Iron in the brain plays a crucial role in sustaining homeostatic activity by participating in different physiological processes such as mitochondrial respiration, myelin formation, neurotransmitter synthesis, DNA synthesis, and metabolism [1]. A high iron level can harm brain cells by generating radicals, leading to oxidative damage. Abnormal iron homeostasis seems critical in potentiating neurodegeneration in disease conditions like traumatic brain injury, Alzheimer’s, Parkinson’s, Huntington’s diseases, and multiple sclerosis [2]. Neurological manifestations of disturbances in iron homeostasis are described in increased prevalence of headache [3], movement disorders [4], secondary dementia [5], epilepsy [6], multiple sclerosis [7], human immunodeficiency virus dementia [8], Freidrich ataxia [9], and Alzheimer and Parkinson diseases [10]. Iron accumulation can be visualized in MRI, and iron deposition in deep brain nuclei has been described as differing between patients with multiple system atrophy and progressive supranuclear palsy [11]. In a few rare neurodegenerative disorders, iron accumulation has been observed in various parts of the brain. The examples are Woodhouse–Sakati syndrome (iron deposition in the small pituitary gland), Kufor–Rakeb disease (deposition in the basal ganglia), Coenzyme A synthase protein-associated neurodegeneration (increased pallidal iron content), SCPx syndrome (accumulation in the thalamus, the pons, and the occipital region), and GM1 type3 gangliosidosis (accumulation in lobus pallidus with progression to the substantia nigra, the red nucleus and the subthalamic nucleus). However, in a few instances, hypo-intensities for iron content in MRI have been reported in diseases including restless legs syndrome and periodic limb movements in sleep [5], adaptor protein complex-4 deficiency, DDHD Domain Containing 1 gene (DDHD1) pathogenic variants, and GTP Binding Protein 2 (GTPBP2) pathogenic variants [12].

Neurodegeneration with brain iron accumulation (NBIA) condition was first named Hallervorden-Spatz syndrome, and it is associated with high quantities of basal ganglia iron and axonal spheroids. Pantothenate kinase-associated neurodegeneration (PKAN), caused by mutations in the PANK2 gene, is the utmost reason for NBIA [13, 14]. Presently, identifying various genes associated with NBIA and recognizing the growing phenotypes within this group has made significant progress. Several NBIA disease genes are involved in processes critical for brain health, including membrane lipid homeostasis regulation. The membrane potential required for normal neurotransmission and the synaptic connection is created and maintained by the membranes of neurons, which play a critical role in their function. Many of these genes are also important for mitochondrial health and appropriate cellular functions [15]. Discovering how the common networks among these genes and proteins intersect, considering lipid, iron, and energy metabolism, will help us better understand disease pathophysiology and point us in the direction of potential therapeutic targets [14]. As a result, several NBIA illnesses are caused by proteins involved in iron metabolism malfunctioning, interrupting normal cellular iron trafficking and storage. Others are caused by aberrant cellular membrane metabolism and myelin maintenance, which leads to an iron buildup through defective phospholipid and ceramide metabolism. Hence, NBIA is a physiologically and genetically heterogeneous entity with several pathophysiological pathways, varying patterns of iron accumulation, and controversy over whether iron is the major cause of degeneration or only a minor contributing element in each illness [13]. Although a reliable diagnostic criterion for NBIA disorders is unknown, they are rare. NBIA affects roughly 1 in 500,000 persons globally, according to incidences for pantothenate kinase-associated neurodegeneration (PKAN) that contributes to almost 50% of cases [16].

Several studies have been done for mutant genes associated with iron accumulation [17]. Codon preferences are one of the essential factors of a gene associated with NBIA. Codons are triplicates of nitrogenous base and are responsible for synthesizing specific amino acid chains, i.e., peptide or protein. Codons are universal, but due to the degeneracy of codons, sometimes more than one codon code for the same amino acid during protein synthesis is known as synonymous codons. These synonymous codons are not used uniformly in an organism. Thus few codons are preferred over others, and this phenomenon is known as codon usage bias (CUB). CUB study is significant in heterologous gene expression, gene expression level prediction, primer designing, and gene function analysis. Protein synthesis (translation process) is a biological process prone to mistakes. CUB-induced translation mistakes are determined by their impact on protein structure, function, and cellular level. CUB studies were done in various animal models for many diseases. Still, the first work on the nucleotide makeup and CUB of genes related to the central nervous system (CNS) was done by Uddin and Chakraborty [18]. The CUB pattern is unique to each organism and varies from one to another. CUB is thought to be influenced by several parameters, including GC content, expression levels, genetic factors, hydropathicity, and protein aromaticity. It could be used to understand evolutionary phenomena such as adaptive radiation, lateral gene transfer, gene expression level, and codon optimization [19].

