| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Biology, University of York, York YO10 5YW, United Kingdom
ABSTRACT
The determination for early cleavage-stage embryos of noninvasive morphologic and metabolic criteria that are predictive of blastocyst development and/or full-term viability remains an important research target. We describe the derivation of a logistic regression model that predicts the probability of porcine blastocyst formation in vitro. Pig zygotes, derived by in vitro maturation and fertilization of slaughterhouse oocytes, were cultured in NCSU-23 medium that was supplemented with a mixture of 20 amino acids (NCSU-23aa). On Day 1, at 21, 23, 25, 27, 29 and 31 h postinsemination, cleaving embryos were evaluated morphologically in terms of the: i) number of blastomeres, ii) evenness of division, and iii) degree of fragmentation. These embryos were then placed in 1.5-µl drops of NCSU-23aa for 24 h, after which time the three morphologic criteria were re-evaluated and 1.2 µl of spent medium were removed for analysis by HPLC, in order to determine the net rates of amino acid depletion and appearance. Embryos were then cultured singly in NCSU-23aa by placing them between the filaments of a woven polyester mesh until Day 6, in order to permit the identification of individual embryos. Of 256 cleaved embryos, 28.7 ± 6.2% (n = 5 replicates) developed into blastocysts. Discriminant analysis was used to select a subset of amino acids (threonine, valine, lysine, and phenylalanine) that discriminated optimally between embryos that became blastocysts or degenerated. These discriminant scores were entered into the logistic regression. Significant univariate relationships were established between the probability of blastocyst development and amino acid score (odds ratio [OR] 0.53, 95% confidence interval [CI] 0.40–0.69, P < 0.001), cleavage time (OR 0.79, 95% CI 0.71–0.87, P < 0.001), degree of fragmentation on Day 1 (OR 0.55, 95% CI 0.35–0.84, P = 0.009) and Day 2 (OR 0.53, 95% CI 0.35–0.78, P = 0.002), evenness of division on Day 2 (OR 0.66, 95% CI 0.46–0.96, P = 0.028), and categorical values of blastomere number on Day 2 (all P < 0.02), although no single variate could accurately predict blastocyst formation. However, multivariate analysis of the cell numbers on Day 1 and Day 2 correctly classified 51.9% of the predicted blastocysts. The inclusion of cleavage time in the regression analysis raised this rate to 63.5%, which was increased to 66.2% by the addition of evenness of division and degree of fragmentation. Finally, the full logistic regression model, which incorporated amino acid score together with all the other morphologic and kinetic variables, correctly classified 80.8% of the predicted blastocysts. This represented 51.2% of the observed blastocysts. Our data are novel in that they not only define in a quantitative manner the influence of previously undescribed predictors of porcine blastocyst formation, but they also provide a simple model of preimplantation development with reasonable predictive accuracy. The present study also provides a basic model for the examination and incorporation of additional early morphologic and metabolic correlates of developmental competence and could potentially be applied to the selection of human embryos for transfer in clinical IVF.
developmental biology, embryo, in vitro fertilization
The ability to predict reliably and noninvasively which early embryos have the ability to develop to full-term would constitute a major breakthrough in embryo-based biotechnologies. This would not only virtually eliminate multiple pregnancies following human assisted conception and the associated high rates of morbidity and mortality [1], but could, in a major way, benefit domestic animal embryology by reducing nonreturn rates and allowing predictive parameters to be used as criteria for the selection of superior culture media. Early identification of embryos that exhibit signs of continuing viability would also obviate the need for extended periods of in vitro culture, which may be associated with a gradual reduction in developmental capacity [2] and disruption of the genetic and epigenetic constitutions [3].
Since the early days of human IVF [4] and the in vitro production of cattle embryos [5], a positive correlation has been recognized between the speed of early embryo development and subsequent viability, although the earliest cleaving embryos are not always those that possess the greatest developmental capacity [6, 7]. Relationships between embryo cytokinetics and viability have also been established in the hamster [8], rhesus monkey [9], and mouse [7]. In addition to kinetic parameters, other easily quantifiable prognostic markers include the symmetry of cell division and the extent of cytoplasmic fragmentation [10]. Indeed, these factors, in association with blastomere number, are commonly used in human fertility clinics as indicators of viability [11]. With an increasing number of predictive factors available for the grading of embryos, it has become evident that rationalization of these parameters into a single viability score for individual human embryos is necessary [12, 13]. The concept of a viability index has subsequently been developed into more sophisticated scoring strategies, into which further determinants of viability, such as pronuclear number and morphology, multinucleation, and cytoplasmic granularity, have been incorporated [14, 15]. Correlates of viability are not limited to morphologic characteristics, since prognostic markers can also include embryonic metabolism or the release of growth factors [16, 17]. One particular facet of metabolism that shows promise in this respect is the quantification of the net rates of amino acid depletion and appearance by the embryo. Application of this technique to early human embryos has established that the release or uptake pattern of particular amino acids correlates with both embryo viability in vitro [18] and pregnancy rates following transfer [19].
Although cumulative embryo scoring is an established if not universally applied method in human fertility clinics, the evaluation of the contributory weightings of individual morphologic features to developmental success by logistic regression has never been applied to the embryos of domestic species. Furthermore, in the latter group, analyses of the net rates of amino acid depletion and appearance as discriminatory markers of in vitro developmental potential have not been performed. In addition, information on the relationship between the timing of the first cleavage and blastocyst development is not available for porcine embryos. Consequently, we have investigated the influence of time of cleavage, symmetry of division, degree of fragmentation, and net rates of amino acid depletion and appearance in early in vitro-produced pig embryos, in order to determine the value of each of these noninvasive factors as prognostic indicators of blastocyst formation. The present study also presents a novel culture system that consists of a monofilament mesh and permits the localization and identification of individual embryos throughout in vitro culture periods, while retaining the developmental benefits of group culture. The mesh culture system allows neighboring embryos to be separated from one another by a distance that has previously been determined to generate maximal blastocyst yields [20].
