Abstract
Background Caudal regression syndrome (CRS) or sacral agenesis is a rare congenital disorder characterized by a constellation of congenital caudal anomalies affecting the caudal spine and spinal cord, the hindgut, the urogenital system, and the lower limbs. CRS is a complex condition, attributed to an abnormal development of the caudal mesoderm, likely caused by the effect of interacting genetic and environmental factors. A well-known risk factor is maternal-insulin-dependent diabetes mellitus.
Results In this pilot study, exome sequencing and copy number variation (CNV) analyses of 5 CRS trios implicate diabetes related genes, including MORN1, ZNF330, CLTCL1 and PDZD2. De novo mutations were identified in SPTBN5, MORN1 and ZNF330 and inherited damaging mutations in PDZD2 (homozygous) and CLTCL1 (compound heterozygous). In addition, a compound heterozygous mutation in GLTSCR2, a direct regulator of the CRS-related gene PTEN, was identified.Pathway based tests suggested the involvement of both pancreatic cancer (p < 1 × 10−4), and an immunity-related (p < 1 × 10−4) KEGG pathways.
Two CNV deletions, one de novo (3q13.13) and one homozygous (8p23.2), were detected in one of our CRS patients. These deletions overlapped with CNVs previously reported in patients with similar phenotype.
Conclusion Despite the genetic diversity and the complexity of the phenotype, this pilot study identified genetic features common across CRS patients including mutations in genes associated with diabetes.
BACKGROUND
Caudal Regression Syndrome (CRS;Caudal Dysgenesis Syndrome, Caudal Dysplasia Sequence, Congenital Sacral Agenesis; OMIM 600145) is a rare (1 in 7,500-100,000 births[1, 2]) congenital disorder characterized by varying degrees of spinal column agenesis associated with anomalies of central nervous, genito-urinary, cardiac, respiratory and gastro-intestinal systems [3] with anorectal malformations (ARMs) being the most common.
CRS has been attributed to an abnormal fetal development of the caudal mesoderm that occurs before the fourth week of gestation [4]. During the abnormal gastrulation, prospective notochordal cells that are wrongly specified in terms of their rostrocaudal positional encoding are eliminated. Eventually, fewer cells or even no cells will be available to form the notochord at a given abnormal segmental level. The consequences of such a segmental notochordal paucity are manifold and affect the development of the spinal column and spinal cord as well as of other organs that rely on the notochord as their inductor. If the prospective notochord is depleted a wide array of segmental vertebral malformations develop including segmentation defects, indeterminate or block vertebrae, or absence of several vertebrae. Because of lack of neural induction and absence of a floor plate, fewer prospective neuroectodermal cells will be induced to form the neural tube. The resulting malformation essentially depends on the segmental level and the extent of the abnormality along the longitudinal embryonic axis, with subsequent interference on the processes of primary and/or secondary neurulation [5]. However, what triggers such abnormal events is not known.
The caudal spinal abnormalities are considered the defining characteristics of CRS. Cama et al. [6] and Pang et al. [7] classified the disorder into 5 categories according to the degree of caudal spine involvement:Type I) total sacral agenesis with normal or short transverse pelvic diameter and some lumbar vertebrae possibly missing; Type II) total sacral agenesis without involvement of lumbar vertebrae; Type III) subtotal sacral agenesis or sacral hypodevelopment; Type IV hemisacrum and Type V) coccygeal agenesis.
As with many congenital disorders, a well-known risk factor for CRS is maternal insulin-dependent diabetes mellitus (type 1 diabetes)[8]. CRS has thye highest known relative risk (252) of any congenital diosorder [9]. While animal studies have shown that embryos exposed to higher levels of glucose develop growth anomalies, hyperglycemia has not been associated with abnormal fetal development in humans [10]. The exact mechanism by which maternal diabetes affects fetal development in humans remains unclear [11]. During normal pregnancies insulin sensitivity is reduced at the start of the third trimester in order to provide metabolic fuel for both mother and the developing fetus. However, since insulin is unable to cross the placenta, the fetus starts producing its own insulin in order to metabolize nutrition. It has been suggested that insulin, antibodies to insulin, or some other abnormality of carbohydrate metabolism could affect the development of a genetically susceptible fetus.
