15 Jul 2015

Our DNA makes us who we are, yet our genetic code is not the same from cell to cell. Errors during replication and cell division can cause sequences to shift or pieces to expand or get lost. Several studies show that our brains in particular are patchworks, with DNA content varying widely from one neuron to the next. Does this variation have consequences for normal function and disease? Until recently, this question was too difficult to address, but now, advances in genetic techniques allow researchers to spot small DNA variants within a given person. Though it is early days for this research, recent reports have made a case that somatic mosaic mutations may cause some instances of neurodevelopmental disease, such as epilepsy or malformation of the brain. The picture for neurodegeneration is less clear, but a handful of papers are beginning to hint that mosaicism might play a role here, too. In one recent study, prefrontal cortical neurons from sporadic Alzheimer’s patients contained an average of two extra copies of the APP gene. Inherited APP duplications cause early onset AD, hinting that these somatic duplications might kick off amyloid pathology.

While these early findings remain to be replicated and extended to larger studies, some researchers in the field already foresee big implications if somatic mutation turn out to be a robust factor in brain disorders. “It would be a complete transformation of how we think about disease,” Christos Proukakis at University College London told Alzforum. Under such a scenario, neurodegeneration might be set off by a few rogue neurons that went bad either during prenatal development or late in life under the pressures of aging, rather than being due to inherited factors. Such a model could suggest that familial AD mechanisms such as Aβ overproduction might play a greater role in sporadic disease than has been thought, or focus attention on ways to suppress aberrant DNA replication in neurons, researchers said. 

From Population to Person
Genetic studies to date have mostly looked for inherited mutations by focusing on large populations or family lineages. In the case of Alzheimer’s disease, such studies have turned up some 30 genes linked to the condition, which together account for about 30 percent of the risk (see Apr 2015 conference news). The situation is similar for other neurodegenerative diseases. Where does the missing genetic risk lurk? Some researchers speculate that somatic mosaicism could account for a portion of it.

Several studies indicate that mosaicism is common even in the healthy brain. In 2001, Jerold Chun at the Scripps Research Institute, La Jolla, California, reported that mouse neuronal precursor cells frequently make chromosomal sorting errors during mitosis, resulting in abnormal numbers of chromosomes in about one-third of cortical neurons (see Rehen et al., 2001). This group extended its work to the postmortem human brain, finding aberrant numbers of chromosome 21 in about 4 percent of neurons, much higher than the 0.6 percent rate in human lymphocytes (see Rehen et al., 2005). In newer work, Chun and colleagues used cell sorting of nuclei labeled with a DNA-binding dye to measure variations in the total amount of genetic material between cells from healthy controls. While lymphocytes and cerebellar neurons appeared fairly uniform, frontal cortex neurons contained an average of 250 megabases of extra DNA, or an increase of about 8 percent of the total (see Westra et al., 2010). Likewise, Rusty Gage at the Salk Institute in La Jolla recently reported that up to 40 percent of frontal cortical neurons from healthy adult donors contained at least one large insertion or deletion (see Nov 2013 news).

“The normal brain contains cells with distinct genomes,” Chun concluded. Notably, the amount of variation differs among regions of the brain and between individuals.

Could this variation contribute to disease? For some neurodevelopmental disorders, the answer appears to be yes. Christopher Walsh at Boston Children’s Hospital reported that about a third of mutations for brain malformations identified in a small cohort had a somatic mosaic origin (see Gleeson et al., 2000; Jamuar et al., 2014). Another report pinpointed a mosaic brain mutation in the mTOR gene as a cause of epilepsy due to cortical dysplasia, an abnormal growth, in a single patient (see Leventer et al., 2015). Other studies have analyzed neurons from schizophrenic brains and found abnormal levels of chromosome number changes or rearrangements, which preferentially affected genes known to be involved in the disorder or in synaptic function (see Yurov et al., 2001; Bundo et al., 2014). 

Among neurodegenerative diseases, those caused by expansions of a repeating triplet in the DNA are perhaps the likeliest to be affected by mosaicism. Such repeats are known to be unstable and prone to amplify in particular cells. For example, the Huntington’s CAG repeats expand preferentially in the striatum, the region most affected by the disease (see Goula et al., 2012). Mosaic repeat expansions have been reported in ataxias and Fragile X disorders, as well as frontotemporal dementias and amyotrophic lateral sclerosis due to the C9ORF72 hexanucleotide repeat (Cancel et al., 1998; Lokanga et al., 2013; Beck et al., 2013).