Cumulus-oocyte complexes (COCs) were aspirated from 2–6-mm-diameter follicles from slaughterhouse-derived ovaries taken from prepubertal pigs. The COCs were selected for the presence of an intact and compact cumulus investment that was several cellular layers deep and a homogenous ooplasm. The COCs were matured in groups of 50 in 100 µl of TCM-199 that was supplemented with 0.1% (w/v) PVA, 2.8 mM glucose, 0.68 mM glutamine, 0.91 mM pyruvate, 0.57 mM cysteine, 10 ng/ml murine epidermal growth factor, 0.5 µg/ml FSH, and 0.5 µg/ml LH. After 44 h of IVM, COCs were washed three times with mTBM [21] that contained 1.5 mM caffeine and transferred in groups of 35 COCs to 100 µl mTBM. Frozen-thawed semen from a single boar (kindly provided by GTC Scotland, PIC Sygen, UK), was overlaid on a two-layer (90%/45%) Percoll (Pharmacia, Uppsala, Sweden) gradient. After centrifugation at 700 x g for 30 min, the pellet was resuspended in 4 ml mTBM. Following recentrifugation at 350 x g for 5 min, the pellet was diluted in mTBM and used to inseminate the oocytes at a final concentration of 6 x 104 spermatozoa/ml. The day of fertilization was designated as Day 0. Six hours after insemination, cumulus cells from the presumptive zygotes were removed by vortexing in modified NCSU-23 [22], designated NCSU-23aa, which contained 20 amino acids [23], the concentrations of which were based on those measured in tubal fluid produced during vascular perfusion of the human Fallopian tube using TCM-199 supplemented with 4% BSA [24]. The embryos were then washed twice in fresh medium before being placed in groups of 20 in 20 µl NCSU-23aa.
On Day 1, at 21, 23, 25, 27, 29, and 31 h postinsemination, embryos that had cleaved to the 2-cell stage, as well as those that had divided into three cells or even attained the 4- or 5–8 cell stage between the 2-h observation time-points, were removed from the culture dishes. These embryos were washed five times in 100-µl drops of NCSU-23aa before being transferred individually in a minimal volume using a narrow-bore glass capillary (approximately equivalent to the embryo diameter) to 1.5-µl drops of NCSU-23aa under oil in Falcon Petri dishes, where they remained for the 24-h amino acid profiling procedure. The glutamine concentration of the NCSU-23aa in the drops during this 24-h incubation period was reduced to 0.2 mM, in order to improve the sensitivity of detection of this amino acid by HPLC. Prior to amino acid profiling, the embryos were evaluated morphologically in terms of the: i) number of blastomeres, ii) evenness of division, and iii) degree of fragmentation, so that each embryo received a three integer code according to a scoring scheme modified from that of Ziebe et al. [10]. Our scoring system is illustrated in Figure 1. For example, an evenly cleaved 2-cell embryo with no fragmentation received a score of 2,2,0, while a 4-cell embryo with slightly unevenly cleaved blastomeres and approximately 15% fragmentation was scored as 4,3,2. Estimates of evenness of division and degree of fragmentation were quantified by visual appraisal. The term ‘cyto-numerically deviant' was applied to embryos that did not conform to the pattern of cleavage that is believed to represent normality, i.e., porcine embryos would be expected to cleave to 2-cells on Day 1 and be at the 4-cell stage by Day 2 [25, 26].
|
After 24 h (Day 2), the medium within the 1.5-µl drop was gently mixed, and the embryo was removed using a narrow-bore glass capillary in a manner similar to that described for its addition, in order to minimize volume changes. Control 1.5-µl drops were located alongside those that contained the embryos, to control for nonspecific changes in amino acid depletion and appearance during the incubation and sample storage periods (e.g., the breakdown of amino acids to ammonium). The drop dishes were frozen at –80°C until amino acid analysis by HPLC. Immediately prior to recovery from the 1.5-µl drops, the embryos were re-evaluated in terms of the three morphologic criteria (Fig. 1), such that each embryo ultimately received two embryo scores in terms of three integers: one before (Day 1) and one immediately at the end of the 24-h incubation period (Day 2).
Culture and Localization of Individual Porcine Embryos
Recovered embryos were placed individually between the filaments of a piece (approximately 2 mm x 2 mm) of nontoxic woven polyester mesh (SefarPetex; Sefar, Switzerland) in a grid fashion in groups of approximately 16 in 20 µl NCSU-23aa (Fig. 2). This novel system allowed the identification of individual embryos and permitted a rate of blastocyst development equivalent to that attained in our conventional culture system of growing embryos in groups of 20 in 20 µl of medium. The size of the mesh opening (160 µm) was sufficient for the embryos to lie snugly between the filaments, while the filament diameter (84 µm) was within the optimal distance range (81–160 µm) for neighboring embryos to benefit from autocrine/paracrine growth effects [20, 27]. On Day 6, the embryos were classified according to whether they had developed to the blastocyst stage or had degenerated.
|
All chemicals were supplied by Sigma Chemical Co. (St. Louis, MO), unless otherwise stated. IVM, IVF, and IVC were all performed in pre-equilibrated drops of media under mineral oil in a humidified atmosphere of 5% CO2 in air at 39°C in Petri dishes (Falcon, Becton Dickinson, NJ).
An estimate of the degree of polyspermia was carried out 12 h after the beginning of IVF, using 
35 zygotes/replicate. Zygotes were fixed in acetic acid:ethanol (1:3) under coverslips for 5 days, stained with 1% (wt/vol) orcein, and observed under a phase-contrast microscope.
Analyses of Amino Acid Depletion and Appearance
From each 1.5-µl drop, 1.2-µl aliquots were removed and diluted in 23.8 µl purified water (ELGA Purelab; Elga, UK) in HPLC tubes. Any variations in pipetting during sample recovery were annulled by mathematical correction in relation to the assayed amount of the nonmetabolizable amino acid D-
-amino-n-butyric acid added to the NCSU-23aa.