Evidence for a genetic cause is provided by the existence of familial segregation as well as animal models. Indeed, while the most severe forms of CRS present sporadically, milder CRS forms can be transmitted within families in a dominant manner with reduced penetrance and phenotypic variability [4]. Thus far, patients presenting association of sacral agenesis type IV , presacral mass, and ARM , a status known as Currarino syndrome, have been associated with mutations in MNX1 gene [12–⇓⇓⇓⇓17]. Yet MNX1 mutations account for only 50% of sporadic and 90% of familial cases [17]. Although private mutations in genes such as VANGL1 [18], HOXD13[19] and PTEN [20] have been described in sporadic cases with caudal dysgenesis and/or vertebrae anomalies, no firm genetic association has been established.
A CRS-like phenotype can be induced by administration in animals of retinoic acid (RA), lithium, cadmium, sulphamide, or organic solvents [21, 22].Several mutated genes including Cyp26a1, Hoxd13 [23], Wnt-3a [24], Acd, Ptf1a, and Pcsk5 underlie a CRS-like phenotype in mice [11, 25], yet mutations in the human orthologs have never been identified in CRS patients. Interestingly, the reverse is also true,Mnx1 (formerly Hxlb9)mutant mice do not present Currarino syndrome features[26]. These exceptions to the human-mouse phenotype correlation suggest differences in genetic etiology between humans and experimental organisms [11].
In order to search for genetic risk factors for CRS we have exome sequenced five sporadic CRS cases and their respective healthy parents. Due to the sporadic nature of the disease we have focused on de novo or recessive inherited damaging genetic variants. In addition, we investigated the relative burden of rare mutations in pathways. We also used a SNP chip assay in order to identify rare and de novo CNVs.
METHODS
Subjects
The records of patients treated between 1995-2010 at the Neurosurgery Department of Giannina (Genoa, Italy) and at the AbaCid-Genetica, Grupo HM Hospitales (Madrid, Spain) for congenital anomalies of the spine were reviewed. For all patients family history, cardiac, respiratory and endocrine status were collected. Neurological, neurophysiological (Somatosensory evoked potential, SEP), radiological, neuroradiologic (MRI), orthopaedic, physical, urological (urodynamic, cystography) and surgical assessments were performed for each case . For this pilot study, we selected four Italian trios (CR5, CR17, CR41, CR46) as well as one trio from Spain (CURR20). Please note that we amended either an A, B or C to the patient ID to indicated father, mother or child respectively. Selected cases had lower spine agenesis with additional anomalies of axial skeleton and internal organs in common. All cases were sporadic and only one children had a mother with diabetes type I. The local ethical committees approved the study and written consent was obtained from all patients and parents.
Bioinformatics Processing
Capture, alignment and base-calling
Whole exome sequencing (WES) was performed at the Centre of Genomic Sciences of the University of Hong Kong. All five trios were exome sequenced using Illumina HiSeq PE100 and captured with TrueSeq Exome Enrichment kit (Illumina Inc.). Alignment was done using BWA MEM [27] against Human Genome HG-19. Duplicated reads were flagged with Picard-tools [28]. The GATK tool set was used to realign indels, perform base recalibration, remove duplicates, perform indel and SNP calling, and was used for genotype refinement to improve accuracy of genotype calls [29]. A hard filter was used to remove variants with insufficient quality, the GATK recommended criteria were used (see supplementary methods). Relatedness of our participants was investigated with PLINK [30]. We then screened variants for their potential pathogenicity and frequency, retaining for further analysis only variants that were rare. We considered a variant to be rare if its minor allele frequency was ≤1% in each of several public databases (see supplementary methods). We considered a variant to be potentially deleterious according to a score obtained from KGGSeq [31]. KGGSeq’s prediction algorithm makes use of available biological information (the mutation’s effect on the gene, i.e. stop gain or loss, frameshift, splice site, missense), as well as scoring from other publicly available prediction algorithms (PolyPhen-2, SIFT and others).