Evidence for mosaicism in other neurodegenerative diseases remains scant. Up to three or four years ago, most researchers in the field had little interest in pursuing this question. They thought tracking down mosaicism would be prohibitively expensive or difficult, Proukakis told Alzforum. Now, interest in this area is picking up, with several reports appearing in the last two years. This is largely due to faster, cheaper genetic methods that make the question tractable.

New Genetics Techniques Open Door to Mosaic Studies
The best methods for finding mosaicisms vary depending on the type of mutation under investigation, geneticists said. They enumerate five general ways in which the genetic code can morph: point mutations, changes in gene copy number, insertions or deletions of chunks of code, chromosome gain or loss (aneuploidy), and movements of retrotransposons (sometimes called “jumping genes”). Some of these changes are easier to detect than others, and they may occur at different points across the lifespan.

Initially, researchers most easily found changes in chromosome number, since they could light up these bodies with fluorescent in situ hybridization (FISH). An early Alzheimer’s study using this method reported that a significant fraction of hippocampal pyramidal and basal forebrain neurons in the AD brain are tetraploid, implying they re-entered the cell cycle late in life and replicated DNA (see Yang et al., 2001). 

Changes in DNA sequence are harder to discern, since a point mutation present in only a few cells may be indistinguishable from noise in the experiment. Nonetheless, in 2004, researchers at University College London identified a mosaic point mutation in the presenilin 1 gene of a woman with sporadic early onset dementia. The mutation occurred in about 14 percent of cortical cells, at about the limit of detection for the classical Sanger sequencing they used (see Beck et al., 2004). Single nucleotide changes likely happen during early embryonic development and thus are inherited by whole lineages of cells, allowing them to be present in large cell populations.

Proukakis recently used a somewhat more sensitive method to hunt for mosaic mutations in the α-synuclein gene of 28 people with Parkinson’s (see Proukakis et al., 2013). He employed high-resolution melting curve analysis, which measures how quickly double-stranded DNA dissociates at high temperatures. Since this is a function of sequence, alterations will shift the melting curve. This method can detect a base change present in 5 to 10 percent of a sample, Proukakis estimates. He found no somatic mutations in this pilot study. He plans to repeat the experiment using modern deep-sequencing techniques, which greatly enhance sensitivity.

In the April 28 Alzheimer’s & Dementia, researchers led by Bart De Strooper at KU Leuven, Belgium, presented an analysis of just how sensitive this method can be. In deep sequencing, instead of “reading” each genetic sequence in the genome 30 to 40 times, researchers add multiple primers for a few regions of interest and sequence only those stretches from 2,000 to 3,000 times. “The whole throughput of the machine focuses on just the genes of interest,” first author Carlo Frigerio told Alzforum. By mixing a known amount of mutant DNA with wild-type, the researchers found they could distinguish a single nucleotide change present in as little as 1 percent of the sample. In addition, in the last two or three years the cost of the procedure has halved, while the capacity of the machines has increased, Frigerio said.

With this method, the Belgian group sequenced the familial Alzheimer’s genes APP, PS1, and PS2, as well as the gene for tau, in postmortem entorhinal cortex samples from a population of 72 Alzheimer’s patients and 58 controls. They found three mosaic mutations, each present at about 1 percent allele frequency. Two were previously unidentified mutations in the tau gene MAPT that were predicted to disrupt protein function; one of these was near a known pathogenic mutation, heightening the probability that it too was pathogenic. Both MAPT mosaicisms occurred in sporadic AD patients. The third mosaicism was a known pathogenic PS2 mutation (S130L), which cropped up in a person who died at age 90 with normal cognition but had mild AD pathology. In future work, the researchers plan to analyze more patients and additional brain regions and genes, to try to map out just how common such mosaic mutations may be in AD.

The Belgian group also analyzed copy number variants (CNVs), which present additional technical challenges. Because apparent differences in copy number could be due to uneven amplification or sequencing efficiency, these mutations are hard to distinguish from noise. De Strooper and colleagues calculated they could detect CNVs present in as few as 10 percent of cells, but they did not find any in this study. However, like aneuploidies, CNVs may arise later in development and thus be present in only a handful of brain cells.