A reverse-phase HPLC analytical technique was employed, as previously described [28], with minor changes. Briefly, the amino acids were derivatized to fluorescent products by automated reaction of the sample with an equal volume of o-phthaldialdehyde (OPA) that contained 2 µl/ml 2-mercaptoethanol. The Waters 2695 Alliance HPLC system was linked to a Waters 2475 fluorescence detector. The flow rate through the Phenomenex Gemini 5-µm (4.6 x 100 mm) column (Phenomenex, Cheshire, UK) was 1.3 ml/min, with the column temperature controlled at 35°C. The solvents required to generate the elution gradient contained 1:4 and 4:1 (v/v) ratios of methanol:sodium acetate (83 mM, pH 5.9), respectively.
The chemistry of the HPLC method did not permit the detection of cysteine and proline and consequently, 18 amino acids were measurable using this technique. Furthermore, due to their very high concentrations in NCSU-23aa, the levels of taurine (7 mM) and hypotaurine (5 mM) exceeded the upper detectable limits of the assay.
The data are presented as mean ± SEM, and differences between groups were assumed to be significantly different at a level of P 
0.05, unless stated otherwise. The normality of the data sets was evaluated by the Kolmogorov-Smirnov test. Multicollinearity was also investigated; any tolerance values equal to or less than 0.2 were identified, and any condition indices equal to or greater than 15 were examined, followed by comparison of coefficient decomposition between variables. The blastocyst rates between the groups were analyzed by 
2 tests. The statistical analyses were performed using SPSS.
The net rates of amino acid or nitrogen appearance and depletion were calculated as pmol (embryo)–1 h–1. It should be appreciated that the net rates of amino acid depletion and appearance represent the differences in absolute rates of uptake and release, which theoretically could greatly exceed the net rates of depletion and appearance being observed.
The net rates of nitrogen appearance and depletion were determined by multiplying the amino acid net rates of appearance or depletion by the number of nitrogen atoms contained in the respective amino acids [23].
In each group of embryos, total net amino acid appearance and total net amino acid depletion were calculated as the sum of the net rates of all those amino acids being released and disappearing, respectively. In addition, total amino acid balance and total amino acid turnover represented the difference between, and sum of, the total net amino acid appearance and depletion, respectively. The rates of total nitrogen appearance, depletion, balance, and turnover were calculated in a corresponding manner. It should be recognized that the ‘total' rates calculated here are underestimates of the true rates of total amino acid and nitrogen depletion and appearance, since: i) cysteine and proline were not measured, and ii) amino acids may be released intracellularly from endocytosed BSA contained in the medium [29, 30].
Those sets that conformed to the null hypothesis were analyzed by one-way analysis ANOVA. Statistical significance (P < 0.05) was investigated by the Fisher least-squares difference test. Nonparametric datasets were analyzed by the Mann-Whitney U-test. The effect of time of first cleavage on blastocyst rates was analyzed by logistic regression.
Differences from zero in the rates of appearance, depletion, balance, and turnover of amino acids or nitrogen were determined by the Wilcoxon signed ranks test.
Discriminant analysis was applied to the amino acid data to maximize the variance between a priori defined groups and to isolate a subset of amino acids that, in combination, provided the greatest discrimination. The assumption of equal covariance between groups was checked by the Box M test, while any violations of multivariate normality were corrected by appropriate transformation of the data. The discriminant function was derived by both stepwise estimation using Mahalanobis D2 measures and simultaneous estimation. Structure matrix loadings of approximately ±0.3 or greater were considered significant unless collinearity between any such variables reduced the discriminatory power [31]. Discriminatory scores derived by the selected function were included as an independent variable in the logistic regression analysis.
Determination of the degree by which variables measured during early cleavage could be predictive of blastocyst development was analyzed by logistic regression [32]. The dichotomous dependent variable registered (i.e., blastocyst formation or degeneration) was coded as 1 (for blastocyst development) or 0 (for arrested development/degeneration). The independent variable predictors registered on Day 1 and Day 2 were blastomere number, evenness of division, and degree of fragmentation. Time of first cleavage and discriminant analysis amino acid score were also entered as independent variables. The ordinal coding for evenness of division is detailed in Figure 1. Coding for degree of fragmentation for the logistic regression was altered from 0, 1, and 2 (see Fig. 1) to 1, 2, and 3, respectively. Dummy variables [33] were created for the three categories of blastomere number; the names of the dummy variables and their coding were as follows: Day 1 @ 2-cell (1 if 2-cell, 0 otherwise), Day 1 @ 3-cell (1 if 3-cell, 0 otherwise), and Day 1 @ 4-cell (1 if 4-cell, 0 otherwise). According to this categorization system, 5–8-cell embryos would be coded 0. Blastomere numbers of embryos on Day 2 were also converted to dummy variables using the identical coding scheme. These variables were nominated as Day 2 @ 2-cell, Day 2 @ 3-cell, and Day 2 @ 4-cell.
Inclusion of predictor variables in the fitted models was achieved by forward and backward stepwise logistic regression plus judicial selection or elimination of variables. The level of significance defined as a threshold for inclusion was changed from P 0.05 to the more liberal P 0.1, to avoid the exclusion of potentially important variables [34]. Measures of predictive efficiency of the models are presented as the percentage of embryos correctly classified overall, the percentage of blastocysts categorized as a proportion of the number predicted, and the added value (also known as Goodman and Kruskal 
) of the model. The latter is the percentage of correctly classified embryos beyond the number that could be correctly predicted simply by choosing the category of the largest size, i.e., since degenerate embryos represent 74% (depending on the model) of all embryos, chance alone can achieve this same level of predictive accuracy. Outlying observations were identified by standard criteria, as suggested by Hair et al. [31].
Fertilization and Polyspermy Rates
Orcein staining of samples of zygotes 12 h after insemination indicated that the penetration rate, the percentage of polyspermic oocytes, and the number of spermatozoa per penetrated oocyte were 55.7 ± 6.6%, 27.4 ± 8.7%, and 1.4 ± 0.2%, respectively.
