RefSeq is an open access database that contains an annotated and curated non-redundant collection of sequences representing genomes, transcripts, and proteins (Pruitt et al., 2007, 2012). We used PilFind to find all the potential pilins from bacteria in order Desulfuromonadales, which includes many electricigens, in RefSeq (release 70) (Imam et al., 2011). Two files of genomic data were obtained from different download paths. One file contained genome information, such as name, length, and number of the genes in the respective genomes; and the other file contained the complete annotations. In this study, short TFPs with < 100 amino acids (the length of the alpha-helix in full-length pilin) were defined as truncated pilins, among them PilA (GSU1496) from G. sulfurreducens was a typical example. The other TFPs were defined as full-length pilins. We downloaded 2774 bacteria and archaea genome sequences, including 11 Desulfuromonadales strains that had intact annotations, from ftp://ftp.ncbi.nlm.nih.gov/genomes/Bacteria/all.faa.tar.gz to analyze the prevalence of truncated pilins among Desulfuromonadales species and to determine the relationship between the truncated and full-length pilins.

We used the Basic Local Alignment Search Tool (BLAST) in NCBI's non-redundant protein sequences (nr) database to identify pilins that were not in RefSeq. Position-Specific Iterative BLAST (PSI-BLAST) was employed to uncover distant relatives of the GSU1497 protein (Altschul, 1999).

To analyze the characteristics of the truncated pilins in Geobacter, all the truncated and full-length pilins in Geobacter (species options) were collected by BLAST (blastp) searches against the nr database where GSU1496 pilin and Gura_2677 pilin (full-length pilin of G.uraniireducens Rf4) were used as query sequences. The Gura_2677 pilin sequence was adopted since its capability of extracellular electron transfer has already been verified by biological experiments (Tan et al., 2016).

We used PSI-BLAST to find the homologous protein of GSU1497, which is encoded in the same operon as pilA (iterative threshold = 0.001, gap open\extend costs = 11/1). We defined the homologous genes of the GSU1496 and GSU1497 as GSU1496-like and GSU1497-like genes respectively. The proteins encoded by the homologous genes were therefore defined as GSU1496-like pilins and GSU1497-like proteins. GSU1496-like pilins are truncated type IVa pilins (TFaPs). The word “like” means these proteins have similar sequence and structure composition.

Whole genome information could be applied to find pilins and to reveal the location relationships between GSU1946-like and corresponding GSU1497-like genes. Pilins acquired from these genomes with complete genome information in Desulfuromonadales were utilized to conduct multiple sequence alignment and structure analysis. However, to investigate the evolution relationships between truncated and full-length pilins, all possible pilin sequences in the nr database were found by BLAST searches with Gura_2677 pilin as the query sequence. The 100 sequences with the highest similarity to Gura_2677 were downloaded from the BLAST results, redundant sequences were deleted, and sequences with the trans-membrane domain that is a characteristic of TFPs were chosen (Gorgel et al., 2015). All the signal peptide sequences were removed to obtain the final dataset of pilin protein sequences. The corresponding 16S ribosomal RNA gene sequences of these strains were also downloaded from NCBI's 16S ribosomal RNA sequences (bacteria and archaea) database.

The PilFind algorithm is an invaluable tool that has long been used to identify TFPs in bacteria (Imam et al., 2011). Although trained on pilins from Gram-positive bacteria, PilFind can also identify TFPs in Gram-negative bacterium with low false positive rates (Imam et al., 2011). For Gram-positive bacteria, PilFind identified 155 of the 160 curated putative type IV pilins within operons containing TFP biosynthesis genes. For Gram-negative bacteria, PilFind was able to identify 27 of 29 experimentally verified type IV pilins, highlighting the predictive potential of this software (Imam et al., 2011). A pilin-like protein is identified when the returned value of the PilFind algorithm is true.

Clustering TFPs involves grouping pilins in such a way that pilins in the same group (called a cluster) are more similar to each other than they are to pilins in other clusters (Trehard et al., 2016). Cluster algorithm was used to detect a Type IVa cluster from among the pilins predicted by PilFind (Rollefson et al., 2011). To cluster TFaPs, the process was divided into two steps: align sequences using MAFFT (mafft-7.221-win64 with algorithm L-INS-i) (Katoh and Standley, 2013), and built phylogenic tree with MEGA 6.06 (using the neighbor-joining method and bootstrap = 1000; Tamura et al., 2013).

WebLogo was developed by Gavin and Schneider to generate sequence logos that are graphical representations of the patterns within a multiple sequence alignment, and to assist in discovering and analyzing those patterns (Schneider and Stephens, 1990; Crooks et al., 2004). We implemented WebLogo to find conserved sites/area and the distribution characteristics of aromatic and charged amino acids in the predicted pilins. The predicted sequences of the GSU1496-like pilins and GSU1497-like proteins were aligned using the BioEdit software (Hall, 1999).

