Human Reproduction vol.6 no.6 pp.836-849, 1991
`
`Using PCR in preimplantation genetic disease diagnosis
`
`W.Navidi 1 and N.Arnheim2
`
`Departments of Mathematics and 2Molecular Biology, University of
`Southern California, Los Angeles, CA 90089, USA
`1To whom correspondence should be addressed
`
`Preimplantation diagnosis of genetic disease can be
`accomplished by embryo biopsy or polar body analysis using
`in-vitro gene amplification (PCR). PCR analysis of single cells
`is subject to a number of errors which decrease the reliability
`of the diagnosis. Using realistic assumptions about error rates
`based on experimenal data, we analyse some of the practical
`consequences to be faced by whose wishing to use this
`diagnostic procedure. We considered both autosomal
`dominant and recessive diseases. We calculate the probability
`of making mistakes in the diagnosis, assuming a realistic
`range in the magnitude of PCR efficiency, cell transfer, and
`contamination errors. We conclude that, in general, analysing
`blastomeres is subject to less mis-diagnosis than polar body
`analysis, except in the case of dominant diseases which are
`caused by genes which lie extremely close to the centromere.
`We also show that typing multiple blastomeres from a single
`embryo or combining polar body typing with hla'itomere
`analysis results in significantly lower levels of mis-diagnosis
`with unacceptable consequences. The preimplantation
`diagnosis of X-linked diseases based upon Y chromosome
`sequence analysis is also discussed.
`Key words: polymerase chain reaction/gene amplification/
`blastomere typing/polar body typing/preimplantation diagnosis
`
`Introduction
`
`The polymerase chain reaction (PCR; Saiki et al., 1985, 1988;
`Mullis and Faloona, 1987) is a simple method capable of rapidly
`amplifying DNA sequences in vitro. A single DNA segment
`composed of a few hundred base pairs present in a human
`genomic background having a complexity of three billion base
`pairs can be selectively amplified hundreds of millions to billions
`of times . In · this way, the proverbial needle in a haystack is
`converted to a stack of needles. The fundamental principle of
`PCR and its applications to biological and medical science have
`been reviewed elsewhe~e (White et al., 1989; Erlich, 1989;
`Arnheim, 1990; Arnheim et al., 1990; Innis et al., 1990; Erlich
`et al. , 1991). The first application of PCR was to the prenatal
`diagnosis of sickle cell anaemia (Saiki etal., 1985). Since then,
`PCR has been applied to the prenatal diagnosis of many other
`genetic diseases using materials obtained from amniocentesis or
`chorionic villus sampling (CVS) (Kazazian, 1989).
`
`The ability of PCR to be so selective in its amplification is
`accompanied by an exquisite sensitivity. Thus, a single molecule
`of DNA present in a single sperm cell can be amplified and
`analysed (Li et al., 1988, 1990; Cui et al., 1989). The ability
`to study DNA sequences in a single haploid or diploid cell (Li
`et al., 1988; Jeffreys et al., 1988) led geneticists and reproductive
`scientists to propose and experiment with the idea that the
`diagnosis of genetic disease could be established-in human embryo
`produced by in-vitro fertilization prior to implantation (Li et al.,
`1988; Handyside eta/., 1989; Coutelle eta/., 1989). Mouse
`embryos obtained from mated females have also been biopsied
`and the DNA analysed (Holding and Monk, 1989; Bradbury
`et al., 1990; Gomez et at., 1990). Normal pregnancies have been
`demonstrated after human (Handy side et al., 1990) and mouse
`(Gomez et al., 1990; Bradbury et al., 1990) embryo analysis.
`The possibility of diagnosin.g genetic diseases even before
`fertilization using eggs and analysing the first polar body has also
`been examined (Monk and Holding, 1990; Strom et al., 1990).
`Successful preimplantation diagnosis of genetic disease depends
`upon being able accurately to determine the genotype of one or
`perhaps a few cells using PCR. In addition, the manipulations
`of the embryo, or unfertilized egg in the case of polar body
`analysis, must not affect its nom1al development. In this paper,
`we consider only the accuracy of DNA analysis by PCR.
`For most routine applications of PCR, a sample consisting of
`the amount of DNA purified from 150 000 diploid cells (1 ,.g)
`is typical. In this case, not all of the original 300 000 copies of
`the target are required to be amplified during every PCR cycle
`in order to determine the genotype of the DNA accurately. The
`analysis of DNA in a single cell is an entirely different matter.