Five replicate experiments were undertaken using a total of 356 embryos, of which 28.70 ± 6.19% developed into blastocysts. The cleaved embryos recovered on Day 1 were at the 2-, 3-, 4-, and 5–8-cell stages of development. Since all cleaved embryos were removed at 2-h intervals, those embryos recovered at the 4- and 5–8-cell stages had divided rapidly from 1-cell zygotes within these 2-h periods. The progressive development of cleaved embryos between Day 1 and Day 2 is depicted in Figure 3. The blastocyst rates of embryos isolated on Day 1 at the 2-, 3-, 4-, or 5–8-cell stage were not significantly different. However, significantly different (P < 0.001) blastocyst rates were produced from embryos that were at the 2-cell (6.4%), 3-cell (2.4%), 4-cell (47.9%), or 5–8 cell (30.8%) stage on Day 2. The 2-cell embryos collected on Day 1 produced the highest number of blastocysts, although 4-cell embryos isolated on the same day generated the highest percentage of blastocysts. The 2-cell embryos on Day 1 that developed into 4-cell embryos on Day 2 produced the highest number of blastocysts. It is noteworthy that some embryos that were evaluated morphologically on Day 2 possessed fewer cells than had been recorded on Day 1. As the 2-cell embryos (on Day 1) generated the highest number of blastocysts, the morphologic characteristics of these developing embryos on both Days 1 and 2, in terms of blastomere numbers, evenness of cleavage, and degree of fragmentation, are summarized (Fig. 4). The data indicate that the sequential scores of embryos that developed from Day 1 to Day 2 and that produced the first-, second-, and third-highest number of blastocysts were 2,2,0
4,2,0, 2,3,04,3,0, and 2,2,14,2,0, respectively. The logistic coefficients for the categorical values of cell number (vide infra) were significant on Day 2 (all P < 0.02) but not on Day 1 when examined individually (Table 1), although none of these variables in isolation could correctly classify any blastocysts (Table 2).
|
|
|
|
Evenness of Division and Degree of Fragmentation
The relationships between (i) evenness of cellular division, and (ii) degree of fragmentation on Day 1 and Day 2, and blastocyst yield are illustrated in Figure 5. On Day 1, three categories of embryo were analyzed for the effects of these morphologic parameters: a) all embryos that had cleaved on Day 1 regardless of cell number, b) those embryos that had cleaved to two cells (and were tentatively designated as presumptive cyto-numerically normal) and, c) embryos that possessed more than two cells (and could be possibly regarded as presumptive cyto-numerically precocious). Three classes of embryo were also analyzed on Day 2: a) all embryos, b) those that had cleaved to two cells on Day 1 and subsequently developed into 4-cell embryos by Day 2 (presumptive cyto-numerically normal), and c) embryos that comprised fewer or more than four cells on Day 2 (presumptive cyto-numerically deviant). Examination of the data for evenness of division suggests that the relationship between this morphologic parameter of embryo quality and blastocyst yield is not linear; rather, that a slight asymmetry of division (score level 3) is not detrimental to blastocyst yield, whereas severe asymmetry (score level 4) hinders subsequent development. This relationship was consistent on Day 1 and Day 2 for all the embryo categories portrayed in Figure 5, except for embryos that had three or more cells on Day 1. In contrast, linearity of response between the degree of fragmentation and blastocyst yield was clearly established for all classes of cleaved embryos on Day 1 (Fig. 5). This relationship was expressed similarly in 4-cell embryos on Day 2 that had developed from the 2-cell embryos on Day 1 (presumptive cyto-numerically normal embryos), was weaker in the group that comprised all cleaving embryos on Day 2, and was negligible in embryos that possessed three or fewer cells or five or more cells on Day 2. The data from all the embryos regarding evenness of cell division and degree of fragmentation were also analyzed by logistic regression. This established that the negative slopes of the individual logistic regression equations were significant for degree of fragmentation on Day 1 and Day 2 (both P < 0.01) and for evenness of division on Day 2 (P = 0.028) but not on Day 1 (Table 1), although none of these parameters per se could predict blastocyst development (data not shown).
|
Timing of Cleavage and Blastocyst Rates
An inverse relationship was recorded between the timing of the first division of all cleaved embryos and subsequent blastocyst yields (Table 3); a relationship that was similarly apparent on examination of those embryos that had divided to two cells only (presumptive cyto-numerically normal) (P < 0.001) or to three or more cells (presumptive cyto-numerically deviant embryos) (P = 0.011). In concordance with this relationship, a significant (P < 0.001) negative logistic regression coefficient was recorded when this parameter was examined in isolation (Table 1), although this model was not adequate to classify accurately any blastocysts (data not shown). The blastocyst yields at each time-point, expressed as percentages of the total number of blastocysts produced, were 23.3 ± 5.4%, 33.1 ± 4.2%, 25.9 ± 5.6%, 9.2 ± 2.1%, 7.4 ± 2.9%, and 1.2 ± 1.1% at 21, 23, 25, 27, 29, and 31 h postinsemination, respectively. Consequently, 56% and 82% of the total blastocyst yields were generated by the second and third collection times, i.e., by 23 h and 25 h postinsemination, respectively. The numbers of cleaved embryos at each collection time, expressed as the percentages of the total number of embryos collected, were 12.1%, 25.8%, 28.7%, 15.7%, 12.1%, and 5.6%, respectively, indicating that the frequency of cleavage was substantially higher after 21 h and up until 25 h postinsemination compared to the periods outside of this range.