The QUARK, I-TASSER (Iterative Threading Assembly Refinement), and PHYRE2 algorithms have been developed for protein folding and protein structure prediction (Xu and Zhang, 2012; Kelley et al., 2015; Yang et al., 2015a). The structures of the GSU1496-like pilins, GSU1497-like proteins, and full-length pilins were predicted using QUARK (for pilins shorter than 200 amino acids) or I-TASSER (for pilins longer than 200 amino acids). PyMOL is a molecular graphics system for the visualization of 3D chemical structures (Hart et al., 2015).

Unrooted tree topology based on multiple alignments of the pilin amino acid or 16S ribosomal RNA gene sequences was obtained using the Maximum Likehood method in MEGA 6.06. Consistency of branching was tested using a bootstrap analysis with 1000 resamplings of the data using MEGA 6.06 (Tamura et al., 2013).

DOOR (database for prokaryotic operons) contains computationally predicted operons of all the sequenced prokaryotic genomes and is a useful bacterial resource (Mao et al., 2009). To determine whether GSU1496-like and corresponding GSU1497-like genes share one operon, the operons of these genes were searched with locus tag in this database.

The 11 sequenced Desulfuromonadales genomes were obtained from the RefSeq database (release 70) (Supplementary Table S1). PilFind detected 183 candidates TFPs in the 11 genomes, including TFPs, Type II Secretion System (T2SS) major pilins, and T2SS minor pilins (Supplementary Figure S1). A phylogenic tree was built (neighbor-joining, bootstrap = 1000) for the candidates TFPs, and only proteins annotated as TFaP or proteins that clustered with the annotated TFaPs were considered as TFaPs (Supplementary Figure S1) for the subsequent analysis. The TFaP cluster contained 11 pilins, six were annotated as pilin-related and the others as hypothetical (Table 1). There was only one type IV pilin in each genome. However, in other class, for example P. aeruginosa, there was a pilin island that includes a major pilin gene cluster. GSU1496 and seven truncated pilins (length < 100 amino acids, Table 1) were included in the TFaP cluster (Supplementary Figure S1). Hence, truncated pilins were prevalent (8/11) in the known genomes of Desulfuromonadales (Table 1).

In Geobacter, the truncated pilins of G. sulfurreducens and G. metallireducens, which are capable of coupling the complete oxidation of organic compounds to the reduction of iron and other metals, have been verified experimentally (Lovley et al., 1993; Caccavo et al., 1994; Lovely, 2012; Lovley, 2012). Additionally, it has been suggested that the pili of G. sulfurreducens KN400, which are assembled by truncated pilins, might serve as biological nanowires that can transfer electrons from the cell surface to the surface of Fe(III) oxides (Reguera et al., 2005). The implication is that truncated pilins may be related to the capacity of extracellular electron transfer in these bacteria. However, the relationship between truncated and full-length pilins has not been reported previously.

To find all Geobacter pilins, including those in sequenced, and un-sequenced genomes, GSU1496 pilin and Gura_2677 pilin were used as query sequences in BLAST searches against the nr database to identify truncated and full-length pilins respectively. Four pilins (from G. soli, G. pickeringii, Geobacter sp. OR-1, and G. bremensis), which are not listed in Table 1, were found in Geobacter species (Figure 1A) with GSU1496 as the query sequence. Besides the pilins found using PilFind, no additional full-length pilins were found with Gura_2677 as the query sequence. Therefore, 11 truncated and two full-length pilins were identified in Geobacter species. For the comparison of full-length pilins in Geobacter, we included two additional pilins (from Desulfomicrobium baculatum DSM 4028 and Desulfuromonas sp. WTL) that are close homologs of Gura_2677. All the selected protein sequences were aligned using BioEdit and sequences logos were built with WebLogo (Hall, 1999).

The multiple alignment shows that all the pilins (full-length and truncated) had the conserved N-terminal signal sequence with a highly conserved hydrophobic helix (Figures 1A,B), which is consistent with them both being TFPs (Gorgel et al., 2015). The distinctive transmembrane domain (the red dotted box in Figure 1B) is also a hallmark of TFPs, which also acts as a conserved structure domain in protein interactions (Giltner et al., 2012). The main role of this transmembrane domain is to fix the membrane protein into the cell membrane. Moreover, pilins have a conserved peptidase splice site (red arrow in Figure 1A) upstream of the transmembrane domain. To form a mature pilin, the signal sequence is cleaved off at this peptidase splice site (Strom and Lory, 1991).

