Time-lapse microscopy and image analysis in basic and clinical embryo development research

Transcription

1 Reproductive BioMedicine Online (2013) 26, REVIEW Time-lapse microscopy and image analysis in basic and clinical embryo development research C Wong a, AA Chen a, B Behr b, S Shen a, * a Auxogyn, Inc., 1490 O Brien Drive, Suite A, Menlo Park, CA 94025, USA; b Department of Obstetrics and Gynecology, School of Medicine, Stanford University, Stanford, CA 94305, USA * Corresponding author. address: sshen@auxogyn.com (S Shen). Connie Wong is a cell, molecular and developmental biologist who has a passion in technology development. During her post-doctoral training in Dr Renee Reijo Pera s laboratory at Stanford University, she developed an experimental system that enabled concurrent time-lapse microscopy and high-throughput single-cell gene expression analysis of preimplantation embryos. Her research interests include the development of biomarkers, predictive parameters and molecular diagnostics tools for assisted reproduction and other clinical areas. Abstract Mammalian preimplantation embryo development is a complex process in which the exact timing and sequence of events are as essential as the accurate execution of the events themselves. Time-lapse microscopy (TLM) is an ideal tool to study this process since the ability to capture images over time provides a combination of morphological, dynamic and quantitative information about developmental events. Here, we systematically review the application of TLM in basic and clinical embryo research. We identified all relevant preimplantation embryo TLM studies published in English up to May 2012 using PubMed and Google Scholar. We then analysed the technical challenges involved in embryo TLM studies and how these challenges may be overcome with technological innovations. Finally, we reviewed the different types of TLM embryo studies, with a special focus on how TLM can benefit clinical assisted reproduction. Although new parameters predictive of embryo development potential may be discovered and used clinically to potentially increase the success rate of IVF, adopting TLM to routine clinical practice will require innovations in both optics and image analysis. Combined with such innovations, TLM may provide embryologists and clinicians with an important tool for making critical decisions in assisted reproduction. RBMOnline ª 2012, Reproductive Healthcare Ltd. Published by Elsevier Ltd. All rights reserved. KEYWORDS: cell dynamics, image analysis, predictive parameters, preimplantation embryo development, time-lapse microscopy Introduction Mammalian preimplantation embryo development encompasses a complex series of morphological and molecular processes. Beginning with the fusion of spermatozoon and egg, the resulting zygote proceeds to epigenetically reprogram and combine the gametic pronuclei (PN). The embryo undergoes a series of cleavage divisions which, up until the point of embryonic genome activation, are largely dependent on the mrna and protein reserve from the egg. The cleavage-stage embryo subsequently compacts to form the morula, and the morula undergoes its first differentiation to form the blastocyst that consists of the inner cell mass and trophectoderm. It is appreciated that these /$ - see front matter ª 2012, Reproductive Healthcare Ltd. Published by Elsevier Ltd. All rights reserved.

2 Time-lapse microscopy in early embryo development research 121 preimplantation processes need to be executed accurately for proper embryo development, yet we are only now beginning to understand the exact timing and sequence of embryo events involved. Time-lapse microscopy (TLM) is an ideal tool to study the dynamic biological processes of early embryo development, as it provides morphological, dynamic and quantitative timing data in a non-invasive manner. In this review, we will discuss the benefits of TLM over traditional time-point analysis, as well as the technical difficulties and solutions involved in the application of TLM technology to early embryo development research. We will also systematically review both basic and clinical embryo research that utilized TLM. Finally, as clinical research with TLM has begun to successfully derive predictive markers of embryo developmental competence, we will highlight these findings and discuss both advantages and practical challenges in their clinical feasibility for improving assisted reproduction. TLM as a tool to study early embryo development Benefits of TLM TLM offers many advantages over traditional time-point microscopy. Using TLM, biological samples are cultured directly on an imaging device that captures images at defined intervals over a specific period of time. Individual captured images can then be processed into a time-lapse sequence, and from the video sequence, morphological, dynamic and quantitative data can be extracted. In contrast, traditional time-point microscopy acquires images at distinct time points. Such time points are often selected