A 6-year-old girl with refractory hypercholesterolemia was referred to our institution for the development of a suitable therapeutic regimen. Before the study, atorvastatin had been administered to the patient for more than 2 years, but her plasma total cholesterol level was decreased to only ∼200 mg/dl and large platelets persisted. Additionally, the patient appeared to have atherosclerotic plaque (6 mm × 7 mm) in her left common carotid artery (Figure 1A, right) upon referral. The patient’s initial presenting clinical and laboratory features included profoundly increased levels of total cholesterol (638 mg/dl) and low-density lipoprotein cholesterol (LDL-C, 527 mg/dl); a reduced level of high-density lipoprotein cholesterol (HDL-C, 46 mg/dl) and a slight elevation of plasma β-sitosterol (3.26 mg/dl). Although there was an elevation of β-sitosterol, criteria for the diagnosis of sitosterolemia were initially felt not to be met (Salen et al., 1985). In addition, the patient had abnormally large platelets (Figure 1A, left); over 41% of her platelets had a size over 30 fl. Clinically, Bernard-Soulier syndrome, of which giant platelets are a feature, was excluded as a plausible diagnosis through examination of clotting time and platelet aggregation, which were all normal in the proband (Table 1).

To track genetic variants in the full pedigree, we also collected blood samples from maternal and paternal relatives (I-1,2,3,4, and II-1,4). DNA was extracted from 400 μl whole blood per sample using the (Qiagen, Hilden, Germany), and all blood samples were stored at −80°C before usage. For screening for candidate ABCG5 variants in a general population, we also enrolled individuals over 2 months in 2019, the Second Affiliated Hospital of Nanchang University, Beijing Anzhen Hospital of Capital Medical University, and Peking University Third Hospital. The samples were examined in the Second Affiliated Hospital of Nanchang University. In all, 30 patients had abnormal platelet size distribution and underwent ABCG5 gene sequencing.

We designed a custom panel of DNA oligonucleotide primers covering 216 amplicons using the Ion AmpliSeqTM platform (Supplementary Table S1, Life Technology, Thermo Fisher, USA). These amplicons covered all coding exons of four familial hypercholesterolemia (FH) causal genes (APOB, LDLRAP1, LDLR, and PCSK9). Sequencing Libraries were barcoded (IonXpress Barcode Kit, Life Technologies) and equalized (Ion Library Equalizer Kit) to a final concentration of approximately 100 pM. Emulsion PCR was performed using the OneTouch DL instrument, and template-positive Ion Sphere particles were enriched using the OneTouch ES instrument according to the manufacturer’s instructions. Sequencing was performed on a 318 chip on the Ion Torrent PGM following the recommended protocol. Reads were aligned to hg19 and variants were called using the TorrentSuite (version 4.0.2).

WGS was conducted in the proband and her parents. Each sequencing library with an average insert size of 250 bp was loaded into an Illumina HiSeq X ten. Roughly 90 Gb of high-quality data in 150 bp pair-end reads was obtained for each library, reaching an average of 30-fold genome coverage for each individual. The quality evaluation was performed using FastQC software, and sequences from adapters or having low Q-score were removed with cutadapt software (Martin, 2011). The sequencing reads were mapped to the reference human genome (hg19) by the BWA algorithm (Li and Durbin, 2009). PCR duplicates were removed with Picard software. Processed bam files of humans were processed via local indel realignment and base-quality recalibration using the Genome Analysis Tool Kit (GATK) (McKenna et al., 2010). Subsequently, sorted BAM files were used for SNV calling and the UnifiedGenotyper method based on a Bayesian genotype likelihood model was applied with GATK. All detected SNVs were annotated using ANNOVAR. VarSelect in TGex platform was applied to filter for variants in WGS data of the family trio, to identify variants most likely responsible for the phenotypes of the proband (Dahary et al., 2019).

The blood lipid measurements were routinely performed by the second Affiliated Hospital of Nanchang University. β-Sitosterol and other plant sterols were analyzed by liquid chromatography-mass spectrometry (AB SCIEX Triple Quad 4500) according to previous studies(Kasama et al., 1987; Hidaka et al., 1990). For each sample, a volume of 20 μl was injected into Agilent Eclipse Plus reversed-phase column (C18, 2.1 × 50 mm). The column temperature was maintained at 35°C. The mobile phase was acetonitrile-methanol (4:1, v/v) at a flow rate of 0.6 ml/min. The mass spectrometer was operated in positive ion polarity mode in the extended dynamic range (1,700 m/z, 2 GHz) with the following parameters: Curtain Gas (GUR) 40 psi; Collision Gas (CAD) 9; IonSpray Voltage (IS) 5,500.0 V; Temperature (TEM) 350°C; Ion Source Gas 1 (GS1) 50 psi; NC 3V. Blood samples from 10 healthy individuals were examined as control samples.