Several conserved aromatic amino acids (at positions +1, +24, +27, +32, +57) were detected in the truncated protein sequences (Figure 1B). Highly conserved sequences are generally considered to have specific functions, including stability, reproduction, and microbial functions (Robson et al., 1988). Aromatic amino acids [phenylalanine (F), tyrosine(Y)] are known to be required for pili conductivity and long-range extracellular electron transfer in G. sulfurreducens (Vargas et al., 2013). TFPs assemble to form symmetrical helical structures (Burrows, 2014). In a polymeric pilus structure, truncated monomers may assemble tightly together so that the conserved aromatic amino acids from different monomers are closely packed, resulting in pi-pi interactions that contribute to pili conductivity. There are also many non-conserved aromatic amino acids between positions +50 and +57, including F51 of G. sulfurreducens PCA/KN400, Y50, Y51, and F56 of G. metallireducens GS-15, and F43, F50, F54, and Y61 of G.bemidjiensis Bem. The conserved and non-conserved aromatic amino acids are mainly distributed in two areas: +24 to +32 and +50 to +61 (Figure 1). The conserved aromatic amino acids may induce the conductivity of pili, while the non-conserved aromatic amino acids may introduce small distances between the aromatic rings, which may contribute to the extracellular electron transfer capacity of some pili.

In the truncated pilins of Geobacter, the conserved amino acids include not only the aromatic amino acids, but also charged amino acids [glutamic acid (E), aspartic acid (D), lysine (K), and arginine (R)] (Figures 1B,C). The conserved charged amino acids (E5, R28, K30, D39, K44, E48, and D53) possibly form salt bridges when the monomers assemble to form polymers. Although the electronic interactions in salt bridges are relatively weak, small stabilizing interactions can add up and make an important contribution to the overall stability of a conformer. Furthermore, the conserved charged amino acids are distributed in the same regions as the aromatic amino acids (+24 to +32 and +50 to +61). Salt bridges in this region might affect the configuration of the aromatic residues and their reduction potential, thus influencing the rates of electron transfer.

By comparing the truncated and full-length pilin sequences, we found that phenylalanine (F) occupied the +32 site in almost all the full-length pilins, while tyrosine (Y) occupied the +32 site in the truncated pilins; further, Y57, K30, K44, E48, and D53 were not conserved in the full-length pilins, but conserved in the truncated pilins (Figure 1C). Phenylalanine and tyrosine differ only in one OH group, which could be easily modified by glycerophosphate that plays a significant role in the process of adsorption and combination with ferric iron ions (Reguera et al., 2005). Moreover, the reorganization energy that contributed to the electron transfer activation energy of tyrosine was reported to be stronger than it was for phenylalanine (Yan et al., 2015). When pi-pi interactions are forming, the more electron transfer activation energy there is, the better the conductive effect the aromatic amino acids may have (Zhuang and Wang, 2015). Feliciano et al. (2015) reported that two mutations (Y32A (alanine), and Y57A) in the pilin amino acid sequence severely affected the function of pili as an electronic conduit in vivo. Therefore, the two amino acids at positions +32, and +57 probably generate pi-pi interactions with other aromatic amino acids, and then contribute to long-range extracellular electron transfer with the conserved charged amino acids in truncated pilins. In the previous study (Feliciano et al., 2015), the type, configuration, and distances of the aromatic amino acids in aromatic contacts and the pilus surface properties were different in an aspartic acid (D) 53A mutant. Hence, it can be inferred that the conserved charged amino acids, K30, K44, E48, and D53, in truncated pilins of Geobacter may play important roles in monomer assembly and the formation of pi-pi interactions.

The truncated pilins in Geobacter have similar amino acid sequence lengths and the truncation sites (blue arrow in Figure 1A) are all in the loop region that is downstream of the alpha-helix. It can be inferred that the fission site in full-length pilins is located in this loop area, assuming truncated pilin evolved from full-length pilin. The sequences downstream of the alpha-helices in full-length pilins are not conserved (Figure 1, Supplementary Figure S2), which may explain why only truncated pilins have been found to be conductive, up to now.

The eight GSU1496-like pilins listed in Table 1 were obtained by combing PilFind with cluster analysis. The eight GSU1497-like proteins were found by PSI-BLAST with the GSU1497 amino acid sequence as the query, because the GSU1497 gene is located directly downstream of GSU1496 gene and in the same operon (Methe et al., 2003; Holmes et al., 2006). The tertiary structures of GSU1496-like pilins and GSU1497-like proteins in Geobacter were predicted by QUARK (<200 amino acids) and I-TASSER (>200 amino acids). We found that all the GSU1496-like pilins contained an alpha-helix, while all GSU1497-like proteins contained a group of consecutive anti-parallel beta-sheets. The full-length pilins all had a complete alpha-helix and a C-terminal globular domain, regardless of whether the structure was obtained from biological experiments or ab initio calculations (Figure 2, Supplementary Figure S3). For example, the full-length pilins of G. daltonii FRC-32 and G. uraniireducens Rf4 both have the N-terminal alpha-helix and the C-terminal globular domain (Figure 2). The consecutive anti-parallel beta-sheets that were characteristic of the GSU1497-like proteins were also part of the C-terminal globular domain of these two full-length pilins.