One way to improve sensitivity is to analyze DNA in single cells. Chun and colleagues reported on such an approach in the Feb 4 Elife. They concentrated on APP copy number, since duplications of this gene are known to trigger Alzheimer’s pathology in familial cases and Down’s syndrome. They compared DNA from about 150 single neurons taken from the prefrontal cortices and cerebella of three AD patients and three controls using quantitative PCR. Cortical neurons from AD patients, but not control or cerebellar cells, contained an average of four copies of APP, with some cells ranging up to 12. The researchers validated the results in separate experiments using FISH, which gave similar results. The results hint at a role for APP expansions in sporadic AD.

Others have used single-cell methods to examine the number of neuronal retrotranspositions, where a section of DNA copies itself into a new location. By some estimates, mobile retrotransposons make up as much as a third of human DNA, suggesting that rearrangements could be a common occurrence (see Cordaux and Batzer, 2009). Researchers in Australia recently estimated that about 14 such somatic transpositions occur in each hippocampal neuron (see Upton et al., 2015). However, other studies place this number at less than one per neuron (see Coufal et al., 2009; Evrony et al., 2012). Differences in methodology likely account for these discrepancies, Proukakis noted. More studies are needed to determine how likely this type of genetic change is to trigger disease.

Researchers in the field believe forthcoming genetic methods will provide greater sensitivity to find small variations. Proukakis plans to study CNVs in α-synuclein and other genes in Parkinson’s brains using droplet digital PCR, which reduces noise and allows detection of small changes in copy number (see Shiroguchi et al., 2012; Mazaika and Homsy, 2014; and Miotke et al., 2014). As another option, FISH in single cells may improve sensitivity because it involves no amplification, allowing researchers to simply count the number of dots where the probe binds, Proukakis said.

Another method under development is nanopore sequencing, which detects changes in sequence by pushing strands of DNA through protein nanopores and reading the resulting interference in the electric field (see Laszlo et al., 2014). This method promises to be high-throughput and allow researchers to sequence strands thousands of bases long. “That will be revolutionary,” Frigerio predicted.

Do Mosaic Mutations Trigger Disease?
To put in the effort and funds needed to find mosaic mutations, researchers have to believe they are important contributors to disease. Currently, the field is divided on this question. Annemieke Rozemuller, a neuropathologist at VU University Medical Center, Amsterdam, noted that many clinicians and geneticists remain skeptical that mosaicisms are worth the hunt. With only a handful of papers to date demonstrating their existence, and even fewer studies linking them to disease, some researchers doubt these mutations will be a major factor. For her part, Rozemuller believes that somatic mutations might explain some of the early onset sporadic cases of dementia. “I think it is important to do these studies,” she wrote to Alzforum (see full comment below).

Some researchers agree. “We think there is a strong possibility that somatic mutations can play a role in neurodegenerative disease,” Gage and collaborator Michael McConnell at the University of Virginia School of Medicine, Charlottesville, wrote to Alzforum. They suggested it will be most fruitful to look for them in pure neuronal populations or individual cells. Frigerio noted that at the EMBL symposium on Mechanisms of Neurodegeneration, held June 14-17 in Heidelberg, Germany, Kári Stefánsson of deCODE genetics said the study of somatic mutations and their role in brain function are the wave of the future.

Others note that simply finding mosaicisms will not be enough. “[The Frigerio et al.] study shows that mosaic variants do exist in AD-related genes. However it remains challenging to understand the role of these changes in Alzheimer’s disease, particularly since the pathogenicity of the variants found is not clear,” Rita Guerreiro at University College London wrote to Alzforum (see full comment below).

To link mutations to pathology, researchers will need to express them in cell cultures and animal models and understand what they do, Frigerio said. Because abnormally folded proteins can spread through the brain, a mutation that triggers protein aggregation in just a few cells may be sufficient to start a slow snowball effect toward neurodegeneration, as seen in some animal models (see Feb 2012 news). Proukakis speculated that mosaic mutations may have an additive effect, and the accumulation of several mutations in numerous neurons might trigger disease. At present, no one can predict how many cases of sporadic disease could be due to such mechanisms, but the field appears poised to start finding out.—Madolyn Bowman Rogers

Comments

1. • Rita Guerreiro
     Van Andel Institute
   • Posted: 15 Jul 2015

   This is a very interesting area of research, as has been recently shown by Acuna-Hidalgo R. et al. when reporting that post-zygotic point mutations are an under-recognized source of de novo genomic variation (Acuna-Hidalgo et al., 2015). Mosaicism is a known event but difficult to study, especially when the proportion of cells with a different genotype is low.

