Mutations in protein kinase C substrate 80K-H (PRKCSH), encoding for the protein hepatocystin, cause autosomal dominant polycystic liver disease (PCLD), which is clinically characterized by the presence of multiple liver cysts. PCLD has been documented in families from Europe (Netherlands, Belgium, Finland) as well as from the United States. In this article, we report results from extensive mutational analysis of the PRKCSH gene in a group of 14 PCLD families and 65 singleton cases of Dutch and Finnish descent with multiple simple liver cysts. We identified PRKCSH mutations in 12 families and in 3 sporadic cases. In 8 of 10 Finnish families we detected the 1437+2delTG splice-site mutation. In Dutch families, we found 2 other mutations that affect correct splicing of PRKCSH: 292+1 G>C (2 families) and 1338-2 A>G (1 family). In another Dutch family, we detected a novel deletion (374-375delAG) in exon 6, predicting an abnormal shortened protein. Investigation of the carrier haplotypes identified a common founder chromosome in unrelated individuals in each of the 3 identified splice-site mutations. In 2 Finnish families with dominantly inherited PCLD, and in 62 of 65 sporadic cases with multiple simple liver cysts, we failed to demonstrate any PRKCSH mutation. This corroborates the notion that autosomal dominant PCLD is genetically heterogeneous. In conclusion, we propose that, on the basis of our results, genetic screening for PRKCSH gene mutations should be limited to patients either with a positive family history for PCLD or who have severe polycystic liver disease. (HEPATOLOGY 2004;39:924–931.)
Autosomal dominant polycystic liver disease (PCLD) describes a condition in which numerous cysts are scattered throughout the liver parenchyma.1 It is clearly different from autosomal dominant polycystic kidney disease type 1 (ADPKD-1), and type 2 (ADPKD-2), in which patients have polycystic kidneys and often a polycystic liver.2 In PCLD, cysts arise from overgrowth and subsequent dilatation of intralobular bile ductules. Symptoms in PCLD arise from the mass effect of the cysts and include abdominal distension, early satiety, dyspnea, and back pain due to hepatomegaly. Other complications of PCLD, such as intracystic hemorrhage or rupture of cysts, can cause acute abdominal pain. Extrahepatic signs of PCLD may include intracranial aneurysms3 and mitral valve abnormalities.1 Women generally have a larger number of cysts than affected males, and, in general, the number and size of liver cysts increase with age. PCLD is, in most instances, left untreated, but in those cases with severe hepatomegaly, cyst fenestration, partial hepatectomy, and even liver transplantation may be warranted.1, 4
One of the first suggestions that PCLD might represent a separate hereditary condition stems from a report by Peltokallio.5 He reports that 29 out of a series of 117 PCLD patients had relatives with the disease and that analysis of the pedigrees was consistent with autosomal dominant inheritance with variable penetrance. Indeed, clinically, PCLD appears to be a distinct hereditary disorder, because renal cysts, typical in ADPKD-1 and ADPKD-2, seem to be lacking.6 The first evidence that PCLD differed genetically from ADPKD-1 or ADPKD-2 stems from linkage studies, which excluded the PKD1 or PKD2 locus in a number of PCLD families.7, 4 In 2000, the PCLD susceptibility gene was localized to a 12.5-centimorgan (cM) interval on chromosome19p13.2-13.1.8 Subsequently, protein kinase C substrate 80K-H (PRKCSH)9 was identified by two independent groups as the PCLD susceptibility gene.10, 11 Hepatocystin, the protein product of PRKCSH, is a 527-amino acid protein of 59 kd that has been discovered in a search for protein kinase C substrates.12 Later studies identified it as the noncatalytic beta subunit of glucosidase II.12 Northern blotting experiments suggest a wide tissue expression of the PRKCSH transcript.10 Computational analysis reveals that the protein contains a signaling sequence, a low-density lipoprotein domain, 2 putative calcium-binding EF hands, and a Hisp-Asp-Glu-Leu (HDEL) sequence at the C-terminal end that suggests a localization in the endoplasmic reticulum (ER).10
The initial studies indicated that the germline mutations in PCLD patients were spread from exon 4 through exon 16 of the PRKCSH gene (Fig. 6). We detected 2 different missense mutations affecting splicing, while others discovered mutations that affected splicing or changed the reading frame.10, 11 Up to now, a total of 8 different PRKCSH mutations have been described. It is not known for certain whether genetic heterogeneity plays a role. In a previous linkage study using Finnish PCLD families, it was suggested that a number of families were unlinked to the PCLD gene locus.6 So far, mutations have only been detected in individuals from large PCLD families. It is unknown whether patients without a family history but harboring multiple liver cysts carry PRKCSH germline mutations. To this end, we designed the current study with 3 major aims: (1) to define the underlying molecular basis in a cohort of apparently unrelated families affected by PCLD; (2) to determine if PRKCSH mutations are present in patients with polycystic livers without a positive family history for the disease; and (3) to detect possible founder effects. In this article, we report the mutational analysis of PRKCSH in a group of Finnish and Dutch PCLD patients. We identified 4 mutations, 3 of which were splicing mutations. We have also investigated the carrier haplotypes associated with the mutations observed, and our data are consistent with a founder effect in 3 of the detected mutations. Finally, we report evidence for genetic heterogeneity of PCLD, with 2 PCLD families not linked to chromosome 19p13.2 and sporadic PCLD cases without demonstrable mutations in the coding region of PRKCSH.
