Abstract
Intraocular pressure (IOP) is the primary risk factor for developing glaucoma, yet little is known about the contribution of genomic background to IOP regulation. The present study leverages an array of systems genetics tools to study genomic factors modulating normal IOP in the mouse. The BXD recombinant inbred (RI) strain set was used to identify genomic loci modulating IOP. We measured the IOP in a total of 506 eyes from 38 different strains. Strain averages were subjected to conventional quantitative trait analysis by means of composite interval mapping. Candidate genes were defined, and immunohistochemistry and quantitative PCR (qPCR) were used for validation. Of the 38 BXD strains examined the mean IOP ranged from a low of 13.2mmHg to a high of 17.1mmHg. The means for each strain were used to calculate a genome wide interval map. One significant quantitative trait locus (QTL) was found on Chr.8 (96 to 103 Mb). Within this 7 Mb region only 4 annotated genes were found: Gm15679, Cdh8, Cdh11 and Gm8730. Only two genes (Cdh8 and Cdh11) were candidates for modulating IOP based on the presence of non-synonymous SNPs. Further examination using SIFT (Sorting Intolerant From Tolerant) analysis revealed that the SNPs in Cdh8 (Cadherin 8) were predicted to not change protein function; while the SNPs in Cdh11 (Cadherin 11) would not be tolerated, affecting protein function. Furthermore, immunohistochemistry demonstrated that CDH11 is expressed in the trabecular meshwork of the mouse. We have examined the genomic regulation of IOP in the BXD RI strain set and found one significant QTL on Chr. 8. Within this QTL, there is one good candidate gene, Cdh11.
Glaucoma is a diverse set of diseases with heterogeneous phenotypic presentations associated with different risk factors. Untreated, glaucoma leads to permanent damage of axons in the optic nerve and visual field loss. Millions of people worldwide are affected (European Glaucoma Prevention Study et al. 2007; Leske et al. 2007) and it is the second leading cause of blindness in the United States (Medeiros et al. 2003). Adult-onset glaucoma is a complex collection of diseases with multiple risk factors and genes with differing magnitudes of effects on the eventual loss of RGCs. The severity of the disease appears to be dependent on the interaction of multiple genes, age, and environmental factors (Herndon et al. 2004). There are also a number of phenotypic risk factors for POAG including: age, ethnicity, central corneal thickness and axial length (Gordon et al. 2002). The primary risk factor is an elevated intraocular pressure (IOP) (Klein et al. 2004). There are known genetic mutations that affect IOP that result in inherited glaucoma (Stone et al. 1997; Wiggs 2007). The prime example is MYOC, a protein secreted by the trabecular meshwork and mutations in this protein cause ER stress which results in a decrease in the function of the trabecular meshwork and an elevation in IOP (Joe et al. 2003; Kasetti et al. 2016). We know a considerable amount about the regulation of IOP from the production of aqueous humor to the outflow pathways. IOP is a complex trait affected by different tissues in the eye each of which is regulated by multiple genomic loci. Interestingly, there are very few studies (Springelkamp et al. 2014; Nag et al. 2014; Chen et al. 2014; Choquet et al. 2017; Ozel et al. 2014; Springelkamp et al. 2017) that have identified genomic loci in humans modulating normal IOP.
In the present study, we are using the BXD recombinant inbred (RI) strain set that is particularly suited for the study of genetics and the effects on the severity of glaucoma. This genetic reference panel presently consists of over 202 strains (Peirce et al. 2004), and we are now in the unique position of being able to study the eyes in mice with shuffled genomes from the two parental strains, C57BL/6J and the DBA/2J. There are over 7,000 break points in our current set of BXD strains. For this study, our group has measured IOP of 506 eyes from 38 strains to identify genomic loci modulating IOP. A systems genetics approach to glaucoma is a relatively new branch of quantitative genetics that has the goal of understanding networks of interactions across multiple levels that link DNA variation to phenotype (Mozhui et al. 2008). Systems genetics involves an analysis of sets of causal interactions among classic traits such as IOP, networks of gene variants, and developmental, environmental, and epigenetic factors. The main challenge is the need for comparatively large sample sizes and the use of more advanced statistical and computational methods and models. We finally have a sufficiently large number of strains to use this approach (Geisert et al. 2009; Freeman et al. 2011). Our goal is now to combine data across several levels from DNA to ocular phenotype and analyze them with newly developed computational methods to understand pre-disease susceptibility to glaucoma along with the genetic networks modulating the response of the eye to elevated IOP.