   The Frigerio et al. paper uses a combined approach to start looking at this issue in Alzheimer’s disease. The authors went to great lengths to validate and optimize their pipeline, which seems to be very efficient but somehow not very specific (with a low rate of variants being validated).

   Technically my only concern is associated with the study of somatic CNVs given the reported systematic biases in loci copy number originating from DNA isolation methods (van Heesch et al., 2013). This study gives us the technical possibility to detect mosaic variants in the brain and shows that these variants do exist in AD-related genes. However, it remains challenging to understand the role of these changes in Alzheimer’s disease, particularly since the pathogenicity of the variants found is not clear.

   References:

   Acuna-Hidalgo R, Bo T, Kwint MP, van de Vorst M, Pinelli M, Veltman JA, Hoischen A, Vissers LE, Gilissen C. Post-zygotic Point Mutations Are an Underrecognized Source of De Novo Genomic Variation. Am J Hum Genet. 2015 Jul 2;97(1):67-74. Epub 2015 Jun 6 PubMed.

   van Heesch S, Mokry M, Boskova V, Junker W, Mehon R, Toonen P, de Bruijn E, Shull JD, Aitman TJ, Cuppen E, Guryev V. Systematic biases in DNA copy number originate from isolation procedures. Genome Biol. 2013 Apr 24;14(4):R33. PubMed.

   View all comments by Rita Guerreiro
2. • Annemieke J.M. Rozemuller
     VU University Medical Center
   • Posted: 15 Jul 2015

   I have tried several times without success to raise interest and funding for a study like that published by Bart De Strooper. Clinicians and geneticists at VUMC thought mosaicism was not present. I am convinced that mosaicism would explain at least some non-familial early cases. We have many presenile AD cases without any known mutations in the presenilins or APP as per blood test. I still would like to do this research and confirm this important study by De Strooper's group with our material from the Netherlands Brain Bank. We have more than 1,000 AD brains, both frozen and paraffin-embedded tissue, serum, sometimes skin, and in many cases DNA from blood.

   I think it is important to do these studies. We can select the most pure cases and get this study done; however, we need financing and interest by geneticists.

   View all comments by Annemieke J.M. Rozemuller
3. • Christos Proukakis
     UCL
   • Posted: 07 Jul 2017

   In response to the comment by Rita Guerrero above, we have now demonstrated a similar bias in DNA isolation from human brain, which could lead to a false impression of CNV mosaicism, see Nacheva et al., 2017. 

   References:

   Nacheva E, Mokretar K, Soenmez A, Pittman AM, Grace C, Valli R, Ejaz A, Vattathil S, Maserati E, Houlden H, Taanman JW, Schapira AH, Proukakis C. DNA isolation protocol effects on nuclear DNA analysis by microarrays, droplet digital PCR, and whole genome sequencing, and on mitochondrial DNA copy number estimation. PLoS One. 2017;12(7):e0180467. Epub 2017 Jul 6 PubMed.

   View all comments by Christos Proukakis

Make a Comment

To make a comment you must login or register.

References

News Citations

1. New Genetics Frontiers: Finding Modifiers, Making Sense of Pathways 10 Apr 2015
2. Brain a Jumble of Genetically Distinct Neurons? 1 Nov 2013
3. Mice Tell Tale of Tau Transmission, Alzheimer’s Progression 3 Feb 2012