PCLD, autosomal dominant polycystic liver disease; ADPKD, autosomal dominant polycystic kidney disease; cM, centimorgan; PRKCSH, protein kinase C substrate 80K-H; ER, endoplasmic reticulum; PCR, polymerase chain reaction; lod, logarithm of odds; SSCP, single strand conformation polymorphism.
Patients and Methods
We drew blood samples from members of Finnish or Dutch families with PCLD and from singleton cases after obtaining written informed consent. A brief medical history was taken and all persons were subjected to ultrasonography, computer tomography, and/or magnetic resonance imaging of liver and kidneys. A standardized set of criteria for the diagnosis PCLD was used throughout the course of the studies.8, 11 Singleton cases were diagnosed as having multiple simple liver cysts when more than 4 liver cysts were present.
To study whether a PRKCSH mutation could be found in patients not fulfilling all these criteria, we also collected and analyzed blood samples from 2 Finnish families (#14 and #15; Fig. 1), both of which were excluded from a previous study because some family members also had kidney cysts.6
Dutch PCLD families and singleton cases with multiple simple liver cysts were collected at the Department of Gastroenterology of University Medical Center St. Radboud, Nijmegen, The Netherlands. To recruit patients, we screened medical records of all patients seen between 1980 and 2002 at our outpatient clinic. To investigate whether germline PRKCSH mutations are important in the development of single hepatic cysts in the general population, we also selected a population of Dutch patients who underwent screening abdominal ultrasonography for other reasons, and who on examination had either 1 or 2 liver cysts. Finnish patients were recruited as outlined elsewhere.6 Briefly, PCLD patients were recruited from patients treated in the Helsinki University Hospital during the years 1987 through 1999 and by contacting the patient organization of PCLD patients in Finland. The study was approved by the local medical ethics committee (CMO [Commissie Mensgebonden Onderzoek] Regio Arnhem-Nijmegen) and by the ethics committee of the National Public Health Institute (Helsinki, Finland).
Genomic DNA was purified from peripheral blood by standardized methods. Fine mapping of chromosome 19p was performed with markers from the Généthon and Marshfield marker sets (Genethon: www.genethon.fr; Marshfield: http://research.marshfieldclinic.org/genetics/sets/combo.html). Marker order and intermarker distances were based on the existing Cooperative Human Linkage Center linkage map (http://gai.nci.nih.gov/CHLC). We genotyped DNA samples by polymerase chain reaction (PCR) amplification of genomic DNA in the presence of [α32P (32 orthophosphate)]dCTP [2′-deoxycytidine 5′-triphosphate], electrophoresis on a 6.5% polyacrylamide gel, and exposure to radiographic film.