for user convenience rather than biological relevance, and they are acquired at a significantly lower frequency than TLM sequences. For example, in embryo grading and embryo research, images may be collected daily, while TLM enables images to be captured at 20-min, 5-min or even 10-s frequencies (Meseguer et al., 2011; Wong et al., 2010; Ajduk et al., 2011). In an example where TLM information aided early embryo research, Wong et al. (2010) demonstrated that two morphologically similar embryos that would have been mistakenly classified to the same developmental stage in a time-point analysis were actually products of completely different developmental processes. One embryo was an 8-cell embryo that followed the normal developmental course, while the other embryo was the result of a series of abnormal cell divisions and fragmentation revealed by TLM (Wong et al., 2010). Separately, Hardarson et al. (2002) and Van Blerkom et al. (2001) used TLM to monitor embryos which had the ability to re-internalize fragments. Because phenomena such as tripolar division, reverse cleavage, fragmentation, internalization and other abnormal divisions may occur between time points, it is extremely difficult to capture these critical biological processes using conventional time-point imaging. Other dynamic data such as the formation and fusion of gametic PN, the exact timing between each subsequent cell division and the duration of cytokinesis can be quantified using TLM (Payne et al., 1997; Meseguer et al., 2011; Mio and Maeda, 2008; Lemmen et al., 2008; Wong et al., 2010). TLM image data may thus present a more complete picture of the biological process being studied and lead to more accurate scientific and clinical morphological interpretations. Another benefit of TLM is the ability to maintain samples in an optimal culture environment during the entire time of data acquisition. Variations in culture ph and temperature can have adverse effects on biological samples and may be especially detrimental to sensitive samples such as embryos. Evidence from TLM studies with embryos suggests that maintenance in optimal culture conditions facilitates normal embryo growth and development (Kirkegaard et al., 2012a,b). In addition, maintenance in optimal culture conditions minimizes sample movement between culture incubator and microscope platforms, which can lead to full or partial displacement of the samples from their original positions. Finally, by reducing excess sample handling, TLM may decrease chances of sample error and improve sample tracking and data analysis. TLM is also accompanied by unique capabilities to be integrated with image analysis software. It is important to note that software for automating the microscope mechanics is a required and inherent feature of all time-lapse systems. Such image capture software enables optical alignment, focusing, image capture and image storage. Recently, software for automating image analysis through computer vision has also emerged to further enable real-time and quantitative assessment of embryo development from time-lapse videos. Computer vision is a branch of science and engineering that, broadly, extracts information from images in order to model and interpret image events (Danuser, 2011). Since time-lapse videos are embedded with substantially more data than static images, emerging advances in computer vision will be critical to efficiently monitoring whole embryo and blastomere parameters such as area, perimeter, diameter, elongation, texture, degree of fragmentation and changes in parameters over time. Finally, TLM can be used in conjunction with molecular tools that allow researchers to visualize and quantify embryo behaviour and their molecular underpinnings. These approaches include utilizing embryos derived from animals expressing a fluorescent protein-tagged antigen (Torres-Padilla et al., 2007) or microinjecting mrna encoding a tagged protein into the embryo (Yamagata et al., 2005; Yamazaki et al., 2007). Further research is needed to validate the molecular phenomena observed using these techniques and understand the implications of such dynamic embryo behaviour. Technical challenges of TLM Despite its many benefits, TLM also poses many technical challenges, especially when applied to the study of mammalian embryos. Mammalian embryos, and in particular human embryos, are extremely sensitive to fluctuations and suboptimal culture environment. Researchers have successfully employed both cleverly homemade technologies (Payne et al., 1997; Hardarson et al., 2002; Mio and Maeda, 2008; Wong et al., 2010) and commercially available devices (Lemmen et al., 2008; Meseguer et al., 2011; Cruz et al., 2011) to maintain proper culture conditions during embryo TLM studies. Current homemade and commercial

