Investigation of Homocysteine Levels in Jordanian Siblings of Autistic Children

2.1. Blood samples

Blood samples were collected from 24 volunteers, including 9 controls and 15 siblings of autistic individuals, comprising 11 females and 13 males. All volunteers were in a fasting state prior to sample collection to minimize postprandial effects on homocysteine levels. Controls were recruited from the general paediatric population in Amman, Jordan, with no personal or first-degree family history of ASD or neurodevelopmental disorders. Age- and sex-frequency matching was applied at the group level to align the control cohort (ages 3–10 years) with the sibling group, ensuring comparability without pairwise matching. Blood samples were centrifuged at 7000 rpm for 10 minutes, and plasma was stored at −80 °C prior to analysis.

2.2. Chemicals and materials

All reagents used in this work were of HPLC-grade. 2-Methyl-L-cysteine hydrochloride (MeCys) as an internal standard (IS) was obtained from Biosynth Carbosynth (Switzerland). Acetic acid was purchased from (AnalaR NORMAPUR, EC), acetonitrile was obtained from Avantor (India), Ethylenediaminetetraacetic acid (EDTA) was supplied by Fisher BioReagents (USA), where potassium dihydrogen phosphate (KH₂PO₄) was purchased from VICKERS (UK), sodium hydroxide (NaOH) was obtained from Guangdong Guanghua (China), hydrochloric acid (HCl) was supplied by Scharlab (EU), methanol (MeOH), L-homocysteine, 1,1’-thiocarbonyldiimidazole (TCDI), Tris(2-carboxyethyl) phosphine (TCEP) and Na₂HPO₄·7H₂O, were supplied by Sigma-Aldrich (France). Ultrapure water (18 MΩ·cm) produced by a Milli-Q Plus purification system (Millipore, Billerica, MA) was used for preparing aqueous solutions and for dilution purposes.

2.3. Sample Preparation

Stock solutions were prepared as follows: TCEP and KH₂PO₄ were prepared in pure water at 50 mM and 0.4 M, respectively. For EDTA, 10 mM was prepared in water, with pH adjusted to 8.0 using NaOH (0.1 M). According to TCDI, 100 mM was prepared in acetonitrile. HCl, MeCys, as well as L-homocysteine (20 mM) solutions, were prepared in 10 mM HCl. solutions at concentrations of 1.45 mM and 2.9 mM were similarly prepared in 10 mM HCl. A 5 mM Na₂HPO₄·7H₂O (pH 7.4) was prepared, and pH was adjusted with NaOH.

2.4. Reduction of disulfide bonds

Plasma samples of 75 µL were spiked with 25 µL of MeCys (1.45 mM) and 75 µL of 5.0 mM phosphate buffer (pH 7.4). Disulfide bond reduction was achieved by adding 20 µL of TCEP (50 mM) and incubating at 24 °C for 30 minutes in a water bath shaker. The reduced samples were then treated with 25 µL KH₂PO₄ (0.4 M), 25 µL EDTA (10 mM, pH 8), and 20 µL NaOH (0.1 M), followed by derivatization with 20 µL TCDI (100 mM) at 37 °C for 20 minutes. Subsequently, 50 µL HCl (1.0 M) was added, and the mixture was centrifuged at 7000 rpm for 10 minutes.

2.5. Solid-phase extraction (SPE)

SPE cartridges were conditioned sequentially with 10 mL of pure MeOH followed by 10 mM HCl. Then, the reduced sample prepared in section 2.4 was loaded, washed with a 10 mL mixture of 10 mM HCl in 100 mL of MeOH/H₂O (20/80, v/v), and eluted with 10 mL MeOH/H₂O (90:10, v/v). The eluate was evaporated to dryness and subsequently reconstituted in 1.0 mL of a solution containing 0.5% acetonitrile in 2% acetic acid prior to HPLC analysis.