The spatial structures of the GSU1496-like pilins, GSU1497-like proteins, and full-length pilins from Geobacter are shown in Figure 2 and Supplementary Figure S3. Obviously, the structural composition of the GSU1496-like pilins (Figure 2A) and GSU1497-like proteins (Figure 2B) are similar to the N-terminal alpha-helix and C-terminal consecutive anti-parallel beta-sheets regions of the full-length pilins (Figure 2C), respectively. Moreover, we obtained a hypothetical full-length pilin by joining the GSU1496 major pilin sequence (61 amino acids) and the GSU1497 protein sequence (111 amino acids with the signal peptide removed). Then, we utilized PHYRE2 (Kelley et al., 2015) to predict its structure (Figure 2D). The structural composition (alpha-helix and consecutive anti-parallel beta-sheets) and the length of the hypothetical protein are similar to those obtained from biological experiments of full-length pilins from N. gonorrhoeae, P. aeruginosa and Vibrio cholera (Figure 2C and Supplementary Figure S3). These results indicate that a GSU1496-like pilin and a corresponding GSU1497-like protein possibly originated from one full-length pilin gene.

It is known that the structure of a protein determines its biological function. The diversity of type IV pilin structures determines their functional diversity. Truncated and full-length pilins have distinct surface characteristics, resulting in interaction with different molecules. The assembly of pilin monomers into type IV pili could create a polymer machinery which mediates diverse cellular functions, including cell signaling, host-cell adhesion, DNA transfer, surface motility, microcolony, and biofilm formation, macromolecule degradation, electron transport, and pathogenesis (Craig et al., 2004). For example, N. gonorrhoeae pili can evokes the host immune response and is potential drug and vaccine targets. Additionally, the C-terminal global domain of full-length pilin is an important region for pilus interaction with other pili, host cell receptors and immune components (Forest et al., 1999). For instance, the surface motility of C-terminal of N. gonorrhoeae pilin (Figure 2C) can help bacteria across semi-solid surfaces such as the mucosal epithelia (Zaburdaev et al., 2014).

However, the truncated pilin have no global domain in C-terminal. This structure character could be result in special biochemical property of pilin. For example, the pilus assembled by truncated pilin of G. sulfurreducens KN400 has metallic-like conductivity (Malvankar et al., 2011). This metallic-like conductivity relies on the structures of the G. sulfurreducens pilus in which aromatic amino acids may promote long-distance electron transport (Malvankar et al., 2015). Pilus is a kind of polymer consisting of thousands copies of pilins. When a microorganism uses ATP-powered machinery to assemble a polymeric pilus structure, truncated pilins may be assembled tightly together so that aromatic amino acids from different monomers could construct pi-pi interactions that contribute to pili conductivity. This mechanism for long-rang electron transport is well known as a potential reason for pili conductivity.

To study the origin of the truncated pilin, we analyzed the genetic relationship between truncated and full-length pilins. Gura_2677 was employed in BLAST searches to find all known truncated and full-length pilins because, with GSU1496 as the query sequence, only truncated pilins will be obtained and no pilins were obtained by using GSU1497. Additionally, in three full-length pilins (Table 1), Gura_2677 and Geob_3699 are belong to Geobacter which is a typical genus with conductive pilins. Pili assembled by Gura_2677 pilin are capable of extracellular electron transfer, which has already been verified by biological experiments. There are biased of final list of pilins, when we utilized Geob_3699 pilin sequence to search for homologs. However, the bias in the final list of pilins was acceptable, because two searched results shared identical set of sequence in the top 60 similar sequences and have few difference (<13%) in top 100 similar sequences. The 100 most similar protein sequences were downloaded for phylogenetic analysis. After deleting redundant sequences, 80 pilin sequences (41 truncated major pilins and 39 full-length pilins) were obtained. The 11 truncated and two full-length pilins sequences from Geobacter were included in the 80 pilin sequences for the analysis.

From the structural and phylogenetic analyses, we inferred that GSU1496-like and GSU1497-like genes probably evolved from the N-terminal and C-terminal encoding regions of a full-length pilin gene, respectively. To analyze the process of gene fission, we studied the location relationship between GSU1496-like and GSU1497-like genes. Whole genome information is available for only eight Geobacter strains (Table 1). Eight GSU1496-like and eight GSU1497-like genes were extracted from the eight genomes using PilFind and PSI-BLAST.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.