Variation of the neurofilament medium KSP repeat sub-domain across mammalian species: implications for altering axonal structure

Neurons communicate through propagation of action potentials along the axon to their targets. Current flow into an axon results in a local change in the axon's voltage. As current enters the axon, this local depolarizing event propagates down the axon by the passive spread of charge from the input source, referred to as electrotonic conduction. The conduction velocity of action potentials down the axon is related to the length constant (λ), which varies directly with respect to membrane resistance (R_m) and inversely with respect to axial resistance (R_a) (Hodgkin, 1947; Hodgkin and Rushton, 1946). Whereas R_m is largely determined by the number of leak channels in the membrane, R_a is influenced strongly by axonal diameter (Hodgkin, 1947; Hodgkin and Rushton, 1946). Therefore, axons with larger diameters have faster conduction velocities (Hursh, 1939; Rushton, 1951) due, in part, to more efficient electrotonic conduction via reduction of R_a. Indeed, invertebrates have large rapidly conducting axons (Hodes, 1953). However, mammalian nerves consist of thousands of individual axons. If increasing axonal diameter were the only mechanism to increase conduction velocity, axons would require extremely large diameters, resulting in nerves that are prohibitively large for the vertebrate nervous system. The evolution of myelin resulted in increased conduction velocity in relatively small axons.

While myelination increases conduction rates by preventing charge loss, it also induces cell biological changes in axons so that they expand their diameter, which increases the rate of electrotonic conduction (Huxley and Stampfli, 1949). Myelin-dependent expansion of axonal diameter is referred to as radial axonal growth. Although the mechanism remains unknown, radial axonal growth is dependent upon both the formation of compact myelin (de Waegh et al., 1992) and upon the axonal accumulation of neurofilaments (NFs) (Ohara et al., 1993; Zhu et al., 1997). Within axons, NFs are responsible for establishing and maintaining the three-dimensional array of axoplasm. Therefore, if mammals were under selective pressure to increase neuronal conduction velocities, then it is probable that NF cytoskeletal architecture, myelination, or both, were likely to be evolutionary targets.

NFs are type-IV intermediate filaments (IFs) (Fuchs and Weber, 1994), composed of neurofilament light (NF-L), neurofilament medium (NF-M) and neurofilament heavy (NF-H) subunits. Mature NFs are obligate heteropolymers composed of NF-L, NF-M and NF-H (Ching and Liem, 1993; Lee et al., 1993) with an ∼4:2:1 (NF-L: NF-M: NF-H) stoichiometric ratio (Scott et al., 1985). Recent work suggests that NFs also contain a fourth type-IV IF subunit, α-internexin, in the mature central nervous system (Yuan et al., 2006). Formation of the classic 10 nm filament, characteristic of all IFs, requires the expression of NF-L (Lee et al., 1993). Unlike other IFs, NF heteropolymers have side arms that project from the core of a mature 10 nm filament (Hirokawa et al., 1984). These side arms, which contain a lysine–serine–proline (KSP) repeat region (Lees et al., 1988; Myers et al., 1987), are the carboxy terminal (C-terminal) tail domains of both the NF-M and NF-H protein subunits (Hirokawa et al., 1984; Hisanaga and Hirokawa, 1988; Rao et al., 2002). The C-terminal 426 amino acid tail of the NF-M subunit is a determinant of axonal diameter in large caliber myelinated axons (Garcia et al., 2003; Rao et al., 2003). Furthermore, loss of this 426 amino acid C-terminal tail domain in mice results in a 30% reduction in sciatic nerve conduction velocity (Garcia et al., 2003).