2.6. Chromatographic analysis

Chromatographic analyses were performed using a HPLC Shimadzu system (Japan) equipped with a SIL-20AC autosampler, an SPD-20AV UV/Vis detector, a DGU-20A3R degassing unit, and an LC-20AB binary pump. Chromatographic separation was implemented on a reversed-phase C18 column (15 cm×4.6 mm ID, 5 µm dp) at two different wavelengths of 272 and 283 nm. HPLC conditions were optimized to provide high-resolution separation with minimal analysis time. The final optimized method employed with a mobile phase composed of 0.5% acetonitrile in 2% acetic acid as a mobile phase and 95% methanol in 2% acetic acid as a washing solution. The gradient elution was as follows: 1.0 mL/min for the first 8 min, 1.5 mL/min from 8.1 to 11 min, and 1.3 mL/min from 11.1 to 17 min at an injection volume of 75 µL, and analyses were carried out at 24°C at 283 nm. The injections have been carried out in triplicate, and the average concentrations have been recorded.

2.7. Statistical analysis

Differences in average concentration between the sibling and control groups were assessed using an independent samples Welch’s t-test. Given the small sample size and potential deviations from normality, a non-parametric Mann–Whitney U test was also performed to confirm the robustness of the results. Statistical significance was defined as p < 0.05.

Multivariate analyses, including PCA and HCA were conducted using Chemoface 1.61 (Matlab VR, MathWorks, USA) and XLSTAT (Excel, Microsoft VR).

3.1. Method development and validation

To select the optimized wavelength, L-homocysteine was injected at 272 and 283 nm in order to validate the best wavelength. The results illustrated that 283 nm has the highest intensity in comparison to 272 nm [Figure 1]. Hence, it was selected for the validated HPLC method.

Close

Figure 1:

Chromatogram of L-homocysteine at 272 and 283 nm.

Export to PPT

Linearity was assessed by preparing L-homocysteine standard solutions in water at the following concentrations (Appendix 1): 3.70, 7.40, 22.20, 37.00, 51.80, and 74.00 μmol/L spiked with MeCys (25 µl, 1.45 mM) as IS. Precision was expressed as the coefficient of variation (%CV) for six replicates of low (3.70), medium (37.00), and high concentrations (74.00 μmol/L) analysed over three consecutive days, and the outcomes were 4.01%, 2.02%, and 0.67%, respectively. Recovery values were measured for the three concentrations, and the results showed 101.65%, 94.36% and 100.39%, respectively. Limits of detection (LOD) and limit of quantitation (LOQ) were calculated using the standard error of the intercept from regression at a 95% confidence level and found to be 0.81 and 2.3 μmol/L, respectively.

Close

Appendix 1:

Calibration curve obtained from the replicate analysis of L-homocysteine.

Export to PPT

3.2. Homocysteine quantification

Table 1 presents the average concentrations of homocysteine in the 24 volunteers of both genders based on triplicate analysis per sample. The cohort included 15 siblings of autistic individuals and 9 control children, revealing notable inter-individual variability across both groups.

Hcy concentrations among siblings ranged from 2.59 to 7.62 µmol/L, whereas the control group displayed a narrower range of 2.37- 4.44 µmol/L.

Overall, homocysteine levels were higher in siblings of autistic individuals compared to the controls, suggesting a potential metabolic alteration even in non-autistic relatives. This finding is consistent with previous reports indicating that elevated Hcy or disrupted methylation metabolism may represent a familial biochemical trait associated with autism spectrum disorder.^[32,33]

Within the sibling group, a subset of five children aged 4–10 years exhibited the highest Hcy concentrations (6.88-7.62 µmol/L), suggesting that some individuals may have subclinical hyperhomocysteinemia. These elevations could reflect partial deficiencies in the folate–vitamin B12–methionine metabolic axis, genetic polymorphisms such as MTHFR C677T, or environmental and nutritional factors affecting one-carbon metabolism.^[34-38] Conversely, four siblings aged 4–8 years exhibited low Hcy concentrations (2.59–3.03 µmol/L) of 3 males and one female, highlighting heterogeneity within the group, which may be attributed to differences in diet, vitamin status, or adaptive metabolic regulation.^[39-41]

Regarding the gender impact, female siblings exhibited a higher average Hcy concentration (6.02 µmol/L) than males (4.49 µmol/L). This finding is in contrast with adult patterns, where males typically have higher Hcy levels due to greater muscle mass and creatine turnover.^[42,43] This could be related to the paediatric hormonal and dietary influences are still developing, so such differences may not yet be physiologically stabilized.

