Innovative Technologies: X-Rays Get in Synch

Synchrotrons may have been designed with high-energy physics in mind, but now biologists are starting to see the light too. Jeffrey Gillow, a researcher at Brookhaven National Laboratory, has been making use of the X-ray microscope at the National Synchrotron Light Source (NSLS) in New York to see extremely fine details of bacteria biochemistry in a technique known as X-ray spectromicroscopy. 
 
Gillow’s team, funded by the Department of Energy Office of Science, uses “soft” X-rays (up to 800 electronvolts, a relatively small amount of energy) to study the chemical structure of organic compounds. “It’s great because you get more than just a detailed picture,” says Gillow. “You also get chemical information about your sample.” 
 
Gillow uses the synchrotron to precisely tune the energy of the X-rays, knocking carbon electrons out of their orbitals. The resulting disturbance changes the bonds of molecules, and the researchers can read the spectra to see which elements were bonded to which. 
 
The precise nature of the X-ray microscope allows Gillow to see exacting chemical detail within bacteria. Recently, his team used the 30-nanometer resolution of the NSLS X-ray microscope to observe an immature spore develop within a Clostridium sp. bacterium, something far too minute and hidden within its host for any conventional electron microscope. These findings were published in the June 2005 issue of the Journal of Electron Spectroscopy and Related Phenomena. 
 
Another strength of X-ray spectromicroscopy is that samples require only minimal preparation. Says Gillow, “There is no staining necessary. Basically you just put the sample on the window and away you go.” Without staining or heat fixing, the bacterium maintains its naturally occurring biochemical composition. 
 
However, X-ray spectromicroscopy does require that experiments be conducted in close proximity to a synchrotron. And even though there are currently 40 of these very expensive machines in the world, only a few have the capabilities to conduct this type of research. Further, no live specimens can be studied due to the extraordinary amount of radiation they receive. 
 
Regardless, X-ray spectromicroscopy offers environmental scientists chemical detail and unaltered observations like never before, which is key to understanding the complex biochemical reactions that bacteria undergo in the environment. For example, groups interested in bioremediation can now see on a molecular scale how bacteria alter the chemistry of metals and radionuclides and remove them from soils and waters. 
 
A better understanding of subcellular microorganism chemistry, specifically sporulation, might also help authorities neutralize bioterrorism threats before they become a problem. “Finding ways to interrupt sporulation could stop bioterrorism attacks,” says Gillow. “But I doubt you will ever see a synchrotron at an airport scanning your luggage.”


Background
In the human genome, alleles at two loci on the same chromosome often show stochastic dependence. This nonrandom pair-wise allelic association, or linkage disequilibrium (LD), is both interesting evolutionarily in its own right and a powerful tool in the mapping of genes for complex genetic traits. The development of the technology to genotype large numbers of single-nucleotide polymorphisms (SNPs) provides the means to explore the extent and patterns of linkage disequilibrium in the genome. The two pseudoautosomal regions on chromosomes X and Y (PAR1 and PAR2) are approximately 2.6 and 0.4 Mb respectively, and are located at the tips of the sex chromosomes. Within these regions pairing and recombination may occur between X and Y as it naturally does in the autosomes or between two X chromosomes in females. Dense genetic linkage maps have revealed a high rate of recombination in PAR1 [1], prompting comparisons between LD patterns in PAR1 and the heterosomal region on chromosome X. The purpose of this paper is to explore the rate of decay in LD as a function of distance between males and females both across and within autosomal and sex chromosomes.

Methods
Data used for our analyses were obtained from the Collaborative Study on the Genetics of Alcoholism (COGA). COGA is a large-scale family study designed to identify genes that contribute to the risk for alcoholism. Our analyses used data genotyped by both Affymetrix and Illumina. Four hundred and thirty-four genetic markers on chromosome X were analyzed (310 SNPs by Affymetrix and 124 SNPs by Illumina). SNPs genotyped and analyzed on the 22 autosomes by Affymetrix and Illumina totaled 15,371. Given that the proportion of self-reported non-Caucasian individuals in the COGA sample was small, our results are based on 102 Caucasian pedigrees. These 102 pedigrees were composed of 430 genotyped male and 446 genotyped female participants; 9 genotyped participants whose sex was not reported were not included in these analyses.

