`
`
`Cell
`PRESS
`
`Non-invasive prenatal diagnosis by
`single molecule counting technologies
`
`Rossa W.K. Chiu1'2, Charles R. Cantor3 and Y.M. Dennis Lol'2
`
`1 Centre for Research into Circulating Fetal Nucleic Acids, Li Ka Shing Institute of Health Sciences, The Chinese University of Hong
`Kong, Prince of Wales Hospital, 3032 Ngan Shing Street. Shatin, New Territories, Hong Kong SAR, China
`2Department of Chemical Pathology, The Chinese University of Hong Kong, Prince of Wales Hospital. 30-32 Ngan Shing Street,
`Shatin, New Territories, Hong Kong SAR, China
`aSequenom Inc., 3595 John Hopkins Court, San Diego, CA 92121—1331, USA
`
`Non-invasive prenatal diagnosis of fetal chromosomal
`aneuploidies and monogenic diseases by analysing fetal
`DNA present in maternal plasma poses a challenging
`goal. In particular, the presence of background maternal
`DNA interferes with the analysis of fetal DNA. Using
`single molecule counting methods,
`including digital
`PCR and massively parallel sequencing, many of the
`former problems have been solved. Digital mutation
`dosage assessment can detect the number of mutant
`alleles a fetus has inherited from its parents for fetal
`monogenic disease diagnosis, and massively parallel
`plasma DNA sequencing enables the direct detection
`of
`fetal chromosomal aneuploidies from maternal
`plasma. The analytical power of these methods, namely
`sensitivity. specificity. accuracy and precision, should
`catalyse the eventual clinical use of non-invasive pre-
`natal diagnosis.
`
`Confronting the challenges
`Non-invasive prenatal diagnosis (NIPD) is a long sought-
`afier goal in medical genetics. Currently, fetal genetic
`material must be collected through procedures such as
`amniocentesis and charionic villus sampling to enable
`the definitive diagnosis of fetal genetic diseases. Unfortu-
`nately, these invasive procedures are associated with a risk
`of fetal loss. Although conventional teachings of reproduc—
`tive biology state that such material should not be present,
`NIPD relies on the identification of traces of fetal genetic
`material in the blood of pregnant women. In 1969, Wal-
`knowska et (1!.
`[1,2] first reported the presence of fetal
`lymphocytes in maternal peripheral blood. Decades of
`research then followed in which intact fetal cells, present
`in the maternal circulation, were studied for use in NIPD
`{2]. However, the rarity of circulating fetal cells, typically
`several cells per milliliter of maternal whole blood, has
`prevented their robust detection [2]. At the conclusion of a
`ten—year multicenter study (1994—2003) funded by the
`United States National Institute of Child Health and
`Human Development, male fetal cells in whole maternal
`blood were detected in 41% of pregnancies with male
`fetuses, with a false-positive rate of 11% [3].
`New opportunities for NIPD emerged in 1997 when cell-
`‘free fetal DNA was identified [4] and estimated to comprise
`some 10% of the total DNA in maternal plasma [5,6]. The
`development of several applications immediately followed
`Corresponding author: Lu, Y.MAD.
`(loy1n@cuhk.edu.hk).
`
`324
`
`and translated into clinical use [7,8], including in the
`determination of fetal gender for sex-linked disorders
`[4,9], fetal rhesus D blood group status in rhesus D nega—
`tive women [7,8], and in the detection of paternally inher-
`ited mutations for autosomal dominant diseases [6,10].
`However, the co-existence of a minor population of fetal
`DNA with the major background of maternal DNA in
`maternal plasma has posed challenges for extending the
`NIPD applications beyond those focusing on the detection
`of paternally inherited fetal alleles. Two such challenging
`areas included the achievement of NIPD for fetal chromo-
`somal aneuploidies and monogenic diseases other than
`those caused by unique paternally inherited mutations.
`In this review, we dissect the root causes of the chal—
`lenges faced by NIPD researchers and review the attempts
`that have been taken to overcome them. In particular, we
`focus on the progress and efficacy in adopting the latest
`sophisticated analytical methods, namely digital polymer-
`ase chain reaction (PCR) and massively parallel sequem
`cing, for non-invasive detection of fetal monogenic diseases
`and chromosomal aneuploidies. We conclude with com-
`ments on the present practical
`feasibility of
`these
`approaches and the possibility that prenatal diagnostic
`practices will be changed in the future.