Paper Citations

1. Rehen SK, McConnell MJ, Kaushal D, Kingsbury MA, Yang AH, Chun J. Chromosomal variation in neurons of the developing and adult mammalian nervous system. Proc Natl Acad Sci U S A. 2001 Nov 6;98(23):13361-6. PubMed.
2. Rehen SK, Yung YC, McCreight MP, Kaushal D, Yang AH, Almeida BS, Kingsbury MA, Cabral KM, McConnell MJ, Anliker B, Fontanoz M, Chun J. Constitutional aneuploidy in the normal human brain. J Neurosci. 2005 Mar 2;25(9):2176-80. PubMed.
3. Westra JW, Rivera RR, Bushman DM, Yung YC, Peterson SE, Barral S, Chun J. Neuronal DNA content variation (DCV) with regional and individual differences in the human brain. J Comp Neurol. 2010 Oct 1;518(19):3981-4000. PubMed.
4. Gleeson JG, Minnerath S, Kuzniecky RI, Dobyns WB, Young ID, Ross ME, Walsh CA. Somatic and germline mosaic mutations in the doublecortin gene are associated with variable phenotypes. Am J Hum Genet. 2000 Sep;67(3):574-81. Epub 2000 Jul 27 PubMed.
5. Jamuar SS, Lam AT, Kircher M, D'Gama AM, Wang J, Barry BJ, Zhang X, Hill RS, Partlow JN, Rozzo A, Servattalab S, Mehta BK, Topcu M, Amrom D, Andermann E, Dan B, Parrini E, Guerrini R, Scheffer IE, Berkovic SF, Leventer RJ, Shen Y, Wu BL, Barkovich AJ, Sahin M, Chang BS, Bamshad M, Nickerson DA, Shendure J, Poduri A, Yu TW, Walsh CA. Somatic mutations in cerebral cortical malformations. N Engl J Med. 2014 Aug 21;371(8):733-43. PubMed.
6. Leventer RJ, Scerri T, Marsh AP, Pope K, Gillies G, Maixner W, MacGregor D, Harvey AS, Delatycki MB, Amor DJ, Crino P, Bahlo M, Lockhart PJ. Hemispheric cortical dysplasia secondary to a mosaic somatic mutation in MTOR. Neurology. 2015 May 19;84(20):2029-32. Epub 2015 Apr 15 PubMed.
7. Yurov YB, Vostrikov VM, Vorsanova SG, Monakhov VV, Iourov IY. Multicolor fluorescent in situ hybridization on post-mortem brain in schizophrenia as an approach for identification of low-level chromosomal aneuploidy in neuropsychiatric diseases. Brain Dev. 2001 Dec;23 Suppl 1:S186-90. PubMed.
8. Bundo M, Toyoshima M, Okada Y, Akamatsu W, Ueda J, Nemoto-Miyauchi T, Sunaga F, Toritsuka M, Ikawa D, Kakita A, Kato M, Kasai K, Kishimoto T, Nawa H, Okano H, Yoshikawa T, Kato T, Iwamoto K. Increased l1 retrotransposition in the neuronal genome in schizophrenia. Neuron. 2014 Jan 22;81(2):306-13. Epub 2014 Jan 2 PubMed.
9. Goula AV, Stys A, Chan JP, Trottier Y, Festenstein R, Merienne K. Transcription elongation and tissue-specific somatic CAG instability. PLoS Genet. 2012;8(11):e1003051. Epub 2012 Nov 29 PubMed.
10. Cancel G, Gourfinkel-An I, Stevanin G, Didierjean O, Abbas N, Hirsch E, Agid Y, Brice A. Somatic mosaicism of the CAG repeat expansion in spinocerebellar ataxia type 3/Machado-Joseph disease. Hum Mutat. 1998;11(1):23-7. PubMed.
11. Lokanga RA, Entezam A, Kumari D, Yudkin D, Qin M, Smith CB, Usdin K. Somatic expansion in mouse and human carriers of fragile X premutation alleles. Hum Mutat. 2013 Jan;34(1):157-66. Epub 2012 Oct 4 PubMed.
12. Beck J, Poulter M, Hensman D, Rohrer JD, Mahoney CJ, Adamson G, Campbell T, Uphill J, Borg A, Fratta P, Orrell RW, Malaspina A, Rowe J, Brown J, Hodges J, Sidle K, Polke JM, Houlden H, Schott JM, Fox NC, Rossor MN, Tabrizi SJ, Isaacs AM, Hardy J, Warren JD, Collinge J, Mead S. Large C9orf72 hexanucleotide repeat expansions are seen in multiple neurodegenerative syndromes and are more frequent than expected in the UK population. Am J Hum Genet. 2013 Mar 7;92(3):345-53. Epub 2013 Feb 21 PubMed.