De novo and compound heterozygous
Single nucleotide variants (SNVs) and small indels
Subsequent analysis of de novo and compound heterozygous, as well as homozygous, mutations was performed using KGGSeq [31]. We defined a de novo mutation as a first time genetic alteration of a specific locus in a proband. Compound heterozygous mutations were defined as the condition in which two nonsynonymous recessive alleles (MAF≤1%) affect two different loci within a certain gene and were inherited from separate parents. The expected probability of de novo mutations was evaluated using the framework of Samocha et al. [32] who approximated de novo mutation probabilities for each gene. Since a similar framework was not available for compound heterozygous mutations we made use of the only large control trio dataset publicly available, the Genome of the Netherlands (GoNL) [33]. The GoNL is a population dataset containing 250 unaffected parents-offspring trios. On the basis of this data we prioritized genes with a higher than expected frequency of compound heterozygous mutations. Thus we classed as a candidate risk locus any gene for which a recessive or de novo model could be constructed in any of our trios using the set of rare potentially deleterious variants we had identified. Detected de novo, compound heterozygous and homozygous mutations were validated using Sanger sequencing.
Analysis of kinship revealed misattributed paternity within one family (CR46). Hence the family CR46 was excluded from all family based analyses (de novo, compound heterozygous and homozygous mutation analysis).
Copy number variation
Several groups have reported methodology and software for calling CNVs from exome sequencing data. We tried 3 programs (EXCAVATOR [34] , CoNIFER [35], , and CONTRA [36]) for this and found no consistency between used tools. Confirming previous studies demonstrating limited power of CNV detection tools from exome sequencing data [37].
Thus we investigated copy number variation (CNV) in the families CR5, CR17 and CR41 with Illumina’s HumanCoreExome-24 beadchip. Quality control of the assayed genotypes was performed using GenomeStudio (Illumina Inc.) using the default settings. CNV calling and de novo CNV detection was performed with PennCNV [38]. We then screened CNVs for their potential pathogenicity as follows. We retained for further analysis CNVs which allowed construction of a recessive disease model for any gene in any of our trios. We also retained de novo CNVs and rare CNVs. We deemed a CNV to be rare if it did not overlap with any CNV detected in the 1000 Genome Project. De novo CNVs were validated by quantitative real-time PCR (ABI Prism 7900 Sequence Detection System; Applied Biosystems) using TaqMan®Copy Number Assay. Ensembl’s genome browser was used investigate CNVs on their overlap with genes or regulatory elements [39].
Pathway based test
Analysis was extended by investigating the burden of rare (MAF≤1%) nonsynonymous variants in KEGG [40] defined pathways (174 different pathways). We compared the burden of rare potentially deleterious mutations in patients versus controls in each pathway (see formula below). Controls were acquired from the UK10K project [41]. This control cohort consisted of 929 individuals affected with psychiatric disorders. This dataset was used as control due to its good coverage, making it ideal for investigations of rare variants.
We applied an adjusted burden test to identify pathways with significantly higher burden of rare potentially deleterious mutations in patients versus controls. Due to the differences in population [42] and calling platforms, we modified the classical burden test by standardizing the test statistic U by the mean number of rare nonsynonymous mutations: where n is the number of subjects in the sample, is the sample mean of the number of rare mutations an individual carries across the genome,xi,j = {0,1,2} and is the genotype call for hether the ith subject carries no rare variants (0), a rare variant from one parent only (1), or a rare variant from both parents (2) in the jth position of the pathway. We estimated the null distribution of U via bootstrapping with 10,000 iterations. Hence we randomly choose a set of n=5 controls and computed the test statistic U 10,000 times to obtain a sample distribution of U,against which we compared U obtained from our cases to obtain a p-value. P-values were then controlled for multiple testing by reporting the Benjamini & Hochberg False Discovery Rate (FDR) [43]. Furthermore all HLA related genes were dropped from the pathways prior to the analysis. This was necessary because for our sample we used a different aligner software to that used for the UK10K variant calling. The complexity of the HLA region can present difficulties to aligners thus inducing unwanted technical bias [44]. In addition, we only considered those pathways in which patients and controls had at least three or more rare variants[45].