`
`Challenge 1: NIPD of monogenic diseases
`Paternally inherited fetal alleles that are not shared by the
`maternal genome are distinguishable as fetal-specific in
`maternal plasma. Thus, detection of the presence or
`absence of paternally inherited mutations in maternal
`plasma can be readily applied to the NIPD of paternally
`inherited monogenic diseases. Reported examples included
`the NIPD of achondroplasia (a disorder which results in
`short stature), myotonic dystrophy and Huntington’s dis—
`ease l10]. However, it is much more challenging to achieve
`NIPD of maternally inherited or autosomal recessive
`monogenic diseases [6]. Maternally inherited fetal alleles
`are genotypically identical to the background maternal
`DNA; hence, fetal inheritance of a maternal mutation
`cannot be established by simply detecting its presence in
`maternal plasma. Similarly, for couples sharing identical
`mutations for an autosomal recessive condition, the fetal
`disease status cannot be assessed by mere detection of the
`mutation in maternal plasma.
`Researchers have instead focused on the non—invasive
`prenatal exclusion of autosomal recessive diseases for
`01 (as-9525135 , see front matter o 2009 Elsevier Ltd. All rights reserved. doi:10.101aj.tig.2coe.os.oca Available online 13 June 2009
`
`Ariosa Exhibit 1036, pg. 1
`|PR2013-00276
`
`Ariosa Exhibit 1036, pg. 1
`IPR2013-00276
`
`

`

`couples harboring non-identical mutations [11,12]. It is
`possible to determine either the absence of the paternal
`mutation in maternal plasma or to ascertain the presence
`of the paternal wild-type allele by detecting paternal-
`specific polymorphisms linked to the disease locus [11].
`The aim of either strategy is to determine whether the
`fetus has inherited the paternal wild-type allele as this
`pattern excludes the chance ofmanifesting a disease which
`requires the inheritance of mutations from both parents.
`These strategies have been used for NIPD of B—thalassemia
`[11,13] and cystic fibrosis [12]. It is difficult to detect fetal-
`specific point mutations or single nucleotide polymorph-
`isms (SNPs) with high analytical certainty when using
`standard tools such as real-time quantitative PCR (QPCR)
`[11,13,14]. Therefore, a challenge arises as just one nucleo-
`tide change in the targeted allele must be detected in the
`low fractional concentration of fetal DNA present
`in
`maternal plasma.
`
`Challenge 2: NIPD of chromosomal aneuploidies
`Down syndrome, where affected individuals typically have
`three instead of two copies of chromosome 21 (i.e. trisomy
`21), is the most common aneuploidy, affecting ~1 in 700
`births [15,16]. Other frequent aneuploidies include tris-
`omy 18 (Edward syndrome), trisomy 13 (Patau syndrome)
`and monosomy X in females (Turner syndrome). Identifi-
`cation of trisomy 21 is the most common reason why
`women opt for prenatal diagnosis. Owing to the risk of
`fetal loss associated with invasive testing, methods invol-
`ving ultrasonography and the analysis of maternal serum
`biochemical markers are currently used to risk-stratify
`those pregnancies which require confirmatory testing [17] .
`However, these screening tests detect the phenotypic
`features instead of the genetic pathology (i.e. root cause)
`of trisomy 21. The sensitivity and specificity profile of
`these tests is suboptimal, often requiring the combined
`use of multiple modalities, and they must be conducted
`within strict gestational age windows [17]. It would be
`ideal if direct detection of trisomy 21 could be achieved
`non-invasively.
`Detection of chromosomal aneuploidy is a challenging
`puzzle in NIPD research. Fetal DNA in maternal plasma is
`cell-free. Thus, the dosage of chromosomes in the fetal
`genome cannot be determined as readily by methods such
`as fluorescence in situ hybridization. The high maternal
`DNA background also dilutes the genetic information one
`can obtain for the fetus through maternal plasma analysis.