We examined the nucleotide composition and codon usage of neurodegeneration with brain iron accumulation to recognize the resemblances and alterations in codon usage preferences between the genes. The analysis also identifies the codons that are over-represented and under-represented in NBIA and housekeeping genes.

Molecular evolutionary investigations suggested that codon usage bias is widespread across the genomes and profoundly contributes to genome evolution [33]. Changes in specific codons with synonymous codons have been quantified at the transcriptome level, and alterations are profoundly present in diseased conditions [34]. Codon usage is associated with specific gene functions also. Genes involved in cell growth and proliferation prefer A/T ending codons, while genes involved in differentiation and specialized cell function prefer G/C ending codons [34]. Also, cellular programs like proliferation and differentiations use a different set of preferred codons [35].

Non-optimal codons are used to regulate the rate of translation [36]. During the process of host adaptation, a differential usage of codon has been displayed in three of the HIV1 genes that are suggestive of host-specific codon usage in a given pathogen [37].

Codon usage bias directly affects mRNA stability and expression level, so synonymous mutations are un-doubtfully involved in various kinds of ailments. They have been reported to link with more than thirty different diseases [38]. The etiology behind this effect is alternative splicing. Alternative splicing produces a highly dynamic proteome responsible for modulating regulating machinery with roles in various diseases [39]. For example, one segment in the BRCA1 gene showed silent mutations very rarely and evolved very slowly in both human and rodent counterparts. Such synonymous mutations were present in the splicing enhancer region and were so harmful that their carriers died out [40]. Altered codon usage in autophagy gene IRGM contributes to Crohn’s disease susceptibility [41]. Rare synonymous variants in 4 genes (CDH23, SLC9A3R1, RHBDD2, and ITIH2) were studied in 67 Caribbean Hispanic families affected with Alzheimer’s disease, found at a higher frequency in comparison to reference population of the same ancestry and contribute to Alzheimer etiology [42]. The codon usage might differ in population and elucidate how susceptible a population is to an ailment [43].

Codon usage analysis of the genes involved in neurodegeneration and brain iron accumulation has been envisaged for codon usage analysis and estimating various forces affecting it. Except for methionine and tryptophan, all the amino acids are encoded by two or more two codons known as synonymous codons. Synonymous single nucleotide variants (sSNVs), altered codon encoding for similar amino acids, have been implicated in several genetic disorders due to pre-mRNA splicing, mRNA structure, and miRNA regulation alterations. It also affects the protein translation rate. Effects of synonymous single nucleotide variants have been studied in the Sonic Hedgehog (SHH) gene, where usage of sSNVs results in holoprosencephaly, a congenital brain defect resulting in incomplete forebrain cleavage. Eight sSNVs have been observed in holoprosencephaly patients compared to healthy individuals. Out of eight, five sSNVs have been found to reduce the protein expression level ranging from 5% to 23% in cell assays [44]. It underlines the importance of synonymous codon usage studies in clinical setup and helps understand the biased codon usage-based pathologies in human diseases.

There are mainly compositional constraints, selection pressure, and mutational pressure involved in shaping codon usage bias [27]. Suppose a significant correlation is present between various compositional constraints. In that case, it is indicative of the presence of all the three compositional, mutational, and selection forces against translation efficiency, accuracy, and RNA folding [45], and the same is found here, indicating the presence of these three forces in shaping codon usage.

The average composition of nucleotide C and G was equal with similar composition at the third codon position; also, the composition of C and G nucleotide did not follow the same trend at first and second codon positions. The average nucleotide %T composition, which was least among all nucleotides, was not least at the third codon position. We observed an interesting feature in compositional constraints related to the set of genes. Two out of four nucleotides have similar compositions at all codon positions. For the first codon position, it is %A1 and %C1 (approx. 25%), for the second position %A2 and %T2 (approx. 29.70%) and the third codon position %G3 and %C3 is similar (approx. 30%). However, we do not have any reasonable explanation for this on the current date.