13. Yang Y, Geldmacher DS, Herrup K. DNA replication precedes neuronal cell death in Alzheimer's disease. J Neurosci. 2001 Apr 15;21(8):2661-8. PubMed.
14. Beck JA, Poulter M, Campbell TA, Uphill JB, Adamson G, Geddes JF, Revesz T, Davis MB, Wood NW, Collinge J, Tabrizi SJ. Somatic and germline mosaicism in sporadic early-onset Alzheimer's disease. Hum Mol Genet. 2004 Jun 15;13(12):1219-24. Epub 2004 Apr 28 PubMed.
15. Proukakis C, Houlden H, Schapira AH. Somatic alpha-synuclein mutations in Parkinson's disease: Hypothesis and preliminary data. Mov Disord. 2013 Jun;28(6):705-12. PubMed.
16. Cordaux R, Batzer MA. The impact of retrotransposons on human genome evolution. Nat Rev Genet. 2009 Oct;10(10):691-703. PubMed.
17. Upton KR, Gerhardt DJ, Jesuadian JS, Richardson SR, Sánchez-Luque FJ, Bodea GO, Ewing AD, Salvador-Palomeque C, van der Knaap MS, Brennan PM, Vanderver A, Faulkner GJ. Ubiquitous L1 mosaicism in hippocampal neurons. Cell. 2015 Apr 9;161(2):228-39. PubMed.
18. Coufal NG, Garcia-Perez JL, Peng GE, Yeo GW, Mu Y, Lovci MT, Morell M, O'Shea KS, Moran JV, Gage FH. L1 retrotransposition in human neural progenitor cells. Nature. 2009 Aug 27;460(7259):1127-31. Epub 2009 Aug 5 PubMed.
19. Evrony GD, Cai X, Lee E, Hills LB, Elhosary PC, Lehmann HS, Parker JJ, Atabay KD, Gilmore EC, Poduri A, Park PJ, Walsh CA. Single-neuron sequencing analysis of L1 retrotransposition and somatic mutation in the human brain. Cell. 2012 Oct 26;151(3):483-96. PubMed.
20. Shiroguchi K, Jia TZ, Sims PA, Xie XS. Digital RNA sequencing minimizes sequence-dependent bias and amplification noise with optimized single-molecule barcodes. Proc Natl Acad Sci U S A. 2012 Jan 24;109(4):1347-52. Epub 2012 Jan 9 PubMed.
21. Mazaika E, Homsy J. Digital Droplet PCR: CNV Analysis and Other Applications. Curr Protoc Hum Genet. 2014 Jul 14;82:7.24.1-7.24.13. PubMed.
22. Miotke L, Lau BT, Rumma RT, Ji HP. High sensitivity detection and quantitation of DNA copy number and single nucleotide variants with single color droplet digital PCR. Anal Chem. 2014 Mar 4;86(5):2618-24. Epub 2014 Feb 12 PubMed.
23. Laszlo AH, Derrington IM, Ross BC, Brinkerhoff H, Adey A, Nova IC, Craig JM, Langford KW, Samson JM, Daza R, Doering K, Shendure J, Gundlach JH. Decoding long nanopore sequencing reads of natural DNA. Nat Biotechnol. 2014 Aug;32(8):829-33. Epub 2014 Jun 25 PubMed.

Other Citations

1. S130L

Further Reading

News

1. Next-Generation Sequencing: Boldly Going Where No Geneticist... 5 Nov 2010
2. Genetics Project Update: Over 1,000 Genomes and Counting 6 Nov 2012
3. New Method to Boost Single Cell Genomics 4 Jan 2013
4. Alzheimer’s Whole-Genome Data Now Available From the NIH 6 Dec 2013
5. New Genetic Almanac Should Help Scientists Understand GWAS Hits 5 Sep 2014
6. ENCODE Expands Analysis of DNA Regulation 5 Sep 2014
7. Stream of Genetics Pushes FTD Research Forward 19 Nov 2014
8. Scientists Seek, Find, and Exploit AD Risk Genes 17 Dec 2014
9. Genentech Strikes Deal with 23andMe to Study Parkinson’s Genomes 15 Jan 2015
10. 44 Parent-Child Trios Yield 40 Candidate ALS Genes 25 Mar 2015
11. Massive Icelandic Genome Analysis Offers Clues to Health and Disease, Including AD 26 Mar 2015
12. New Genetics Frontiers: Finding Modifiers, Making Sense of Pathways 10 Apr 2015