`A single diploid cell or polar body contains only two DNA
`molecules representing each single copy gene and therefore the
`accuracy of genotype determination is much more sensitive to
`random fluctuations in the efficiency with which each individual
`molecule is amplified during each PCR cycle. Thus, each of the
`two DNA molecules of the target gene present in a diploid cell
`must be efficiently amplified to a detectable level, and if a cell
`is heterozygous for a gene', both alleles need to be capable ·
`of being identified.
`In the case of X-linked genetic diseases, however, a different
`approach may reduce the difficulty of single cell analysis. If the
`mother is a carrier of an X -linked recessive gene, female embryos
`would not suffer from the disease while male embryos would
`have a 50% chance of being affected. Handyside et al. (1989,
`1990) have used human embryo sex-typing to allow implantation
`of only female embryos. In this procedudre, a DNA sequence
`rep~atcd many times and specific to the Y chromosome was
`
`836
`
`© Oxford University Press
`
`The Johns Hopkins University Exhibit JHU2009 - Page 1 of 14
`
`

`

`analysed hy PCR. Highly repeated gene targets increase the
`probability of detecting PCR product from a single cell.
`The analysis of DNA sequences in single diploid cells using
`PCR has other associated problems. The single cell needs to be
`reliably transferred to the PCR reaction tube and contamination
`of the PCR reaction tube with exogenous DNA must be kept to
`a minimum since the target itself consists of only two molecules.
`We have analysed the accuracy of DNA diagnosis on one or a
`few cells taking into consideration the effects of amplification
`efficiency , the reliability of cell transfer, and contamination by
`either maternal cell contributions or by PCR products present
`in the laboratory from previous experiments. We consider the
`effect of these types of errors on blastomere and polar body
`analysis of single copy genes in the case of both dominant and
`recessive autosomal diseases. We also consider errors associated
`with the diagnosis of X-linked disease.
`
`Assumptions
`
`Denote the two alleles at the locus of interest by A and a,
`where A is dominant and a is recessive. We consider two cases.
`In the first case, the disease is recessive and both parents are
`heterozygous. In this case, an embryo has 25% probability of
`being of genotype AA, 50% of being of genotype Aa and 25%
`of being of genotype aa . Since the disease is recessive, only
`embryos of genotypes AA and Aa are suitable for implantation.
`Thus, an untyped embryo has 75% probability of being suitable.
`In the second case, the disease is dominant and one parent is
`heterozygous while the other is homozygous aa. In this case, an
`embryo has 50% probability of being of genotype Aa and 50%
`probability of being of genotype aa. Since the disease is dominant,
`only embryos of genotype aa are suitable for implantation, so
`an untyped embryo has 50% probability of being suitable. In
`either case, using blastomere typing to select an embryo, or using
`polar body typing to select an oocyte for fertilization, considerably
`increases the probability of selecting a suitable embryo. The
`probability does not rise to I 00%, however, because of the
`possibility of typing error. lt is of interest to compute the
`probability that an embryo has a genotype which makes it suitable
`for use, given that a blastomere cell has been typed as having
`such a genotype, and the probability that a fertilized oocyte has
`a genotype making it suitable for implantation given that it has
`been deduced to be suitable through typing of the corresponding
`polar body.
`We present some calculations of such probabilities, based on
`some assumptions about the typing procedure. We consider
`three sources of error . First, the chance that the blastomere
`or polar body may fail to be placed in the reaction tube, so
`that the tube contains no relevant DNA. Second, the chance
`that an allele present in the tube may fail to be amplified to a
`detectable level. Third, the chance that the reaction may be
`contaminated, resulting in the false detection of an allele. We
`make the following assumptions:
`(i) Each allele from the polar body or blastomere which is present
`in the tube has probability r of being amplified to a detectable
`level, independently of each other allele.
`(ii) Each cell has probability d of being placed in the reaction
`tube, and probability 1 -d of failing to be placed in the tube.
`
`Using PCR in preimplantation diagnosiS
`
`(iii) There is probability c that the reaction tube is contaminated
`and that the contaminant will be detected. At most one
`contaminating allele will be present. Each of the two alleles
`is equally likely to be a contaminant. When contamination is
`present, the number of contaminating molecules is approximately
`the same as the number of target molecules, so that the presence
`of contamination will not affect the detection of the target
`molecules. Other models of contamination are also plausible. We
`might assume, for example, that simultaneous comamin tion of
`two distinct alleles is possible , or that the amount of contamination
`can he o great as to prevent detection of the target molecules.