Two ml of peripheral blood was collected using standardized tubes (INSEPACK ST serials, Beijing, China) and all tubes were stored at room temperature. Within 30 min, these samples were analyzed on SysmexXE-2100 Haematology System (Sysmex Corporation, Kobe, Japan) for lipid profile and coagulation tests following the methods previously developed protocols (Table 1) (Barsam et al., 2011; Depoorter et al., 2015). Large platelets are identified based on the volume (>30 fl). For each sample, we prepared two blood films dyed by Wright’s dye for platelet morphology analysis using a BX53 microscope (OLYMPUS, Japan). The criteria for large platelets is a diameter greater than 4 microns. All blood films were reviewed by two independent examiners.

Var2 is located at a splice region of ABCG5, and we performed the splicing pattern assay to verify its functional effect. Genomic DNAs including reference (WT) or ABCG5 variant were isolated from peripheral whole blood of all individuals in the study with Blood Genomic DNA Mini Kit (CWBIO, CW 2087S) and PCR was performed using PrimeSTAR® MaxDNA Polymerase (TAKARA, R045A). The 947 bp PCR fragments of ABCG5 gene, spanning from part of exon 7 (44 bp) to part of exon 9 (37 bp) with exon8 (214 bp) and its flanking intron 7 and 8, were amplified by the primer pair, ABCG5-minigene_Forward (5′-GCG​GTA​CCG​CGG​AAA​TGC​TTG​ATT​TCT​T-3′) with KpnI restriction site and ABCG5-minigene_Reverse (5′-GCCTCGAG TTA​AAG​GAG​GAA​CAA​ACC​CAT​GA-3′) with XhoI restriction site. The genomic fragments (both WT and variant, MT) containing the intron of interest were cloned into the pcDNATM-3.1 (+) vector and then all cloned plasmids were verified by sequencing to confirm whether the insertion contained the WT or MT.

HepG2 cells were cultured in DMEM (Gibco) supplemented with 10% FBS (Gibco). After seeded in 6-well plates for 24 h, HepG2 cells were transiently transfected with 1ug prepared vectors with the allele of the WT or MT and corresponding empty vectors respectively, using Superfectin II In Vitro siRNA Transfection Reagent (Shanghai Pufei Biotech). After 48 h, total RNA was extracted by TRIzol reagent (Invitrogen) and 1 mg total RNA was used for reverse transcription using PrimeScript RT reagent kit with gDNA eraser (Takara) according to the manufacturer’s instructions. The primer sequences of quantitative RT-PCR were: ABCG5-MGQ_Forward (5′-CGG​TTA​CCC​TTG​TCC​TGA​AC-3′) located in exon 7 and ABCG5-MGQ_Reverse (5′-TGC​CAG​CTT​ATT​TCT​CAC​CA-3′) located in exon 9. Quantitative RT-PCR was performed with SYBR Green dye using ViiA7Real-Time PCR System (Applied Biosystems). The relative mRNA expression was calculated by the comparative Ct method using GAPDH as a control. PCR reactions were performed in triplicate. The RT-PCR products were separated on a 2% agarose gel and detected with Chemidoc Xrs Gel Doc Xr (Bio-Rad Universal Hood Ii 2, USA).

ABCG5/ABCG8 double knockout rat model was created by Beijing Biocytogen. In brief, a 19 Kb region was knocked out using a CRISPR/Cas9 system, with two sgRNAs targeting one site in the intron 4 of ABCG5, and the other in the intron 6 of ABCG8. The design was adapted from the protocol developed by Yu L, et al. The sgRNA activity was evaluated by the UCA^TM (Universal CRISPR Activity Assay), developed by Biocytogen. By zygote microinjection, transferred zygotes of SD rats were obtained and the founders were positively confirmed by PCR product sequencing. The genotyping primers were Forward: 5′-cta​ggt​cca​cca​agc​cat​gtg​aac​a and Reverse: 5′-att​ttc​tgg​gca​ccc​tgt​gtt​cca​c. The animal study was approved by the Ethics Committee of the Second Affiliated Hospital of Nanchang University (SYXK-20150001).

Exon amplification was performed using LA Taq (Takara, Osaka, Japan). Supplementary Table S2 listed all the primers for amplification of 13 exons of ABCG5 extending 50 bp towards both upstream and downstream to cover the splice regions and corresponding PCR protocols for each primer pair. All reactions were performed on a PTC-200 Peltier Thermal Cycler (MJ Research, MA, USA). PCR products were sequenced by Sangon Biotech (Shanghai, China), and the results were manually checked using SeqMan (DNAstar 5.0, WI, USA).

The t-test was used to compare the mean difference of the blood lipid levels in the rat model using R 3.4(R Core Development Team, 2018).

