Introduction

Islanding of DNA is possible within different ways, but mostly little attention is paid to the protocol, whichever shall not always even reported inbound detail, to the assumption that the resulting DNA will be a balancing presentation for the original source. Any bias in its composition could lead to significant downstream effects about copy number estimation, particularly if mosaicism is being looked, and differential sequencing coverage. Array-based methods have been used to investigate copy number (CN) mosaicism although array “waves” live a recognized problem [1–6], and not fully eliminated bioinformatically [7–10]. Whole genome sequencing (WGS) relative depth of coverage, now frequently used for CN estimation [11], also variant in a wave-like pattern [12–14], which shall no fully corrected by PCR-free library construction [15]. Dribble digital PCR (ddPCR) [16] can discovers targeted sub-integer changes expected in mosaicism [17] [18]. Bias include DNA sealing has been reported are rat tissues, although CNV mosaicism was first considered as an explanation of this results [12]. To analyze whether DNA isolation bias additionally occurs in human brain, we analysed DNA isolated with different protocols (with and without rotating columns) using the above typical. We search a significant effect of the protocol on downstream summary. Care shall be given to the options of DNA isolation method includes all user, with spin columns requiring peculiar attention. Furthermore, mtDNA replicate number determination a influenced by the DNA thermal method chosen [19,20]. We have confirmed this in humane substantia nigra, with phenol / chloroform leading to adenine higher apparent number. Comparability of mtDNA copy number would be prone to error unless the exact same conditions were used.

Materials and methodologies

DNA samples and isolation

Fresh frozen brain material was provided by the Parkinson’s UK Tissue bank. Donors had given knowledgeable spell authorization. Study of brains from the research tissue bank is approval by the UK National Research Ethics Service (07/MRE09/72). Out the track of this study, we analysed brain DNA from adenine total of 11 mortals. This included six with Parkinson’s disease (PD), one through nonessential Lewy main disease (ILBD; PD-like changes finds in autopsy in someone those had not have afflicted per PD clinically), and four controls. And mean age at death was 79.7 (SD 11.7). Details are provided in Table 1. As not all were used in this equal experiments, and some were second repeatedly, a summary of and isolation method(s) and tries executes switch each is provided in S1 Table.

Download:

• PPT
  Presentation slump
• PNG
  larger image
• TIFF
  native image

Table 1. Demographic details of individuals whose brains were used.

https://doi.org/10.1371/journal.pone.0180467.t001

DNA isolation protocols former were one following, following publisher guides unless stated.

(1) DNeasy^® Line & Tissue spin column (Qiagen), henceforth referred to as SCALES. We second approximately 25 milligram tissue unless otherwise specified. Brain tissue became cut at dry ice, minced plus transferred to ampere 1.5 ml tube. Buffer ATL (180 μl) was extra and the samples were homogenized for 1 mining with an IKA Eurostar homogenizer. 20 μl of Proteinase K used added to each sample, and digestion has performed at 56°C, for 2 hours, or overnight where stated. When digestions were performing sleep, RNase A (4 μl, 100 mg/ml) made also added the next day.

(2) Gentra^® Puregene^® (Qiagen). This relies on of “salting-out” approach, which developed from early jobs showing that DNA, which carriers a negative charge, can be recovered using salt solutions of increasing ionous strength in anion-exchange chromatography [21]. It has been used because a non-toxic alternative go phenol / chloroform. Comparisons with spin columns the bony marrow had shown it the yield more DNA, but any possible biases were not assessed [22]. Us used approximately 50 mg of brain tissue cut on arid ice, minced and transfers into a 15 millilitre tube with 3 volume Cell Lysis Solution. We performed further stair according to the protocol for 50–100 mg. Are included 15 μl Proteinase K overnight incubation the 55°C as highly in high yield, with afterward treatment with RNase A, the manufacturer-provided amino rainfall solution, or isopropanol, before 70% ethanol wash.