Linkage analysis and calculation of pairwise logarithm of odds (lod) scores between the disease locus and each individual marker was performed using the MLINK and ILINK programs of the LINKAGE package (Laboratory of Statistical Genetics at Rockefeller University, New York, version 5.1).14 Statistical analysis was performed assuming an autosomal dominant inheritance trait. In view of the relative rarity of the disease, the frequency of the abnormal allele was set at 0.001. We used specific allele frequencies adopted from the Généthon and Marshfield databases. Each allelic profile was verified, and genotypes were checked for Mendelian segregation, using PedCheck software.15 Because of the late onset of the disease, the penetrance was set at 90%. Recombination frequency was assumed to be equal for males and females. Based on all 2-point lod scores, we calculated a summary multipoint map using the program FASTMAP (Laboratory of Statistical Genetics at Rockefeller University, New York).16 Haplotypes were constructed by manual inspection so as to minimize the number of crossovers in each family. After identification of a recurrent PRKCSH mutation, allele-sharing analysis was done in the families to determine the possible existence of a founder effect.
DNA Sequencing and Mutation Detection.
Mutation screening was done for the 17 coding exons that constitute the PRKCSH open reading frame. Genomic DNA was amplified by PCR with oligonucleotide primers complementary to flanking intronic sequences. Primers were designed using the Oligo 4.0 program (Table 1). The 50 μL reaction mixture contained 200 ng of genomic DNA, 10 mmol Tris-HCl (pH 9.0), 50 mmol KCl, 0.1% Triton, 2 mmol MgCl2, 0.25 mmol dNTP's, 100 ng of forward and reverse primers and 3.0 U Taq –DNA –polymerase. The PCR conditions were 5 minutes at 95 of the used primer set, 1 minute at 72) according to the manufacturer's manual and the primers used in the PCR reaction. Sequences were analyzed on an ABI3700 capillary sequencer (Perkin Elmer Applied Biosystems).
Table 1. PCR Primers Used for Mutation Detection of PRKCSH
NOTE. Primers were designed to generate a 200–400 base-pair large amplicon and to include exon/intron boundaries.
PRKCSH intron 3R
PRKCSH intron 3F
PRKCSH intron 5R
PRKCSH intron 5F
PRKCSH intron 6R
PRKCSH intron 6F
PRKCSH intron 7R
PRKCSH intron 7F
PRKCSH intron 9R
PRKCSH intron 9F
PRKCSH intron 12R
PRKCSH intron 12F
PRKCSH intron 14R
PRKCSH intron 14F
PRKCSH intron 17R
Single Strand Conformation Polymorphism (SSCP).
For SSCP, PCR products (4μL) were diluted 1:1 with loading buffer (95% formamide, 20 mmol ethylenediaminetetraacetic acid, 20 mol NaOH and 0.05% bromophenol blue/Xilexe Cyanol), denatured at 95. 28>C, cooled on ice, and loaded on a 12% polyacrylamide gel with 10% glycerol. Electrophoresis was carried out on a vertical electrophoresis apparatus (Hoefer Pharmacia-Biotech Inc., San Francisco, CA) for 16 hours at 200 V at 4 28>C. Strands were visualized using ethidiumbromide.
The 292+1 G>C and the 1338-2 A>G mutations were studied using PCR followed by restriction fragment length polymorphism. The primers used for the PCR were PRKCSH intron 3 F and PRKCSH intron 5 R for the 292+1 G>C mutation and PRKCSH intron 14 F and PRKCSH intron 17 R for the 1338-2 A>G mutation (Table 1). A 393 or 603 base-pair product was amplified and subjected to digestion with restriction by either the enzyme DdeI or BanI (New England Biolabs, Beverly, MA). Digested samples were run on a 3% agarose gel (Biozym, Landgrant, The Netherlands) and stained with ethidiumbromide. The wild type at base pair 292+1 produced only the 393 base-pair fragment, while the heterozygous variant showed 393, 207, and 186 base-pair fragments. The wild type at base pair 1338-2 produced 428 and 175 base-pair fragments; the heterozygous variant showed 428, 383, 175, and 45 base-pair fragments.
Mutational Analysis Strategy.
Mutational analysis was performed according to the following strategy. First, we subjected all Finnish and Dutch familial cases to haplotype analysis using 7 to 13 highly polymorphic microsatellite markers surrounding the PCLD locus. Next, we analyzed PCR amplicons of each of the 17 coding exons of PRKCSH by direct sequencing from selected affected family members. DNA from all singleton cases were investigated by (1) restriction enzyme analysis of mutations detected in the course of this study, and (2) SSCP analysis on all PRKCSH exons. For confirmation of the SSCP results, samples with conformational changes in the SSCP analysis were subjected to direct sequencing. We sequenced PRKCSH completely in 12 of 65 Dutch and Finnish singleton cases.