3 122 C Wong et al. technologies maintain an optimal embryo culture environment by either implementing a small incubator on the stage of a full-size imaging system (Payne et al., 1997; Hardarson et al., 2002; Mio and Maeda, 2008) or housing a small-size imaging system inside a large incubator (Lemmen et al., 2008; Meseguer et al., 2011; Cruz et al., 2011). Both strategies have merit, as the ideal environment should provide embryos with stable culture conditions over the course of the imaging study (Lemmen et al., 2008; Meseguer et al., 2011; Wong et al., 2010). Overall, TLM usage should not compromise required culture conditions including temperature, ph and humidity. An additional challenge to using TLM is in maintaining samples on the culture dish and within the field of view of the microscope throughout imaging. Embryos are non-adherent and may move in culture dishes, especially during manipulation or dish movement. Because standard embryo culture commonly requires media changes during incubation, it can therefore be extremely challenging to keep track of an individual embryo s identity during the course of a study. This problem could be addressed by culturing embryos in individual media droplets under oil and using a motorized system to continuously move different embryos into the field of view during TLM observation (Meseguer et al., 2011). Alternatively, embryos may be cultured in groups under a shared media drop using a multiwell culture dish that holds individual embryos in close proximity to each other. A close-spacing microwell design allows embryos to be kept separate and tracked throughout the TLM experiment, even during media changes (Pribenszky et al., 2010a,b). Importantly, microwell spacing enables individual well tracking under a single optical field of view, which minimizes the need for moving parts such as a motorized stage or moving optics (Vajta et al., 2000; Sugimura et al., 2010; Cruz et al., 2011; Meseguer et al., 2011; Hashimoto et al., 2012). Group culture may also be preferred by many laboratories since studies have shown higher blastocyst formation rates in group versus individual human embryo culture, possibly due to positive paracrine signalling between embryos (Blake et al., 2007; Papanikolaou et al., 2006, 2005). Embryos cultured closely in a multiwell dish under a single drop of common media are allowed to exchange such paracrine signals. Users of TLM must also consider the potential phototoxicity effect to embryos. It is well known that light can cause damage to DNA, induce localized heating or generate free-radical species inside cells (Frigault et al., 2009). Visible light is less disruptive than high-energy light such as ultraviolet light, focused infrared light or the excitation of fluorescent proteins; thus TLM systems for embryo imaging have commonly employed visible light capabilities including bright-field imaging (which takes advantage of sample light absorbance to create contrast) and dark-field imaging (which takes advantage of sample light scattering to create contrast). Importantly, the total light exposure time of a multiple-day experiment can be limited to exposure lengths of only minutes by minimizing the duration of image acquisition used during a TLM experiment (Meseguer et al., 2011; Wong et al., 2010). This light exposure is significantly less than that used in IVF laboratory practice for routine time-point embryo morphology analysis. Furthermore, modifications to the illumination system itself can help minimize the light intensity, a key and overlooked component of some light exposure analyses (Meseguer et al., 2012). Wong et al. (2010) utilized low-power LED light in a dark-field illumination system and also reduced the amount of light used by projecting a hollow cone of light onto the samples instead. Mouse zygotes cultured under this TLM system achieved similar blastocyst formation rate as control embryos and exhibited no significant difference in their gene expression profiles. Meseguer et al. (2011) employed low-intensity red light (635 nm) in their Hoffman illumination system to avoid exposing the embryos to damaging short wavelength light. With a carefully optimized experimental set up, preimplantation mouse embryos subjected to even higher-energy fluorescent excitation during a TLM study could develop normally and became healthy live pups (Yamagata et al., 2009). Numerous TLM studies performed in both human and other mammalian preimplantation embryos have further shown that TLM does not cause any significant differences in the percentage of high-quality embryos (Nakahara et al., 2010; Lemmen et al., 2008), developmental kinetics (Grisart et al., 1994; Gonzales et al., 1995; Barlow et al., 1992), developmental potential (Kirkegaard et al., 2012b) (Lemmen et al., 2008; Payne et al., 1997), blastocyst formation rate (Kirkegaard et al., 2012b; Cruz et al., 2011; Gonzales et al., 1995; Grisart et al., 1994; Pribenszky et al., 2010a), blastocyst viability (Sugimura et al., 2010; Cruz et al., 2011), fertilization rate (Nakahara et al., 2010; Payne et al., 1997), pregnancy rate (Kirkegaard et al., 2012b; Barlow et al., 1992; Cruz et al., 2011; Mio and Maeda, 2008) and gene expression profile (Wong et al., 2010). Moreover, human