RESULTS
After extensive quality control and MAF (MAF≤1%) filtering we retained 127,344 variants of which 91.5% were known in dbSNP137. Out of these, 25,487 missense, 365 frameshift, 498 nonfram e-shift, 54 splicing, 273 stop-gain and 52 stop-loss variants in 18,212 different genes were analyzed in regards to de novo, compound heterozygous and homozygous mutations.
De novo variants
We identified two missense and one frameshift de novo mutations in three different genes:MORN1, SPTBN5, and ZNF330 in patients CR41C, CR5C, and CURR20 respectively (Table 2). MORN1 encodes MORN (membrane occupation and recognition nexus) repeats [46]. The exact function of this gene is not known, however, in Toxoplasma gondii it is known to be involved in nuclear cell division [47]. Furthermore MORN repeats are known to be part of a number of genes, including junctophilins [48] which are involved in cardiomyopathy [49]. Notably, MORN1 was reported to be produced by insulin producing cells (IPCs) derived from pancreatic stem cells [50]. The estimated probability for a de novo mutation to occur in MORN1 is 0.8%. The probability of having a damaging de novo mutation, is lower than that of MORN1 for 59% of the genes analyzed within the de novo framework, is lower than that of MORN1 [32]. Pathogenicity analysis by KGGSeq suggests that the de novo mutation is damaging.
SPTBN5 (OMIM:605916) is a beta-spectrin encoding protein. It plays an important role in linking proteins, lipids, and cytosolic factors of the cell membrane to the cytoskeletal filament systems of the cell [51]. SPTBN5 is expressed in the cerebellum, pancreas, kidney, and bladder, as well as in a number of others systems. The gene has not been associated with any disease or disorder. The gene-specific probability of a de novo mutation is 1.8% and 99% of genes have a lower probability making the de novo mutation less likely to be causally related. Further, KGGSeq’s pathogenic prediction algorithm suggests that this variant is benign.
ZNF330 (OMIM:609550) is a zinc finger protein with no known disease association and is mainly present in the nucleus during interphase as well as at the centromeres during mitosis [52]. Interestingly, this gene is differentially expressed in pancreatic the Islets of Langerhans and in peripheral blood mononuclear cells [53]. The approximated gene-based de novo mutation probability is 0.6%, relatively low but still within the 28th percentile of all genes.
Thus, given the pathogenic nature of the two de novo variants and their expression pattern in pancreatic cells, MORN1 and ZNF330 are potential candidate genes for CRS.
We detected one de novo CNV deletion on 3q13.13 in CR5C (Table 3). The deletion does not seem to encompass any gene or functional element, yet it overlaps with CNVs previously reported in patients with a similar phenotype. In particular, a documented de novo deletion in a Japanese patient with OEIS (omphalocele, exstrophy of the cloaca, imperforate anus, spinal defects) complex who also had also sacrum malformation (DECIPHER:971) overlaps with the de novo CNV identified in CR5C.
Homozygous and Compound Heterozygous Mutations
In total we identified 8 compound missense heterozygous and two homozygous (one missense and one non-frameshift) mutations (PDZD2, SYNGR1) which passed the described filtering criteria (detailed in Table 2). Strikingly, and as in the de novo variants, mutations in genes related to diabetes were detected in two patients. None of the affected genes were recurrent. The two genes associated with diabetes were PDZD2 and CLTCL1 and were found mutated in CR5C and CR17C respectively.
PDZD2 has been shown to be an important promoter of fetal pancreatic progenitor cell proliferation [54, 55]. Ma et al. [56] showed that expression of PDZD2 is specific to pancreatic beta cells. Furthermore, higher concentrations of secreted PDZD2 in rat insulinoma cell lines were correlated with higher rate of cell proliferation and inhibited transcription of INS, an insulin promoter.
CLTCL1 is involved in the intracellular trafficking of glucose transporter GLUT4 Intracellular trafficking of the glucose transporter GLUT4 from storage compartments to the plasma membrane is triggered in muscle and fat during the body's response to insulin [57].