`To overcome these issues, background maternal DNA
`interference can be minimized by the detection of molecu-
`lar signatures that are present in maternal plasma but are
`contributed almost completely by the fetus. Circulating
`fetal DNA is derived predominantly from the placenta,
`whereas maternal DNA in plasma derives from maternal
`blood cells [18,19]. Genes that demonstrate differential
`DNA methylation [18] or expression profiles [20] between
`placental tissues and maternal blood cells have been devel—
`oped as universal fetal nucleic acid markers for maternal
`plasma detection [10]. For example, serpin peptidase
`inhibitor, clade B (ovalbumin), member 5 (SERPINB5),
`also known as maspin [21,22],
`is hypomethylated in
`placental tissues but almost completely methylated in
`
`Trends in Genetics Vol.25 No.7
`
`maternal blood cells [18]. Genotype analyses confirm that
`hypomethylated SERPINB5 molecules
`in maternal
`plasma originate from the placenta or fetus.
`SERPINBE is located on chromosome 18. Tong et al. [23]
`achieved a chromosome 18 dosage comparison between
`trisomy 18 and euploid pregnancies by determining the
`ratio between polymorphic alleles ofhypomethylated SER-
`PINB5 molecules in the maternal circulation. The ration-
`ale was based on the expectation that in a heterozygous
`trisomy 18 fetus, the ratio between the SERPINB5 alleles
`would be 2:1 or 1:2 instead of 1:1 as in a heterozygous
`euploid fetus. Termed the epigenetic allelic ratio approach,
`it is the first strategy reported for the direct detection of
`chromosomal aneuploidies by cell—free DNA analysis from
`maternal plasma [23]. Extensive searches for chromosome
`21 loci which demonstrate differential methylation be-
`tween placental tissues and blood cells have been con—
`ducted [24—26] to extend the approach to the NIPD of
`trisomy 21. The low abundance of fetal DNA poses the
`main constraint on the practical feasibility of epigenetic
`based approaches [23].
`Similarly, fetal-specific, placentally expressed mRNA
`molecules are detectable in maternal plasma. We devel—
`oped an RNA—SNP allelic ratio test for the NIPD oftrisomy
`21 by determining the ratio between polymorphic alleles of
`placenta-specific 4 (PLAC4) mRNA, a transcript on
`chromosome 21, in maternal plasma [20]. PLAC4 mRNA
`is more abundant than fetal DNA in maternal plasma and
`genotyping confirms that it is of fetal origin. Deviation of
`the PLAC4 mRNA SNP allelic ratios were observed in
`plasma of trisomy 21 pregnancies compared with the
`expected 1:1 ratio in heterozygous euploid fetuses. 90%
`sensitivity and 96% specificity for the non—invasive detec—
`tion oftrisomy 21 were achieved by using the PLAC4 RNA—
`SNP test alone; this is comparable to many of the currently
`used multi—modality screening tests.
`Instead of simply targeting fetal—specific genetic signa-
`tures in maternal plasma, Dhallan et al. [27] attempted to
`reduce the maternal DNA background interference, thus
`resulting in a relative enrichment of fetal DNA. Formal-
`dehyde was used as a cell stabilizing agent to minimize
`DNA release from maternal blood cells. Fetal aneuploidy
`was then detected by assessing statistically significant
`differences between polymorphic SNP ratios in chromo-
`somes with and without involvement in the aneuploidy.
`However, controversies exist regarding the effectiveness of
`formaldehyde treatment because those findings could not
`be replicated consistently [28—32].
`These approaches enable the direct detection of fetal
`chromosomal aneuploidies, but they are only applicable to
`fetuses with certain genotypes. For example, RNA—SNP
`tests are only informative for heterozygous fetuses; thus, a
`panel of coding SNPs is required to increase the population
`coverage of those tests.