embryos subjected to TLM studies have successfully yielded live births, suggesting that TLM does not cause any irreversible and detrimental effect to early embryo development (Mio and Maeda, 2008; Pribenszky et al., 2010b). Thus, when performed using appropriate considerations for light intensity and exposure, TLM does not appear to have a detrimental effect on embryo development. TLM in mammalian preimplantation embryo development research Due to its unique capabilities, TLM has been used in preimplantation embryo research to investigate a variety of unknown developmental questions. A significant question in early mammalian embryo development, for example, is whether the polarity of a mammalian embryo is preestablished in the egg like their non-mammalian counterparts. Hiiragi and Solter (2004) made progress in addressing this question by monitoring early mouse embryo development using TLM. With TLM, it was possible to observe that the first cleavage plane of the mouse embryo did not appear to be predefined in early interphase, but rather specified by the position of the 2 PN in the zygote. Further, unlike non-mammalian embryos which mark a stationary animal pole, the second polar body in a mouse embryo appeared to be mobile (Hiiragi and Solter, 2004). In support of these findings, Motosugi et al. (2005) employed TLM to observe that the two blastomeres of the 2-cell mouse embryo did not differ in their developmental potential, suggesting that

4 Time-lapse microscopy in early embryo development research 123 a predetermined axis may not exist in mouse embryos. During mouse blastocyst formation, the blastocoele appeared to originate from different points within either or both clonal areas of the two blastomeres daughter cells (Motosugi et al., 2005). Together, these data contributed new understanding of an extremely complex biological process (Gardner, 1997, 2001; Piotrowska et al., 2001; Piotrowska and Zernicka-Goetz, 2001; Fujimori et al., 2003). In current and future research, additional TLM studies will validate and establish further understanding of axis establishment in the mammalian embryo. TLM has also been used to track the kinetic changes of molecular mechanisms during early embryo development. Yamazaki et al. (2007) studied the dynamics of epigenetic reprogramming of PN in the mouse zygote, using mouse embryos expressing EGFP (enhanced green fluorescent binding protein)-labelled MBD1 (methyl CpG binding protein). The fluorescently tagged EGFP-MBD1 protein binds specifically to methylated DNA. Coupled with TLM, the fluorescent tool allowed for the observation of dramatic configuration changes in methylated chromatin between the 2- and 4-cell stages of embryo development (Yamazaki et al., 2007). Using a similar technique, Yamagata et al. (2005) used EGFP-labelled tubulin (a microtubule protein) to monitor spindle formation and HcRed-labelled Histone H4 to monitor nuclear formation in mouse embryos during second meiosis and first mitosis. Yamagata et al. (2009) also succeeded in performing a long-term, multicolour and multisample TLM experiment on preimplantation mouse embryos using spinning-disk confocal microscopy. The experiment lasted 70 h, and 51 images were acquired in the z-axis at 7.5-min intervals. The embryos were subsequently transferred into pseudopregnant mothers and produced live pups that grew up to be healthy adults (Yamagata et al., 2009). Altogether, evidence from TLM studies has greatly enhanced our knowledge of early embryo development and suggests that TLM is an ideal tool for early embryo development research. TLM in human assisted reproduction clinical research More than three decades following the first successful IVF procedure, the overall success rate of IVF hovers around only 30% ( SummaryReport.aspx), while the multiple birth rate due to multiple embryo transfer is still alarmingly high. This is in large part due to a lack of robust and reliable markers that can predict an embryo s growth potential and enable the selection of the best embryo(s) for transfer. Many morphology-based parameters and composite scoring systems have been proposed for embryo selection, yet data show that these parameters have very limited predictive value in IVF outcome (De Placido et al., 2002; Alpha, 2011; Guerif et al., 2010). Therefore, new prediction markers that may improve identification of the best embryo for transfer, and therefore improve IVF success rates, are of extreme interest and importance to the field of clinical reproductive medicine. Recently, advances in TLM technologies have led to the discovery of novel, non-invasive, prediction markers of embryo quality for human assisted reproduction. Underlying these discoveries, the proper timing of embryo development has long been recognized as a critical factor in assessing embryo quality. Several studies have qualitatively correlated early cleavage with higher number of cells in the embryo and higher blastocyst formation rate (Shoukir et al., 1997; Sakkas et al., 1998; Lundin