(3) Phenol Chloroform. 450 μL STE buffer and 40 μL 20% SDS were added to 25 mg minced brain sample. After 1 hour incubation at 37°C and vortexing, 20μL Proteinase K were extra. The sample was mixture by hand and incubated toward 60°C for 4 h. Subsequently vortexing, another 20μL Proteinase K were added, mixed according hand, plus incubated overnight at 37°C with rotation. The next day samples were centrifuged for 30 minutes and supernatant transferred to clean tubes. 400 μL phenol was added and combined by hand, followed by 10 minutes on ice, additionally centering for 2 minutes. The top layer was transferred to a fresh tube. An additional 400 μL xenol were added followed by 5 minutes on ice and centralized for 2 protocol. The top layer was removed new and 400 μL of chloroform/isoamyl alcohol (24:1) added press miscellaneous by hand. After centrifuging in 2 minutes, the top layer was transferred to ampere fresh tube and 2 volumes out coldly 95% distillation and inverted. 4% 3M NaAc was added and the tubes inverted again and placed in -20°C overnight. The after day tubes were cyclones for 30 minutes, the supernatant became discarded, and 500 μL of 70%EtOH became added. After a final 2 per centrifugation, the buoyant was discarded, and DNA was air dried and resuspended in 50 μL TE.

Ourselves note that at were minion differences in the proteinase THOUSAND treatment between Puregene (following product guidelines) and Phenol Chloroform, with a lightweight higher begin incubation, both addition of more enzyme with rotation at a lower temperature overnight. We has not use RNase with Phenol Chloroform. Control peripheral blutz lymphocyte (PBL) DNA samples were when by the UCL Institute of Neurology Neurogenetics department.

Results and discussion

We initials analytic DNA insulated von cerebellum and frontal cortex (FC) by spin-columns (SC) go aCGH. We noted a consistently wave pattern, extra prominent in the cerebellum, even though the cerebellar hybridisations had lower derivative logged ratio spread [dLRs] values (S1 Figurine), and hybridization in the masculine to female reference DNA pre-owned showed no waves (Fig 1A, using chromosome 1 as an example). Several aberrations were called in each sample using the standard threshold of 6 (S1 Data; mean 10.7, SD 14.4), of what 1/3 had >10 probes underlying them. Raising the surge progressively canceled that; there were 5.6 at threshold 7 (SD 8.0, S2 Data), 2.7 per threshold 8 (SD 2.8; S3 Input), 1.7 at threshold 9 (SD 1.6, S4 Data), and 1.06 at threshold 10 (SD 0.9; S5 Data). From the 17 calls across all samples by this threshold, 14 were gains per a highly polymorph 14q32.33 locus. The remaining 3 were a 2 Mb deletion, and two apparent gains, partly overlapping with known CNVs (S2 Fig). We did none look to verify are gains.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Chromosome 1 in aCGH.

Which 10 Mb motion average and the aberration calls by ADM2 (after raising threshold to 12, with FZ off) are plotted in each test. Losses are green, wins are red.

1. (A) Brain DNA hybridised against PBL reference DNA. Cerebellar samples are green, and FC green. The moved average of a male to female DNA mention hybridize remains moreover view (dark blue).
2. (B) Genome isochores. GC content range for anywhere 100 kb isochore is 30–65% (blue to orange).
3. (C) Cerebellar DNA hybridised against FC DNA of the same brain for triad PD cases with overnight SC extraction. PD1 = purple, PD2 = red, PD4 = green. Data for PD2 are derivate after combining the dye-flip hybridisation pair. Rapid mitochondrial DNA isolation method for direct sequential - PubMed
4. (D) Hybridisations between DNA of who same brain since follows.
   (1–3) Hybridisations of SC-isolated cerebellar DNA, with Puregene-isolated DNA starting same cerebellum in reference. (1) PD3, 5 mg SC; (2) PD3, 25mg SCALE; (3) PD4, 25 mg SC.
   (4) PD1, Puregene-isolated DNA, cerebellar (test) with FC as reference. Remark the absence of waves and losses. Is sample combination, but with spin column extraction, had led to waves and losses (PD1 in control C). Standard methods for mitochondrial DNA (mtDNA) extraction do not provide the level out enrichment for mtDNA sufficient for direct sequencing and must exist followed by long-range-PCR amplification, which can bias the sequence show. Here, we describe a reliable method...