In the double-stranded DNA, the GC molar content ranges between 13% and $\approx$ 75%, and the GC content exhibit a strong correlation with mean values for GC1, GC2, and GC3 [46]. Genome GC content is linked with amino acid and codon usage [47]. The higher the overall GC composition, the higher the GC1, GC2, and GC3. A significant positive correlation and a considerable variation in the GC content will result in greater CUB [48]. However, in our results, overall GC content had a significant positive correlation with %GC1 and %GC3 (p 0.001), while no correlation with %GC2, which indicated that significant enhancement or reduction in the overall GC component would not affect the %GC2 component as well as bias.

ApC, ApG, CpA, GpT and TpG dinucleotides have been over-represented in pp1a, and pp1ab genes of infectious bronchitis virus [49]. Contrary to the finding of Franzo et al. [11], Kunec and Osterrieder found GpT dinucleotide as the third-lowest relatively abundant dinucleotide in human viruses. Along with CpG and TpA, GpT has been found underrepresented, indicating that both in viruses and humans, the dinucleotide abundance could follow a similar pattern. We also can speculate that the genes involved in neurodegeneration with iron accumulation in the brain originated from viral genes that are incorporated and became part of the human genome and gained functionality over a long period of evolution. Underrepresentation of CpG and TpA can be considered an effect of selectional pressure [50] since TpA dinucleotide is present in two canonical stop codons (TAA and TAG). Also, TpA dinucleotides are prone to degradation through cellular RNases. Reducing TpA will help eliminate the risk of mutations and subsequent protein truncation and degradation through RNases. CpG suppression could be explained based on the phenomenon of cytosine methylation at the 5th carbon of cytosine and subsequent deamination that results in the C$\to$T transition. As a result, CpG may convert into TpG/CpA [51].

Contrary to the underabundance of CpG in other organisms, in Mycobacterium lepromatosis, CpG dinucleotide possesses a high odds ratio, possibly owing to its association with higher proinflammatory cytokine production and NF-$\kappa$B activation that eventually leads to high pathogenicity of M. lepromatosis [40]. TpG dinucleotide had a maximum odds ratio (1.327), indicating its over-representation. TpG dinucleotide is a part of the CTG codon, and the CTG codon is over-represented in genes common to cancer and primary immunodeficiencies [52]. We found the same phenomenon, and CTG was over presented in 80% of the genes envisaged. The codons which were underrepresented in 80% of genes were ATA, CTA, TCG and GCG. Underrepresentation of these codons can be understood because these codons contain either TpA or CpG dinucleotides as a canonical part, and the same is reflected in codons.

CAI value determines gene expression and ranges between 0 and 1 [53]. Higher CAI values exhibit a higher expression level of a gene and are determined by comparing its codon usage frequency with the reference set. For any synthetic construct to be expressed in a heterologous host at a higher level, several algorithms are available that maximize the CAI to obtain a higher level of expression [54]. For E. coli genes highest CAI value of 0.85 is for the most abundant lpp gene, which encodes for an outer membrane lipoprotein. The pnp gene also used a biased codon with moderate CAI 0.63 [55]. We observed CAI values 0.683 (gene TBCE) to 0.833 (gene PLA2G6), suggestive of the presence of CUB ranging from moderate to high.

Similar to CAI, ENc is also a measure of codon usage. Considering the ENc, in Drosophila, female-biased genes showed lower ENc and higher CUB while male-biased genes showed high ENc and lower codon bias. ENc also indicates host-specific distribution in the LT-Ag gene of polyomaviruses in aves, fishes and mammalian hosts [56]. ENc indicated a range of 39.36 (ATP13A2) to 56.06 (TBCE), indicating moderate to less bias in codon usage, further supporting the findings obtained from CAI analysis. Similarly, lower CUB is observed in genes involved in Krabbe disease [57] and the central nervous system [18]. The bias is increased in longer genes to use optimal codons to prevent non-sense errors in energetically costly lengthy genes [58]. Highly expressed genes show greater bias than other genes [59]. During the process of evolution, codons undergo adaptive selection to obtain a required level of protein synthesis. Housekeeping and other proteins like DNA binding proteins must use more favored codons in higher amounts.