Methods
Mice
This study measured the IOP in the 36 BXD strains of mice along with the parental strains, the C57BL/6J mouse strain and the DBA/2J mouse strain. None of the BXD strains included in this study carried both mutations (Tyrp1 and Gpnmb) known to cause the severe glaucoma phenotype observed in the DBA2/J strain. All of the mice in this study were between 60 and 120 days of age, a time before there is any significant elevation in IOP due pigment dispersion (Anderson et al. 2002). The data presented in this paper is based on measurements from 506 eyes with roughly equal numbers of male and female mice. All breeding stock was ordered from Jackson Laboratories (Bar Harbor, ME) and maintained at Emory. Mice were housed in the animal facility at Emory University, maintained on a 12 hr light/dark cycle (lights on at 0700), and provided with food and water ad libitum. IOP measurements were made between 0900 and 1100. Both eyes were measured and the data from each eye was entered into the database. An induction–impact tonometer (Tonolab Colonial Medical Supply) was used to measure the IOP according to manufacturer’s instructions and as previously described (Saleh et al. 2007; Nagaraju et al. 2009). Mice were anesthetized with Avertin (334 mg/kg) or ketamine/xylazine (100,15mg/kg). Three consecutive IOP readings for each eye were averaged. IOP readings obtained with Tonolab have been shown to be accurate and reproducible in various mouse strains, including DBA/2J (Wang et al., 2005). All measurements were taken approximately 10 min after the induction of anesthesia. These IOP measurements were made on mice prior to two different experimental procedures, blast injury to the eye (Struebing et al. 2018a) or elevation of IOP by injection of magnetic beads into the anterior chamber (Struebing et al. 2018b). All of the measurements were made on normal eyes. Previous studies have found significant effects of anesthesia on IOP (Ding et al. 2011; Qiu et al. 2014). When we compared the IOP of all mice anesthetized with Avertin to those anesthetized with ketamine/xylazine over the entire dataset there was no significant difference between the two groups. We specifically examined the C57BL/6J mouse strain, with 11 mice anesthetized with Avertin (mean IOP 10.2, SD 0.15) and 27 mice anesthetized with ketamine/xylazine (mean IOP 11.2, SD 2.9). Although there was a trend for the ketamine/xylazine anesthetized mice to have higher IOP, there was no statistically significant difference between the two groups using the Mann-Whitney U-test.
Interval Mapping of IOP Phenotype
The IOP data will be subjected to conventional QTL analysis using simple and composite interval mapping along with epistatic interactions. The means for each BXD strain were calculated and these means were used to generate genome-wide interval maps. All of these data are available on GeneNetwork.org. Genotype was regressed against each trait using the Haley-Knott equations implemented in the WebQTL module of GeneNetwork (Chesler et al. 2005; Rosen et al. 2003; Carlborg et al. 2005). Empirical significance thresholds of linkage are determined by permutations (Churchill and Doerge 1994). We correlated phenotypes with gene expression data for whole eye and retina (Geisert et al. 2009; King et al. 2015; Templeton et al. 2013).
Immunohistochemistry
For immunohistochemical experiments mice were deeply anesthetized with a mixture of 15 mg/kg of xylazine (AnaSed) and 100 mg/kg of ketamine (Ketaset) and perfused through the heart with saline followed by 4% paraformaldehyde in phosphate buffer (pH 7.3). The eyes were embedded in paraffin as described by Sun et al. (Sun et al. 2015). The eyes were dehydrated in a series of ethanol and xylenes changes for 20 min each (50% ETOH, 70% ETOH, 90% ETOH, 95% ETOH, two changes of 100% ETOH, 50% ETOH with 50% xylenes, two changes of 100% xylene, two changes of paraffin). The eyes were then embedded in paraffin blocks. The eyes were sectioned with a rotary microtome at 10µm and mounted on glass slides. Paraffin was removed from the sections and the sections were rehydrated. The sections were rinsed in PBS, and then placed in blocking buffer containing 2% donkey serum, 0.05% DMSO and 0.05% Triton X-100 for 30 min. The sections were rinsed in PBS, and then placed in blocking buffer containing 2% donkey serum, 0.05% DMSO and 0.05% Triton X-100 for 30 min. The sections were incubated in primary antibodies (1:500) against Cadherin 11 (Thermofisher, Cat. #71-7600, Waltham, MA) overnight at 4°. After rinsing, the sections were incubated with secondary antibody conjugated to AlexaFluor-488 (donkey anti-rabbit, Jackson Immunoresearch Cat #711-545-152, Westgrove, PA), (1:1000), for 2 hr at room temperature. The sections were then rinsed 3 times in PBS for 15 min each. Then they were counterstained with TO-PRO-3 iodide purchased from Invitrogen (T3605, Invitrogen, Eugene OR). The slides were flooded with Fluoromount-G (SouthernBiotech Cat #. 0100-01, Birmingham, AL), and covered with a coverslip. All images were photographed using a Nikon Eclipse TE2000-E (Melville, NY) confocal microscope and images were acquired by Nikon’s EZ-C1 Software (Bronze Version, 3.91).