`The effect of such alternative contamination assumptions will
`be discussed below.
`In a PCR analysis of > 700 single spermatozoa, Cui et al.
`(1989) estimated that the value of c was < 5%, the value of d,
`i3% , and the value of r was
`using micromanipulation, was -
`-95%. In other unpublished experiments, the value of r was
`usual! y found to range from 80 to 95%, c from 0 to 7%,
`and d from lO to 20%. Our calculations take these ranges
`into consideration .
`The parameters r, d and c used for calculation are unlikely
`to be constant from laboratory to laboratory, or even from time
`to time within the same laboratory. For this reason, we include
`calculations across a range of values. In any particular laboratory
`where this work is to be carried out, laboratory-specific parameter
`values should be estimated empirically. The results given here
`can then provide a guide to estimating the accuracy which
`may be expected.
`Errors in typing can be divided into three categories, reflecting
`varying degrees of severity. The least seriou error is one which
`results from using an embryo or oocyte which bould have been
`used anyway, or in not using one which should not have been
`used. An example of such an error is typing an Aa cell as AA ,
`in a situation where only aa cells should be used. Such t!ri'OI',
`will be classified as 'acceptable ' . Given the limited number of
`embryos or oocytes available, a somewhat more undesirable sort
`of error is one which results in not using an embryo or oocyte
`which could have been used. Since in general no great harm
`is done by such errors, they are classified as 'tolerable'. The
`most serious error is one which results in implanting an embryo
`which should not be implanted. Such errors are classified as
`'unacceptable'.
`Using assumptions (i- iii), we calculated the probabilities that
`a cell which has been typed as being of a given genotype will
`in fact be of some other given genotype, and from these we
`computed the probabilities of error for each of the three categories
`mentioned above . The re. ults are given below. Detail of the
`calculations are given in the Appendix.
`
`Results and discussion
`
`Blastomeres, autosomal recessive disease
`When the disease is recessive, embryos with genotypes AA and
`Aa are usable, while those with genotype aa are not. We assume
`both parents are heterozygous, ·o any mbryo has probability
`75 % of being u. able. Table I ategorizes the possible typing
`errors in this case. The column labelled ' primary cause( ) of
`
`837
`
`The Johns Hopkins University Exhibit JHU2009 - Page 2 of 14
`
`

`

`W.Navilfi and N.Arnheim
`
`Table I. Errors on blastomere typing for autosomal recessive disease~
`
`True
`genotype
`
`Observed
`genotype
`
`Use'l
`
`Error
`category
`
`Primary
`causc(s)
`of error
`
`Conditional prohability of error
`
`(I)
`r o: O.ll
`d = 0.8
`c = 0.05
`
`(2)
`I' = 0.8
`d=l
`c = 0
`
`(3)
`'= 0.9
`d = 0.8
`(" = 0.5
`
`(4)
`r = 0.9
`d = 0.9
`c = 0.5
`
`AA
`AA
`A a
`A a
`a a
`a a
`
`A a
`a a
`AA
`aa
`AA
`A a
`
`Yes
`No
`Yes
`No
`Yes
`Yes
`
`Acceptable
`Tolerated
`Acceptable
`Tolerated
`Unacceptable
`Unacceptable
`
`c
`d,c
`r
`
`d,c
`c
`
`0 018
`0.0057
`0.26
`0.26
`0.0057
`0.018
`
`0
`0
`0 .25
`0.25
`0
`0
`
`0.015
`0.0056
`0 16
`0.16
`0.0056
`0.015
`
`0.015
`0.0026
`0.16
`0 . 16
`0.0026
`0.015
`
`(5)
`r = 0.9
`d=l
`c = 0.5
`
`0
`0.00022
`0. 15
`0 . 15
`0.00022
`0.015
`
`(6)
`r = 0.9
`d=l
`c = 0
`
`(7)
`r ~ I
`d = O.H
`c = 0.5
`
`(Ill
`r =
`d = I
`(' = 0
`
`0
`0
`0. 15
`015
`0
`0
`
`0.012
`0.0063
`0 013
`0.013
`0.0063
`0.0 12
`
`0
`()
`0
`0
`0
`0
`
`error' tells which of the parameters r (amplification efficiency),
`d (cell placement) and c (contamination) plays the greatest role
`in determining the frequency at which the given error occurs .
