increase in the fraction of G-to-T and C-to-A candidate variants for all callers, with Mutect2 (17)
showing a significant (P < 0.05) difference in distributions (fig. S8). Mutect2 variant profiles displayed a 9% average increase in the fraction of
variants being either G-to-T or C-to-A in data
sets predicted to be heavily damaged compared
with weak or no damage data sets, suggesting
that large numbers of variants called with high
confidence are derived from artifactual damage. This result is predicted to affect the accurate identification of individual loci and may
lead to incorrect diagnostic conclusions in those
To distinguish true from artifactual somatic
variants, standard strategies include increasing
sequencing coverage, setting stringent variant
frequency thresholds, and applying postprocessing
computational filters to derive high-confidence
variant calls. These stringent criteria can minimize the effect of damage detected genome-wide,
as seen for the TCGA variant profiles. Applying
stringent criteria, however, does not guarantee
the elimination of all errors from damage and,
more important, can increase the false-negative
rate. For example, variant-calling algorithms can
include strand bias to eliminate artifacts, but
when faced with limited numbers of variant reads
there is a reasonable chance that all evidence
reads derived from the same strand orientation,
even for genuine variants. Thus, filtering steps
are de facto inferior substitutes to preventing
mutagenic DNA damage from occurring in the
In this work, DNA repair has been used to
specifically eliminate oxidative damage in our
experimental setup for the purpose of evaluating
the GIV score and understanding how damage
affects variant calling. Additional work will be
required to properly identify conditions that will
be effective in eliminating damage from TCGA,
1000 Genomes Project samples, and sequencing
samples in general.
REFERENCES AND NOTES
1. I. Martincorena, P. J. Campbell, Science 349, 1483–1489
2. L. B. Alexandrov et al., Nat. Genet. 47, 1402–1407
3. M. R. Stratton, Science 331, 1553–1558 (2011).
4. D. A. Landau et al., Cell 152, 714–726 (2013).
5. K. Anderson et al., Nature 469, 356–361 (2011).
6. B. Ewing, P. Green, Genome Res. 8, 186–194 (1998).
7. H. Do, A. Dobrovic, Clin. Chem. 61, 64–71 (2015).
8. A. W. Briggs et al., Nucleic Acids Res. 38, e87 (2010).
9. A. M. Newman et al., Nat. Biotechnol. 34, 547–555
10. M. Costello et al., Nucleic Acids Res. 41, e67 (2013).
11. Materials and methods are available as supplementary
12. K. C. Cheng, D. S. Cahill, H. Kasai, S. Nishimura, L. A. Loeb,
J. Biol. Chem. 267, 166–172 (1992).
13. J. Tchou, A. P. Grollman, Mutat. Res. 299, 277–287
14. A. Auton et al., Nature 526, 68–74 (2015).
15. J. Bendl et al., PLOS Comput. Biol. 12, e1004962 (2016).
16. D. C. Koboldt et al., Genome Res. 22, 568–576 (2012).
17. K. Cibulskis et al., Nat. Biotechnol. 31, 213–219 (2013).
We would like to thank S. Kaiser, A. Messelaar, T. Vincze, and
C. Lin for information technology and compliance with the TCGA
hosting guidelines; L. Mazzola, J. Bybee, and D. Rivizzigno for
sequencing; and R. Roberts, W. Jack, T. Carlow, A. Gardner, H. Runz,
E. Dimalanta, and S. Russello for critical comments. The results
shown here are in part based on data generated by the 1000
Genomes Projects and the TCGA Research Network: https://
cancergenome.nih.gov. This research was supported by New
England Biolabs Inc. L.E., T.C.E., L.C., and P.L. are inventors on
U.S. provisional serial number 62/376,165, submitted by New
England Biolabs Inc., which covers improved sequence accuracy
determination of a nucleic acid sample. Sequencing data have been
deposited at the European Nucleotide Archive under accession
number PRJEB16681. The algorithm for the GIV score is available
at https://github.com/Ettwiller/Damage-estimator. Disclosure
declaration: All authors are employees of New England Biolabs Inc.
Ethics statement: DNA samples in this study were collected
under the BioChain Institute Inc. Institutional Review Board (IRB).
