1 Department of Reproductive Medicine, St Mary's Hospital, Manchester M13 OJH, 2 Department of Biology, University of York, P.O.Box 373, York Y010 5YW, 3 Manchester Fertility Services, Manchester BUPA Hospital, Russell House, Russell Road, Manchester M16 8AJ and 4 Biostatistics Group, School of Epidemiology and Health Sciences, University of Manchester, Manchester M13 9PT, UK
5 To whom correspondence should be addressed. Email: daniel.brison{at}man.ac.uk
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: amino acid/developmental potential/human preimplantation embryo selection/IVF
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
Embryo culture and selection
Seventeen to 18 h post insemination, a maximum of six normally fertilized pronucleate zygotes were selected for further culture, with any supernumerary zygotes cryopreserved. Zygotes were cultured individually in 4 µl drops of pre-equilibrated medium containing a physiological mixture of amino acids for 24 h (Houghton et al., 2002). Embryo-free control drops were also incubated alongside the embryo-containing drops to allow for any non-specific amino acid degradation/appearance. Embryos were transferred to and from the drops using a very fine bore pipette with the minimal transfer of medium (<100 nl). At the end of the 24 h period the culture dish containing the spent 4 µl drops and a number of control drops was immediately frozen at 20°C and shipped on dry ice (70°C) to the University of York for analysis. On day 2 after oocyte retrieval, two (exceptionally one or three) embryos were selected for transfer to the patient using conventional morphological criteria including the number of blastomeres, their regularity and extent of cellular fragmentation (Steer et al., 1992
). Embryo cell number varied from 2 to 6. Embryo morphological grade was scored from 1 for an embryo with irregular blastomeres and/or >50% fragmentation, to 4 for an embryo with regular blastomeres and no fragmentation. The product of these two yielded an embryo score, from 1 to 24. Embryos were selected for replacement based on this score, with the higher-scoring embryos replaced in preference to those with lower scores. Amino acid data were not available at the time of embryo replacement. Pregnancy following embryo transfer was detected by serum hCG 15 days post implantation and confirmed by the observation of a gestational sac and the presence of a fetal heart at 5 weeks.
Amino acid analysis
Embryo culture dishes described above were stored in York at 80°C until analysis. Following thawing, an aliquot (2 µl) was removed and diluted 1:12.5 in high performance liquid chromatography (HPLC) grade water. The amino acids were analysed by reverse-phase HPLC as previously described (Houghton et al., 2002). These data were then related retrospectively to the outcome of the treatment cycle.
Statistical analysis
In order to control for variability in the data, a two-step normalization procedure was utilized. Firstly, the concentration of each amino acid in each drop was expressed as a ratio relative to the total concentration of all 18 amino acids in the drop. Secondly, to correct for variations in the control amino acid concentrations, these ratios for embryo-containing drops were divided by the average of the ratios in the control drops cultured in the same dish.
All comparisons between pregnancy outcomes are based on recipient level data, using the mean values of the potential prognostic factors from all the transferred embryos. Simple comparisons between cycles yielding a pregnancy, and those which did not, were based on MannWhitney U-tests. We used logistic regression, stratifying by the number of embryos transferred, to model the probability of pregnancy and to investigate the association of amino acid measurements after allowing for the known predictors, embryo cell number and morphological grade. Correlations between predictors of pregnancy outcome were assessed using Spearman correlation coefficients. As the amino acid data consisted of 18 separate measurements, which were not statistically independent of each other, we utilized a principal components analysis to reduce the dimensionality of the dataset and provide an objective amalgamation of the amino acid data. In addition we recomputed the probability values for the individual amino acids using an adjustment for multiple testing. For the MannWhitney U-tests we used the step-up permutation-based test of Troendle (1995) which explicitly allows for the correlations between the tests. Ten thousand permutations were sampled. This was not practical for the logistic regression analyses so here we used the simpler, more conservative, Holm (1979)
test. For presentation of the logistic regression we have normalized the data by the central 90% range of the distribution of each parameter. This enables comparisons between endpoints to be made on the basis of the range of values observed in the population. Analysis of variance (ANOVA) and variance decomposition were used to assess the variance between and within patients, using the data on all embryos.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We analysed by HPLC the change in concentration over a 24 h period of 18 amino acids in the culture medium of individual human embryos and related these data retrospectively to the outcome of the treatment cycle. The entire dataset was analysed at the level of embryo transfer recipient (patient), with outcome (clinical pregnancy and live birth) analysed as a function of the transferred embryo properties for each patient. Embryo properties included amino acid parameters and the known outcome predictors of embryo cell number, morphological grade and their product, embryo score.