`For example, the last row of the table refers to the error of typing
`a cell Aa when its true genotype is aa. For this error to occur,
`the reaction must be contaminated with the A allele. Thus the
`value of c has more effect on the frequency of this error than
`the values of r and d. The second to last row refers to the error
`oftyping a cell AA when its true genotype is aa. For this to occur,
`not only must the reaction be contaminated with the A allele,
`but both a alleles must escape detection . The likelihood of the
`former event depends on the parameter c. The latter event may
`be due either to failure to place the cell in the tube, or to failure
`of PCR to amplify either of the two copies of the allele. For
`reasonable values of d and r, failure to place the cell in the tube
`is more likely to happen than failure of PCR to amplify either
`allele. Thus in general, the likelihood of both alleles escaping
`detection depends primarily on the parameter d, with r exerting
`a mild influence. Thus the parameters d and c are most impor(cid:173)
`tant in determining the frequency of this error. As a final exam(cid:173)
`ple, the third row of the table refers to the error of typing a cell
`AA when the true genotype is Aa . In general, this happens when
`the A allele is amplified to a detectable level but the a allele is
`not. Thus the value of r is most important in determining the
`frequency of this error. It is also possible for this error to be
`caused by a failure to place the cell in the tube, combined with
`contamination by a fragment contajning the A allele. This com(cid:173)
`bined error is probably much less likely than a lack of amplifica(cid:173)
`tion, so the effect of the parameters c and dis relatively minor.
`The right-most columns of Table I (labelled Conditional
`probability of error') give probabilities for each possible error
`in the case of blastomeres where the disease is recessive, for
`several values of r, d and c. The probabilities are conditional
`on the predicted genotype. Thus, for values of r = 0.9, d := 0.8
`and c = 0.05, then a blastomere which has been typed Aa has
`probability of - 1. 5% of in fact being of genotype aa, thus
`producing an unacceptable result. A blastomere which has been
`typed AA, however, has probability of only -0.56% of having
`true genotype aa. Thus under these conditions we can expect that
`- 1.5% of implantations of b1astomeres typed Aa and -0 .56%
`of implantations of blastomeres typed AA will yield unacceptable
`results . Examination of Table I shows that in each row the
`parameters primarily
`responsible
`for
`fluctuations
`in
`the
`probabilities are indeed the ones listed as 'primary causes of
`
`838
`
`error' . For example, in the last row, comparing columns 3, 4
`and 5 shows that increasing the value of d from 0. 8 to 1 causes
`little or no improvement in the frequency with which cells typed
`Aa will turn out actually to be aa. Comparing columns l , 3 and
`7 shows that increasing the efficiency of r from 0. 8 to 0. 9 to
`1 results in only a small improvement. Comparing the columns
`where the contamination rate cis 0.05 with those where c = 0
`shows that reducing the contamination rate dramatically reduces
`the error rate. A similar analysis can be made in each row
`of the table.
`Careful perusal of the tables shows what appears to be an
`anomalous result. In the second line of Table I, comparing
`colurrms 3 and 7 shows that if d = 0.8, c = 0.05, the probability
`that a cell typed aa will actually be of type AA increases slightly
`as the efficiency r increases from 0.9 to 1. This is discussed
`further in the Appendix. This anomaly appears again in line 5
`of Table I and also in Tables III and Vll.
`
`Polar bodies, autosomal recessive disease
`When the disease is recessive, onJy oocytes with genotype AA
`are suitable for fertilization. The polar bodies corresponding to
`these oocytes have genotype aa. The proportion of oocytes which
`are suitable depends on the recombination fraction 0 between the
`centromere and the locus of interest. The proportion of oocytes
`of genotypes AA, Aa and aa is V2 -e. 20 and th -0 respectively.
`We assume both parents are heterozygous, so any oocyte has
`probability 25% of being usable. Table II categorizes the possible
`typing errors in this case. The errors in the second and fourth
`rows are classified as 'potentially unacceptable'. In these cases,
`an oocyte containing the disease allele a is used, and whether
`the resulting embryo is of the unacceptable genotype aa is a matter
`of chance. If the oocyte is of genotype aa, the probability that
`the embryo will have genotype aa is 50%, and if the oocyte is
`of genotype Aa, the corresponding probability is 25%, which
`is identical to the probability if no preimplantation diagnosis
`is attempted.