|
The effects of cleavage time on the evenness of division and degree of fragmentation were also examined. Embryos were divided into fast and slow cleavers based on those cleaving in two equal 3-h periods, namely, between 21–25 h and between 27–31 h postinsemination, respectively. Examination of all embryos on Day 1 (i.e., regardless of whether they subsequently became blastocysts or degenerated) established that the distributions of embryos according to the three classes of degree of fragmentation and between fast and slow cleavers were significantly different (P < 0.01) (Fig. 6B), in that slower cleaving embryos exhibited greater degrees of fragmentation than those cleaving within 21–25 h postinsemination, causing the distribution of these embryos to be shifted to the right in Figure 6. This difference in distribution between fast and slow cleavers was not evident when the degree of fragmentation was examined on Day 2 (data not shown). However, if only those embryos that developed to blastocysts were included in the analysis, different distributions in degree of fragmentation between fast and slow cleavers were established on Day 1 (P < 0.01) and Day 2 (P < 0.05) (data not shown), which were similar in profile to those shown in Figure 6B. The distributions of embryos in terms of evenness of division between fast and slow cleavers were also investigated. Examination of all embryos (those subsequently developing to blastocysts or degenerating) indicated a difference in distribution in evenness of division between fast and slow cleavers on Day 1 (P < 0.01) but not on Day 2 (data not shown). This difference in distribution was observed as a shift from score 2 to score 3 in slower cleaving embryos, thereby representing an increase in asymmetry of division. Examination of only those embryos that developed to blastocysts indicated no differences in distribution when evenness of division was evaluated on Day 1 and Day 2 (data not shown).
|
The net rates of appearance or depletion of amino acids, together with the total net rates of appearance, depletion, balance, and turnover of amino acids and nitrogen, by fast and slow cleavers are presented in Table 4. All the rates were significantly different (P < 0.001) from zero. The net rate of appearance of methionine was lower (P < 0.001), the net rate of appearance of asparagine was higher (P = 0.004), and the net rate of depletion of arginine was higher (P = 0.04) in fast cleavers than in slow cleavers. Overall, the total net rates of depletion (P = 0.013) and turnover (P = 0.027) of amino acids were higher in those embryos that cleaved before 25 h postinsemination. Similarly, total net nitrogen depletion (P < 0.001) and turnover (P = 0.005) were higher in fast cleaving embryos. The higher total net depletion rate of nitrogen by fast cleaving embryos produced a larger balance (P = 0.024) in total net nitrogen between these and the slower dividing embryos.
|
The net rates of appearance or depletion of amino acids between those that developed to blastocysts and those that degenerated or arrested are illustrated in Figure 7. Embryos that progressed to blastocysts expressed higher net rates of appearance of glycine (P = 0.0228) and depletion of threonine (P < 0.001) than those that degenerated. In contrast, lower net rates of isoleucine (P = 0.0087), valine (P = 0.0322), and lysine (P = 0.0402) were recorded in the embryos that achieved blastocyst development than in those embryos that degenerated. Higher total net rates of amino acid depletion (P = 0.0086) were observed in those embryos that developed to blastocysts, causing a greater total net balance (P = 0.0432) in amino acids in these embryos (Fig. 8). These differences were not reflected in the total net rates of nitrogen depletion (–4.57 ± 0.25 vs. –4.31 ± 0.15) and balance (–1.79 ± 0.13 vs. –1.52 ± 0.11) nor in the appearance (2.78 ± 0.15 vs. 2.79 ± 0.09) or turnover (7.35 ± 0.38 vs. 7.10 ± 0.21) between developing and degenerating embryos, respectively.
|
|
Stepwise discriminant analysis identified threonine, valine, isoleucine, lysine, leucine, and phenylalanine as possessing high structure matrix loadings. However, isoleucine and leucine were excluded from the function, since they exhibited collinearity with the amino acids already included, thereby reducing the discriminatory power. The discriminant scores on their own yielded a significant logistic regression (P < 0.001) (Table 1), although it could identify correctly only 6.9% of the observed blastocysts (Table 2). Furthermore, the reduction in the percentage of correctly predicted degenerate embryos reduced the added value to below that of chance.
In addition to the logistic regression analysis of potential predictor variables examined univariately (Table 1), combinations of variables were selected to determine their predictive accuracy as presented in the classification matrices (Table 2). The use of blastomere numbers on Day 1 or Day 2 or of blastomere numbers on Day 1 in combination with the simultaneous determination of evenness of division and degree of fragmentation was of no value in terms of defining determinants of subsequent blastocyst formation. Associating blastomere numbers on Day 1 with the time that the embryos cleaved correctly identified 5.4% of blastocysts but simultaneously misclassified 2.3% of the degenerate embryos, thereby reducing the success rate of prediction to below that of chance. Incorporating cleavage time and blastomere numbers on Day 1 with both evenness of division and degree of fragmentation on Day 1 did not improve the success rate of prediction, which indicates that combinations of these three estimates of morphology and cleavage time on Day 1 cannot be used to predict blastocyst development. In contrast, evaluation of morphologic parameters on Day 2, in the absence of the cleavage time data, classified correctly 49.4% of the observed blastocysts, representing 56.4% of the predicted blastocysts. Using this model, an 11.2% increase in added value over chance was achieved. In contrast to the lack of predictive accuracy of utilizing blastomere numbers on Day 1 or Day 2 in isolation, combining these categorical variables in the logistic regression successfully identified over 60% of the observed blastocysts, although the substantial (19.2%) misclassification of degenerate embryos only raised the added value to 4.5% above chance. A more successful model involved adding cleavage time to the cell number data recorded on Day 1 and Day 2, whereby 63.5% of the predicted embryos became blastocysts, representing a 19.8% increase in added value. Further incorporation of evenness of division and degree of fragmentation of embryos on Day 1 and Day 2 into the latter model raised the percentage of blastocysts out of those predicted to 66.2% and the percentage of correctly classified embryos above that attainable by random allocation to 27.3%. Inclusion of the amino acid scores in the analysis to form the final model (Pseudo R2 = 0.266) also augmented the success rate, such that 80.8% of the predicted embryos were correctly classified as blastocysts. Furthermore, 51.2% of the observed blastocysts were successfully categorized in this model, generating an overall correct classification rate for the prediction of both classes of embryo of 84.2%, which was 39% higher than the number expected by chance. The variables and interaction terms included in this full model are listed in Table 5.