PCR Validation
Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) were used to validate the mRNA expression level of Cdh11 and Cdh8 and Myoc in whole eyes of C57BL/6J mice. Primers were designed for Cdh11, Cdh8 and Myoc using Primer BLAST-NCBI so that predicted PCR products were approximately 150bp long. Sequences of the PCR primers are listed in Table 1. PCR reactions were carried out in triplicate in 10μl reactions containing 5μl of 2x QuantiTect SYBR Green PCR Master Mix (Qiagen, Cat #204141 Hilden, Germany), 0.5 μl of forward primer (0.5μM), 0.5 μl of reverse primer (0.5 μM), 2μl of template cDNA(10ng) and 2μl of RNA free H2O. PCR of mouse genes was performed using a program beginning at 95° for 15 min, followed by 40 cycles of reaction with denaturation at 94° for 15 sec, annealing at 59° for 30 sec and extension at 72° for 30 sec of each cycle. The cycle threshold (Ct) values were normalized to a mouse housekeeping gene peptidylprolyl isomerase A (Ppia) to generate Delta Ct values (ΔCt) for each gene. Fold changes (FC) were calculated from the differences of each gene when compared with Cdh11(ΔΔCt) using the formula FC= 2^(-ΔΔCT). Data were represented as mean ± SE of mean. Four biological independent samples were used for statistical analysis using Mann-Whitney-U-test (Figure 4).
Results
The overall goal of the present investigation was to determine if specific genomic loci modulate IOP in the BXD RI strains. IOP was measured in 506 eyes from 38 BXD RI strains and the two parental strains C57BL/6J mouse and DBA2/J mouse. To create a mapping file the strain averages and standard errors were calculated (Figure 1). The mean IOP measured across the 38 strains was 13.2 mmHg and the standard error of the mean was 1.5 mmHg. The strain with the lowest IOP was DBA2/J, with an average IOP of 10.9 mmHg. The strain with the highest IOP was BXD48 with an average IOP of 17.1 mmHg. The IOP of the parental strains was 11.6 mmHg for the C57BL/6J and 10.9 mmHg for the DBA2/J. This is a substantial amount of genetic transgression across the BXD RI strain set. This type of phenotypic variability is a clear indication that IOP is in fact a complex trait. These data can also be used to calculate the heritability of IOP. Figure 1 reveals a considerable variability in the IOP from strain to strain and the standard error for each strain is rather small. This type of data suggests that the genetic variability has a greater effect than the environmental variability. These data can be used to calculate the heritability H2 of IOP. H2 is the genetic variance (Vg) of the trait divided by the sum of genetic variance plus the environmental variance (Vg +Ve). The genetic variance can be estimated by taking the standard deviation of the mean of IOP for each strain (Vg = 1.7 mmHg). The environmental variance can be estimated by taking the mean of the standard deviation across the strain (Ve = 3.2 mmHg). Using the formula for heritability, H2 = Vg/(Vg + Ve), the calculation of 1.7 mmHg /(1.7 mmHg + 3.2 mmHg) reveals that H2 = 0.35. Thus, IOP is a heritability trait in the BXD RI strain set.
The distribution of IOP across the BXD strains is illustrated in a bar chart with means and Standard Error of the Mean In the 38 strains of mice the IOP ranged from a low of 10.9 mmHg to a high of 17.1 mmHg.