`in
`Table III gives probabilities of unacceptable results
`polar body typing for recessive diseases for several values
`of the parameters, and lists the parameters whose values
`must be improved in order to improve noticeably the error
`rate . In polar body typing, the probability of error depends
`on the recombination fraction e as well as on r, d and c. The
`probabilities for the polar body errors are found by calculating
`the probabilities of each of the two errors which are potentially
`
`The Johns Hopkins University Exhibit JHU2009 - Page 3 of 14
`
`

`

`Using PCR in preimplantation diagnosis
`
`Tuble II. Errors in polar body typing tor autosomal recessive diseases
`
`Polar body
`True
`gc:notype
`
`AA
`AA
`A a
`A a
`a a
`a a
`
`Observed
`genotype
`
`A a
`a a
`AA
`aa
`AA
`A a
`
`Oocyte
`True
`genotype
`
`aa
`aa
`A a
`A a
`AA
`AA
`
`Deduced
`genotype
`
`A a
`AA
`a a
`AA
`a a
`A a
`
`Use?
`
`Error category
`
`Primary cause(s)
`of error
`
`No
`Yes
`No
`Yes
`No
`No
`
`Acceptable
`Potentially unacceptable
`Acceptable
`Potentially unacceptable
`Tolerated
`Tolerated
`
`c
`d,c
`,.
`,.
`d,c
`c
`
`Table Ill. Conditional probabilities of unacceptable errors in polar body typing for autosomal recessive diseases
`
`Primary cause(s)
`of error
`
`Polar body, aa observed, 0 = 0
`Polar body, a a observed, 0 = 0.10
`Polar body, aa observed, 8 = 0.20
`Polar body, a a nhserved, 8 = 0 25
`Polar body, aa observed. 8 = 0.30
`Polar body, aa observed, 0 = 0.40
`
`d,c
`r
`r
`r
`I'
`
`I'
`
`(I)
`,. = 0.8
`d = 0.8
`c = 0 5
`
`O.CXl38
`0.023
`0.050
`0.067
`0.088
`0.15
`
`(2)
`/' "" 0 .8
`d = l
`c = 0
`
`(3)
`r = 0 .9
`d = 0 .8
`c = 0.5
`
`0
`0.019
`0,045
`0.063
`0.083
`0.14
`
`0.0033
`0.015
`0.031
`0.043
`0 .059
`0.11
`
`(4)
`I'= 0.9
`d = 0.9
`c = 0.5
`
`0.0016
`0.013
`0.029
`0.041
`0.056
`0.11
`
`(5)
`r = 0.9
`d=l
`c = 0.5
`
`0.00013
`0.011
`0.027
`0.039
`0.054
`0.11
`
`(6)
`r = 0.9
`d =I
`c = 0
`
`(7)
`r = 1
`d = 0.8
`c = 0.5
`
`(8)
`r = I
`d=l
`c=O
`
`0
`0.011
`0.027
`0.038
`0.054
`0.11
`
`0.0032
`0.0039
`0.0052
`0.0063
`0.0078
`0.015
`
`0
`0
`0
`0
`0
`0
`
`Table IV. Number of oocytes or embryos needed to have 95% probability of finding required number both usable and typed usable: autosomal recessive
`diseases
`
`r ':= 0 8, d = 0.8, c = 0.05
`Number required
`
`r = 0.9, d = 0.9, c = 0.05
`Number required
`
`r = I, d = I, c = 0
`Number required
`
`Blastomere
`Polar body, 0 = 0
`Polar body, 0 = 0.10
`Polar body, 8 = 0.20
`Polar body, 8 = 0.25
`Polar body, 0 = 0 30
`Polar body, 8 = 0.40
`
`5
`7
`9
`12
`15
`19
`39
`
`2
`
`R
`II
`14
`19
`24
`30
`61
`
`3
`
`10
`15
`19
`26
`32
`40
`82
`
`2
`
`6
`9
`12
`17
`20
`26
`53
`
`3
`
`8
`13
`16
`22
`27
`34
`71
`
`4
`6
`7
`10
`13
`16
`33
`
`2
`
`5
`8
`10
`14
`18
`22
`46
`
`3
`
`6
`11
`14
`19
`23
`30
`61
`
`3
`5
`6
`9
`11
`14
`29
`
`unacceptable (Table II), multiplying each by its probability of
`leading to the implantation of an unacceptable embryo (50 and
`25%), and summing the results. Table III shows that the greater
`the value of 0, the greater the frequency of unacceptable error.
`This is due to the fact that when the recombination fraction is
`large, a large proportion of oocytes do not have the usable
`genotype. Thus it will more often happen that a non-usable oocyte
`will mistakenly be typed as being usable, and less often happen
`that a usable oocyte will be correctly typed. It follows that when
`an oocyte is typed as being usable, it is more likely to be the
`result of a typing error.