The NF-M gene is composed of three exons and two introns (Levy et al., 1987; Myers et al., 1987). Twenty-two amino acids of the end of the coiled-coil rod domain and all 426 amino acids of the C-terminal tail required for radial axonal growth are generated from exon 3 (Garcia et al., 2003). Sequence analysis of NF-M exon 3 identified a repeated KSP motif, within the C-terminal 426 amino acid tail in mouse (Levy et al., 1987; Myers et al., 1987; Napolitano et al., 1987). Short tandem repeat regions, such as microsatellites, are found throughout the genomes of most vertebrates (Schlotterer, 2004). While most are thought to be inconsequential to normal gene function, recent evidence indicates that alterations in short tandem repeat lengths may have contributed to the evolution of species-specific traits. Variations of repeats within cis-regulatory elements may lead to morphological changes in the evolution of species, due to alterations in gene regulation (Carroll, 2000). Moreover, sequence analyses of several developmentally important genes in dog suggested that alterations in repeat lengths might have robust effects on dog morphology (Fondon and Garner, 2004). For example, alterations in the number of tandem repeats coding for stretches of glutamines and alanines within the Runx-2 gene correlated with changes in snout length and bending of the snout (Fondon and Garner, 2004). Contraction of a microsatellite within the Alx-4 gene was correlated with polydactyly in one breed of dog (Fondon and Garner, 2004). Similar to tandem repeat variations affecting snout morphology in dogs (Fondon and Garner, 2004), it is possible that expansion in the number of KSP repeats evolved as part of a mechanism to increase axonal diameter with increasing axonal length. Because the NF-M C-terminal tail is a determinant of axonal diameter in large caliber motor axons (Garcia et al., 2003), expansion in the number of KSP repeats within this region of NF-M may have been an evolutionary mechanism for increasing nerve conduction velocities to maintain the rate of signal propagation as larger mammals evolved.

Homo sapiens sapiens L. (human) purified DNA from an ethnic diversity panel was obtained through European Collection of Cell Cultures (ECACC). DNA from four inbred strains of Mus musculus L. (mouse) was isolated from tail biopsies from existing mice. Wild-trapped mice tail snips were obtained from specimens collected in northwest and northeast Missouri, USA. Phoca vitulina L. (harbor seal) DNA was isolated from lymphoblasts that were provided as a gift from Dr Mark Milanick. Tissues from Castor canadensis Kuhl (beaver), Sciurus carolinensis Gmelin (gray squirrel), Sus scrofa L. (feral hog), Ovis aries L. (sheep), Synaptomys cooperi Baird (southern bog lemming) and Blarina hylophaga Elliot (short-tailed shrew) were obtained from the vertebrate collection at the University of Missouri. Hydrochaeris hydrochaeris L. (capybara) DNA was provided as a gift from Dr Juan Campos of Kansas State University. Dr Guangping Gao of the University of Massachusetts Medical School provided Pan troglodytes Blumenbach (chimpanzee), Macaca mulatta Zimmerman (rhesus macaque) and Macaca fascicularis Raffles (crab-eating macaque) DNA as a gift. The San Diego Zoological Society, CA, USA, provided DNA from their Frozen Zoo® collection for Loxodonta africana africana Blumenbach (savanna elephant), Giraffa camelopardalis rothschildi Lydekker (giraffe), Ceratotherium simum simum Burchell (Southern white rhinoceros), Hippopotamus amphibious kiboko L. (hippopotamus) and Choloepus hoffmani Peters (Hoffman's two-toed sloth). DNA was isolated from all tissues by standard phenol-chloroform extraction.

Degenerate primers were designed for NF-M exon 3 based on published sequences (Table 1). In these sequences, the NF-M protein ends with either the two amino acids serine and aspartate (SD) or glycine and aspartate (GD). To minimize primer complexity, codons for these amino acids were omitted from the reverse degenerate primer. The forward degenerate primer was designed to the beginning of NF-M exon 3. The approximate locations of exon 3 degenerate PCR primers are indicated in a schematic of NF-M C-terminus (Fig. 1B).