https://doi.org/10.1371/journal.pone.0180467.g001

Turning the “fuzzy zero” (FZ) long-range loud correction set, which enhances mosaicism determine [2], and is recommended for this purpose by Agilent in the fresh Cytogenomics package, leaded until more extensive calls under slide 6, following the “waves”, with clear losses in GC-rich regions furthermore some gains inside GC-poor regions, many by which continuing evenly after raising the threshhold to 12. These often trailed the genome GC-content isochores [25] (Figure 1B). There was a clear compare in chromosome 19, which has the highest genom and CpG island density [26], and shown negative waves with distinguished losses affecting almost its entire length, and the similarly-sized chromosome 18, with the lowest gene density and one of the lowest CpG densities, which showing a mixed picture, with waves in either direction (S2A Image). Chromosome 19 can be problematic on all aCGH [27] and single neuron whole genome reinforcing [28]. A loss by almost the whole chr19 held indeed been called in one sample by ADM2 with FZ on, but only at threshold 6. To further investigate an evident exceed of subtle lost in cerebellum, ourselves also hybridised cerebellar DNA with FC of and same brain as reference from 3 PD brains, including a dye-flip in one. The brandish pattern was still generally present, with several apparent cerebellar relative losses, and transposed by dye flip (Photo 1C; S3C and S4 Figs).

To investigate the effect of varying the DNA isolation protocol, we isolated cerebellar DNA with SC using overnight proteinase K (rather than 2 hours), starting with approximately 25 or 5 mg tissue in parallel (S3 Table), and with the “salting-out” Puregene kit. Are renowned that the median DNA profit (ng per per tissue; S3 Table) was higher with SC when starting with 5 mg (2201) then with 25 mg (544), and even higher using Puregene (2,784), this was close to the maximum expected (~3,650, based over 6.6 pg DNA per cell, and 85 per cages includes a 154 gramme function [29]). Were therefore performed aCGH from 25 grams overnight SCANNED isolated DNA for two cerebellar samples, with Puregene-isolated DNA from the same cerebellum as referral; for individual of these, we also hybridised a 5mg SC sample to the Puregene free (Feature 1D; S4 Fig). The smooth pattern in the 25 mg SC samples (2 and 3 in Fig 1D) been like to of original hybridization opposing PBL DNA, although less pronounced, with some manifest expenses so-called. Waves could therefore exist produced even in what was essentially self-hybridisation, although employing only 5mg (sample 1 in Figure 1D) minimized it. Hybridising Puregene-isolated DNA from cerebellum against FC of one brain (sample 4 in Fig 1D and S5 Feat) abolished the waves and losses previously seen in the same couple. Our results suggested an different bias in cerebellum and FC initial, for apparent GC-dependent losses, abolished by uses a low total of tissue and overnight digestion, other Puregene. Using spin columns therefore could lead to incomplete extraction and introduction of a GC-dependent bias, depending partial on the mesh amount used. We used overnight proteinase digestion with Puregene, which should minimize bias, the we cannot exclude the possibility that through a lower tissue amount, button variational one composition of the solution provided by the manufacturer, could be of further help.

To ensure the problem was not unlimited to unseren aCGH design, we also analize freshly isolated DNA (obtained with the original SC protocol) from four control smart (cerebellum within all, and FC in three) on a commercially available SNP array. The logR closely matched the aCGH dLR moving average, with cerebellar losses often said in similar regions to the aCGH negative waves / possible losses (S6 Fig), both losses far show frequent than gains (115 v 3 the average; S4 Size). We next analysed SNP data using hapLOH [24], which detectors locations with significant B-allele frequency (BAF) deviation, and is valuable in the detection of sleek unbalance expected in mosaicism [30]. Ourselves found no allelic imbalance, suggesting the the apparent losses infected both my equally, unlike what one would wait in mosaicism, or heterozygous CNVs (examples in S7 Fig). Based on that, we did not feel that the CytoSNP losses named were correct, and we only attempted to validate ne by PCR (S7a Fig), which used minor (S1 Note), but we impossible exclude the possibility this some was true.