Genome Wide Mapping
Taking the average IOP from 38 strains of mice we performed an unbiased genome-wide scan to identify genomic loci (QTL) that modulate IOP. The genome-wide interval map (Figure 2) identifies one significant peak on Chr. 8. Examining an expanded view of Chr.8, 90 to 110 Mb (Figure 3), the peak of the IOP QTL reaches significance from the genomic marker rs3661882 (96.2 Mb) to rs13479958 (103 Mb). BXD strains with higher IOPs (Figure 3B) tend to have the C57BL/6J allele (red) and strains with lower IOPs tend to have the DBA2/J allele (green). When the distribution of genes within this region is examined (gene track Figure 3A) the significant portion of the QTL peak covers a region of the genome that is a gene desert. Within this region there are only 4 annotated genes: Gm15679 (predicted gene 15679, Chr.8: 99.01- 99.03Mb), Cdh8 (cadherin 8, Chr.8: 99.03-99.42Mb), Cdh11 (cadherin11, Chr.8:102.63-102.79Mb) and Gm8730 (predicted gene 8730, Chr.8: 102.86- 102.87Mb). Using the tools available on GeneNetwork (genenetwork.org) we are able to identify potential candidates for modulating IOP in the BXD RI strains. The candidate genes can either be genomic elements with cis-QTL or they can be genes with nonsynonymous SNPs changing protein sequence and affecting protein function. Within this region there are only two putative candidate genes. None of the candidate genes had valid cis-QTL in either the eye dataset (Geisert et al. 2009) or the DoD normal retinal dataset on GeneNetwork (King et al. 2015). Two of the genes in this region had non-synonymous SNPs: Cdh11 and Cdh8.
A genome-wide interval map of IOP. The interval map plots the likelihood ratio statistic (LRS) across the genome from chromosome 1 to chromosome X. The light gray line is the suggestive level and the light red line is genome-wide significance (P = 0.05). When the IOP measures were mapped to the mouse genome there was a significant association between IOP and a locus on Chromosome 8.
The interval map for Chr. 8: 90 to 110 Mb is illustrated. A is the gene tract, that identifies the locations of known genes across the genome. B is a haplotype map for the different BXD RI strains listed to the right and ranked from the highest IOP to the lowest IOP. The location of genomic markers is indicated by black vertical lines. C is an expanded version of the interval map for IOP. Finally, the bottom trace (yellow) identified the location of SNPs between the C57BL/6 mouse and the DBA2/J mouse. The genomic location is indicated along this lower trace. Notice that the peak of the QTL in C sits in a region of the genome that contains very few known genes (A).
For the initial evaluation of the two candidate genes we examined their expression level in a microarray dataset hosted on GeneNetwork: the eye database (Eye M430v2 (Sep08) RMA). In the eye dataset, the highest level of expression for a valid Cdh11 probe set (1450757_at) is 10.8 Log2. This dataset is on a Log2 scale with the mean mRNA expression set to 8. Thus, Cdh11 is expressed at approximately eightfold above the mean. For Cdh8 the highest expression of a valid probe set (1422052_at) is 7.3 Log2 or approximately twofold below the mean. These data demonstrate that Cdh11 is expressed at a level over eightfold higher in the whole eye relative to the expression of Cdh8.
To confirm the expression levels of Cdh11 and Cdh8 in the eye, we examined the levels of mRNA in the whole eye by RT-qPCR (Figure 4). In 4 biological replicate samples, we examined the levels of Cdh11, Cdh8 and Myoc (a marker of trabecular meshwork cells, (Takahashi et al. 1998)). Our PCR analysis confirmed the general findings of the microarray data sets. In the 4 biological samples of whole eye, Cdh11 was more highly expressed than Cdh8. The average of the 4 samples demonstrated a more than twofold higher expression of Cdh11 (ΔCt = 5.75 ± 0.27) than Cdh8 (ΔCt = 7.36 ± 0.33). Myoc (ΔCt = 5.27 ± 0.38) was expressed at a higher level than both Cdh8 and Cdh11. All of these data taken together indicate that Cdh11 is the prime candidate for an upstream modulator of IOP in the mouse. Using the RNA-seq data presented in the study by Carnes et al. (Carnes et al. 2018), we examined the expression level of CDH11 and CDH8 in the human trabecular meshwork. CDH11 is highly expressed in adult human trabecular meshwork; while CDH8 is expressed at very low levels if at all. The levels of CDH11 are more than 55 times higher than that for CDH8(FPKMCDH11 = 88.5 vs. FPKMCDH8 = 1.6, FPKM: Fragments Per Kilobase Million). This remains true for fetal human trabecular meshwork (FPKMCDH11 = 270.1 vs. FPKMCDH8 = 4.6).