Amplification using PCR was performed using 50–100 ng of genomic DNA, 10 or 20 pmol each of forward and reverse primer, 2.5 units of Ex Taq (Takara Bio Inc., Otsu, Shiga, Japan), Taq Polymerase (Invitrogen, Carlsbad, CA, USA) or GoTaq (Promega, Madison, MA, USA), 1X of manufacturer's buffer, 50–100 μmol l^–1 of each deoxyribonucleotide triphosphate in a total volume of 25 or 50 μml. PCR reactions were performed using the following parameters: one cycle of 95°C denaturation (5 min); 35 cycles of 95°C denaturation (30 s), 55–60°C annealing (30 s) and 72°C extension (2 min); one cycle of 72°C final extension (5 min). 10% of each reaction was analyzed by agarose gel electrophoresis to confirm amplification. PCR products were in the range of 1.3–1.6 kb (Fig. 1A). The remaining PCR products were purified using either Qiagen PCR purification kits or the Qiagen Gel Extraction kits (Valencia, CA, USA). Purified products were sequenced using the forward and reverse PCR primers, as well as specific internal primers (see Table S1 in supplementary material). DNA sequencing was performed on an ABI 3730 capillary DNA analyzer (Applied Biosystems Inc., Foster City, CA, USA) utilizing Big Dye Terminator chemistry at the DNA core facility at the University of Missouri. Analysis of sequencing was performed using DNAStar programs (DNAStar, Inc., Madison, WI, USA). The previously published sequences of human, Oryctolagus cuniculus L. (rabbit), Equus caballus L. (horse), Bos taurus L. (cow), Canus lupus familiaris L. (dog) and Rattus norvegicus Berkenhout (Norway rat) were obtained from GenBank (Benson et al., 2009). The BLAST tool at the Ensembl Genome Browser was used to identify NF-M exon 3 from Gorilla gorilla Savage (gorilla), Tursiops truncatus Montagu (bottlenose dolphin) and Pteropys vampyrus L. [large flying fox (bat)] reference assemblies (Hubbard et al., 2009). The ClustalW (Slow/Accurate Gonnet) method was used in all protein alignments in MegAlign (DNAStar, Inc.) (Thompson et al., 1994).

As the ‘KLLEGEE’ codons marked the beginning of exon 3 and are highly conserved across all intermediate filaments, we determined the reading frame by identifying this motif within our DNA sequences. We used this sequence to convert DNA sequences to protein sequences for phylogenetic analysis. NF-M exon 3 protein sequences were aligned in Megalign (DNAStar, Inc.) (Thompson et al., 1994). The ClustalW method aligned the sequences using the Slow-Accurate pairwise alignment parameters with a gap penalty of 10.00 and gap length of 0.10. The ClustalW method also used a Gonnet 250 protein weight matrix in the alignment. Construction of consensus trees were performed by MegAlign v.8.0 (DNAStar, Inc.) using the Neighbor Joining Method (Saitou and Nei, 1987) with 10,000 bootstrap trials and a random number generator seed of 111.

An established consensus tree of placental mammalian phylogeny was adapted (Murphy et al., 2001) to map NF-M KSP repeats and highlight independent KSP expansion events. Only species that were sequenced for NF-M exon 3 were indicated and species without sequencing were omitted for simplicity. Branching points for omitted species are also collapsed for simplicity. Species that were sequenced for NF-M exon 3 but were not used to generate the phylogeny were incorporated at appropriate points based on other mammalian phylogenies (Bininda-Emonds et al., 2007) and are indicated by dashed lines (Fig. 3).

To compare NF-M KSP number with animal size, linear regression analysis of KSP numbers with head body length (Macdonald, 1984; Myers, 2006) were performed using Sigmaplot v.11 (Systat Software Inc., Point Richmond, CA, USA). The PredictProtein Server (http://www.predictprotein.org) was used to predict secondary structure of the NF-M tail domain alignment (Rost et al., 2004).

Mice and rats were perfused intracardially with 4% formaldehyde in 0.1 mol l^–1 Sorenson's phosphate buffer, pH 7.2, and post-fixed overnight. The fifth lumbar root was isolated as previously described (Garcia et al., 2003).

Cows were killed by lethal injection of pentobarbital into the carotid artery. The fifth lumbar root was dissected and immediately submerged in 4% formaldehyde in 0.1 mol l^–1 Sorenson's phosphate buffer, pH 7.2. The tissue was allowed to fix overnight at 4°C. Samples were treated with 2% osmium tetroxide, washed, dehydrated and embedded in Epon-Araldite resin. Thick sections (0.75 μmm) for light microscopy were stained with p-phenylene diamine. Axonal diameters were measured using the AxioVision Digital Image Processing Software (Carl Zeiss MicroImaging, Oberkochen, Germany). Axon diameters were grouped in 0.5-μm bins as previously described (Garcia et al., 2003).