Krabbe disease is a rare neurodegenerative disorder characterized by a mutation in the GALC gene, resulting in the deficiency of the enzyme galactosylceramidase. It is a lysosomal storage disorder and causes the accumulation of cytotoxic metabolites, viz. galactosylsphingosine or psychosine, in the lysosome [60]. CUB studies in four isoforms of the GALC gene were performed by Das et al. [57], and the mean ENc for the four isoforms was 56.98, which suggests a low CUB of the GALC gene.

Compositional bias tends to affect the CUB, and it is considered that GC bias in any genome affects CUB. Palidwor et al. [61] observed in species of prokaryotes, plants, and human genes that G/C ending codon usage also will be increased with an increment in genomic GC content and vice versa for A/T ending codons [62]. In the genes associated with brain iron accumulation, a preference of GC ending codons was observed over AT ending codons. Our observation differed from Das et al. [57]. They reported preference of AT ending codons in the GALC gene of humans, responsible for rare neurodegenerative Krabbe disease. In the genes associated with neurodegeneration and iron accumulation, it was observed that ENc has a significant positive association with many of the A/T ending codons, and a significant negative association with G/C ending codons signifies that G/C ending codons are associated with higher bias. In concordance to the results of Palidwor et al. [61], with the increase in the GC content of the gene, an increase in usage of G/C ending codons also is observed. However, we found an exception to this general phenomenon. G ending codon TTG encoding for leucine was observed to have positive association with ENc, indicating that if the genomic GC content is increased, the usage of leucine encoding is increased codon TTG will be decreased.

Unlike our result, Palidwar et al. [61] observed two G-ending codons, AGG (arginine) and TTG (leucine), showed a decrease in overall usage with an increment of GC component [61]. In the genes related to Alzheimer’s and other neurodegenerative diseases, Yang et al. [62] found that C ending codons are preferred in all the highly expressed genes. In contrast, G ending codons either remained constant (AGG, coding for Arg) or showed a marked fall, even disappearing (TTG, encoding Leu). Similar to the finding of Yang et al. [62], in genes associated with iron accumulation in the brain, a positive association of expression level and C ending codons was observed. However, TGC (encoding Cys) and GTC (encoding for Val) were two codons, which did not influence the gene expression [62].

On the one hand, where G ending codons either remained constant (AGG, coding for Arg) or exhibited a marked fall (TTG encoding Leu), in Alzheimer’s and other neurodegenerative disease genes, we observed no association between gene expression and CGG and AGG (both encoding for Arg) in our case. However, for codon TTG (encoding for Leu), a negative influence on gene expression has been observed, similar to Yang et al. [62]. The results suggest a typical feature adapted by genes related to neurodegenerative disorders. The driving force for codon usage bias could be mutation, genetic drift, and/or biased gene conversion, and selection forces like the section for translational efficiency and/or accuracy. Therefore, a biased codon might reflect the nucleotide disproportion in local base composition [63].

We found a negative relationship between AT skew and purine skew, while a positive relationship between CUB and GC skew indicates the impact of nucleotide disproportion of CUB. AT, GC and purines Skews had been found linked to aliphatic index and hydrophobicity, indicating the effects of nucleotide disproportion of physical properties of the protein. The results concordance with Khandia et al. [52], who also observed an association between nucleotide skew and protein properties [52]. Effects of nucleotide skew on codon position were observed, and it was observed that T is preferred over A and C is preferred over G at the third codon position as depicted by parity analysis. Similar results were observed in T. saginata, where C and T were used more frequently than G and A [50]. Differed to parity analysis results, which indicated nucleotide T and C, GC ending codons are preferred as per RSCU analysis. The difference in parity analysis and RSCU analysis results, as depicted in Kumar et al. [64], also found dominance of A and G in parity analysis and Dominance of G/C in RSCU analysis.

Similarly, in the equine Influenza virus coding sequences, all polymerase genes had AT bias GC content at the third codon position in fourfold degenerate codon families [65]. The bias also indicated the role of selectional force. In a regression plot between %GC3 and %GC12, it was evident that relative neutrality was less (12.58%), and therefore the role of selection pressure is dominant. Similar results have been obtained by He et al. [66], where they found selectional force as the dominant force in shaping codon usage in Ginkgo biloba, one of the most ancient tree species [66]. Contrary results had been obtained by Nasrullah et al. [67], who found mutational force as the dominant force over the selection on the genome of the Marburg virus [67]. The mutational force was not the major force in shaping the codon usage pattern.

