAamp DNA extraction protocol (Qiagen, Hilden, Germany). DNA concentration was determined fluorometrically (Bio-Rad Laboratories, Hercules, CA), employing the dye PicoGreen (Molecular Probes, Leiden, Netherlands).
For genotyping, we selected functionally relevant nonsynonymous coding single-nucleotide polymorphisms (SNPs) of the ABCG5/G8 genes (Fig. 1): ABCG5rs6720173 = c.1810G>C (p.E604Q); ABCG8rs11887534 = c.55G>C (p.D19H), rs4148211 = c.161A>G (p.Y54C), rs4148217 = c.1199C>A (p.T400K), and rs6544718 = c.1895C>T (p.A632V).22, 23, 25, 28, 29 All SNPs were genotyped using solution-phase hybridization reactions with 5′-nuclease and fluorescence detection (TaqMan assays) in a 7300 Real-Time polymerase chain reaction (PCR) system (Applera, Norwalk, CT). PCR reactions contained 20 ng of genomic DNA, 1× TaqMan Universal Master Mix, 900 nM of each primer, and 200 nM of VIC-labeled and FAM-labeled probes in 25-μL reactions. Amplification conditions were 95°C for 10 minutes, 40 cycles of 92°C for 15 seconds, and 60°C for 1 minute. Probes and assay conditions are available at www.appliedbiosystems.com.
Because SNPs rs6720173 and rs11887534 are of the G↔C variant type, the identity of the minor allele is critical. The probes used identify the C-alleles as minor alleles for both rs6720173 and rs11887534. Results of TaqMan assays were ascertained by direct BigDye termination cycle sequencing with a ABI PRISM 310 Genetic Analyzer (Applera); in addition, individuals carrying c.55C alleles were genotyped in duplicate.
ASP Linkage and Association Studies.
Allele and genotype frequencies were determined and tested for consistency with Hardy-Weinberg equilibrium using an exact test. Linkage disequilibrium measures were estimated between pairs of loci using Lewontin's D′ and r2 values30 and were measured using the GOLD program (www. sph.umich.edu/csg/abecasis/GOLD).31
Nonparametric linkage (NPL) analyses were performed using GENEHUNTER-MODSCORE v2.0.1 (www.staff.uni-marburg.de/∼strauch/software. html).32 The NPL statistics used estimate the significance of shared alleles identical by descent among all affected family members; calculation is based on estimation of the identical-by-descent distribution and comparison of a scoring function for linkage averaged over all inheritance patterns with the standard normal distribution.26 With the 5 ABCG5/G8 SNPs are at the same chromosomal locus as known marker spacings (c.1810G>C 25,846 bp c.55G>C 5,496 bp c.161A>G 27,690 bp c.1199C>A 5,492 bp c.1895C>T), we performed single-point and multipoint NPL analyses.
Exact inference of allele sharing depends on parental genotypes, but because of the late onset of gallstone disease of most patients (median age 55 years; Table 2), data on these genotypes are inevitably missing in our study except for 2 affected mothers. Therefore, we performed a sensitivity analysis for SNPs with significant NPL scores, using variable minor allele frequencies. NPL scores are reported for the allele frequencies estimated for the present cohort, as calculated using the genotypes of all individuals from the 84 families. All allele frequencies were within the ranges reported in the Entrez SNP database (www.ncbi.nih.gov) and in previous publications.23, 25, 28–29, 33–35 In addition, we calculated allele-sharing logarithm of the odds ratio (LOD) scores for exponential models according to Kong and Cox,36 using GENEHUNTER-PLUS.37
Table 2. Clinical Characteristics of ASPs, Cases and Controls
For the association (case-control) study, we randomly selected one affected sibling from each pedigree as a case. Allele and genotype frequencies were compared between cases and controls by Pearson's goodness-of-fit χ2 test and Armitage's trend test, respectively (http://ihg.gsf.de/ihg/snps.html). For multiple logistic regression analysis, we used SPSS v12.0 (SPSS Inc., Chicago, IL). The population-attributable fraction (PAF) was calculated employing PARC (www.miner.rochester.edu/cpm/education/mach/productspubs.html).
Affected Sib Pair Analysis Identified ABCG8 Gene Variant as Susceptibility Factor for Gallstones.