`Table III shows that the primary cause of error in polar body
`typing is PCR inefficiency, except when the recombination
`fraction is quite small, in which case contamination and failure
`to place a cell in the reaction tube are the primary causes. The
`reason for this is as follows. As shown in Table II, potentially
`unacceptable results in polar body typing for recessive diseases
`come about when an oocyte of genotype aa or Aa is mistakenly
`typed as having genotype AA. In general, given reasonable values
`of r, d and c, it is much more common for an oocyte of genotype
`
`Aa to be mistyped as AA than for an oocyte of genotype aa to
`be so mistyped. Thus most unacceptable errors rc ult from the
`usc of an oocyte of genotype A a. When the recombination fraction
`is quite small, however, very few oocytes of genotype Aa exist,
`so mo ·t unacceptable errors result from usc of oocytes with
`genotype aa. A · shown in the second and fourth rows of Table
`Il, the frequency of pOLentially unacceptable err r
`involving
`mistyping oocytcs of genotype aa is determined primarily by the
`values of d and c, while the frequency of unacceptable errors
`involving mistyping oocytes of genotype Aa is determined
`the values of d and c exert
`primarily by the value of r. Thu
`primary influence over the frequency or unacccptnble errors when
`the recombination fraction is very small, while Lhe alue of r is
`most importanl otherwise.
`
`Comparing blastomere typing with polar body typing, autoso11Ull
`recessive disease
`Comparing the probabilities in the last two rows of Table I with
`the probabilities in Table III shows that for recessive diseases,
`
`839
`
`The Johns Hopkins University Exhibit JHU2009 - Page 4 of 14
`
`

`

`W.Nuvidi and N.Arnheim
`
`Table V. Errors in blastomere typing for autosomal dominant diseases
`
`Tnte
`genotype
`
`Observed Use? Error category Primary cause(s) Conditiona I probability of error
`genotype
`of error
`( I)
`r = 0.1!
`d = 0.8
`(' = 0 05
`
`(2)
`r = 0 8
`c/= 1
`(' = 0
`
`(3)
`r = 0. 9
`d = 0. 8
`(' = 0.05
`
`(4)
`r = 0.9
`d = 09
`c = 0.05
`
`(5)
`r = 0.9
`d=l
`(' = 0.05
`
`(6)
`r = 0 9
`d= I
`(' = 0
`
`(7)
`r = l
`d = 0.8
`(' = 0.05
`
`(8)
`r = I
`d = I
`(' = ()
`
`A a
`a a
`
`a a
`A a
`
`Yes
`No
`
`Unacceptable
`Tolerated
`
`r
`c
`
`0.15
`0.037
`
`0.14
`0
`
`0.088
`0.033
`
`0.086
`0.029
`
`0 .084
`0.027
`
`0.083
`0
`
`0.0063
`0.030
`
`[)
`0
`
`Table VI. Errors in polar body typing for autosomal dominant diseases
`
`Polar body
`
`True
`genotype
`
`AA
`AA
`A a
`A a
`a a
`aa
`
`Observed
`genotype
`
`A a
`a a
`AA
`a a
`AA
`A a
`
`Oocyte
`True
`genotype
`
`aa
`a a
`A a
`A a
`AA
`AA
`
`Deduced
`genotype
`
`A a
`AA
`a a
`AA
`a a
`A a
`
`and the same values for r, d and c, unacceptable results occur
`less frequently with blastomere typing than with polar body typing
`unless the recombination fraction is quite small , or unless both
`PCR efficiency and the contamination rate are rather high.
`Two factors are at work here, one favouring blastomere typing
`and one favouring polar body typing . Favouring blastomere
`typing is the fact that 75% of all embryos are usable, while the
`percentage of oocytes which are usable is 50% or less, depending
`on the recombination fraction between the centromere and the
`locus of interest. The factor favouring polar body typing is that
`selection of an oocyte which should not be used does not always
`result in an embryo of an unacceptable genotype. Fertilizing an
`oocyte of genotype Aa will yield an unacceptable embryo only
`25% of the time, and selecting an oocyte of genotype aa will
`yield an unacceptable embryo 50% of the time.