No consistent linear correlation was observed between age vs. Hcy concentration within the sibling cohort. For example, the highest Hcy level was found in a female sibling at 5 years old, followed by an 8-year-old. While the lowest levels were found in 4-year-old males and 8-year-old females, with 2.59 and 2.81 µmol/L, respectively. These findings suggest that age may not be the factor influencing Hcy metabolism.^[43,44] Instead, genetic predisposition and micronutrient status likely play greater roles.

The control group exhibited relatively lower Hcy concentrations than siblings, with no age-related impact on the Hcy levels. The presence of several low-Hcy individuals highlights metabolic homeostasis in controls, in contrast to the more variable distribution seen among siblings. The mean Hcy concentrations were 2.69 and 3.77 µmol/L for female and male control samples, respectively. These outcomes are consistent with reported paediatric reference ranges in healthy populations that showed higher levels in males than females.^[45] In contrast to the sibling samples, males control samples exhibited higher mean homocysteine concentrations than the females, although notable variability was still present among the males, with some showing low levels and others showing higher values.

Collectively, these findings suggest that siblings of autistic individuals may display subtle metabolic imbalances in homocysteine regulation, potentially reflecting shared genetic or nutritional vulnerabilities within ASD-affected families. Such variations, even among neurotypical siblings, may provide insight into epigenetic and one-carbon pathway disruptions that contribute to ASD risk.

3.3. Parametric and non-parametric statistical comparison

To determine whether the average concentration levels differed between sibling and control volunteers, statistical comparisons were conducted using both parametric and non-parametric approaches (Table 2). This dual-analytic strategy was employed to account for the relatively small sample sizes and the possibility of non-normal data distributions.

As shown in Table 2, the average concentration levels were significantly higher in the sibling group (n = 15; 5.21 ± 1.85 μmol/L) compared with the control group (n = 9; 3.29 ± 0.87 μmol/L). An independent samples Welch’s t-test demonstrated a statistically significant difference between groups (t = 3.43, p = 0.0027). Consistent with this finding, a non-parametric Mann–Whitney U test also revealed a significant difference between groups (U = 109, p = 0.0145). Collectively, these results indicate that concentration levels were significantly elevated in sibling volunteers relative to controls, and this effect was robust across both parametric and non-parametric analyses.

In summary, these data suggest that elevated and variable homocysteine levels among siblings of autistic individuals may represent a biochemical endophenotype associated with ASD, underscoring the importance of metabolic screening and nutritional support in at-risk families.

3.4. Multivariate analysis of volunteer samples using unsupervised clustering methods

HCA and PCA are performed to confirm whether the plasma samples obtained from 24 different healthy and sibling volunteers would be grouped together based on the chromatographic outcomes of homocysteine levels vs. gender and age.

Quantitative samples classification by PCA and HCA

In this study, 24 plasma samples are the original variables (24 dimensions) in PCA. By calculating the covariance matrix between these dimensions, PCA can generate PCs that are orthogonal to each other and can explain 100% of the total variance of the orthogonal data. Each PC is correlated with the original 4 variables.