In an affected sib pair study, samples of sib pairs both affected with the disease are collected, and the allele frequencies at a specific gene locus are compared with the null hypothesis of no linkage.26 If the disease is linked to the gene locus, there is a higher probability that both siblings share the same alleles at this locus. For the present study, we recruited 178 siblings with gallstones from 84 families. Table 1 illustrates that most pedigrees (89%; 75 of 84) consisted of 2 affected siblings, but in addition, 8 pedigrees with 3 affected siblings and 1 pedigree with 4 affected siblings were identified.
ABCG5 and ABCG8 variants were successfully genotyped in all siblings. Figure 1 summarizes the allele and genotype distributions of the SNPs in the cases and controls. The genotype distributions in cases and controls did not deviate from Hardy-Weinberg equilibrium (exact tests, all P > 0.05). The frequencies of all alleles (Fig. 1) were in line with allele frequencies previously reported in the Entrez SNP database as well as in other studies.23, 25, 28, 29, 33–35 Furthermore, linkage disequilibrium in this region was indicated by a maximum D′ of 0.92 in the combined cohort of index cases and controls (Table 3; Supplementary Fig. 1 online).
Table 3. Linkage Disequilibrium (Pairwise D′ and r2 Values) between selected SNPs of ABCG5 and ABCG8 Genes
The genetic linkage analysis of the ASPs revealed highly significant linkage between the presence of gallstones and carrying the D19H variant of the ABCG8 gene (NPL score = 7.1). Of note, this NPL score corresponded to a P value of 4.6 ×10-13. Because most of the pedigrees lacked parental genotypes, the NPL scores computed in the linkage analysis were sensitive to the allele frequencies used in the analysis. Therefore, we performed a sensitivity analysis for the D19H variant (Fig. 2, Tables 4 and 5). This analysis resulted in highly significant NPL scores for the total range of 19H allele frequencies (f = 0.03-0.13) previously reported.23, 25, 28–29, 33–35 Figure 2A illustrates that an NPL score of 7.1, calculated using the marker allele frequency estimate in our sample (f = 0.11), is a conservative estimate because the allele frequency is a result of selective sampling, higher than in other populations; lower frequencies would result in even higher NPL scores (Fig. 2A; Table 4).
Table 4. Sensitivity Analysis with GENEHUNTER-MODSCORE (GHM)— NPL Score for ABCG8 Variant D19H (Single-Point Analysis)
Frequency of 19H (c.55C) allele
Origin of data
1.10 × 10−22
1.35 × 10−21
1.86 × 10−17
Entrez SNP database
1.25 × 10−14
1.55 × 10−13
84 ASP families
4.60 × 10−13
1.23 × 10−10
2.03 × 10−8
1.03 × 10−6
Table 5. Sensitivity Analysis with GENEHUNTER-MODSCORE (GHM)— NPL Score for ABCG8 Variant D19H (Multipoint Analysis)
Frequency of 19H (c.55C) allele
Origin of data
4.23 × 10−19
4.19 × 10−12
5.68 × 10−7
Entrez SNP database
84 ASP families
Of note, the NPL statistics we used are conservative, that is, they have less statistical power when marker information is missing.38 Accordingly, significant results were obtained when calculating the more powerful allele-sharing LOD scores according to Kong and Cox36 for the range of 19H allele frequencies (all allele-sharing scores > 17.0; data not shown) as well as in a multipoint NPL analysis of all 5 ABCG5/G SNPs together (Fig. 2B; Table 5). To account for potential bias in multipoint linkage analysis with SNPs that are in linkage disequilibrium,39 we also calculated NPL scores after removal of markers that were in strong linkage disequilibrium with another marker (as indicated by D′ > 0.7; r2 > 0.1), which still yielded significant results (all NPL scores > 3.3, or P < 9.7 × 10−6). Finally, with 23% of cases phenotyped by ultrasound (Table 2), we determined linkage with ABCG8 D19H exclusively in cases confirmed by cholecystectomy (symptomatic patients), obtaining a single-point NPL score of 4.3 (P = 1.78 × 10−8).
In contrast to ABCG8 D19H, no significant single-point linkage was detected for the other ABCG5/G8 SNPs (p.E604Q, p.Y54C, p.T400K, p.A632V).
Genetic Association Study Confirmed Lithogenic Effects of ABCG8 Variant D19H.
To obtain additional evidence that ABCG8 D19H represents a lithogenic risk factor, we performed an association (case-control) study. We randomly selected a single case from each of the 84 families in the ASP study and recruited 70 stone-free controls (Table 2).