`
`Digitizing cell-free fetal DNA analysis
`Most issues confounding circulating fetal DNA analysis are
`related to the interference caused by the high maternal DNA
`background. The approaches described rely on removing the
`influence of the background nucleic acids through the
`analysis of fetal-specific mutations or nucleic acid species
`
`325
`
`Ariosa Exhibit 1036, pg. 2
`|PR2013-00276
`
`Ariosa Exhibit 1036, pg. 2
`IPR2013-00276
`
`

`

`Box 1. Digital PCR
`
`In digital PCR, template DNA is diluted to average concentrations of
`<1 molecule per well and analysed in hundreds to thousands of
`replicates [33]. Some PCR wells will be positive, whereas others will
`be negative for the targeted amplicon. Because most positive wells
`contain just one template molecule, counting the positive wells
`enables the absolute quantification of the original template DNA.
`Such quantification does not require the use of calibration standards
`or other gene targets for normalization; therefore, digital quantifica—
`tion is more accurate and precise than conventional QPCFl [34,35], in
`addition, by segregating template nucleic acid molecules into
`individual compartments, the amplification and detection of each
`template would not be affected by other templates with similar
`sequence context. As a result, the analytical power of digital PCR
`has been exploited in wide-ranging applications: the qualitative
`detection of trace molecular signatures, such as cancer mutations in
`heterogeneous biological samples [50]; quantitative imbatances
`between loci such as in loss of heterozygosity [51]; and copy
`number variation [52]. The performance of digital PCR is tradition-
`ally tedious and laborious, but disadvantages have been overcome
`by the introduction of nanofluidics devices for digital PCR analysis
`[35.52.53].
`
`in maternal plasma. However, such approaches do not
`provide general solutions. Methods need to be developed
`to extract fetal genetic information from circulating fetal
`DNA analysis in spite of co—existing maternal DNA mol-
`ecules. Digital PCR presents one option [33,34] (Box 1).
`By compartmentalizing individual template DNA mol-
`ecules, digital PCR enables fetal and maternal DNA
`molecules in maternal plasma to be analysed separately
`without cross-interference. Specific detection of fetal
`alleles, for example point mutations or SNPs, in maternal
`plasma is therefore possible. Lun et al. [35] used a micro-
`fluidics digital PCR platform to detect fetal-derived Y-
`chromosomal DNA in maternal plasma. The fractional
`concentration of fetal DNA could be determined more
`accurately by this digital method than by conventional
`(analog) QPCR. The zinc finger protein homologs present
`on chromosomes Y (ZFY) and X (ZFX) were co-amplified
`using the same primer set to quantify the fetal and total
`DNA, respectively. The ZFY and ZFX amplicons differ by
`only two nucleotides;
`they are discriminated readily
`using duplex fluorogenic probes. In another study, Lun
`et at. [36] used duplex fluorogem'c probes in a digital PCR
`assay to discriminate the wild-type hemoglobin, [3 (H313)
`allele from a paternally inherited HBB point mutation,
`namely hemoglobin E, in maternal plasma. Previously,
`QPCR—based discrimination of fetal and maternal alleles
`that differed by just one or a few nucleotides was not
`specific or sensitive enough and required the use of more
`complex tools such as mass spectrometry [11,13,141].
`Thus, the use of digital PCR could overcome the challenge
`of detecting fetal point mutations or SNPs in maternal
`plasma.
`
`Quantitative power of molecule counting
`In digital PCR (Box 1), direct counting of the wells with
`positive amplification of the target amplicon enables
`absolute quantification of the template DNA without
`the need for quantitative calibration standards. The ana-
`log and exponential nature of QPCR becomes a ‘1’ and ‘0’
`signal in digital PCR [34]. Hence, digital counting plat.-
`326
`
`Trends in Genetics Vol.25 No.7
`
`Box 2. Trisomy 21 detection
`Fetal DNA co-exists with maternal DNA in maternal plasma. A
`trisomy 21 fetus,
`in comparison to a euploid fetus. adds extra
`amounts of chromosome 21 sequences into maternal plasma, in
`direct proportion to fetal DNA concentration. For example, a
`maternal plasma sample from a euploid pregnancy containing 100
`genome-equivalents lGEllmI of total DNA with 10 GEirrII DNA
`contributed by the fetus (Le. 10% fetal DNA which is the typical
`median concentration for the first and second trimesters of
`pregnancy [35]) should contain a total of 200 copies (180 maternal
`copies + 20 fetal copies) of chromosome 21 sequences per milliliter
`of maternal plasma. For a trisomy 21 pregnancy, each fetal GE
`would contribute three copies of chromosome 21, resulting in a total
`of 210 copieslrnl (180 maternal copies + 30 fetal copies] of chromo-
`some 21 sequences in maternal plasma. At 10% fetal DNA
`concentration, the amount of chromosome-Z‘i-derived sequences
`in the maternal plasma of a trisomy 21 pregnancy would therefore
`be 1.05 times that of a euploid case. This degree of quantitative
`difference is difficult to discriminate confidently by QPCR [34].