|
The frequency distributions of blastocysts and degenerate embryos in relation to the probabilities predicted by the full logistic regression model are depicted in Figure 9. The embryo categories possess two distinct distributions. Degenerate embryos exhibit a positive skewness, having a median value of 0.11, which indicates the reasonable predictive accuracy of this embryo category (95.7%). In contrast, blastocysts exhibit a flat distribution with predicted probabilities that range from 0.02 to 0.92 (median, 0.51), indicative of poor predictive accuracy (51.2%) but, as already stated, this is sufficient to identify correctly over 80% of the predicted blastocysts. Despite the latter level of accuracy, selection criteria that maximize the numbers of observed blastocysts out of those predicted are relevant for embryo selection purposes. To this end, the results shown in Figure 10 suggest that increasing the probability threshold beyond which embryos are selected to more than the normal level of 0.5 has little effect on the percentage of observed blastocysts out of those predicted (see also Fig. 9), since the distributions of the degenerate embryos and those that form blastocysts are relatively flat between the probability values of 0.5 and 0.85. Moreover, raising the probability threshold merely reduces the number of observed blastocysts in the predicted sample. A homogeneic sample of blastocysts could only be realized using a probability threshold of 0.9, at which level only 2.4% of the total number of observed blastocysts would be isolated.
|
|
Receiver Operating Characteristic (ROC) Curve
A ROC curve for the fitted model is depicted in Figure 11. This shows the fraction of true positive results as a function of those that are false positives. The area under the curve is a global measure of the predictive accuracy of the model, which in this instance is 0.839 (95% CI 0.78–0.89; P = 0.001), which thereby rejects the null hypothesis and suggests that the predictive accuracy of the model is good. The probability value at which the sum of the sensitivity and specificity was maximized was 0.33, corresponding to sensitivity and 1-specificity of 0.72 and 0.19, respectively (Fig. 11). Using this probability as a threshold in Figure 10, the positive predictive values for blastocyst formation and degeneracy are 70.7% and 81.6%, respectively. However, information derived from Figure 10 suggests that a more pragmatic cut-off point is 0.5 (vide supra).
|
Probability Estimates Using The Full Logistic Regression Model
Probability estimates of selected combinations of predictor variables were calculated (Table 6) using the full logistic regression model (Table 5) according to the formula:
|
| 77 |
|
and
The selected predictor variables illustrated in Table 6 comprise the cell numbers on Days 1 and 2, evenness of division on Day 2, cleavage time postinsemination, and amino acid scores. Fast cleaving and slow cleaving embryos represent those embryos that cleaved at 21 h and 27 h postinsemination, respectively. The low and high amino acid scores are derived from the discriminant function analysis and represent the values of the first and third quartile of the score distribution, such that embryos that exhibit low amino acid scores have a higher probability of developing to blastocysts. The values for evenness of cell division and degree of fragmentation on Day 1 are classified in Table 6 as 2 and 0, respectively, while the degree of fragmentation on Day 2 was classified as 0. The estimated probability values equal to or greater than 0.5 that are underlined in Table 6 represent combinations of variables that are likely to be conducive to blastocyst development. Examination of the data in Table 6 indicates that the estimated probability values in relation to sequential cell numbers on Day 1 and Day 2 generally vary in good agreement to those observed (see Fig. 3). Compliance between these estimated probability values is particularly good for embryos that developed from the 2-cell stage on Day 1, which represented the majority (69.7%) of embryos. Evenness of division was featured in the interaction terms associated with 2- and 4-cell embryos on Day 2 and accordingly, reduced probability values were estimated for such embryos that exhibited asymmetry of division (see also Fig. 5 and Table 1). The negative relationship between cleavage time and probability of blastocyst development (see Tables 1 and 3) was reproduced in the full model and is illustrated in Table 6 as the reduced estimated probabilities of embryos that cleaved at 27 h compared to those that divided at 21 h postinsemination. The use of the amino acid score as another contributory variable to predict blastocyst development (Table 1) is also faithfully incorporated into the full model, whereby lower amino acid scores derived from the discriminant analysis generate higher estimated probability values.
A reliable noninvasive embryo scoring system undertaken during early cleavage, to predict accurately which embryos are either capable of reaching the blastocyst stage or possess full developmental potential, would be of considerable benefit for the advancement of embryo-based biotechnologies in domestic animals and assisted conception in humans. This is particularly the case for the embryos of domestic animals, specifically those of pigs and cattle, which present special challenges for noninvasive morphologic investigations owing to the large amount of cytoplasmic lipid [35], which effectively precludes the observation of nuclei and hinders or prevents the characterization of cytoplasmic structures, rendering apparently useful markers of viability in human embryos [36] inapplicable to the pig and cow [37]. Faced with these restrictions, and in the absence of further porcine-specific markers of viability, the simplest system by which to gauge embryo quality in the pig would appear to rest on the symmetry of division, degree of fragmentation, and time of cleavage, all of which are quality factors that have been borrowed from other species and have rarely been investigated in the pig [26], especially for embryos that are produced entirely in vitro.
There is increasing awareness that the timing of certain events during the earliest stages of development may be linked to subsequent embryonic viability [38]. One such association, evident in a variety of species, is the inverse relationship, as observed here, between the interval between insemination and first cleavage and morula/blastocyst formation [7, 39–41] or one in which either precocious or delayed initial division equates to reduced developmental capacity in terms of blastocyst formation [7, 42]. To the best of our knowledge, this is the first report to document this interrelationship in pig embryos derived by in vitro fertilization, and is consistent with the recent report that the timely attainment of specific stages of porcine embryonic development is related to blastocyst formation in embryos fertilized in vivo [26]. This developmental superiority established by early cleavage is clearly profound, since it is also predictive of pregnancy and birth following transfer of single human embryos [43–45] and, as can be concluded from the fertilization of human oocytes by intracytoplasmic sperm injection, is apparently independent of variations in the timing of gamete fusion [43]. The latter may not be irrelevant, as the kinetics of nuclear maturation in pig oocytes in vitro are notoriously asynchronous [46], a factor that could broaden the time window over which gamete fusion occurs and, at the same time, influences development, since, in cattle, a link between early polar body extrusion and superior in vitro embryonic growth has been reported [47].