Noteworthy, a compound heterozygous mutation in GLTSCR2 (Glioma Tumor Suppressor Candidate Region Gene 2) was identified in patient CR5C. GLTSCR2, is expressed at high levels in pancreas and heart, is a tumor suppressor gene and a direct regulator of PTEN. Mutations of PTEN have been previously identified in a patient affected with VACTERL (Vertebral anomalies, Anal atresia, Cardiac defects, Tracheoesophageal fistula and/or Esophageal atresia, Renal & Radial anomalies and Limb defects) [20] which has commonalities with CRS [25].
Additional compound heterozygous mutations were identified in DNAH10 (CR41), an inner arm dynein heavy chain. Dynein proteins are implicated in many disorders such as motor neuropathies, cortical development diseases, as well as congenital malformation such as heterotaxia, situs inversus. Moreover, cytoplasmic Dyneins have been reported to interact with Kinesin (KIF1A, mutated in patient CR17) for Interkinetic nuclear migration in neural stem cells [58].
We detected a homozygous CNV deletion encompassing part of 8p23.2 in patient CR5C (Table 3). The CNV does not overlap with known genes but is contained within a duplication found in a patient with abnormal sacrum (DECIPHER:271204). This documented patient, while also harboring another deletion (7q34-7q36.3), displayed a great variety of phenotypes including central hypertonia, hypermetropia, long thorax, narrow mouth, seizures, and strabismus, as well as deeply set eyes. Additional detected rare CNVs (see supplementary Material) were overlapping with a number of other genes. However, none were directly related to CRS.
Pathway-based tests
KEGG pathways were tested for an excess of rare nonsynonymous mutations in cases versus controls. Table 4 lists the 10 most significantly burdened pathways. After adjusting for multiple testing by controlling for a False discovery Rate (FDR) of 0.05, two pathways remained significant; the KEGG pathway for pancreatic cancer (p < l × 10−4) and graft versus host disease (p < l × 10−4). Inspection of the QQ-plot (Figure 1) indicates overall robustness of the test.
Although the above reported candidate risk loci are not part of either of these two pathways, the enrichment of mutations in genes of the pancreatic cancer pathway is in line with our finding of mutations in diabetes-related genes and with previously reported results.
DISCUSSSION
We have identified a number of potential genetic causal mechanisms for CRS. Here we discuss the relevance and implications of these findings, and will outline possible directions for future studies. In addition to the already known genetic risk factors [11], we have been able to identify a number of novel risk loci potentially connected to CRS.
Foremost, all four patients were affected by a homozygous, compound heterozygous mutations or de novo variant in a diabetes-relevant (CLTCL1 and PDZD2) or pancreatic expressed (MORN1 and ZNF330) gene. This result is in line with the increased risk of CRS in children born to diabetic mothers. In addition, one de novo (3q13.13) and one homozygous CNV (8p23.2) overlap with CNVs reported in patients with similar phenotype. Identification of overlapping CNVs in patients with similar phenotype is the central aim of DECIPHER. Since many patients with rare diseases harbor novel or extremely rare variants, it is crucial to accumulate evidence across different patients in order to foster our genetic understanding of the disease and its mechanism. Furthermore, we identified a heterozygous mutation in GLTSCR2, a direct regulator of PTEN. PTEN has been previously associated with CRS-like phenotypes (VACTER) [11]. Finally, we would like to point out that our patients have a significantly greater number of rare deleterious mutations in the pancreatic cancer and graft-versus-host disease pathways compared to controls. The excess of mutations in pancreatic cancer genes is a result which is consistent with previous research showing similar alteration of the Hedgehog protein (Hh) in patients affected with CRS compared to those suffering from pancreatic cancer [59] and it is tempting to speculate that this may somehow relate to insulin production. Furthermore, PTEN is known for its involvement in pancreatic cancer [60], indicating a connection between identified heterozygous mutation in the PTEN regulator GLTSCR2 and the higher burden of rare deleterious mutation within the pancreatic cancer pathway. Equally interesting but not so easily justifiable, is the involvement of the immunology-related graft-versus-host disease pathway, a rare disease characterized by an aggressive immune response of donated tissues against the host organism [61].