Table 6 demonstrates that the ABCG8 D19H variant was significantly associated with the presence of gallstones in cases: the 19H (c.55C) allele was significantly more common in cases than in controls (OR = 3.018; P = 0.017; Table 6). Carriers of the 19H allele whether homozygous or heterozygous were significantly overrepresented in the cases compared with the controls (OR = 3.118; P = 0.019). Consistent with the mouse model,16 heterozygosity for the lithogenic ABCG8 allele is associated with gallstones in humans: 21.4% (18 of 84) of all gallstone patients carried the heterozygous D19H genotype, as compared with 8.6% (6 of 70) of the controls (OR = 2.954; P = 0.026). However, no association was observed for the other 4 ABCG5/G8 SNPs under investigation.
Table 6. Distributions of Alleles and Genotypes for ABCG8 D19H in Cases with Gallstones and Stone-Free Controls and Tests for Association
ABCG8 D19H (c.55G>C) allele/genotype
Count (frequency) of alleles/genotypes
Cases (2N = 168)
Controls (2N = 140)
χ2 Statistics, P values, odds ratios (OR), and confidence intervals (CI) were determined using contingency table statistics. ORs were calculated relative to the low-risk allele c.55G.
Test for association
Allele frequency difference test
Armitage's trend test
[C] ↔ [G]
[CC + CG] ↔ [GG]
[CG] ↔ [GG]
Because controls were significantly younger (P = 0.024) and leaner (P < 0.001) than the affected patients, we analyzed the effects of the known lithogenic risk factors age, sex, and BMI as well as the D19H variant separately by logistic regression analysis in cases and controls. In univariate analysis, only age (OR = 1.033; P = 0.026; 95% CI 1.004-1.063), BMI (OR = 1.244; P < 0.001; 95% CI 1.113-1.366), and D19H (OR = 3.167; P = 0.021; 95% CI 1.188-8.438) are significant risk factors for gallstones. We included these 3 risk factors as covariates in multiple logistic regression analysis and observed a significant effect for BMI (Exp[B] = 1.244; P < 0.001; 95% CI 1.133-1.366), a trend for D19H (Exp[B] = 2.651; P = 0.063), and no effect for age.
From the results of the association study, we calculated the population-attributable fraction (PAF), the proportion of the incidence of a disease in a population that is a result of the gene variant. The PAF corresponds to the proportion of the incidence that would be eliminated if exposure to the variant were absent. The PAF of ABCG8 D19H is 0.08, that is, the D19H variant contributed 8% to the total gallstone risk. This PAF is exactly in the range expected for polygenic traits such as gallstones, which are the result of the interaction between multiple genes and environmental factors.8
This combined linkage and association study provides strong evidence that the D19H variant in the first exon of the hepatocanalicular cholesterol transporter gene (ABCG8) on chromosome 2 contributes to gallstone formation in humans. Consistent with the results of our study, there is supporting evidence from animal and human genetic studies that the ABCG5/G8 transporter is a true lithogenic risk factor: (1) Wittenburg et al.16 mapped a susceptibility locus for gallstone formation to the murine orthologue of the ABCG5/G8 genes, using quantitative trait locus analysis in experimental crosses of inbred mouse strains, and (2) a recent linkage study of 715 individuals from 39 Mexican American families with gallstones provided suggestive evidence of linkage to gallbladder disease on chromosome 2p21,40 the site of the ABCG5/G8 genes.
Despite increasing evidence from mouse studies for gallstone susceptibility genes,15 none of the murine Lith loci have yet been confirmed in a human population.41 Thus, our study represents a successful translational study that confirmed an association of a mouse Lith locus (Lith9 on chromosome 17) as a susceptibility factor for cholelithiasis in humans or a human LITH gene. Rosmorduc et al.12 provided evidence that rare loss-of-function mutations in the ABCB4 gene encoding the hepatocanalicular transporter required for biliary phosphatidylcholine secretion lead to gallstone formation. These findings in humans are consistent with the spontaneous occurrence of gallstones in Abcb4-knockout mice.42 Taken together, these studies demonstrate that an identical set of genes might be critical in the regulation of the gallstone trait across species. Because the defect described by Rosmorduc et al.12 has been named GBD1 (gallbladder disease 1) in the OMIM database (www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM), and 2 additional loci yet to be cloned (GBD2 and GBD3) have been localized in a recent genomewide scan in Mexican American families,40 we name the ABCG8 variant identified here GBD4, as approved by the HUGO Gene Nomenclature Committee (01-05-07: www.gene.ucl.ac.uk/nomenclature/). In fact, the significant linkage results for the ABCG8 D19H variant in symptomatic cases only (as defined by history of cholecystectomy) suggest that the variant is linked to the presence of both gallstones and gallbladder disease.