`
`forms should enable more precise and accurate quantifi-
`cation [33,37]. For example, conventional QPCR can
`readily discriminate a difference in one or more threshold
`cycles (i.e. the number of cycles required for a reaction to
`reach a predetermined fluorescent threshold). As the
`amplicon concentration approximately doubles in each
`PCR cycle, the minimum quantitative difference that
`QPCR can easily discriminate is approximately a twofold
`change of template DNA concentration, which would be
`inadequate for detecting the increase in chromosome 21
`DNA concentrations in maternal plasma of trisomy 21
`pregnancies [38] (Box 2).
`it should be
`Using precise quantification methods,
`possible to directly detect fetal aneuploidy by determining
`if the total (maternal +fetal) amount of the aneuploid
`chromosome (e.g. chromosome 21 for trisomy 21) is over-
`represented or under-represented compared with other
`chromosomes in maternal plasma. Digital PCR was per-
`formed for an amplicon located on chromosome 21 and
`another amplicon not on chromosome 21, that is, 3. refer-
`ence chromosome [38,39]. The relative amounts of the two
`amplicons were compared in a strategy termed digital
`relative chromosome dosage (RCD) analysis [38]. Although
`over-representation of the chromosome 21 amplicon is
`expected in trisomy 21 fetuses, the degree of over-repres-
`entation relies on the fetal DNA concentration (Box 2) and
`is smaller at low fetal DNA concentrations. Therefore,
`higher numbers of digital PCR analyses are required to
`ensure adequate statistical power to determine with con—
`fidence the presence or absence of chromosome 21 over-
`representation [38,39]. Our analysis of mixtures of placen-
`tal and maternal blood cell DNA samples obtained from
`euploid and trisomy 21 pregnancies showed that NIPD of
`trisomy 21 could be accurately detected or excluded in 97%
`of cases by performing 7680 PCR analyses when the
`sample contained 25% fetal DNA [38]. Using mixtures of
`cell line DNA, Fan et al. [39] also demonstrated the use of
`digital PCR to detect chromosome 21 over-representation
`in trisomy 21.
`Our study revealed several key parameters that affect
`digital RCD analysis of maternal plasma DNA [38].
`Because the median concentration of circulating feta]
`DNA is usually (25% in first and second trimesters of
`
`Ariosa Exhibit 1036, pg. 3
`|PR2013-00276
`
`Ariosa Exhibit 1036, pg. 3
`IPR2013-00276
`
`

`

`
`Trends in Genetice Vol.25 No.7
`
`(a)
`
`6!)
`
`DNA molecules
`
`Maternal
`genotype
`
`..
`
`ms X
`—>
`...._..._l_‘.....
`
`Digital PCl—‘l I
`
`TRENDS in Genetics
`
`identification
`Product
`
`I
`
`+
`
`l3
`
`Interpretation
`
`
`
`Figure 1. Schematic illustration of digital RMD and digital MASS. Digital FlMD measures the relative amounts of the maternal mutant and wild-type alleles in maternal
`plasma to determine the inherited dosage of the mutant allele by the fetus. Digital NASS is a digital PCB-based method that enables the preferential analysis of short DNA
`molecules without physical size fractionation of DNA molecules. Lori at at. [36] applied the combined use of digital RMD and NASS to demonstrate the feasibility for the
`NIPD of fetal monogenic diseases. la] When a pregnant woman and her fetus are both heterozygous for a gene mutation, the amounts of the mutant allele M [shown in
`orange) and wild-type allele N [Shown in blue) are in allelic balance in maternal plasma. When the fetus is homozygous forthe wild-type or mutant allele, there will be an
`under-representation or oversrepresentation of the mutant allele, respectively. Digital HMD determines if the mutant and wild-type alleles in maternal plasma are in allelic
`balance or imbalance. lb) The scheme of the ZFY—ZFXdigital NASS essay is shown. The assay can discriminate between ZFX, denoted by X, and ZFY. denoted by Y, DNA
`molecules. In addition. the assay can distinguish if the ZFXand ZFYDNA molecules are long or short. Digital PCR is performed using two fomard primers (black and red
`arrows pointing right) and one reverse primer (red arrows pointing left], or vice versa, that are oriented to produce a short amplicon overlapping with the long amplicon.