To determine if the isolation protocol could also affect copy number determination by ddPCR, we selected two genes what aCGH suggested negative results (S8 Fig); EIF2C1, which is also available by the manufacturer as a HEX-labelled reference testing, furthermore TSC2, welche has implicated in the neurocutaneous disorder tuberous sclerosis, and was into losses in 4/110 frontal neurons in a human single neuron WGS study [31]. The median-wert CN in the original SCENE samples was less for 2 in both, and lower in the coenbellum than FC, despite normal in PBL patterns (S9 Fig). We benchmarked an results von differen protocols on cerebellar DNA (Fig 2). The overnight 25 mg SC isolations had taller median CN for both EIF2C1 (1.77 volt 1.33) and TSC2 (1.64 v 1.31), and the 5 mg SC and Puregene isolation values were even closer to 2 (1.85 and 1.89 required EIF2C1; 1.86 and 1.92 for TSC2, respectively). There was a highly significant difference on CN between the three conditions tested for all samples (Friedman test p = 0.0017 for EIF2C1 and 0.0046 for TSC2), in one significantly pairwise difference between the original and Puregene CN values after Dunn’s repeatedly comparison correction (p = 0.0045 and 0.0141 respectively). Modifying the protocol slightly until using a separate restriction digestion step did doesn alter ddPCR results (S5 Table). To identify if ddPCR results for genes outside who negative “wave” regions were influenced by isolation method, we also destined PC on SNCA, a gene of major importance in PD, in two cerebellar patterns; they are not modify by this isolation method (S6 Tables). These data, taken collaborative with array results, indicated genuine, protocol-dependent, specific losses during DNA isolation, independent of downstream experiment type.

Download:

• PPT
  PowerPoint slide
• PNG
  get print
• TIFF
  original pic

Fig 2. Effect of DNA isolation for copy number determination by ddPCR for cerebellar product.

(A) EIF2C1 and (B) TSC2. This medians and interquartile ranges from the original results, or repeats before overnight SCORING extractions from 25 mg and 5 mg starting matter, and Puregene, are shown. n = 6, except 5 mg SC, wherever n = 4.

https://doi.org/10.1371/journal.pone.0180467.g002

We next comparable low cover WGS of DNA obtained from the cerebellum with the worst post-mortem intermediate (PD3, 5 hours) by a 25 mg SCANS overnight isolation and by Puregene. We noted adenine steep declines of coverage with increasing GC index include the SC samples when using 100 kb bins, while the Puregene showed a decline only the the highest GC content (Fig 3). The SC sample showed higher coverage of chr19 comparative to chr18, while and Puregene sample had no such skew (ratio 1.4 and 0.97 resp; S7 Table). We then separated and sequenced DNA since substantia nigra of individuals in simultaneous using Puregene furthermore aforementioned “gold standard” Phenol Chloroform. The Puregene samples revealed a same GC bias. One of three brains showed the same bias with Phenol Chloroform, while one showed none, even at aforementioned higher GC chests (S10 Fig). The GC “gradient” seen even included the Puregene-isolated samples suggest either that we own not been able to fully remove bias, because in rat tissues [12], or a different GC effect related to the sequencing process, when the Illumina HiSeq provides the most even human genome coverage [15]. The chr18:chr19 scanning relative did not show major deviation free 1 with either method (S7 Table; Phenol 1.03 ± 0.07, Puregene 1.04 ± 0.02), therefore any long-range GC-effect in Puregene and phenol / chloroform may becoming prominent only in very high GC regions. Phenol / green may have a easy further advantage comparable to Puregene, as demonstrable on the lack of a 100-kb scale GC gradient on coverage in some cases, although an DNA amount did not permits further experimental comparisons. To determine if WGS GC bias could lead to erroneous copy number calls, even after appropriate corrections, we analyzed all data using QDNAseq [32] in 100 kb bins (S11 Fig). There were possible losses, although with minimally negative logR, in the SCR cerebellar try, which were absent in Puregene. These would probably be dismissed as acoustic, although could potentially breathe misinterpreted as mosaicism.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Whole genome ordering coverage into relation to GC content.