Expression of Cdh11, Cdh8 and Myoc in whole eye. The mRNA expression levels of Cdh11, Cdh8 and Myoc in mice whole eye are shown as fold changes normalized to Cdh11. Significant changes in gene expression were observed using a Mann-Whitney U-test.
To determine if any of the non-synonymous SNPs in either Cdh8 or Cdh11 could affect protein function, we ran a SIFT (Sorting Intolerant From Tolerant) analysis (Kumar et al. 2009). The SNP in Cdh8 (rs37017336 cysteine to arginine) had a SIFT value of 0.19 indicating that this mutation in the protein would be tolerated, not affecting protein function. The two SNPs in Cdh11 (rs30742273 valine to methionine, and rs33464298 glutamic acid to aspartic acid) both had SIFT values of 0.03, predicting that each of these SNPs would not be tolerated and would be deleterious to protein function. All of these data point to Cdh11 as the one good candidate gene.
Distribution of Cadherin 11 in the Eye
To determine if Cadherin 11 is found in structures associated with the control of IOP, we stained sections of the eye for Cadherin 11. In these sections, there was antibody-specific staining of several structures (Figure 5A) as demonstrated by the lack of staining in control sections stained with secondary antibody only (Figure 5B). There is extensive labeling of all layers of the cornea. The epithelium of the ciliary body is also heavily labeled as well as labeling of pars plana. There is also light labeling of the retina. At higher magnification (Figure 5C), clear labeling of the trabecular meshwork (arrow) is observed. These data reveal that the expression pattern of Cadherin 11 in the mouse eye and its specific expression in the cells of the trabecular meshwork is appropriate for a protein involved in regulation of IOP.
The distribution of cadherin 11 in the limbal area of the eye is illustrated. The section in A was stained with an antibody specific to cadherin 11 (green) and for DNA (blue). This staining is specific to the primary antibody for it is not observed in a section stained with the secondary antibody alone (B). The staining pattern of the trabecular meshwork is shown at higher magnification in C (arrow). A and B are taken at the same magnification and the scale bar in panel B represents 25 µm.
Discussion
The normal regulation of intraocular pressure is a balance between production in the ciliary body and outflow (Brubaker 1991; Goel et al. 2010). In the human, IOP can range from a relatively low pressure to an extremely high one that occurs in acute angle closure glaucoma. It is generally accepted that the “normal” range for IOP in humans is from 12mmHg to 22mmHg (Hollows and Graham 1966; Renard et al. 2010). In addition, monitoring throughout the day reveals IOP is pulsatory and has a diurnal variability (Aptel et al. 2016). These findings tell an interesting story about the regulation of pressure in the eye; however, the primary driving force behind the intense investigation of IOP in humans is the fact that it is the primary risk factor for developing glaucoma (Sommer et al. 1991). Furthermore, all of the current treatments for glaucoma center around lowering IOP either by pharmacological approaches or surgery (Cohen and Pasquale 2014; Gedde et al. 2013).
The known association between elevated IOP and glaucoma has driven most of the study of IOP in human populations (Gordon et al. 2002; Ojha et al. 2013; Ozel et al. 2014). Most of these studies involve the study of glaucoma, but a few have a primary focused on the regulation of IOP. These studies have found that IOP is a heritable trait with estimates of heritability ranging from 0.39 to 0.64 (Klein et al. 2004; Chang et al. 2005; Carbonaro et al. 2009; van Koolwijk et al. 2007). In the present study, we found that the heritability of IOP in the BXD RI strains was 0.35. Thus, the mouse strains demonstrated a heritability near the lower end of the human populations. The interest has prompted studies to identify genes regulating IOP. In a genome-wide association study of IOP involving 11,972 subjects, significant associations were observed with SNPs in two genes, GAS7 and TMCO1 (van Koolwijk et al. 2012). Both of these genes are expressed at high levels in the ciliary body and trabecular meshwork (Liton et al. 2006) and both of the genes interact with known glaucoma risk genes (van Koolwijk et al. 2012). TMCO1 is also known to be associated with severe glaucoma risk (Burdon et al. 2011).