`When a single DNA molecule at least as long as that specified by the long amplicon is captured in the reaction well, both the long (green lines] and short (orange lines) PCR
`products are generated. When a single DNA molecule shorterthan the span of the long amplicon is captured, only the short amplicon is generated. The presence ofthe long
`andi'or short amplicons can be detected by strategically located hybridization probes or extension primers. An extension primer. L [blue box). is designed to detect the
`presence of the long amplicon. A separate extension primer is located within the short amplicon, and the extension products are used to discriminate the ZFX and ZFY
`alleles (shown as boxed X [yellow] and boxed V [pink], respectively). The identities of the DNA molecules can be determined by counting the products present within each
`well. +, present; —, absent. Figure adapted, with permission, from Ref. [36] (Copyright, 2008; National Academy of Sciences, U.S.A.).
`
`pregnancy [35], either fetal DNA enrichment or additional
`PCR analyses are required. Indeed, if the fetal DNA con-
`centration is halved, four times as many digital PCR
`analyses are needed [38]. Because maternal plasma
`DNA concentrations are typically thousands of copies!
`mL, tens of milliliters of maternal blood are needed to
`perform tens of thousands of digital PCRS. Alternatively,
`given that plasma DNA exists as small fragments, instead
`oftargeting one chromosome 21 amplicon, digital detection
`of multiple multiplexed amplicons on chromosome 21
`would effectively increase the number of digital PCR data-
`points within the same fixed maternal plasma volume [38].
`Further studies are required to evaluate the efficacy of
`such options.
`
`From relative chromosome dosage to relative mutation
`dosage
`Principles similar to digital RCD have been developed for
`NIPD of monogenic diseases. Instead of targeting only
`paternal mutations, it is possible to compare the relative
`amounts of the maternal mutant and wild-type alleles in
`maternal plasma to determine the inherited dosage of the
`mutant: allele. Termed the relative mutation dosage (RMD)
`approach [36,40],
`this application is most clinically
`relevant for pregnant women who are heterozygous for a
`
`given mutation (Figure la). If the fetus has not inherited
`the mutation, under-representation of the mutant allele is
`expected. If the fetus is homozygous for the mutation (i.e. a
`second mutant copy was contributed by the father), over-
`representation of the mutant allele is expected. Lastly, if
`the fetus is heterozygous for the maternal mutation, the
`mutant and Wild—type alleles should be in allelic balance.
`Lun et al. [36] used digital RMD maternal plasma analysis
`to determine the fetal inheritance of hemoglobin E and B-
`thalassemia mutations in mothers who are carriers of
`either mutation. The principles of digital RMD are feasible
`but, as with digital RCD, large numbers of digital PCR
`analyses are required for samples containing low fetal
`DNA concentrations. To render digital RMD more practi-
`cal, it was combined with a fetal DNA enrichment strategy
`[36]. Fetal DNAs are shorter than maternal DNAS [41]
`and, therefore, size fractionation of short DNA molecules
`can enrich fetal DNA. Instead of using physical methods of
`size fractionation, such as gel electrophoresis [13], Lou
`er al. [36] used a digital method, termed digital nucleic acid
`size selection (NASS), to derive information from short;
`DNA molecules. Digital NASS uses a duplex digital PCR
`assay targeting overlapping amplicons of different sizes
`(Figure 1b). During NASS analysis, only wells showing the
`presence of short DNA molecules are counted for RMD
`
`327
`
`Ariosa Exhibit 1036, pg. 4
`|PR2013-00276
`
`Ariosa Exhibit 1036, pg. 4
`IPR2013-00276
`
`

`

`assessment. In maternal plasma samples, the combined
`use of digital NASS and RMD enables the fetal genotype to
`be discernible in cases in which RMD alone would be
`insufficient [36]. With these new developments, NIPD of
`paternally or maternally contributed or autosoma] reces—
`sive monogenic diseases can be achieved.