A convincing explanation for our observation that slower cleaving embryos exhibit greater degrees of asymmetry and fragmentation remains elusive, although cytoplasmic maturity [48–52] and chromosomal abnormalities [53, 54] secondary to polyspermia [55] may be involved. Van Soom et al. [42] have similarly reported that fast cleaving bovine embryos contain fewer fragments, albeit at the morula stage, and early cleaving human embryos are superior in quality (as determined by cell number and other morphologic features) compared to those that are retarded [43–45, 53].
Together with the speed of development, both the symmetry of cleavage and the degree of fragmentation are morphologic features that are recognized as determinants of viability [10, 11, 56]. Even a slight unevenness of division can substantially compromise the viability of human embryos [10, 11, 53], with greater extremes of asymmetry further reducing posttransfer rates of developmental success [11]. Albeit when measuring a different endpoint, namely, the rates of blastocyst formation, the pig embryos described in the present study were apparently resistant to minor deviations in cleavage symmetry, although unsurprisingly, they were developmentally compromised by greater extremes of unevenness. The amount of fragmentation up to which viability is not compromised has been estimated at 9–10% [10, 11] or 15–20% [36, 53] of the total cytoplasmic volume in the human and 5% in the in vivo-fertilized pig embryo [26]. Our observations that 10% fragmentation compromises development in the in vitro-fertilized pig embryo suggest that this species is more sensitive to this phenomenon. In addition to the absolute volume of the fragments, their spatial patterning has also been linked to viability in the human [36]; an association that could not be corroborated in the present study, since the fragments were nearly exclusively localized as clusters. This type of patterning might be explained by the earlier stages of development being recorded in the present study compared to those observed by Mateusen et al. [26] in in vivo-fertilized pig embryos. Asymmetry of division and degree of fragmentation are likely products of insufficient in vivo maturation and/or the effects of inadequate culture conditions that compromise cytoskeletal patency [48, 50]. In addition to these problems, porcine in vitro production systems suffer from polyspermia [55] which further alters the normal pattern of cytokinetic events and can affect viability in an unpredictable manner [57].
A cinematographic analysis of the cytokinetics of the early in vitro-produced pig embryo has shown the durations of the 2- and 4-cell stages to be 10.4 h and 32.6 h, respectively [58]. Thus, any of the embryos in the present study that divided to either more than two cells during their initial cleavage or five cells or more 24 h later should be categorized as either fragmenting or cytokinetically abnormal. Indeed, Wang et al. [50] found that by 24 h postinsemination, 72%, 5%, 17%, and 7% of cleaved IVP pig embryos were at the 2-, 3-, 4-, and 5–8 cell stages (largely in agreement with the data presented here), and that 27%, 50%, 86%, and 100% of these embryos contained blastomeres that were anucleate and/or binucleated. In the present work, abnormal embryos that contained more than 2 cells at the first cleavage were remarkable in that, although representing only 30.3% of the cleaved embryos on Day 1, the 3-cell embryos on Day 1 produced a blastocyst production rate of 27%, which was equivalent to that produced by 2-cell embryos (which are regarded as cytokinetically normal), while embryos that cleaved initially to four cells on Day 1 achieved a 40% blastocyst yield. In the absence of time-lapse images in the current study, we cannot know if these cytokinetically abnormal embryos were derived by one or several rapid sequential cleavage divisions, although Han et al. [25] consider that such porcine embryos divide directly to three or more cells from the 1-cell stage, and that the origin of these abnormal cleavage patterns is polyspermic fertilization. These same authors have also reported that blastocyst rates are not impaired by polyspermia, which implies that this fertilization error is the source of the abnormal cleavage observed for a proportion of the embryos in the present study.
One further potential predictor of blastocyst formation included in our analysis was the pattern of amino acid appearance and depletion rates. Together with our previous studies on IVP pig embryos [23], the results of Houghton et al. [18] and Brison et al. [19] indicate that developmental changes occur in the patterns of depletion and appearance of individual amino acids in preimplantation embryos. In addition, amino acids that exhibit significantly different patterns of depletion or release between arresting and developing embryos may differ according to the stage of embryo examined. For these reasons, it is apparent that identification of one or a set of amino acids that are consistently predictive of viability throughout in vitro development in one species, let alone between species, is likely to remain elusive. Consequently, it is of little surprise that of the predictive amino acids observed in our current study, only glycine corresponded to those observed by Houghton et al. [18] and Brison et al. [19]. However, the only amino acids (methionine, asparagines, and arginine) in the present study that were differentially released or depleted between slow and fast cleaving embryos were identical to three out of five amino acids whose depletion/appearance differed between arresting and developing human embryos on Days 2–3 [18]. The question as to how delayed cleavage and the associated inferior morphologic quality can modify metabolism or, more specifically, the patterns of amino acid uptake and release remains to be resolved. However, if polyspermia is involved, our previous research has already established that multiple sperm penetration can alter the rates of amino acid appearance and consumption [23].