We amplified (Fig. 1A) and sequenced the NF-M C-terminal tail from species representing several Orders of the Class Mammalia. The alignment of translated DNA sequences revealed two regions of high sequence identity flanking a highly divergent KSP repeat region. Based on amino acid percentage identity and insertion events among mammals, the NF-M tail domain is divided into three sub-domains: the amino, KSP repeat and carboxy sub-domains (Fig. 1B). After sequencing, we quantified and classified the KSP repeats of all mammalian species analyzed (Table 2). Additionally, the NF-M C-terminal tail domain amino acids were abundant in glutamate and lysine. Structural analysis of the C-terminal tail domain failed to predict any obvious secondary structure (Rost et al., 2004).

Previously, sequence analysis of the NF-H C-terminal tail determined that intraspecies variations in the number of KSP repeats occurred in both dogs (Green et al., 2005) and humans (Al-Chalabi et al., 1999; Figlewicz et al., 1993; Tomkins et al., 1998; Vechio et al., 1996), albeit loss of KSP repeats in NF-H are associated with motor neuron disease (Al-Chalabi et al., 1999; Green et al., 2005). To determine if intraspecies variation in the number of KSP repeats occurred within the NF-M tail domain, exon 3 of the NF-M gene was sequenced from 127 mice (Table 3). These mice consisted of three inbred mouse strains (C57/Bl6, FVB and CD1) and 10 wild-trapped mice. Moreover, 11 out-bred rats and 76 humans, derived from an ethnic diversity panel, were analyzed (Table 3). Sequence analysis revealed no intraspecies allelic variations in the number of KSP repeats in any of the analyzed species (Table 3), suggesting stabilizing selection. Moreover, no amino acid variations were identified within exon 3 of any of the species, with the exception of published single nucleotide polymorphisms (SNPs) in mouse (NCBI SNPs rs31130946, rs30628515, rs30748232, rs31130043, rs52626242). No human SNPs were observed but previous studies have observed rare SNPs within human NF-M exon 3 (Garcia et al., 2006; Vechio et al., 1996).

We performed linear regression analyses of NF-M KSP repeat number with head–body length of each species (Fig. 2). Linear regression analyses indicated a bimodal distribution of species. The upper distributions (Fig. 2) suggested a positive correlation of NF-M KSP repeat number with mammalian size. However, the lower distributions (Fig. 2) suggested a weak negative correlation of NF-M KSP repeat number with mammalian size. The R² value was calculated from the linear regression of the red-grouped species only.

The NF-M C-terminal sequences generated a consensus tree that resulted in similar phylogeny to the adapted placental mammalian tree (Fig. 3A,B). Although there were differences in superordinal branching between the trees, grouping of animal Orders appeared accurate in the NF-M tree as members of Rodentia, Primates, Artiodactyla, Carnivora and Perissodactyla were monophyletic (Fig. 3B). Moreover, ancestral lineage indicates NF-M KSP expansion appears to have occurred independently several times within mammals. If we assume that mammalian ancestors of present-day mammals had fewer KSP repeats, then expansion of KSP repeats occurred across several clades. For example, dolphin, feral hog, hippo, giraffe, cow and sheep are all members of the Superorder Cetartiodactyla, yet they diverge into four separate clades. Dolphin and hippo are more closely related and retained the ancestral low KSP repeat number while giraffe, cow and sheep, members of Bovidae, diverged and their KSP repeats expanded to 19–22 (Fig. 3A,B). Similarly, feral hog, member of Suidae, diverged at an earlier point into another clade and its KSP repeats expanded to 19. This is supported by the differential patterning of the KSP repeats as feral hog expanded KXSPs and KXXSPs, while Bovidae members expanded almost exclusively KSPs (Table 2). Similar expansions appear to have occurred throughout the mammalian phylogeny, including in Rodentia, Primates and Proboscidea (elephant).

Upon further inspection, the bimodal distributions appear to be sub-divided into three groups based upon their positions within the graphs (Fig. 2). Moreover, individual members of these sub-groups patterned their KSP repeats in a similar manner (Fig. 4). While the blue and green group species (Fig. 4A,B) appear to have patterned their KSP repeats similarly, the blue group uniquely patterned the first two KSP repeats within the repeat sub-domain (Fig. 4A). The red group contained the most divergence in KSP repeat patterning (Fig. 4C). An alignment of KSP repeat sub-domain of species from each of the three sub-groups illustrates the distinguishing patterns of each group (Fig. 4D). It is interesting that the species within the red group appeared to have patterned their KSP repeats by adding repeats almost exclusively after the fourth consensus KSP repeat (Fig. 4D).