There are, however, a number of limitations to our study. First, the sample size is small, but to be expected given the rarity of the disease. Second, we were not able to identify recurrent affected genes across different patients. Third replication of our pathway burden test results is desirable given the differences in QC pipeline between cases and controls.
The diversity of identified potential disease mechanisms matches that of previous studies [11, 62–⇓64] and also reflects the phenotypic diversity associated with CRS [62]. Furthermore, we [65] have shown that given a complex genetic disorder, one should expect a large genetic heterogeneity across patients. Thus the number of candidate genes identified is not surprising and is similar to that reported for other complex rare genetic disorders [66]. Further the increasing amount of candidate genes as well as those reported by others [25, 67–⇓⇓⇓71] suggests that CRS might be caused by a multitude of private genetic risk factors. This makes it difficult to identify a common underlying genetic architecture. Differences in the genetic etiology between humans and experimental organisms makes it challenging to investigate the exact causal mechanism. In addition, some principal aspects of the disease are still unknown. A challenge, for example, is the uncertainty of the overall frequency of the disease. While some studies have estimated that 1 in 7,700 children might be affected by CRS [1], others suggest it might be as rare as 1 in 100,000 births [2]. This further complicates estimation of the expected number of disease causing mechanisms [72].
Conclusion
Despite the complexity of the phenotype, we were able to identify common genetic characteristics across patients, potentially causally related to the present phenotypes. Furthermore our data, although limited to a small group of patients, support a multigenic model for CRS. Future studies should consider larger accumulated samples across multiple centers in order to identify common genetic characteristics via whole genome or whole exome sequencing.
List of abbreviations
CRS, Caudal Regression; CNV, Copy Number Variation; ARM, Anorectal Malformation; SEP, Somatosensory Evoked Potential; WES, Whole Exome Sequencing; SNP, Single Nucleotide Polymorphism; SNV, Single Nucleotide Variation; MAF, Minor Allele Frequency; FDR, False Discovery Rate; IPC, Insuline Producing Cells; Hh, Hedgehog protein; QC, Quality Control
Competing interests
The authors declare that they have no competing interests.
Availability of data and materials
The corresponding sequencing data, on which this study is based on, can be accessed through the European Genome-phenome Archive (EGA).
Author’s contributions
RMP, DDC, SSC, GC and MMGB analyzed the data and drafted the paper. VC, EM, PDM and LR played a major role in collecting the samples and phenotyping the patients. PCS and PKT reviewed the study proposal, provided feedback on the study progress and mansucript. MMGB proposed the study idea.
Acknowledgements
We thank all participants who made this study possible.
Seed Funding Programme for Basic Research, University of Hong Kong, Project Code:201410159002 to MMGB.
Small Project Funding, University of Hong Kong, Project Code 201209176012 to DC. EM; PDM, VC would like to acknowledge Ricerca Corrente Ministero della Salute Italia 5X Mille, Aletti-Volpati Trust Onlus and private funding resources. We also thank A.S.B.I. (Associazione Spina Bifida Italia).
Footnotes
Email Addresses: Robert M. PORSCH(robert.porsch{at}hku.hk), Elisa MERELLO (ElisaMerello{at}ospedale-gaslini.ge.it) , Patrizia DE MARCO (PatriziaDeMarco{at}gaslini.org), Guo CHENG (guochenghnny{at}gmail.com), Laura RODRIGUEZ (laurarodriguez{at}abacid.es), Paul K. TAM (paultam{at}hku.hk), Stacey S. CHERNY (cherny{at}hku.hk), Pak C. SHAM (pcsham{at}hku.hk), Valeria CAPRA (valeriacapra{at}ospedale-gaslini.ge.it), Maria-Mercè GARCIA-BARCELO(mmgarcia{at}hku.hk), Desmond D. CAMPBELL (ddc123{at}hku.hk)
Maria-Mercé GARCIA-BARCELO / Room 1-05D, 1/F The Hong Kong Jockey Club Building for Interdisciplinary Research 5 Sassoon Road, Pokfulam, Hong Kong SAR
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.
- 74.
- 75.
- 76.
- 77.
- 78.
- 79.
- 80.