Sequence-dependent cluster analysis of biomineralization peptides

1 Introduction

Biomineralization is a nature-inspired method of inorganic nanomaterial synthesis that relies on peptides [1]. The process is based on the inherent ability of most organisms to biosynthesize inorganic nanostructures that will aid them in carrying out vital functions [2]. For instance, the magnetotactic bacteria produce magnetite nanoparticles which serve sensing functions [3]. Recently, biomineralization has been widely utilized to create functional nanomaterials due to the several advantages this biomimetic strategy offers [4]. Some of these advantages include ambient synthetic conditions, diversity of metal and peptide combinations available, among others. Biomineralization peptides (BMPep) occupy a central role in this process since they are responsible for guiding the growth of the inorganic nanostructure. BMPeps accomplish this through different mechanisms such as capping [5], regulation of nucleation and reduction [6]. As a result, current research focusses on understanding, controlling and improving this process either through the BMPep [7] or through the environment [8]. Thus, selection of an appropriate BMPep is the critical first step in achieving these goals. In the absence of an organized and reliable method of grouping BMPep, the abundance of known BMPep sequences can make the selection process difficult and confusing. Currently, BMPeps are conveniently classified according to their inorganic substrate. However this practice is unreliable since most BMPep can bind to numerous inorganic substrates. A good example is the R5 peptide which was isolated from Cylindrotheca fusiformis [9]. This 19-residue peptide natively binds to silica but reports have shown that this BMPep can be used to form nanostructures of Ti [10], Pd [11] and Au [12]. Therefore creating credible and statistically sound clusters will help researchers choose which BMPep is suitable for their system based on a chosen clustering variable. This will facilitate a more efficient and systematic approach in the selection of the BMPep. In addition, establishing relationships among BMPeps on the basis of common and dissimilar features will aid in the development of a general understanding regarding the properties of BMPeps. In this paper, we established a 2-cluster solution for the 27 reported BMPep for different metals. The clustering variables were based from the primary structure since the sequence influences numerous properties of the peptides.

2 Methods

The names and sequences of the 27 BMPep used in this study are shown in Table 1. The clustering variables are based on the amino acid composition of the BMPep. The clustering variables are the BMPep length, molecular weight (MW), isoelectric point (pI),% heterogeneity, % aliphatic residues, % aromatic residues, % polar residues, % acidic residues, % basic residues, and% sulfur-containing residues. The MW and pI were calculated using expasy (http://web.expasy.org/compute_pi/) whereas the other variables were calculated by counting the number of amino acid residues corresponding to each category divided by the length of the BMPep. For the % heterogeneity, this corresponds to the absolute number of amino acids within the BMPep sequence without counting the repetition. Using these variables, a two-cluster solution using k-means method was calculated using Statistica. K-means clustering is a form of nonhierarchical grouping method which starts with a predefined number of clusters [30]. Two divisions were chosen in this study in order to attain simplicity. The significant clustering variables were identified through analysis of variance (ANOVA). A variable that possesses the highest p-value which is greater than the confidence level was discarded and clustering was repeated using the remaining variables. This iterative procedure was repeated until the remaining variables all had p-values lower than 0.05. All statistical analyses, including analysis of variance and t-test were conducted at a 5% confidence level.

3 Results and discussion

We carried out the classification of BMPep on the basis of their primary structure. The primary structure of the BMPep has tremendous effects on the resulting nanostructure wherein single residue substitutions within the peptide can produce different morphologies of the nanostructure [31]. While the diversity of BMPeps is high, several BMPeps share common motifs and conserved residues. For example, the three silver binding peptides discovered by Naik et al. [16] have identical lengths and the majority of their residues are conserved for all three BMPeps. Thus, classifying BMPeps on the basis of the similarities and differences of their primary structures is ideal since discrimination on the most fundamental level of peptide structure can be achieved. Out of the ten computed clustering variables, only four were deemed significant after a step-wise elimination of insignificant variables (Table 2). This means that only the length, MW, % heterogeneity and % aliphatic residues are significant variables that can differentiate and cluster together the BMPeps into two groups. More specifically, molecular weight carries the most weight in terms of significance in clustering since it has the lowest p-value.