Still, we were anxious to know the amplitude of mutation pressure on CUB. Mutational pressure affected A and T nucleotide positively and C and G nucleotide negatively (Regression plot between ENc and nucleotide composition at third codon position), and the mutational force on A and T nucleotides equally affected CUB, while in the case of G and C nucleotides, mutational force on C nucleotide, affected codon usage least. Since ENc is a non-directional measure of codon usage [68], the negative regression coefficient for C and G indicated that these two nucleotides would tend to increase bias if their content is increased [18]. Unlike ENc, ICDI is a directional measure of CUB [69], and ICDI trends indicated low bias.

Principal component analysis revealed that first and second components are associated with protein hydrophobicity, and basic and amino acid frequencies indicate selective forces choosing protein properties [70]. Also, few genes were grouped with A/T ending codons while others grouped with G/C ending codons. It indicates the effects of composition on codon usage. Upon PCA, almost all genes gathered near the X-axis and displayed less bias.

Codon usage of any gene is mainly driven by mutational, selectional, and compositional forces. A significant correlation between compositional constraints suggests that the above forces shape codon usage patterns in the envisaged genes. Average C and G nucleotide composition was equal in the genes. Still, the same is not observed at all codon positions, indicating the presence of forces responsible for skewness in nucleotide composition. CpG, TpA, and GpT dinucleotides were underrepresented with the overrepresentation of TpG dinucleotide. Like the dinucleotide frequency, ATA, CTA, TCG, and GCG codons encompassing TpA and CpA dinucleotides were underrepresented, and CTG codon encompassing TpG dinucleotide is over-represented, showing effects of compositional parameters on codon usage. G/C ending codons are prevalent in the GC-rich genome, while in AT-rich sequences, A/T ending codons are prevalent. However, a single exception of this phenomenon was codon TTG, which had a positive association with ENc, suggesting a decrease in TTG codon frequency with an increase in GC content. Based on moderate to high CAI values, moderate to high-level gene expression indicates translation efficacy and translational selection. The results of CAI are supported by the results of ENc, which is indicative of moderate to high CUB. Indices of compositional skews like AT, GC and purines skews are associated with some of the protein’s physical properties. T and C nucleotides are preferred over A and G nucleotides at the third codon position, as indicated by parity analysis. The relative neutrality value of 12.58% refers to the minor role of mutational force over selectional force. Mutational force acted equally on G, A, and T nucleotide, while acted least on nucleotide C. PCA analysis revealed a low level of codon bias and role of selection pressure. Cumulatively looking at all analyses performed, it was observed that compositional, mutational, and selectional forces were applied on the set of genes considered to participate in neurodegeneration with iron accumulation with selection force as dominant.

CAI, Codon adaptation index; RSCU, Relative synonymous codon usage; ENc, effective number of codon; SCS, scaled chi-square; ICDI, Intrinsic codon bias index.

Conceptualization—RK, TA, MAAlg and NP; methodology—TA, RK, NP, AMA, MAAlg and MAAlm; software—RK, TA, NP, and AMA; validation—RK, TA, and MAAlg and MAAlm; formal analysis—RK, TA, and MAAlg, MAAlm; investigation—RK, NP, MAAlg, AMA and MAAlm; resources—RK, TA, and AMA. MAAlg; experimentation, data curation—RK, NP, TA, MAAlg and MAAlm; writing—original draft preparation—RK, NP, AMA, and MAA. MAAlg and MAAlm; writing—review and editing—RK, TA, and AMA, MAAlg and MAAlm; visualization—RK, TA; supervision—RK; project administration—RK, TA, NP and AMA. MAAlg and MAAlm; funding acquisition—TA, AMA, MAAlg and MAAlm. All authors have read and agreed to the published version of the manuscript. All authors read and agreed on the final version of the manuscript.

Not applicable.

All authors acknowledge their respective institutes and organization for providing support.

Personal funding.

The authors declare no conflict of interest.

See Table 3.