et al., 2001; Hesters et al., 2008; Montag et al., 2011; Filho et al., 2010). For example, transfer of early cleavage embryos was shown to yield a higher pregnancy rate than the late cleavage group (Van Montfoort et al., 2004). One of the first clinically focused human embryo TLM experiments was performed by Payne et al. (1997). Using a homemade perspex environmental chamber and a switching box to control a light source and video casette recorder, 50 human oocytes were successfully filmed shortly after intracytoplasmic sperm injection (ICSI) for h at 1-min intervals. This experiment was the first documentation of the exact sequence and timing of events during human fertilization, including the extrusion of the second polar body and the appearance of PN. Furthermore, novel features of human fertilization such as cytoplasmic flare and the periodic waves of granulation within the ooplasm were observed. The authors also noted that good-quality embryos appeared to arise from oocytes with more uniform timing from ICSI to PN abuttal, and that they tended to exhibit cytoplasmic waves at slightly longer periodicity. Unfortunately, the small sample size of the study made the statistical significance of the data difficult to assess and warranted a need for additional validation studies. The success by Payne et al. (1997) in using TLM to derive kinetic information of early human embryo development inspired many other TLM studies. Using a new TLM system that allowed a much longer observation period, Mio and Maeda (2008) recorded the development of 286 human embryos for 2 5 days at 2-min intervals. Three particular TLM videos were discussed in the report. The first video described again the sequence and timing of events during the fertilization process, supplementing the observations of Payne et al. (1997) with finer details such as the development of a fertilization cone in a human embryo during IVF. The second video followed the development of a 2-cell embryo into a hatched blastocyst. The third video captured the splitting of the inner cell mass in two of the three blastocysts that were imaged, leading the authors to postulate that prolonged in-vitro culture might lead to higher risk of monozygotic twin pregnancy. Following the TLM study, 46 embryos were selected to transfer into patients based on routine morphology scoring criteria. Most notably, a clinical pregnancy rate of 21.8%, similar to that of control embryos at their centre, as well as four healthy live births, was observed with these embryos. Another human embryo TLM study that resulted in live birth was performed by Pribenszky et al. (2010b). In the study, the authors subjected all five of 2PN embryos of a single patient to 5 days of TLM at 10-min intervals. Out of the five embryos, two successfully reached the blastocyst stage. One of the blastocysts was chosen to be transferred based on a lack of fragmentation and earlier cleavage compared with the other blastocyst. The transfer resulted in a successful pregnancy and live birth. Collectively, the studies above described the first observations of human embryos with TLM systems and provided evidence that TLM study does

5 124 C Wong et al. not cause any observable detrimental effect in human embryo development. Lemmen et al. (2008) performed a human embryo TLM study in which a total of 102 2PN embryos were imaged for h after fertilization at 5-min intervals. The embryos were chosen randomly from their cohorts for the TLM study and were promptly returned to the incubator after the TLM experiment completed. The imaged embryos and their siblings were then scored by an embryologist according to the clinic s routine procedure, and the embryo with the highest score was subsequently transferred. Using data from all 102 embryos, they found that three kinetic parameters correlated with higher number of blastomeres on day 2: (i) early disappearance of PN; (ii) early first cleavage; and (iii) early appearance of nuclei after the first cleavage. Among the 102 embryos, 19 were transferred as 4-cell embryos and resulted in six pregnancies. Interestingly, the embryos that successfully implanted had significantly more synchronous appearances of the nuclei in both blastomeres after the first cleavage than those that failed to implant. Nonetheless, these observations did not result in a set of directly measurable, non-overlapping parameters that can be easily applied clinically to stratify embryos based on their quality and development potential. Following these promising observational studies in the human embryo, Wong et al. (2010) tracked the development of 242 supernumerary frozen IVF day 1 embryos in a study that combined TLM and high-throughput gene expression technologies. TLM imaging commenced immediately after embryo thawing. Images were taken at 5-min intervals, and small groups of embryos were collected every 24 h for single-cell or single-embryo gene expression analysis. Among the 242 embryos, 100 embryos were cultured up to 5 6 days, and 39% successfully reached