This proband had hypercholesterolemia with atypical changes in the β-Sitosterol level in comparison with previous reports (Supplementary Table S3). To identify pathogenic variants responsible for the hypercholesterolemia of the proband, a customized Gene Panel Test was initially performed to screen for variants within all coding regions of four common familial hypercholesterolemia genes: APOB, LDLR, LDLRAP1, and PCSK9. The sequencing results showed that no family member carried any rare coding variants (<1%, MAF in CHB of 1,000 genomes) in these genes. Next, we performed WGS for the proband and her parents (∼30X coverage on average for each person). Two heterozygous variants of the ABCG5 gene were found in the proband, and no other variant was observed in the known genes related to hypercholesterolemia. The proband inherited Var1 (chr2:44050062G/T, NM_022436.3:c.1337G>A, p. Arg446Gln) from her father and Var2 (chr2:44051355T/G, splice region, c.1118+3G>T) from her mother. These two variants (Var1 and Var2) were further verified by Sanger sequencing (Supplementary Figure S1).

Var1 was reported to be a variant of uncertain significance for sitosterolemia in both Clinvar (Accession: VCV000289811) and gnomAD (2-44050062-C-T, allele frequency: 6.74e-5 and no homozygote). While at the same position, p. Arg446Ter (Accession: VCV000030485), has been previously reported as pathogenic for sitosterolemia and hypercholesterolemia (Wang et al., 2014; Buonuomo et al., 2017; Pek et al., 2018). Var2 was not documented in either Clinvar or gnomAD; it was predicted to be a splice site variant. The blood total cholesterol level of the proband was similar to the level in previously reported individuals with biallelic ABCG5 variants, ranging from 116 mg/dl to 870 mg/dl (Supplementary Table S3).

Subsequently, these two variants were examined in all family members. The results showed that these variants were transmitted across all three generations in heterozygous form (Figure 1B). Var1 was specific to the paternal side, and Var2 was only detected in the maternal lineage. However, individual I-1 did not carry any variants in the ABCG5 gene. In addition, all maternal members heterozygous for Var2 exhibited the large platelet phenotype without dysfunction of blood coagulation (Figure 1C; Table 1; Supplementary Figure S2).

Since Var2 was predicted to be a splice site variant, we examined if Var2 can result in novel splicing transcripts of ABCG5 (Figure 1D). As shown in Figure 1E, the mRNA transcripts from the reference and Var2 fragments were 295 bp and less than 100 bp, respectively. Subsequent Sanger sequencing verified that this “truncated” transcript was 81 bp and excluded exon 8. Therefore, this result suggests that Var2 leads to deletion of exon 8 of ABCG5 during transcription, and the creation of a premature termination codon triggering nonsense-mediated mRNA decay.

Before the study, atorvastatin 10 mg/day was first administered to the proband and her plasma LDL-C level gradually decreased to around 200 mg/dl within 2 years (Figure 2A). Although the proband’s serum LDL-C level was reduced to about 150 mg/dl when atorvastatin was increased to 20 mg/day, she experienced severe adverse effects (Supplementary Table S4). By reducing atorvastatin back to 10 mg/day, the proband’s LDL-C level remained around 200 mg/dl.

Considering the insufficient response to atorvastatin which aims to inhibit cholesterol biosynthesis, we suspected that the proband may have an abnormality in cholesterol excretion instead. At that point, the treating physician suggested stopping cow milk intake (450 ml/day), following a similar infant case with a limited increase in β-Sitosterol where breastfeeding was arrested to ameliorate the symptoms (Rios et al., 2010).

After 1 month, the proband’s LDL-C level was reduced to 130 mg/dl without further decrease. The effect caused by the reduction of cholesterol intake supported the hypothesis of a cholesterol excretion dysfunction since this strategy is not effective for individuals with abnormal cholesterol synthesis. WGS analysis identified compound heterozygous variants in ABCG5 responsible for abnormal cholesterol excretion consistent with the known disorder sitosterolemia, which supported our hypothesis. Following this finding, the physician in our group recommended a combination of atorvastatin (10 mg/day) and Ezetimibe (10 mg/day) be administered to the proband.

Within 2 months, the patient’s LDL-C level was further decreased to an optimal level (<80 mg/dl). After treatment for 39 months, atorvastatin was discontinued but Ezetimibe treatment was continued with no milk intake. Thus far, the patient’s LDL-C level has been maintained at around 80 mg/dl without any obvious adverse effects. An ultrasound scan 2 months following the revised treatment showed that the proband’s atherosclerotic plaque disappeared from her left carotid artery (Figure 2B).