Collection of Samples.
We collected samples from 4 Dutch families (#5, #19, #20, #21; Fig. 1) with a total of 8 members (5 affected), from 50 singleton Dutch cases with multiple simple liver cysts, and from 37 Dutch cases with either 1 or 2 hepatic cysts. There were 10 Finnish PCLD families with a total of 74 members (27 affected) and 15 singleton cases with multiple simple liver cysts in whom an obvious family history was absent (Fig. 1). Clinical and genetic data for Finnish families #6 to #13 have been published elsewhere.6
Lod Score Analysis.
First, we subjected the 10 Finnish families to linkage analysis using 7 polymorphic microsatellite markers. The 4 Dutch families were too small to yield meaningful results. We obtained evidence for linkage with the PCLD locus on chromosome 19p13.2 in 8 of 10 of the Finnish families (#7, #8, #10-#15). Table 2 summarizes the results: the highest 2-point lod score was obtained at θ 0.05 with marker D19S581 (3.06) and D19S221 (3.56).
Table 2. Two-Point Lod Scores at Various Recombination Fractions
NOTE. Scores are for the PCLD locus and for a set of 7 markers on chromosome 19p13.2 in families #7, #8, and #10–15.
Multipoint Lod Score Analysis.
Two Finnish families (#6 and #9; Fig. 1) fulfilled the complete criteria for PCLD, but 2-point linkage analysis with markers surrounding the PCLD gene locus yielded only negative lod scores. We therefore decided to subject the results to multipoint analysis. Seven markers encompassing 4.6 cM on chromosome 19p13.2 were used to generate a multipoint map (Fig. 2) of the region. Both families obtained negative lod scores in the multipoint analysis, making it unlikely that this region harbors the gene for PCLD in these families. The results were corroborated by inspection of the haplotypes, which showed that no single block of markers was shared among affected family members (data not shown) within either family. This suggests that PCLD is genetically heterogeneous and that more than a single gene can be responsible for the disease.
Mutational Analysis in Familial Cases.
Sequence analysis of the PRKCSH gene in Finnish DNA demonstrated a deletion of base pair doublet (TG) at the splice-donor site of exon 16 (Table 2, Fig. 3). This mutation, 1437+2delTG, is predicted to affect splicing, likely results in skipping of exon 16, and encounters a premature stop codon in exon 17. The heterozygous mutation was present in all 27 affected individuals from the 8 linked Finnish families (#7, #8, #10-#15; Fig. 1). This was confirmed by SSCP analysis of PCR products, which showed that the mutation was absent in the 25 remaining unaffected family members (Fig. 4). In one Dutch family (#19), a heterozygous 1338-2 A>G mutation at the splice-acceptor site of exon 16 was present in the single available case. A heterozygous splice-donor-site mutation 292+1 G>C was found in 3 affected individuals from 2 separate Dutch families (#5 and #20; Fig. 1). A single Dutch patient from family #21 carried a heterozygous deletion of AG (374-375delAG) in exon 6 (Fig. 3). This mutation is expected to produce a stop codon after base pair 434 of the PRKCSH gene. All affected individuals from PCLD families carried a single mutation, consistent with cosegregation of the sequence variant with disease under a dominant model. The detected mutations were absent in a set of 200 control chromosomes.
Effect of Genotype on Severity of PCLD.
We then compared the severity of the disease between 2 Finnish linkage-negative families (#6 and #9) and the 8 linkage-positive families. We divided the patients into 2 groups: those with mild or moderate symptoms and affected livers and those with severe PCLD, in which the livers were full of cysts. Patients in the linkage-negative group had a milder phenotype, with only in 6 of 13 (42%) patients classified as having severe disease, compared to 18 of 27 (67%) in the linkage-positive group. One confounding factor may be the age of the patient, given the age-dependent increase of liver cysts. However, the difference in age between these groups cannot account for the observed difference because the patients with mild disease from the linkage-positive group were generally younger (52 ± 22) compared to those from the linkage-negative group (63.7 ± 12.1). The number of females was comparable in both groups (linkage-positive 6/9 vs. linkage-negative 5/7). This suggests that, in the Finnish population, PCLD patients not linked to PRKCSH have a milder phenotype.
Renal Cysts and PCLD.