Epidemiological studies have shown varied prevalence of gallstones among different ethnic groups, and it has been speculated that different populations have multiple distinct genetic susceptibility factors for gallstones.2, 7 In our population, the allele frequencies of all ABCG5/G8 SNPs were within the range reported previously in other white populations. Because marker allele frequency critically influences linkage scores, we performed a sensitivity analysis over a broad range of frequencies. Within the range of reasonable estimates covering the allele frequencies reported by the Entrez SNP database and by previous studies, all resulting NPL scores were well above the level for significance (Fig. 2, Tables 4 and 5).
To assess whether the association results for the D19H variant were affected by confounding risk factors for gallstones, we performed multivariate analysis that included the D19H variant and lithogenic risk factors. In this analysis, only age and BMI remained significant, but the power of this analysis was limited by the sample size of our case-control study. Of note, the findings of the present study are in line with the results of a large genomewide association scan in German gallstone patients that localized susceptibility for gallstone disease to the same ABCG8 mutation.43 The odds ratio for heterozygous 19H carriers in that case-control study (OR 2.2)43 was comparable to that in the present study (OR 3.0), indicating a robust association of ABCG8 with gallstones in different populations.
Importantly, a vast amount of functional data supports the hypothesis that ABCG8 D19H might affect ABCG5/G8 transporter function. Recently Berge et al.22 and Gylling et al.23 showed that the 19H allele is associated with markedly reduced serum levels of the plant sterols campesterol and sitosterol, which are cholesterol absorption markers in healthy and hypercholesterolemic individuals, respectively. The association between serum plant sterol concentration and this nonconservative substitution suggests that the change from aspartic acid to the polar histidine at amino acid 19 alters ABCG5/G8 function.22 Because serum plant sterol levels are lower in 19H carriers, the predicted change would be expected to increase transporter expression or function, although effects on ABCG5/G8 function can only be determined after reconstitution in in vitro sterol transfer assays.44
Our previous study of siblings with gallstones25 and other reports23 have also observed significantly lower serum levels of total and LDL cholesterol in 19H carriers, indicating that D19H represents a gain-of-function variant that increases ABCG5/G8-mediated removal of plant sterols and cholesterol into bile and intestine. Hypersecretion of cholesterol from liver with cholesterol supersaturation of bile represents the common defect in patients with cholesterol gallstones.4, 18, 45 Furthermore, the 19H allele has also been associated with higher serum levels of cholesterol precursors (cholestanol, lathosterol),23 indicating increased cholesterol biosynthesis, as has been observed in obese gallstone patients.46, 47 These findings taken together suggest low intestinal cholesterol absorption and high compensatory cholesterol biosynthesis in carriers of the ABCG8 D19H variant, as hypothesized previously by Gylling et al.23 Of note, the up-regulated cholesterol biosynthesis in these subjects explains the greater efficacy of HMG-CoA reductase inhibitors such as atorvastatin therapy in 19H carriers.28 In fact, statins are reported to decrease the cholesterol saturation index of duodenal bile and to dissolve gallstones in some patients48 but not in gallstone patients in general.49
We have proposed50 that the identification of common gallstone genes (LITH or GBD) should be based on complementary evidence from (1) genetic studies in mice and/or humans, (2) phenotypic characterization of genetically defined animal models (for example, knockout or transgenic mice), and (3) epidemiological studies defining the overall clinical relevance. All these criteria are fulfilled by ABCG5/G8. Given the estimated PAF for the ABCG8 D19H variant of 8% and the estimation that genetic determinants account for 25% to 29% of the phenotypic variation in gallstone disease,10, 11 approximately 30% of the genetic effects might be a result of this specific risk variant. Accordingly, it is likely that other, yet unidentified genes contribute to the overall genetic risk for gallstones in the general population.
In conclusion, we have identified ABCG5/G8 as the first common susceptibility factor for gallstones in humans. Interestingly, because the ABCG8 19H allele is associated with both increased cholesterol biosynthesis and better response to HMG-CoA reductase inhibitors,28 we speculate that statins might be selectively antilithogenic in 19H carriers. In the future, additional genetic risk factors are likely to be identified, with the whole gene ensemble providing novel means of risk assessment, prevention, and therapy of gallstones.
We thank all patients for participating in this study and providing blood samples; Siegfried Matern and Klaus Zerres (Aachen), who initiated and supported this study; and Gudrun Hack and Hildegard Keppeler (Bonn) for excellent technical assistance.