(page number not for citation purposes)
A sample of 172 genotyped unrelated participants was derived from the COGA pedigrees in order to compute LD statistics. To form the sample of unrelateds, a hierarchical approach was used. In pedigrees in which founders were genotyped, all available founders were chosen for analysis. Where founders were not genotyped, the proband (index case) was used if a proband was declared. Otherwise, the individual with the smallest COGA identification number within each family who had been genotyped was chosen. SNPs with minor alleles <0.05 were removed from analysis. The statistic r 2 = D 2 /(p 1 p 2 q 1 q 2 ) was calculated for each adjacent pair of markers across the genome.
Here, D is defined as D = x 11 -p 1 q 1 (x 11 is the frequency of haplotype A 1 B 1 , and p 1 and q 1 are the frequencies of alleles A 1 and B 1 at loci A and B, respectively) [2]. In males phase is known on the heterosomal region of chromosome X. Thus, in males estimates were computed directly from the genotype data using maximum likelihood estimation. For females, we used the software Dprime (available from authors) which employs the EM (expectation maximization) algorithm, a maximum likelihood method, for haplotype frequency estimation [3]. For adjacent SNPs on all chromosomes, we explored the fit of the r 2 results to the exponential, Weibull, Gamma and log- ββ ββ ββ ββ ββ normal distributions, adjusting for distance (cM). Among these distributions, the Weibull distribution provided the best fit and will be reported. Inferences based on the other distributions are comparable. Note that these models do not account for the fact that each individual contributes multiple observations to the analysis (i.e., once for each marker pair.) Chi-squared tests were used to compare the sex-specific Weibull distributions of r 2 1) within the X chromosome, 2) between the pter pseudoautosomal region (PAR1) on X and the heterosomal region on X, 3) between PAR1 and the autosomes, and 4) between chromosome X and the autosomes.

Autosomes
Sex-specific LD analyses across the genome revealed the anticipated rapid decay of LD with increased distance (cM) for both males and females ( -coefficients for distance are = -0.1005, = -0.6457, respectively). Very few adjacent SNP markers exhibited significant LD (r 2 > 0.7), generally supporting the use of these markers in linkage analyses. r 2 measures for adjacent SNPs showed that the LD distribution for males and females on chromosome X were comparable (males were coded as 0 and females coded as 1, p = 0.9573) (Table 1A).

Chromosome X
Only females were used to compare the pseudoautosomal region PAR1 to the heterosomal region of chromosome X (chromosome Y data not available). The first 7 SNPs lie in the PAR1 region, while 265 SNPs lie in the heterosomal region of X (46-130 cM). PAR2 was not analyzed due to lack of SNPs in that region. The 7 PAR1 SNPs exhibited very low LD (r 2 < 0.1). The chi-squared test showed a significant difference in LD for the two regions on X (heterosomal markers were coded as 0 and PAR1 coded as 1, p = 0.0045) (Table 1B and Figure 1A).
Although the autosomes and chromosome X both showed a similar decline in sex-specific LD as intermarker distance increased, the rate of decay proved to be quite different. LD between adjacent SNPs on chromosome X tended to be concentrated more toward the extremes (either very strong LD (r 2 > 0.9) or very weak LD (r 2 < 0.1)), while LD in the autosomes decreased gradually with distance (Tables 1C, and Table 1D, and Figure 1B and Figure 1C). Chi-squared tests confirmed this difference in both males and females (autosomes were coded as 0 and chromosome X coded as 1, p < 0.0001 for both comparisons). The chi-squared test also revealed a difference in LD between PAR1 and the autosomes (p = 0.0019) for females (Table 1E, Figure 1D).

Discussion
The primary results from this study are the observed variations in recombination rates between the autosomes and chromosome X and between PAR1 and the heterosomal regions of chromosome X. These results have obvious analysis implications. Thorough fine mapping of regions with high recombination rates requires significantly more SNPs to delineate fully the LD block structure and capture the haplotypic variation. Recombination of chromosomes tends to increase genetic diversity. More specifically, recombination is not the source of new genes but the source of new combinations of genes. The evolutionary significance as to why PAR1 and the heterosomal region on chromosome X vary in the extent of LD compared to each other and to the autosomes is not immediately obvious. That is, it is not clear whether the observed patterns in these genomic regions are the results of current or past selective pressure on specific genes, or represent an interesting characteristic of the genome void of meaningful evolutionary cause or significance. The resolution of this question likely must wait until the functions of the individual genes in these regions are known and placed in the context of modern evolutionary theory.
A significant limitation of this investigation is that the SNPs are not dense enough to appropriately explore differences in block structures, particularly in the PAR1 region. A more dense set of markers across the genome would provide additional information about LD structures as well as potential "recombinational hot-spots". Furthermore, if sufficient data from pseudoautosomal region 2 were available, a comparison of LD in PAR1 and PAR2 would be of interest. With the advent of genomewide association analysis using 500 K chips, these questions can be explored in detail.

Conclusion
A comparison of sex-specific LD across chromosome X revealed a comparable rate of decay in LD with increased distance for males and females. When compared with the autosomes, however, the rate decay of LD on chromosome X was significantly more rapid. In addition, we observed much lower LD on pseudoautosomal region 1 (PAR1) of chromosome X compared with the heterosomal region of chromosome X. This observation of low LD across PAR1 is consistent with previous results that suggest higher recombination rates on PAR1 when compared with the heterosomal region [4].

Abbreviations
COGA: Collaborative Study on the Genetics of Alcoholism EM: Expectation maximization LD: Linkage disequilibrium SNP: Single-nucleotide polymorphism