`
`Molecule counting by massively parallel maternal
`plasma DNA sequencing
`The low fetal DNA fractional concentration and the low
`absolute concentration of template DNA in maternal
`plasma requires either fetal DNA enrichment or large
`numbers of counted DNA molecules to bring NIPD by
`molecule counting close to clinical use. As plasma DNA
`is fragmented, instead of targeting specific loci in the
`genome, a locus-independent method could be used. One
`copy of chromosome 21 therefore would be sampled and
`counted many times in a locus-independent method,
`instead ofjust once, as for example in locus-specific assays
`(Figure 2). By increasing the number of measurements per
`sample, higher analytical precision can be achieved without
`the need to increase the volume of input maternal plasma.
`The recent availability of massively parallel sequencing
`platforms [42] have been adopted as a tool for maternal
`plasma DNA analysis for the NIPD of trisomy 21 and
`potentially other chromosomal aneuploidies [43,44].
`The rationale is to use massively parallel sequencing to
`count DNA molecules in maternal plasma. When a woman
`is pregnant with a trisomy 21 fetus, an over—representation
`ofthe fractional concentration ofchromosome 2 1 sequences
`in her plasma is expected (Box 2). Therefore, if a random
`representative portion of DNA fragments from a maternal
`plasma sample is sequenced, the frequency distribution of
`the chromosomal origin of the sequenced DNA fragments
`should reflect the genomic representation of the original
`maternal plasma sample. Assuming that the genomic
`
`Trends in Genetics Vol.25 No.7
`
`representations of maternal and feta] DNA in maternal
`plasma are not grossly skewed or uneven across the
`chromosomes, an increased proportion of chromosome 21
`sequences should be present
`in relation to the total
`sequenced reads for DNA in maternal plasma obtained
`from a trisomy 21 pregnancy.
`Both Chiu et at. [43] and Fan etal. [44] demonstrated the
`use of massively parallel plasma DNA sequencing for
`NIPD of trisorny 21 on Illumina’s Genome Analyzer plat-
`form [45]. A short region on one end of each plasma DNA
`molecule is sequenced and aligned computationally to the
`reference human genome to determine the chromosomal
`origin of each DNA fragment. The proportion of sequenced
`reads from chromosome 21 were compared between tris‘
`omy 21 and euploid pregnancies. Fan et al.
`[44] used
`1.3 mL to 3.2 mL plasma from nine trisomy 21 and six
`euploid pregnancies and obtained ~5 million unique reads
`per sample, with up to one nucleotide mismatch. The
`sequence density per 50-kb window for chromosome 21
`was normalized by the median value obtained from the
`euploid cases. The normalized sequence tag densities for
`the trisomy 21 cases were >99% confidence interval bound
`for the euploid cases.
`Chiu et al. [43] used DNA from 5 mL to 10 mL plasma
`from 14 trisomy 21 and 14 euploid pregnancies and
`obtained a mean of ~2 million unique reads per sample,
`without mismatches to the reference human genome. The
`number of reads originating from chromosome 21 was
`expressed as a proportion of all sequenced reads, and z-
`scores, representing the number of standard deviations
`away from the mean proportion of chromosome 21 reads in
`a reference set of euploid cases, were determined for each
`case. A z-score >:l: 3 indicated a 99% chance of a statisti-
`cally significant difference in the assessed parameter for
`the test case compared with the reference group. Thus, a
`high z—score was expected for trisomy 21 cases. The mas—
`
`
`(a) Locus-specific
`
`Targeted amplicon
`
`:2
`
`5 plasma DNA fragments from 1 molecule
`of chromosome 21 DNA
`
`
`Figure 2. Schematic comparison between locusespecific and locus-independent methods for DNA quantification. DNA molecules exist as short fragments in maternal
`plasma [41]. Hence. instead of comparing the relative amounts between specific loci as with conventional DNA quantification methods [5], the amount of quantitative
`information that one could derive with the same amount of plasma DNA input greatly increases with the use of locus—independent quantification methods that treat each
`DNA fragment as an individual target. (a) When using locus—specific assays. five copies of chromosome 21 with the targeted amplicon region intact [depicted by the
`different colored DNA molecules) would need to be physically present to generate a count of five. lbl In the locus-independent method, five fragmented portions originating
`from a single chromosome 21 (depicted by the DNA fragments of the same color] could potentially contribute to a count of five.