Interestingly, the proband’s macrothrombocytopenia nearly completely resolved within 1 month after the addition of Ezetimibe while the β-Sitosterol level was reduced to normal values (0.83 mg/dl). The size distribution of platelets normalized, and only 25% of platelets remained over 30 fl (Figures 2C,D). ABCG5 functional inactivation leads to relatively active NIPIC, targeting the NIPIC by Ezetimibe was shown to restore the balance of cholesterol absorption and excretion maintained by them. These results suggested that the macrothrombocytopenia of the proband was likely caused by the ABCG5 variant. More unexpectedly, all Var2 heterozygotes in the family had large platelets, and after 1 month of Ezetimibe treatment alone, the macrothrombocytopenia of her maternal uncle (II-2, Figure 2) also disappeared (Supplementary Figure S3).

Disruptions of the ABCG5/ABCG8 genes mimic phenotypes of the proband in double knockout rat.

To further determine if heterozygotes of ABCG5/ABCG8 genes could drive hypercholesterolemia and macrothrombocytopenia phenotypes, we created a double knockout rat (Figure 3A), having a growth curve similar to the WT strain without visible abnormalities (Supplementary Figure S4A) and succesful depletion of ABCG5 and ABCG8 expression in liver and intestine (Supplementary Figure S4B). As expected, the G5G8^--/-- rats had a significant elevation of blood lipid levels (Figure 3B, left) with large platelets (4 out of 6 animals) compared with wild-type rats. However, the alterations in blood lipids were not strictly correlated with the occurrence of large platelets (Figure 3B, right), further supporting the observations of phenotypic heterogeneity of ABCG5/ABCG8 variants. In particular, one heterozygous rat (G5G8^{+ +/- -}) also had this biased distribution of platelet size (Figure 3C), similar to the maternal kindred with Var2, suggesting that heterozygotes of this ABCG5 variant might be sufficient to cause macrothrombocytopenia. Considering these functional studies, Var2 may be classified as “Likely pathogenic” according to the standards and guidelines of the American College of Medical Genetics and Genomics (ACMG) (Richards et al., 2015): in this case PS3, our functional studies show a deleterious effect; PM2, absent in population databases. Addtionally, the homozygous KO rats had significantly higher β-Sitosterol level (mean, 7 mg/dl) than WT strain (mean, 2.5 mg/dl) when even being fed by normal diet. We further treated 2 KO rats with Ezetimibe 0.2mg/Kg. Their serum TC reduced from 2.3 and 2.4 mmol/L to 1.7 and 1.9 mmol/L , and β-Sitosterol level reduced from 6.5 and 7.5 mg/dl to 2.6 and 2.7 mg/dl.

We decided to determine whether ABCG5 variants are associated with large platelets in the general population. Routine automated blood cell counting systems differentiate blood cells by their size and do not recognize large platelets as platelets, these instruments may not accurately detect macrothrombocytopenia. The mean platelet volume also does not reflect actual platelet size in the case of large platelets. Platelet count should therefore be determined manually in a calculating chamber or on peripheral blood smears when suspecting the condition. On routine clinical laboratory examination, biased platelet distribution can be observed, with an abnormal “NaN” value instead of specific values. Without other indicators, large platelets may be easily overlooked.

Here, we intended to examine if platelet size could be used to identify ABCG5 heterozygotes (Methods). We collected platelet size distribution results for all individuals having this examination within 2 months in several hospitals. Among 1,180 individuals screened, 30 had evidence of large platelets (>30 fl, Supplementary Figure S5). In ABCG5 sequencing of the 30 individuals (Figure 4, Supplementary Table S5), we identified a known pathogenic variant (p.Arg446Ter, gnomad:2-44050063-G-A; allele frequency:1.70e-4 and no homozygotes) in one individual. In addition, two novel variants, Var3 (hg19-chr2:44039644, NM_022436:c.*610_*611insCCCAGTGATTTTACTGAGGATTA) and Var4 (hg19-chr2:44039650, NM_022436:c.*604_*605insTACAGAGCACCCAGTGATTTTACTGA,hg19), insertions in 3’ UTR region, were also identified in three other individuals, respectively. These are classified as variants of uncertain significance (no record in gnomAD, thus only having PM2 evidence according to ACMG standards and guidelines). Hence, this population-based work along with segregation of an ABCG5 variant in a large pedigree with sitosterolemia identifies heterozygosity for specific ABCG5 variants as an additional genetic cause of macrothrombocytopenia. Our data suggest that 2.5% of individuals requiring complete blood counts may have large platelets, and roughly 10% of those with large platelets may have ABCG5 variants that are causative. Therefore, screening for large platelets may be helpful to identify heterozygotes for ABCG5 pathogenic variants; conversely, sequencing ABCG5 in individuals with large platelets may be useful in identifying patients amenable to Ezetimibe treatment, if it were shown to be effective and beneficial in future large clinical trials.