The diameter distribution and numbers of axons were analyzed from motor axons of the fifth lumbar ventral roots of five mice, three rats and two cows. All three species displayed typical bimodal distribution of axonal calibers (Fig. 5). The peaks in the distribution of the largest motor axons in mouse, rat and cow were 6, 7 and 10 μmm, respectively, suggesting motor axon caliber expansion in mammals (Fig. 5). Mouse and rat at six months of age have likely completed the majority of radial growth of motor axons (Garcia et al., 2003; Garcia et al., 2009; Rao et al., 2003; Rao et al., 2002; Rao et al., 1998; Zhu et al., 1997). However, seven-month-old cows are still relatively immature and may not have completed axonal radial growth. Therefore, it is likely that the peak distribution of the largest group of axons in cow will be shifted to a greater caliber at completion of axonal radial growth. It should be emphasized that the axonal diameter data are limited to three species and more experimentation is warranted.

Evolutionary relationships of the structure of the NF-M C-terminus in mammals were examined to better understand its influence on axonal diameter. Results indicate that the C-terminus contains highly conserved amino and carboxy flanking sub-domains that surround a highly variable sub-domain with characteristic KSP repeats. The high variability of the KSP repeat sub-domain is due not only to higher substitution rates, but also due to 362 non-conserved insertion events of amino acid codons according to the total sequence alignment (see Fig. S1 in supplementary material). By contrast, the alignment contained few insertion events in the amino and carboxy sub-domains that were outside of the KSP repeat sub-domain (see Fig. S1 in supplementary material). The number of KSP repeats correspond directly to the length of the NF-M C-terminus, meaning that the expansion of KSP repeats is the predominant mechanism utilized for lengthening this domain among mammals.

Our initial pooled analysis revealed a general relationship between KSP expansion and head–body length (Fig. 2), suggesting a role for KSP expansion in axon diameter and therefore conduction velocity. However, a finer detailed phylogenetic analysis suggests that KSP expansion may have occurred independently across multiple clades as a convergent mechanism to influence axonal diameter. Mapping KSP repeat numbers onto mammalian phylogenies suggests that ancestral mammals contained relatively few KSP repeats (∼5). Some present-day mammals maintain this number of KSP repeats. For example, we found no evidence of KSP expansion in the Carnivora, Chiroptera and Perissodactyla (Fig. 3A). But within several different clades, including Cetartiodactyla, Rodentia, Primates and Proboscidea, independent KSP expansion events occurred that appear to have been adaptive as the expansions remain in present-day descendants. Moreover, the constructed trees (Fig. 3A,B) suggest that independent KSP expansion occurred in Rodentia at least three times as rat, gray squirrel and capybara have 8, 11 and 14 KSP repeats, respectively. We suggest that gray squirrel and capybara KSP expansion occurred independently, as gray squirrels developed an expansion of KXXSPs, while capybara expanded with KSPs. It is also possible that rat and capybara could be descended from a common ancestor as they both have increased numbers of KSP repeats. However, rat and capybara diverge from one another by several branch points in Rodentia (Fig. 3A,B), so the KSP expansions still may have occurred independently.

Independent expansion of KSP repeats occurring across several mammalian clades suggests that increasing the length of the C-terminus may be a common mechanism utilized to influence axonal diameter and may represent an example of convergent evolution. If expanding the length of NF-M C-terminus through the addition of KSP repeats is part of the mechanism utilized to determine axonal diameter, then, to our knowledge, this would represent the first functional enhancement associated with the expansion of a repetitive DNA element. It has become clear, though, that within individual species NF-M C-terminal expansion is not the only available mechanism for influencing axonal diameter. For example, despite having 22 KSP repeats, cows have a bimodal distribution of motor axons of the fifth lumbar root (Fig. 5B) with relatively small diameter axons, yet all axons contain the same NF-M protein. Therefore, expansion of NF-M C-terminus may only be part of the mechanism utilized to determine axonal diameter.