`When the recombination fraction is small, the percentage of
`usable oocytes is nearly 50%, and the two factors combine to
`provide a lower frequency of error for polar body typing. When
`the recombination fraction is larger than -0.1, and both the PCR
`efficiency r and the contamination rate c are both not high,
`blastomere typing has a lower error rate. If rand care both large,
`the error rate for polar body typing is lower. This is because
`large values of r tend to decrease the frequency of polar body
`typing errors, and large values of c tend to increase the frequency
`of blastomere typing .errors. Thus typing procedures with high
`levels of PCR efficiency and high contamination rates may be
`more accurate with polar hody typing, while under other
`conditions blastomere typing might be more accurate .
`
`Number of oocytes or embryos needed, autosomal recessive
`disease
`Table IV gives the number of oocytes or embryos which must
`be typed in order to have at least 95% probability of finding 1,
`2 or 3 which are usable and are typed as being usable. For
`example, when r = 0.9, d == 0.9 and c = 0.05, if six blastomeres
`
`840
`
`Use?
`
`Error category
`
`Primary cause(s)
`of error
`
`No
`No
`Yes
`No
`Yes
`No
`
`Tolerated
`Tolerated
`Potentially unacceptable
`Acceptable
`Unacceptable
`Acceptable
`
`c
`d,c
`
`d.c
`c
`
`are typed, the probability is at least 95% that at least two of the
`corresponding embryos will in fact be usable (i .e. of genotype
`AA or Aa) and will be typed as such. On the other hand, when
`the recombination fraction is 0.1, 12 oocytes must be typed to
`be equally confident of finding two usable ones typed as such.
`The reason that fewer blastomeres need to be typed is that 75%
`of oocytes are usable, while only 50% or fewer of oocytes are.
`When () is large, the proportion of oocytes which are usable is
`quite small. As shown in Table IV, one cannot be confident of
`finding a usable oocyte unless a very large number are available.
`This is a clear advantage for blastomere typing.
`
`Blastomere typing, autosomal dominant disease
`In the case of dominant disease, only embryos of genotype aa
`are usable. We assume that one parent is of genotype Aa, and
`the other is of genotype aa. An embryo has 50% probability of
`being of genotype aa and 50% probability of being of genotype
`Aa. Table V categorizes the two typing errors that are possible
`in this situation. The probabilities of error are noticeably greater
`than in the case of recessive disease, unless the PCR efficiency
`r is very high. This is partly because only 50% of embryos are
`usable, compared with 75% when the disease is recessive. Also,
`unacceptable errors when the disease is dominant are usually due
`to lack of PCR efficiency, while unacceptable errors when the
`disease is recessive usually involve contamination (cf. Tables I
`and V) . Unless the PCR efficiency is .very high, efficiency errors
`are more common than contamination errors.
`
`Polar body typitJg, autosomal dominant disease
`In the case of dominant disease, only oocytes of genotype aa are
`usable. The proportion of oocytes which are usable is 1h -0. We
`assume that the female is heterozygous and that the male is
`homozygous aa. Table VI categorizes the various typing errors
`in this case. Line 3 of Table VI indicates that using an oocyte
`of genotype Aa is 'potentially unacceptable'. This is because the
`
`The Johns Hopkins University Exhibit JHU2009 - Page 5 of 14
`
`

`

`U~ing PCR in preimplantation diagnosis
`
`Table VII. Conditional probabilities of unacceptable errors in polar body typing for autosomal dominant diseases
`
`Primary causc(s)
`of error
`
`Polar body, AA observed. () = 0
`Pola1 body. AA observed, () = 0.10
`Polar body, AA observed, () = 0.20
`Polar body, AA observed.() = 0.25
`Polar body, AA observed, 0 = 0.30
`Polar body, AA observed. 0 = 0.40
`
`d.c
`r
`r
`,.
`r
`
`(I)
`r = 0.8
`d=l
`c = 0.5
`
`O.<XJ76
`0.047
`0.099
`0.013
`0.18
`0.29
`
`(2)
`r "' 0.8
`d = I
`c· = 0
`
`(3)
`r = 0.9
`d = 0.8
`c = 0.5
`
`(4)
`r = 0.9
`d = 0.9
`c = 0.5
`
`0
`0.038
`0.091
`0.13
`0 .17
`0.29
`
`0.0066
`0 029
`0.063
`0 .086
`0.12
`0 .22
`
`0.031
`0 025
`0.058
`0.081
`0.11
`0.22
`
`(5)
`r = 0.9
`d=I
`c = 0.5
`
`0.00025
`0 .022
`0.054
`0.077
`0.11
`0.21
`
`(6)
`r = 0.9
`d=l
`c = 0
`
`(7)
`r =I
`d = 0.8
`c = 0.5
`
`(8)
`r = I
`d = I
`c = 0
`
`0
`0.022
`0.054
`0.77
`0.11
`0.21
`
`0.0063 .