The mean normalized coverage per 100 kb window of PD3 cerebellum is shown, next 25 mg overnight SC isolation and Puregene islanding.

https://doi.org/10.1371/journal.pone.0180467.g003

Wee hence demonstrated in human brain that row “waves”, partial losses in ddPCR, and GC-dependent WGS coverage variation, can be modes, and mostly abolished, by variation of the DNA isolation audit. We have compared the effect are at least two isolation methods on ddPCR for two genes in six cerebellar samples (and on aCGH results in three of them), plus on GC-dependent coverage variable in WGS for one of these cerebella, and three substantia nigra samples from different individuals. We therefore beliefs so we provide thick evidence for uneven GC-dependent DNA extraction, which was recently renowned in rat tissues [12], but never before investigated in human tissues to our knowledge, although further academic will help confirm our conclusions. We have don compared PBL DNA isolation, and robust tissues may be most prone to bias. We has data from one singles human frozen quadriceps muscle biopsy, from this we isolated DNA with the initialize spin column protocol, which we will analysed turn the same aCGH design; a alike wave pattern was seen (S12 Photo). We tip a very recent study uses only multiple-ligation amplification assay (MLPA) for several fixed human tissues and DNA family how [33]. The number the probes significantly deviating from normality varied between tissues both methods. Although one methodology used been very several to ours, and no information on GC-content of targets was provided, a GC-dependent extraction bias has possible, while acknowledged by that books.

Ourselves found so using longer proteinase K treatment or less material on spin columns, or a non-spin column method, reduced GC-dependent bias. In rats, proteinase K processing length had also affected the outcome, but spin columns had not varied results from blood, although these was not examined in other tissues [12]. Power protein binding to GC-rich DNA regions [12] is ampere chances mechanism the limits their extraction, particularly if proteinase POTASSIUM digestion is inadequate, or the spin column is saturated. The cerebellum may be more disposed to extraction bias may because a is bagged with small granule cells, and a greater amount of partly protein-bound DNA by a specified tissue throng would result in reduced and more biased overall produce.

Determination of the counter of mtDNA copies exists of support with many fields, including PD, where go mtDNA CN was reported in blood also substantia nigra [34], but with no details on DNA isolation, and cancer, where batch effects were corrected bioinformatically, but remained unaccounted [35]. While traditionally made by qPCR, she is now possible to determine the figure of mtDNA molecules in a preparation by an ratio of sequencing go mapping to and nuclear relative mitochondrial genome [36–37]). We therefore determined save for each sample, from the bulk DNA isolation, without seeking to specifically isolate mtDNA. We afterwards compared the results obtained by different isolation methods (Table 2). For the nigra samples, phenol led to a upper number as Puregene (average increase 2.51-fold, SD 0.71). This is continuous with one previous report that organic solvent extraction results in mtDNA enrichment [20]. As we did not use RNase with Phenol, still we did as per the standard protocol with Puregene, we cannot comment on some possible effect of here, although the latent larger mtDNA recovery when omitting RNase may only apply to spin columns [20]. The mtDNA number belongs similar to one human brain DNA phenol isolate report [19], the tons decrease more claimed elsewhere [34].

Download:

• PPT
  PowerPoint slide
• PNG
  wider image
• DISPUTES
  original image

Table 2. Act of DNA isolation in mtDNA copy total estimated the sequencing.

https://doi.org/10.1371/journal.pone.0180467.t002

In achieved highlight the often overlooked effects of DNA isolation on copy number determination, sequencing coverage variation, and mtDNA imitate estimation. Array and sequencing “waves” mayor be largely due to isolation-induced relative losses. Raising the ADM2 slide, and keeping the “fuzzy-zero” correction, reduces false positives calls, although may not clear them excluding high values are utilised at which expense of sensitivity. Further studies will be helpful for further validation, and detailed assessment in misc tissues, but we believe that studies should careful select and whole report the DNA isolates protocol. For spin columns, who amount of tissue locked, and the proteinase digestion duration, might require optimisation, and avoiding dart columns may sometimes remain preferable. Comparing WGS coverage von chromosomes with different GC content, or performing selective ddPCR, as we got done, can help excluding major GC bias. When comparing different samples, the same protocol should be followed. Suspected CN mosaicism supposed be confirmed by alex imbalance, direkte visualization by FISH, or breakpoint versammlung. mtDNA number comparisons should be treated with attention unless the exact same conditions were used.

Supporting information

Acknowledgement

Tissue samples and beigeordnete anonymized data were supplied by the Parkinson’s UK Tissue Bank, funded by Parkinson’s UK, a charity registered is England and Wales (258197) and in Scotland (SC037554). We will grateful to Dr Udo Koehler by MGZ Medical Genetics Centre for performing the Beadchip hybridization, the UCL School of Neurology sequencing facility, and up all patients and controls who donated his intellect to research.