The relative concentrations of the 18 amino acids in the culture medium are shown in Figure 1, with the mean concentrations in those patients with and without a successful pregnancy outcome.
|
In general, the four twin pregnancies had values for the amino acid levels that were intermediate between that of the singleton pregnant and non-pregnant recipients, whereas we would expect a greater difference in mean levels where more than one of the transferred embryos implanted successfully (data not shown). However, there are too few such cases to draw any reliable conclusions from this observation.
Principal components analysis was then used to amalgamate the large number of highly correlated amino acid variables into a small number of components that have a degree of statistical independence and represent the majority of variability in the data. The first principal component (PC1) accounts for 28% of the variance in the amino acid data and the first two components (PC1 and PC2) together account for 45% of the variation. Figure 2 shows the relationship between first two principal components and pregnancy outcome, with PC1 being significantly lower in the patients with a successful outcome. Also shown in Figure 2 are the relationship between the morphological predictors and outcome, with embryo grade and the composite score (grade x cells) both showing a significant association.
|
There were weak, but significant, correlations between amino acid levels and age for just two amino acids (Glu: r=0.35, P=0.01; Arg: r=0.29, P=0.04; all others |r|<0.23), but given the large numbers of potential correlations these are likely to be spurious. There were no significant correlations between amino acid levels and basal FSH.
We used multiple logistic regression analysis to establish the independent association of amino acids after allowing for known predictive factors (Table II). In this analysis we stratified for the number of embryos transferred.
|
The extent to which amino acids varied between patients, rather than between embryos in a particular patient cycle (within patient), was then tested by one-way ANOVA using all 226 embryos. A formal variance decomposition showed that the SD for amino acid data within patients was generally much larger than that between patients (data not shown). For example, the principal component PC1 SD is 2.2 within patients compared to 0.4 between patients.
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Previously, we showed that amino acid turnover could predict the ability of early cleavage stage embryos to form a blastocyst (Houghton et al., 2002). We now extend this work to show, crucially, that amino acid turnover is significantly associated with the ability of an embryo to implant and give rise to a clinical pregnancy and live birth. This association is largely independent of other known indicators of pregnancy outcome such as maternal age, ovarian reserve (measured as basal FSH level), and embryo cell number and morphological grade. Moreover, the data are highly significant even though in most cycles we could not know which embryo of the two replaced had implanted. In cycles in which a single embryo is replaced, the data would be expected to be even more strongly predictive of live birth.
We do not know which aspects of embryo viability are reflected by amino acid turnover. Since our analysis was conducted on embryos of between 1 and 4 cells on day 12, this must reflect oocyte quality, since in the human the embryonic genome is not activated until after the 4-cell stage. Interestingly, the amino acids whose turnover predicted blastocyst formation in vitro (Houghton et al., 2002) are not the same (bar one) as those that predict pregnancy and live birth following transfer (present study). Blastocyst formation from human embryos developing in vitro was predicted by Ala, Arg, Gln, Met and Asn turnover between days 2 and 3 (Houghton et al., 2002
) whereas pregnancy was predicted by Asn, Gly and Leu between days 1 and 2 (present study). This may reflect the fact that not all blastocysts that form in vitro are viable. These differences in amino acids reaching statistical significance will, however, be in part attributable to stochastic variability between studies, and indeed, the overall profiles of amino acid depletion/appearance were qualitatively similar between the two studies. All amino acids which predicted blastocyst formation showed the same trend with respect to pregnancy in the present study; similarly the amino acids significantly predictive of pregnancy showed the same trend for blastocyst formation (Houghton et al., 2002
). Alternatively, amino acid turnover may vary with developmental stage, such that viability is reflected by a different subset on day 1/2 (present study) compared to day 2/3 of development (Houghton et al., 2002
). A direct comparison is further complicated by the fact that Houghton et al. (2002)
studied single embryos, whereas in our study we do not know which of the transferred embryos implanted. Clearly more data are required to establish a definitive comparison. However, the fact that a single summary parameter (PC1) showed significance, and that significance remained after adjustment for multiple testing, confirms with some degree of confidence the potentially predictive power of amino acid analysis.