In HCA, the collected data is displayed in a certain way to emphasize its natural clusters and patterns in a two-dimensional space. The similarity between samples can be evaluated following a suitable distance measure, which is commonly applied in pattern recognition. Euclidean distance d between samples is estimated as:

(1)

$d_{i,k}={\left(\sum _{k=1}^K{(x_{i,k}-x_{j,k})}^2\right)}^{1/2}$

Where K, i, and j are the number of variables measured for samples and indices for samples, respectively. Estimations would be made using the main principal components of the original data after decomposition by PCA. Initially, d_i,k is estimated between all samples (i.e., every sample is to be compared with the remaining samples) to create the distance matrix. The similarity or aggregation between samples is then estimated using the weighted average linkage method.

To classify plasma samples and to specify the parameter responsible for clustering, the HPLC data were subjected to PCA and HCA analysis. The resulted PCA outcome of the 24 plasma samples is provided in Figure 2. The 2PCAs collects 99.82% of the total variance of the orthogonal data. The results revealed two distinct groups, each comprising a mix of sibling and control samples of different ages, but consisting of only one gender type. This suggests that age does not significantly influence the grouping of samples, whereas gender does have an effect. Therefore, HCA was performed separately on the sibling and control groups.

Close

Figure 2:

Bi-plot outcomes for the 24 volunteers (S: sibling, C: control, M: male, F: female and age/year is indicated by number).

Export to PPT

Regarding the HCA outcomes for the sibling individuals [Figure 3a], two clusters were observed. No consistent pattern was found among the sibling volunteers, indicating that gender and age do not influence the classifications, hence, these two parameters cannot be used to categorize Jordanian siblings of autistic children based on their homocysteine levels. This finding supports the results reported in section 3.2 for the sibling samples.

Close

Figure 3:

Dendrogram of the a) sibling and b) control plasma samples.

Export to PPT

Figure 3(b) represents the classification of the control samples; it has been noticed that different patterns are observed in the sibling-male volunteers. In the control, there are 2 main clusters: a cluster with only the males of the highest levels, and another cluster that has 2 sub-clusters of low Hcy levels due to the gender but not the age (i.e., one sub-cluster with males at low values and another sub-cluster with only the female volunteers). This indicates that, for the control samples, gender is the key parameter influencing the clustering of high- and low-value groups.

Several studies consistently report that homocysteine concentrations are higher in males than in females among control populations. Xu et al. analysed data from the Health Management Centre of Tongji Hospital, Wuhan, China, and found that males had higher homocysteine levels and were more likely to exhibit hyperhomocysteinemia than females; their samples were controls as they were neither ASD patients nor siblings of ASD patients.^[46] Similarly, Chia et al. measured blood homocysteine in 183 factory workers in Singapore and observed higher levels in males, with the volunteers serving as controls.^[47] Dong et al. assessed 36 patients with Central Retinal Vein Occlusion (CRVO) and reported lower homocysteine levels in females compared to males, with these patients not being ASD-related.^[31] Nishi et al. used HPLC with fluorometric detection on 380 schizophrenia patients and also found higher homocysteine concentrations in males, with none of the participants being ASD patients or siblings.^[48] Silberberg et al., in a study of 522 individuals, and Zhao et al., with 14,911 participants, both confirmed that males exhibit higher homocysteine levels than females, and all participants were considered controls.^[42,49] Collectively, these studies agree that control females generally have lower blood homocysteine concentrations than males, consistent with the outcomes of this study.

3.5. Bootstrap validation and effect size analysis

Bootstrap is a statistical resampling method used to estimate the variability or confidence intervals of a statistic by repeatedly sampling with replacement from the original data. It helps in assessing the robustness and stability of results, without relying on strict parametric assumptions. By generating a large number of bootstrap samples, one can calculate confidence intervals and evaluate the precision of estimates, making it especially useful for small datasets. Analysis of the Hcy concentrations shows that the sibling group has significantly higher levels than the control group. Welch’s t-test and the Mann–Whitney U test both confirm this difference as shown in section 3.3. Bootstrap validation of the mean difference (1.92 µmol/L, 95% CI: 0.85–2.96) supports the robustness of the result, and the large effect size (Cohen’s d = 1.33) indicates that the difference is not only statistically significant but also practically meaningful.