`0.0079
`0 010
`0.013
`0.016
`0.030
`
`0
`0
`0
`0
`0
`0
`
`Table VIII. Number of oocytes or embryos needed to have 95% probability of finding required number both usable ami typed usable: autosomal dominant
`diseases
`
`r = O.R, d = 0.8. r = 0.05
`Number required
`
`Blastomere
`Polar body, 0 = 0
`Polar body, 8 = 0.10
`Polar body, 8 = 0.20
`Polar body, 0 = 0.25
`Polar body, 8 = 0.30
`Polar body, 0 = 0.40
`
`7
`7
`9
`12
`15
`19
`39
`
`2
`
`II
`II
`14
`19
`24
`30
`61
`
`3
`
`15
`15
`19
`26
`32
`40
`82
`
`r = 0.9, d = 0.9. d = 0.05
`Number required
`2
`
`6
`6
`7
`10
`13
`16
`33
`
`9
`9
`12
`17
`20
`26
`53
`
`3
`
`13
`13
`16
`22
`27
`34
`71
`
`r = l, d = I, c = 0
`Number required
`
`2
`
`8
`8
`10
`14
`18
`22
`46
`
`3
`
`II
`II
`14
`19
`23
`30
`61
`
`5
`5
`6
`9
`11
`14
`29
`
`Table IX, Conditional probabilities of unacceptable errors when two or three blastomeres are typed in one tube: autosomal diseases
`
`Recessive, AA observed, 2 cells
`Recessive, Aa observed, 2 cells
`Dominant, aa observed, 2 cells
`
`Recessive, AA observed, 3 cells
`Recessive, Aa observed, 3 cells
`Dominant, aa observed, 3 cells
`
`(I)
`r = 0 8
`d = 0.8
`(' = 0.05
`
`0.0013
`0.014
`0.075
`
`0.00030
`0.013
`0.034
`
`(2)
`r = 0.8
`d=l
`(' = 0
`
`0
`0
`0.037
`
`0
`0
`0.0079
`
`(3)
`r = 0.9
`d = 0.8
`c = 0 OS
`
`0.0011
`0.013
`0.036
`
`0 .00023
`0.013
`0.013
`
`(4)
`,. = 0.9
`d = 0.9
`c ~ 0.05
`
`0.00029
`0.013
`0.024
`
`0.00033
`0.012
`0.0055
`
`(5)
`r = 0.9
`d=!
`c = 0.05
`
`0.0000025
`0.012
`0.0098
`
`0.00000003
`0.012
`0.0010
`
`(6)
`r = 0.9
`d=1
`c=O
`
`0
`0
`0.0098
`
`0
`0
`0.0010
`
`(7)
`r= I
`d = 0.8
`c = 0.05
`
`0.0011
`0.012
`0 .0011
`
`0.00021
`0.012
`0.00021
`
`(8)
`r = I
`d =I
`c = 0
`
`0
`0
`0
`
`0
`0
`0
`
`use of such an oocyte yields the unacceptable result of an embryo
`carrying the A allele with 50% probability. Line 5 of the table
`indicates that using an oocyte of genotype AA is 'unacceptable'.
`Use of such an oocyte always results
`in an embryo of
`unacceptable genotype.
`Table VII gives conditional probabilities of unacceptable errors.
`The errors are exactly twice as large as in the recessive case.
`This is because while the proportion of oocytes which are usable,
`V2- 0, is the same for both dominant and recessive diseases, the
`frequencies with which oocytes with non-usable genotypes result
`in embryos with unacceptable genotypes are twice as great when
`the disease is dominant, i.e. 50 and 100% versus 25 and 50%.
`
`Comparing blastomere typing with polar body typing, autosomal
`dominant disease
`The two factors affecting the relationship between blastomere
`typing errors and polar body typing errors when the disease is
`recessive are relevant when the disease is dominant, but they are
`
`Table X. Number of embryos needed to nave 95% probability of finding
`required number both usable ami typed usaulc when tW(I or three
`blastomeres are typed in one tube: autosomal diseases
`
`r = 0.8
`d = 0 .8
`c == 0.05
`No. required
`
`2
`
`5
`9
`
`5
`8
`
`3
`
`7
`12
`
`7
`11
`
`3
`5
`3
`5
`
`r = 0.9
`d