The association of amino acid turnover with embryo viability is likely to be true in principle for a wide variety of culture conditions including media formulations. However, this has been demonstrated so far only with the medium of Houghton et al. (2002) and will require independent studies with other media since the particular subset of predictive amino acids may vary depending, for example, on the medium electrolyte and nutrient composition, and macromolecule supplements. In general, since amino acid turnover is independent of embryo morphological predictors, the prospective use of all these factors in combination should increase the implantation rate per embryo over and above what is currently achievable in any particular IVF programme. The predictive power of the method will vary with the number of embryos made available for selection, and with the number replaced. In our study, up to six embryos were available with generally two replaced, and the predictive power would be expected to be substantially higher if all embryos generated were profiled but only the best single embryo replaced. Currently it takes 45 mins per embryo to complete an amino acid profile; therefore embryos cultured from day 1 to day 2 could be replaced late on day 2, or on day 3, using prospective amino acid selection. We are currently working to reduce the length of time required by refining the HPLC technology to enable embryo transfers immediately following the period of profiling. The method may also be used to select embryos for cryopreservation, or for selection post-thaw in frozen embryo replacement cycles. We are currently extending our studies into this area.
Other methods of embryo selection require prolonged culture of embryos, for example to the blastocyst stage by day 5 or 6 after fertilization. While this can produce embryos of high viability for transfer (Gardner and Schoolcraft, 1999), using some culture systems it has also been associated with developmental abnormalities in animals (Young et al., 1998
; Ecker et al., 2004
). As a result, there may be an argument for limiting the length of time that human embryos are exposed to potentially adverse conditions in vitro. We have now shown that measurement of amino acid turnover could potentially increase significantly our ability to select the most viable embryo for transfer in clinical IVF, without the need for extended culture.
![]() |
Acknowledgements |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Ecker DJ, Stein P, Xu Z, Williams CJ, Kopf GS, Bilker WB, Abel T and Schultz RM (2004) Long-term effects of culture of preimplantation mouse embryos on behavior. Proc Natl Acad Sci USA 101, 15951600.
ESHRE Campus Course Report (2001) Prevention of twin pregnancies after IVF/ICSI by single embryo transfer. Hum Reprod 16, 790800.
Gardner DK and Leese HJ (2000) Assessment of embryo metabolism and viability. In Trounson AO and Gardner DK (eds) Handbook of In Vitro Fertilisation, 2nd edn. CRC Press, Boca Raton, pp. 347371.
Gardner DK and Schoolcraft WB (1999) Culture and transfer of human blastocysts. Curr Opin Obstet Gynecol 11, 307311.[CrossRef][ISI][Medline]
Hamberger L and Hazekamp J (2002) Towards single embryo transfer in IVF. J Reprod Immunol 55, 141148.[CrossRef][ISI][Medline]
Holm S (1979) A simple sequentially rejective multiple test procedure. Scand J Statist 6, 6570.[ISI]
Houghton FD, Hawkhead JA, Humpherson PG, Hogg JE, Balen AH, Rutherford AJ and Leese HJ (2002) Non-invasive amino acid turnover predicts human embryo developmental capacity. Hum Reprod 17, 9991005 [Corrigenda (2003) 18, 1756-1757].
Human Fertilisation and Embryology Authority (2002) Patients' Guide to IVF Clinics.
Lieberman B (1998) An embryo too many? Hum Reprod 13, 26642666.
Lieberman BA, Falconer D and Brison DR (2001) Presentation of in-vitro fertilisation results. Lancet 357, 397.[CrossRef]
Mugford M and Henderson J (1995) Resource implications of multiple births. In Ward HR and Whittle M (eds) Multiple Births RCOG, London, pp. 334345.
Racowsky C, Jackson KV, Cekleniak NA, Fox JH, Hornstein MD and Ginsburg ES (2000) The number of eight-cell embryos is a key determinant for selecting day 3 or day 5 transfer. Fertil Steril 73, 558564.[CrossRef][ISI][Medline]
Steer CV, Mills CL, Tan SL, Campbell S and Edwards RG (1992) The cumulative embryo score: a predictive embryo scoring technique to select the optimal number of embryos to transfer in an in-vitro fertilization and embryo transfer programme. Hum Reprod 7, 117119.[Abstract]
Templeton A (2000) Avoiding multiple pregnancies in ART: replace as many embryos as you likeone at a time. Hum Reprod 15, 1662.
Troendle JF (1995) A stepwise resampling method of multiple hypothesis testing. J Am Statist Assoc 90, 370378.[ISI]
Van Steirteghem AC, Liu J, Joris H, Nagy Z, Janssenswillen C, Tournaye H, Derde MP, Van Assche E and Devroey P (1993) Higher success rate by intracytoplasmic sperm injection than by subzonal insemination. Report of a second series of 300 consecutive treatment cycles. Hum Reprod 8, 10551060.[Abstract]
Young LE, Sinclair KD and Wilmut I (1998) Large offspring syndrome in cattle and sheep. Rev Reprod 3, 155163.
Submitted on June 8, 2004; accepted on June 16, 2004.