In 2 Finnish 1437+2delTG-positive families (#14 and #15), we detected patients who had, apart from their polycystic livers, 1 or more renal cysts. In family #14, a 77-year-old woman with a very large polycystic liver also had 3 renal cysts under 1 cm in diameter on ultrasound examination. A 43-year-old woman from the same family had a large (6 cm in diameter) single renal cyst and a polycystic liver. In another family (#15) a 64-year-old woman had 3 renal cysts and a polycystic liver, and a 67-year-old male had 10 renal cysts along with a polycystic liver. In total, 4 of 15 singleton Finnish patients with multiple simple liver cysts also had multiple simple renal cysts, and PRKCSH mutations were absent in these patients. It should be noted that the predominant feature in all of them was the polycystic liver.
PRKCSH Mutations in Patients With Multiple Single Liver Cysts.
We genotyped DNA from 65 adults with multiple simple liver cysts but without an obvious positive family history of the disease. Using restriction endonuclease digestion we detected the 1437+2delTG mutation in 3 of 15 Finnish samples. None of the 50 Dutch patients with multiple simple liver cysts had any of the previously detected mutations. SSCP of separate PRKCSH exons did not show differences with control alleles, making it unlikely that we missed a previously unrecognized PRKCSH mutation. Haplotype analysis was performed in 10 of 50 Dutch patients. Inspection revealed that one patient (#565) shared the founder haplotype of the 1338-2 A>G mutation, although segregation of the haplotype could not been determined to complete certainty because the patient's immediate family had died. Nevertheless, we failed to detect this mutation either by bidirectional sequencing or by restriction endonuclease digestion. Subsequent sequencing of all 18 individual exons of PRKCSH in this particular patient did not reveal novel sequence abnormalities. We were unable to detect pathological PRKCSH variants in the 37 samples of patients with either 1 or 2 hepatic cysts with restriction endonuclease digestion and SSCP analysis.
Haplotype Analysis of PRKCSH Mutations.
Since the 1437+2delTG allele was detected only in patients of Finnish origin, we suspected that this allele derives from a common founder. To resolve this issue, we constructed haplotypes of the 7 highly polymorphic markers by inspection of segregation within families (Fig. 5). All markers are located on a 4.9 cM stretch on chromosome 19 contig NT_011295. Markers D19S884, D19S586, D19S583, D19S581, and D19S584 are located centromeric of PRKCSH; D19S906 and D19S221 are located telomeric of the gene. We found a common disease haplotype with allele numbers arbitrarily designated 7-2-1-6-2-3-7 at markers tested in families 13 and 14 (Fig. 5). Recombination events within the various families limited the common haplotype to a single common allele (number 2) for marker D19S584; this was evident for families 10, 12, and 15. All patients with the heterozygous 1437+2delTG mutation, including the singleton cases, shared allele number 2 at marker D19S584. The marker distances define common inheritance of a shared region of chromosome 19 as maximally 1.2 cM in size. This suggests that the 1437+2delTG variant arose in a common ancestor. Previous analysis demonstrated that alleles carrying 1338-2 A>G, shared another founder haplotype.10 Our current data also suggest that carriers with the 292+1 G>C mutation share an ancestral haplotype. We detected that the 2 292+1 G>C-positive families (#5 and #20) in this study share the complete 2.9 cM haplotype between marker D19S586 and D19S221, while a third family (#4) reported earlier,10 only shared the haplotype defined by marker D19S906 (data not shown).
We present 4 PRKCSH mutations found among a series of patients and families of Finnish and Dutch origin. The new data are consistent with the previous observation that dominantly inherited PCLD can be caused by structural changes in PRKCSH, but we now have a broader view of the mutations that can cause this phenotype.
We provide evidence that a single splice-site mutation is responsible for PCLD in the linkage-positive Finnish families. Haplotype analysis suggests that the 1437+2delTG variant stems from a common (most probably Finnish) founder (Fig. 5). The relatively short length of the chromosomal region in which linkage disequilibrium is conserved suggests ancient origin of the mutation. A similar genotype (IVS16_1delGT) has been reported in 1 PCLD family by another group11 (Fig. 3). As this family is of Finnish origin, (its ancestors moved to the United States at the beginning of the 20th century; Dr. V. E. Torres, written communication, October 10, 2003), the variants are most likely identical at the genomic level and probably arose from the same founder.