We hypothesized that as mammals evolved larger body plans across different clades, selective pressures resulted in increased conduction velocities to maintain the rate of signal propagation. Furthermore, we proposed that expansion of the KSP repeats within the NF-M C-terminus may have been an evolutionary mechanism that resulted in increased axonal diameters. Direct measurements of nerve conduction velocity are extremely difficult to obtain from many of the species analyzed in this study. However, analysis of individual nerve fibers of adult cat hind-limb demonstrated a linear relationship of axonal diameter with conduction velocity, with the largest diameter fibers having the fastest conduction velocities (Hursh, 1939). Subsequent studies (Arbuthnott et al., 1980a; Arbuthnott et al., 1980b; Boyd and Kalu, 1979; Waxman, 1980; Westbury, 1982), as well as theoretical results (Rushton, 1951; Smith and Koles, 1970), support this finding. Moreover, direct measurements of nerve conduction velocities in human (Chang et al., 2006) and mouse (Garcia et al., 2003; Garcia et al., 2009) suggest that larger mammals have faster rates of conduction. Our results suggest that increased number of KSP repeats within NF-M is associated with larger axonal diameter (Fig. 5). If the relationship between axonal diameter and conduction velocity holds for other mammals, then it is likely that an expanded NF-M C-terminus is correlated with faster conduction velocity for a subset of mammals.

It has long been thought that myelin-dependent phosphorylation of NF KSP repeats was required for radial axonal growth (de Waegh et al., 1992). However, the generation of a mouse expressing full-length, KSP phosphorylation-incompetent NF-M has challenged the role of KSP phosphorylation in establishing axonal diameter (Garcia et al., 2009). These new results were not inconsistent with our proposal. It is possible that the overall length of NF-M C-terminus determines axonal diameter, as axonal diameter and conduction velocity were altered only in mice missing the entire NF-M C-terminus (Garcia et al., 2003). Recently, mathematical models of NF side arms suggest that human NF-M C-termini project the farthest from the NF core relative to NF-L and NF-H C-termini (Chang et al., 2009). The distance that NF-M side arms project appears to be largely dependent upon adjacent C-termini stoichiometry, amino acid sequence, and the charge of the residues (Chang et al., 2009). Since the amino acid compositions and overall charges of NF-M C-termini are similar across mammals (see Table S2 in supplementary material), it is reasonable to predict that lengthening the NF-M C-terminus would allow for it to project farther. Mechanistically, it has been proposed that NF C-terminal tails form cross-bridging structures with adjacent NFs or microtubules (Hirokawa et al., 1984; Hisanaga and Hirokawa, 1988). Alterations in the length of the KSP repeat sub-domain may have evolved as a means of extending the length of the cross-bridge, allowing for longer-range interactions that may provide greater structural stability in larger diameter axons. Our data suggests that increasing the overall length of the NF-M C-terminus by adding additional KSP repeats may be a potential mechanism for increasing axonal diameter.

NF-M KSP repeats may not be considered classic tandem repeats as they consist of tri- tetra- and pentapeptide repeats (KSP, KSD, KXSP, KXXSP) and are not typically adjacent to one another but interspersed along the sub-domain. Nevertheless, the number and patterning of KSP repeats was highly divergent between mammalian species (Table 2 and Fig. 4D). Moreover, the KSP repeat sub-domain is located within a region of NF-M that is required for radial growth (Garcia et al., 2003). We propose that much like tandem repeat variation affecting dog snout morphology (Fondon and Garner, 2004), changes in KSP repeat number within this highly divergent sub-domain of NF-M C-terminus appear to be important in altering axonal morphology and may have been a vital part of a mechanism for increasing conduction velocity to maintain the rate of signal propagation as subsets of larger mammals evolved.

We thank Ms Monica Burgers for assistance with tissue preparation for light microscopic and morphometric analysis, and Mr Brian Reigle for writing the scripts utilized in automated determination of numbers of axons in assigned groups for axonal caliber measurements.

 
• IF

  intermediate filaments

 
• KSP

  lysine–serine–proline

 
• NF

  neurofilament

 
• NF-H

  neurofilament heavy

 
• NF-L

  neurofilament light

 
• NF-M

  neurofilament medium

 
• R_a

  axial resistance

 
• R_m

  membrane resistance

 
• λ

  length constant