This IRB is registered with the Office for Human Research
Protections (OHRP), registration number IRB00008283, and has
been issued with Federal Wide Assurance (FWA), FWA00017355,
for the Protection of Human Subjects for Institutions within the
United States by OHRP of the U.S. Department of Health and
Human Services. The DNA was collected using an informed
consent form approved by Biochain’s IRB. The nature and possible
consequences of the studies were explained in the informed
Materials and Methods
Supplementary Text 1 to 5
Figs. S1 to S8
Tables S1 and S2
23 August 2016; accepted 23 January 2017
Vitamin B3 modulates mitochondrial
vulnerability and prevents
glaucoma in aged mice
Pete A. Williams,1 Jeffrey M. Harder,1 Nicole E. Foxworth,1 Kelly E. Cochran,1
Vivek M. Philip,1 Vittorio Porciatti,2 Oliver Smithies,3 Simon W. M. John1,4,5*
Glaucomas are neurodegenerative diseases that cause vision loss, especially in the elderly.
The mechanisms initiating glaucoma and driving neuronal vulnerability during normal
aging are unknown. Studying glaucoma-prone mice, we show that mitochondrial
abnormalities are an early driver of neuronal dysfunction, occurring before detectable
degeneration. Retinal levels of nicotinamide adenine dinucleotide (NAD+, a key molecule in
energy and redox metabolism) decrease with age and render aging neurons vulnerable to
disease-related insults. Oral administration of the NAD+ precursor nicotinamide (vitamin B3),
and/or gene therapy (driving expression of Nmnat1, a key NAD+-producing enzyme), was
protective both prophylactically and as an intervention. At the highest dose tested, 93% of
eyes did not develop glaucoma. This supports therapeutic use of vitamin B3 in glaucoma and
potentially other age-related neurodegenerations.
Glaucoma is a group of complex, multi- factorial diseases characterized by the progressive dysfunction and loss of ret- inal ganglion cells (RGCs), leading to vision loss. Glaucoma is one of the most common
neurodegenerative diseases worldwide, affecting
more than 70 million people (1). High intraocular
pressure (IOP) and increasing age are important
risk factors for glaucoma (2, 3). However, specific
mechanisms rendering RGCs more vulnerable to
damage with age are unknown. Here, we address
how increasing age and high IOP interact to
drive neurodegeneration using DBA/2J (D2) mice,
a widely used model of chronic, age-related, inherited glaucoma (4).
We used RNA-sequencing (RNA-seq) to elucidate age and IOP-dependent molecular changes
within RGCs that precede glaucomatous neurodegeneration. We analyzed RGCs of 9-month-old
D2 mice [in a stage termed early glaucoma—high
IOP and molecular changes but lacking neurodegeneration (2)]; 4-month-old D2 mice (in a
stage preceding high IOP); and age-, sex-, and
strain-matched D2-Gpnmb+ controls [which do not
develop high IOP or glaucoma (4)] (Fig. 1). RGCs
were isolated (fig. S1), and their RNA was sequenced at a depth of 35 million reads per sample. Unsupervised hierarchical clustering (HC)
allowed molecular definition of early glaucoma
stages among samples that were still morphologically indistinguishable from age-matched D2-
Gpnmb+ or young controls. HC identified four
distinct groups of 9-month-old D2 samples (groups
1 to 4). Group 1 clustered with all of the control
samples and represents D2 RGCs with no detectable glaucoma at a molecular level. Although
groups 2 to 4 were all early stages, increasing
group number reflects greater glaucoma progression at a transcriptomic level (Fig. 1A and fig. S2,
756 17 FEBRUARY 2017 • VOL 355 ISSUE 6326
1The Jackson Laboratory, Bar Harbor, ME 04609, USA.
2Bascom Palmer Eye Institute, University of Miami Miller
School of Medicine, Miami, FL 33136, USA. 3Department of
Pathology and Laboratory Medicine, University of North
Carolina, Chapel Hill, NC 27599, USA. 4Department of
Ophthalmology, Tufts University of Medicine, Boston, MA
02111, USA. 5The Howard Hughes Medical Institute, Bar
Harbor, ME 04609, USA.
*Corresponding author. Email: email@example.com