In an effort to understand the regulation of IOP and its effects on the retina, many research groups have used inbred mouse strains (Savinova et al. 2001; Struebing and Geisert 2015; Sappington et al. 2010; Cone et al. 2010; Samsel et al. 2011; Chintalapudi et al. 2017). IOP varies widely across different strains of mice (Savinova et al. 2001; Wang et al. 2005), ranging from a low of 11mmHg in the BALB/c mouse strain to a high of 19mmHg in the CBA/Ca mouse strain. In the present study, the average measured IOP across the 34 strains was 13.2mmHg. The lowest measured IOP was 10.9mmHg in the DBA/2J strain and the highest was 17.1mmHg in the BXD48 strain. All of these studies are complicated by the difficulties of measuring IOP in the mouse (Wang et al. 2005; Ding et al. 2011). The present study examines IOP in the anesthetized mouse and there is substantial evidence that anesthesia can affect IOP (Ding et al. 2011; Qiu et al. 2014). That being said the present study relies on the variability across the BXD RI strains. All of the current data suggests that anesthesia affects the absolute IOP value; however, the relative differences in IOP between BXD strains is sufficient to map a QTL. Thus, we were able to use these measures to map a single significant QTL on Chr. 8 that modulates IOP in the mouse. The peak of the QTL was in a gene desert and within this region there were only two potential candidate genes that could be modulating IOP in the BXD strain set. Based on expression of mRNA in the eye microarray dataset, the findings of real time PCR and functional analysis of SNPs, Cdh11 appears to be the best candidate. Cdh11 is expressed approximately eightfold higher in the eye than is Cdh8. A recent study of human trabecular meshwork using RNA Seq found that CDH11 is expressed at relatively high levels and at least 55-fold higher that the expression of CDH8 (Carnes et al. 2018). Furthermore, CDH11 is highly expressed in cultured human trabecular meshwork cells (Paylakhi et al. 2012). We found that Cadherin 11 is expressed in the trabecular meshwork of the mouse using indirect immunohistochemistry. All of these data suggest that the expression of Cadherin 11 in the trabecular meshwork modulates IOP across the BXD RI strain set.
How is it possible that a cadherin can modulate IOP in the mouse eye? IOP is regulated by fluid resistance at the trabecular meshwork and Schlemm’s canal (Ethier et al. 1986; Brubaker 1975). The stiffness of these structures is determined by the extracellular matrix within the trabecular meshwork and Schlemm’s canal (Zhou et al. 2012). The contractile nature of the cells within the inner wall is considered to play an important role in regulating outflow resistance (Bradley et al. 1998; Johnson 2006; Vranka et al. 2015). The dysregulation or poor organization of extracellular matrix may increase the fluid resistance, leading to an elevation of the IOP. Cdh11 was recently revealed to be a novel regulator of extracellular matrix synthesis and tissue mechanics (Row et al. 2016), and it is also found to be highly expressed in cultured human trabecular meshwork cells (Paylakhi et al. 2012). It is possible that the IOP can be regulated by Cdh11 and related pathways by altering the extracellular matrix structure of the trabecular meshwork. Future studies about the role of Cdh11 in the trabecular meshwork may give insights into the mechanism of IOP modulation.
Acknowledgments
We thank Robert W. Williams and his group for providing a wealth of bioinformatic resources on GeneNetwork.org. We thank Chelsey Faircloth for her assistance in immunostaining. This study was supported by an Unrestricted Grand from Research to Prevent Blindness, NEI grant R01EY017841 (E.E.G.), Owens Family Glaucoma Research Fund, P30EY06360 (Emory Vision Core) and DoD CDMRP Grant W81XWH-12-1-0255 from the USA Army Medical Research & Materiel Command and the Telemedicine and Advanced Technology (E.E.G.).
Footnotes
Communicating editor: D. Threadgill
Supplemental Material is available online at www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200190/-/DC1.
- Received November 16, 2017.
- Accepted February 28, 2018.
- Copyright © 2018 King et al.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.