The 1338-2 A>G and 292+1 G>C PRKCSH gene mutations have been described previously.10 These mutations too, are founder mutations; this is evidenced by haplotype studies. Indeed, all of the families carrying these mutations originated from 2 separate regions, the central and eastern regions of The Netherlands. We also detected a novel mutation (374-375delAG) that deletes 2 base pairs from the PRKCSH exon 6 sequence. This mutation is predicted to produce a stop codon after base pair 434 of the PRKCSH gene and lead to a shortened and abnormal protein.
The data from the present study show a very low mutational detection rate in sporadic cases. This suggests that PRKCSH screening is only worthwhile in families showing clear autosomal dominant inheritance. Although our detection tools have some technical limitations—both SSCP and direct sequencing may miss mutations—we believe that the prevalence of germline PRKCSH mutations is much lower in sporadic than in familial cases. On the basis of our results, we propose that genetic screening for PRKCSH gene mutations should be limited to patients with a positive family history or to those who present with severe polycystic liver disease.
The mutations identified in this study are predicted to have a profound effect on the resultant RNA sequence. Most likely, the mutations result in a shortened and abnormal hepatocystin. This corroborates the concept that, despite its autosomal dominant inheritance, PCLD arises from a loss of function of this protein. In general, PRKCSH mutations can lead to a loss of function either through a dominant-negative effect, haploinsufficiency, or a 2-hit model. The latter has been hypothesized to cause ADPKD, in which somatic mutations and a loss of heterozygosity have been detected in renal as well as liver cysts.17 Along the same lines, we favor a 2-hit model over the other mechanisms for several reasons. In the 2-hit model, patients with the inherited form of PCLD are born with 1 mutant PRKCSH allele and 1 wild-type allele, and they lose the second allele through a somatic mutation, ultimately leading to complete loss of PRKCSH expression. Only the progeny of these cells develop into cysts. In contrast, patients with the sporadic form of PCLD are born with 2 wild-type copies of the PRKCSH gene and have to acquire 2 independent somatic mutations before hepatic cysts develop. One result is that these sporadic cases tend to occur in patients who are older and less likely to have multiple cysts than their familial PCLD counterparts. This observation is in line with the fact that the prevalence of hepatic cysts in the general population increases with age.18
In the present sample, we did not detect an appreciable difference in phenotype between patients losing either a small part (≈15%; 1338-2 A>G) or a large part (≈81%; 292+1 G>C) of wild-type hepatocystin. However, both mutations will lead to loss of the carboxy-terminal domain, which contains an important Hisp-Asp-Glu-Leu (HDEL) peptide sequence. It is thought that this sequence is important for the protein to be retained in the ER. Mutation or deletion of the carboxy-terminal domain could disrupt binding to the ER and affect protein function. Studies are currently underway to investigate the effects of these mutations on protein function.
It is possible to make a brief comment about the presence of a possible genotype-phenotype correlation. Our data suggest that Finnish PCLD families with a 1437+2delTG mutation have a more severe phenotype than those without a mutation. However, it is difficult to establish a relation between the mutation on the one hand and phenotype on the other because we have identified only 3 mutations so far. It will be important to assess samples from more families that have been identified at both the molecular and clinical level.
The aim of the present study was to sequence and characterize the PRKCSH gene in a large cohort of PCLD patients. We were able to detect a heterozygous PRKCSH mutation in almost all cases with a positive family history for PCLD. We failed to do so in 2 families, and this, in addition to linkage data, suggests genetic heterogeneity, i.e., another locus for PCLD. A positional cloning effort to detect the genetic locus is currently underway. 3
Table 3. Distribution of PRKCSH Mutations Among PCLD Patients Sorted by Ethnic Group and Family History
NOTE. The 37 cases with 1 or 2 hepatic cysts are not listed.
1437 + 2delTG
292 + 1 G > C
1338-2 A > G
The authors thank Renate Smink (University Medical Center, St. Radboud, Nijmegen, The Netherlands) for her help in scrutinizing the medical records and collecting blood samples; Dr. Outi Vierimaa, (Oulu University Hospital, Oulu, Finland) for sending a DNA sample; and Dr. Juan S. Bonifacino (NICHD, Bethesda, MD for his support during the course of this study. We are grateful to the patients and families who participated in this study.