The advantage of being able to predict blastocyst formation is dependent upon the intended application. In some situations, it might be crucial to isolate only those single or groups of embryos that possess full-term viability, e.g., for embryo transfer in monotocous species or for research purposes, such as gene expression studies. Alternatively, as in the present study, the intention is to maximize the percentage of blastocysts correctly classified as a proportion of those predicted, while optimizing the percentage of embryos correctly classified as blastocysts. Embryo viability evaluation systems in human IVF commonly grade embryos on the basis of three morphologic features (blastomere number, asymmetry of division, and degree of fragmentation) and combine these grades into a single score [13, 14]. The score is often constructed quite simply to facilitate both the ease of calculation and practicality in busy IVF clinics, although it may lack the robustness of logistical regression models in which more precise weightings of often numerous individual parameters can be assembled and computed. Combining prognostic factors by multivariate computation is a far more powerful technique for predictive purposes than could be achieved univariately. In this regard, using our own data, we found that cell number, evenness of division, and degree of fragmentation on Day 2, together with cleavage time, were univariately related to blastocyst formation, even though no factor in isolation could be used to predict whether an individual embryo was destined to become a blastocyst. In contrast, a logistical regression model that simply comprised blastomere numbers, evenness of division, and degree of fragmentation on Day 2 correctly classified 49% of the available blastocysts, equivalent to 56% of the blastocysts in the predicted population. The ability to predict blastocyst formation was also considerably different when using the morphologic data gathered from either Day 1 or Day 2, since, in contrast to the Day 2 data, only the degree of fragmentation had been shown to be related to blastocyst formation on Day 1 by univariate analysis. Surprisingly, a relatively effective model for the prediction of blastocyst formation was based simply on blastomere numbers on Days 1 and 2, namely, the interaction between 2-cell embryos on Day 1 and 4-cell embryos on Day 2 and that between 4-cell embryos on Day 1 and Day 2. This system identified over 60% of the observed blastocysts, generating an approximate 50% split between degenerate embryos and blastocysts in the predicted group. Similarly, in human embryos, the number of blastomeres present at specific time-points is one of the prime determinants of implantation potential [10, 13, 14, 59], although correlations between embryo morphology on Day 3 and blastocyst formation may be poor [60, 61]. However, improvements in the predictability of outcome in human studies can be achieved by simply performing multiple observations of the developing embryos [11, 15, 62, 63], a concept that we have successfully incorporated into the design of our experiments. The importance of time of cleavage as a predictor of porcine blastocyst formation was highlighted by the beneficial effect of adding this factor to a model that contained only the blastomere numbers on Day 1 and Day 2, such that 11.5% more predicted embryos constituted blastocysts, although the proportion of blastocysts correctly identified was reduced. The addition of symmetry of division and degree of fragmentation recorded on Day 1 and Day 2 to this model again improved the predictive power of the analysis, emphasizing the value of these morphologic features when combined interactively with other prognostic variates. The final factor to be amalgamated into the model was amino acid score, which was of particular interest owing to the correlation between this variate and the implantation rate of human embryos [19]. In support of the prognostic value of this variate, our results establish that, univariately, amino acid score is significantly related to blastocyst formation and can substantially improve the accuracy of blastocyst prediction, albeit only when assimilated into a model that contains other relevant variables. Incorporation of amino acid score into the analysis to generate the final model improved the proportion of accurately predicted blastocysts to more than 80%, representing an improvement of over 14% on the previous model.
One final aspect of the present study is the novel application of a woven polyester mesh, which is composed of monofilaments between which embryos can be placed such that neighboring embryos are separated by the optimal distance (84 µm) for maximal blastocyst formation, as defined by Stokes et al. [20] and Gopichandran and Leese [27]. This system confers the double benefit of group culture whilst enabling the tracking of individual embryos throughout in vitro development. The mesh system has been shown by us (unpublished results) to generate blastocyst rates equivalent to those recorded for both traditional group culture and the "well of the well" (WOW) system [64], although the culture system described by Stokes et al. [20], in which embryos are attached to a substratum coated with Celltak at the optimal distance apart, remains superior. One may surmise that the monofilaments act as a partial barrier to paracrine growth enhancers [65]. Once this barrier is overcome by simply increasing the porosity of the mesh, the principle of this simple system should find widespread use for the culturing of human embryos and those of domestic animals.
The characteristics that an embryo exhibits during early development and its consequent viability following monospermic fertilization in vitro are generally considered to depend largely on the product of the quality of the oocyte and the spermatozoon [66, 67], excluding any iatrogenic affects of the culture environment. If this is a realistic hypothesis, then scrutiny of the gametes prior to fertilization plus evaluation of any predictive morphologic and cytokinetic features of the zygote and embryos up to the second or perhaps just the first cleavage division, could provide a sufficient estimate of implantation potential. Indeed, Scott and Smith [62] have applied this principle by demonstrating the high predictability of human embryo implantation by generating a cumulative score for pronuclear morphology, time of cleavage, and Day-1 morphology. The absence of any difference in implantation rate between Day-2 and Day-3 transfers [68] and between Day-3 and Day-5 transfers in the human [69], as deduced by Cochrane review, would appear to support the notion that sufficient morphologic features are present by the time of the second cleavage division to select the most viable embryos. Even our data, which attempt to model the development of pig embryos in which cytokinetic abnormalities are common, support this notion. Our data also reinforce the concept, as indicated in human embryo research [11, 15, 63, 70], that reliance on any single variate as a predictor is unlikely to prove effective, whereas the synergy of quantifiable morphologic features, when measured sequentially, coupled with kinetic (ideally cinematographic) and metabolic data [16, 17, 19] will undoubtedly prove the most prognostic of the methods currently available.
ACKNOWLEDGMENTS
We gratefully acknowledge Grampian Country Pork for the supply of pig ovaries, and GTC Scotland (PIC Sygen, UK), particularly Mr. R. Holmes, for the provision of frozen boar semen. We also thank Mrs. J. Hawkhead for technical assistance and Mr. C. Bingham for collection of the tissues.
FOOTNOTES
3Current address: Leeds General Infirmary, Reproductive Medicine Unit, Clarendon Wing, Great George Street, Leeds LS1 3EX, United Kingdom. 
1Supported by the UK Biotechnology and the Biological Sciences Research Council.
Correspondence: 2P.J. Booth, Assisted Conception Unit, Gledhow Wing, Level 6, St. James's University Hospital, Beckett Street, Leeds LS9 7TF, United Kingdom. FAX: 44 0113 2064812; e-mail: paul.booth{at}leedsth.nhs.uk
Received: 20 May 2007.
First decision: 16 June 2007.
Accepted: 10 July 2007.
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  C. Racowsky, L. Ohno-Machado, J. Kim, and J. D. Biggers Is there an advantage in scoring early embryos on more than one day? Hum. Reprod., June 2, 2009; (2009) dep198v1. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. G. Sturmey, J. A. Hawkhead, E. A. Barker, and H. J. Leese DNA damage and metabolic activity in the preimplantation embryo Hum. Reprod., January 1, 2009; 24(1): 81 - 91. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
30%, respectively. To illustrate the system used for scoring evenness of division and degree of fragmentation, the example shown utilizes a 2-cell embryo.