The second most significant clustering variable is the length. This is expected since the peptide length is closely associated with the molecular weight. Peptide heterogeneity is also a significant clustering variable which implies that sequence diversity is an important point of difference among the reported BMPep. The least significant clustering variable is the % aliphatic residues. Among the other variables included which take into consideration the kind of amino acid, only the % aliphatic residues was deemed significant. The importance of aliphatic residues in a given peptide potentially lies in solvent interaction. Since BMPeps are relatively small and short compared to actual proteins, it is expected that all residues are exposed to the solvent. The role of aliphatic residues arises in regulating the interaction of the peptide with the aqueous environment. The other criteria were deemed insignificant probably due to their very frequent occurrence or erratic appearance. For example, the acidic and basic amino acids are common to all BMPeps, since these residues are responsible for metal complexation [32] and capping. BMPeps regulate nanostructure growth by means of capping wherein the peptide attaches itself to the growing nanoparticle at very specific facets in order to arrest further formation. Since capping is a shared characteristic for all BMPeps, the type of capping amino acid was not determined to be a discriminatory variable due to its commonality. The presence of polar residues are also common to all since they help make the peptide more hydrophilic given that aqueous systems are always utilized. Finally, the appearance of aromatic residues is unpredictable while that of the sulfur-containing residues is rare. If ever they are present in a given BMPep, aromatic residues such as tryptophan and tyrosine help in the reduction of the metal ions [33]. The erratic appearance of aromatic residues indicates that not all BMPeps exhibit the ability to reduce metals. This is consistent with the practice of adding reducing agents in order to convert the metal ions into their zero valent state. Common reducing agents used are ascorbic acid, sodium borohydride, among others. Recently [34] concluded that the type of reducing agent added influenced the morphology of the nanostructures yielded from biomineralization. Thus, it is difficult to discriminate and classify BMPep on the basis of aromatic residues due to their erratic and unpredictable occurrence. This is reflected from our analysis wherein aromatic residues were not deemed to be a significant clustering variable. On the other hand, sulfur-containing amino acids such cysteine exert their influence in the secondary structure of the peptides by forming disulfide bonds. Therefore finding a connection among the other BMPep using these criteria might be difficult. Based from the determined significant clustering variables, the 27 BMPeps were then divided into two groups. The first group contains 18 members whereas the second cluster contains 9 members as presented in Table 3.

The discriminatory ability of the identified significant clustering variables was validated by conducting a t-test between the means of the two clusters (Table 4). As expected, the previously identified clustering variables were significantly different for each group as indicated by their respective t and p values. This further means that the other properties used as clustering variables are not significantly different among the 27 BMPeps.

Based from the descriptive statistics of each group, cluster 1 possesses a longer sequence, higher molecular weight, more diverse with respect to the amino acid composition and contains more aliphatic amino acids (Table 5). On the other hand, members of the second cluster had lower values for these variables (Table 6).

Analyzing the members of each cluster reveals that both clusters possess diverse peptides with respect to their inorganic substrate. For example, both clusters contain Au, Ag and Pd binding BMPeps. This indicates that for a given inorganic substrate, two types of BMPep are available. While type-1 BMPeps are more commonly found, the shorter type-2 also exist which are more attractive in terms of cost. Reducing the length of a peptide by several residues will have a drastic effect on the efficiency of the synthesis. For example, both the Pd4 and q7 are palladium biomineralization peptides. Both BMPep are capable of forming sub-5 nm crystalline nanoparticles. The q7 BMPep however is more attractive due to its shorter length and less heterogeneous character. The q7 BMPep is more cost effective to synthesize since it is shorter by 5 residues. Moreover, it only needs 6 types of amino acids compared to Pd4 which requires 9. In a similar manner are the Midas2 and Z1 peptides for Au. Our findings suggest and encourage discovering more type 2 BMPeps, which are shorter, lighter and less heterogeneous which will translate into a more cost-effective nanostructure production. Truncation studies can be carried out wherein a long BMPep sequence can be systematically reduced into a shorter fragment without compromising its ability to direct nanostructure growth. Generally, increasing the known members of type-2 BMPep will further broaden the applicability of biomineralization as a tool for nanomaterial synthesis. The utilization of type-2 BMPeps is more practical since the synthesis is more straightforward at a considerable lower cost.

4 Conclusion

In summary, we have established a reliable classification of biomineralization peptides based on their length, molecular weight, heterogeneity, and aliphatic residues. Type-1 BMPeps are more commonly found and exhibit higher values for these significant clustering variables. In contrast are the type-2 BMPeps which have lower values for these parameters and are less common. Our findings suggest and encourage discovering and developing more type-2 BMPep since these peptides are more cost-effective to prepare. Increasing the known sequences of type-2 BMPep will widen the applicability of biomineralization as a method to prepare inorganic nanomaterials. Through our clustering analysis, a more efficient and systematic approach in BMPep selection is possible since previous methods of BMPep classification are unreliable.