`32B
`
`Ariosa Exhibit 1036, pg. 5
`|PR2013-00276
`
`Chromosome 21
`
`5 plasma DNA fragments
`from 5 molecules of
`chromosome 21 DNA
`
`(b) Locus-independent
`
`Chromosome 21
`
`
`
`Ariosa Exhibit 1036, pg. 5
`IPR2013-00276
`
`

`

`
`
`Trends in Genetics Vol.25 No.7
`W
`
`Chromosome
`3 5 4
`
`
`1
`
`2
`
`
`
`
`
`2~score O
`
`
`
`11
`
`16
`
`12
`
`1?
`
`13
`
`18
`
`14
`
`19
`
`15
`
`20
`
`
`
`
`2“score 0
`
`21
`
`27
`24
`21
`18
`
`1'5
`
`z-score
`
`
`22
`23
`
`Key:
`‘ Euploid male fetus
`fl Euploid female fetus
`
`I Trisomy 21 female letus
`
`I Trisomy 21 male fetus
`
`
`
`TRENDS in Genetics
`
`Figure 3. Detection of fetal trisomy 21 by massively parallel sequencing. In this approach, a random representative ponion of DNA molecules in maternal plasma is
`sequenced. The chromosomal origin of each sequenced read is identified by bioinformatics analysis, The mean and standard deviation of the proportion of reads from each
`chromosome of a reference sample set comprising pregnancies with euploid male fetuses are determined. Z-scores, representing the number of standard deviations from
`the mean of the reference sample set. of the percentage chromosomal representation for each maternal plasma sample are calculated. Z-scores beyond 1: 3 suggest a >99%
`chance of the presence of chromosomal over— or under-representation compared with the reference group. Here, plots of z-scores for each chromosome for maternal
`plasma samples from 14 trisomy 21 and 14 eupioid pregnancies are shown. Each ofthe 28 bars shown for each chromosome corresponded to the z-scores lor one of the 28
`maternal plasma samples. Samples 1 to 23 are shown consecutively from left to right. Figure adapted, with permission, from Ref. [43] lCopvright, 2008; National Academy
`of Sciences, U.S.A.).
`
`sively parallel sequencing approach was reliable and
`robust: in all cases, z-scores <i 3 were obtained for all
`chromoaomes except 21 and X (Figure 3). Z—scores of
`chromosome 21 were beyond +5 for all 14 trisomy 21 cases
`but within i 3 for all euploid cases. Because pregnancies
`with male fetuses were used as the reference sample set,
`z—scores for the X-chromosome were increased in all preg—
`nancies with female fetuses.
`
`Both studies demonstrated that massively parallel
`sequencing can randomly count and identify DNA frag—
`merits in maternal plasma in a locus-independent manner
`(Figure 2) to detect small quantitative perturbations in
`genomic distribution of plasma DNA [43,44]. The large
`number ofmeasurements done per sample enables ahighly
`precise estimation of the proportion of chromosome 21
`sequences; hence, its over-representation in trisomy 21
`
`329
`
`Ariosa Exhibit 1036, pg. 6
`|PR2013-00276
`
`Ariosa Exhibit 1036, pg. 6
`IPR2013-00276
`
`

`

`can be robustly detected. The robustness of the approach
`further suggests that the genomic distributions ofmaternal
`and fetal DNA molecules in maternal plasma are unlikely to
`be grossly skewed. In spite of the low abundance of fetal
`DNA in maternal plasma [5,35], feta