the blastocyst stage. The authors subsequently extracted and quantified six different morphological and dynamic parameters from the TLM data and found three parameters which collectively predicted blastocyst formation with a sensitivity and specificity of 94% and 93%: (i) duration of the first cytokinesis within 14.3 ± 6.0 min; (ii) time between the first and second mitosis within 11.1 ± 2.2 h; and (iii) time between (i.e. synchronicity) the second and third mitosis within 1.0 ± 1.6 h. These are the first quantitative cell division parameters shown to be directly correlated to embryo development potential, and the first to also be linked to the underlying gene expression and molecular health of human embryos. Embryos that arrested prior to embryonic gene activation appeared to have an abnormal gene expression profile, indicating problems with the maternal mrna reserve (Wong et al., 2010). Most notably, all three events occurred before the embryonic gene activation at the 8-cell stage. This collection of studies indicated that embryo fate may be predetermined very early in development, perhaps as an inheritance from the egg and that blastocyst prediction could be achieved using biologically validated parameters and image analysis techniques, potentially by day 2. Subsequently, Meseguer et al. (2011) independently confirmed two out of the three parameters identified by Wong et al. (2010) in fresh embryos, by showing that early morphokinetic markers also correlate with embryo implantation in a retrospective, single-centre clinical study. TLM analysis was performed on 522 2PN embryos at 15-min intervals up to 64 h before transfer (Meseguer et al., 2011). Only data from 247 embryos that resulted in either full or no implantation were included in the study analysis. Among the morphological and kinetic parameters extracted from the TLM data, the authors concluded that parameters most predictive of subsequent implantation success were: (i) time of division to 5-cell embryo h after ICSI (ii) time between the first and second mitosis 11.9 h; and (iii) time between (i.e. synchronicity) the second and third mitosis 0.76 h. Meseguer et al. (2011) also reported significant differences in the implantation rate between embryos that reached the 2-cell, 3-cell and 4-cell stages within a certain time range, though further statistical analysis showed that these parameters are not as predictive as the three described above. In a following study, Hashimoto et al. (2012) showed that embryos that developed into better-quality blastocysts required significantly shorter time to complete the second and third cell divisions, further illustrating the correlation between the timing of embryo development and their developmental potential. Notably, two of the parameters reported in Meseguer et al. (2011) and one reported in Hashimoto et al. (2012) were identical to those reported by Wong et al. (2010). This congruence demonstrates repeatability of the parameters across independent studies, for different outcome measures reflecting embryo viability and developmental potential. Parameter repeatability is a critically important step that has been difficult to achieve in the field of human embryo assessment; it also suggests that measurement of these published parameters may be worth integrating into clinical practice alongside reliable and real-time image analysis capabilities. Embryo TLM study in animals can also contribute to the improvement of assisted reproduction in human. For example, in order to assess the safety of blastomere biopsy, Ugajin et al. (2010) removed a blastomere from 4-cell mouse embryos and followed their subsequent development using TLM. Blastocyst hatching in embryos subjected to blastomere removal was delayed by more than 10 h on average compared with control embryos. The frequency of the blastocyst expansion and contraction cycles also appeared to be much higher in the embryos with removed blastomeres (Ugajin et al., 2010). These data suggest that blastomere removal and the subsequent impact on embryo hatching might potentially lead to decreased efficiency in embryo implantation. Similar results were obtained in independent studies in mouse and human (Terada et al., 2009; Kirkegaard et al., 2012c). More recently, a TLM study focusing on cytoplasmic movements in the mouse embryo determined that cytoplasmic waves can be detected by particle image velocimetry methods at 10-s image intervals. These waves were attributed to fertilization-induced Ca 2+ oscillations that go on to trigger contractions of the actomyosin cytoskeleton. The mouse study and follow-on work in the human suggested that the timing and pattern of cytoplasmic movements may be used for early prediction of the developmental potential of the human zygote (Ajduk et al., 2011; Swann et al., 2012). Furthermore, a recent study in cattle demonstrated that combining TLM with measurement of embryo metabolism parameters such as oxygen consumption promoted pregnancy success to 78.9% (Sugimura et al., 2012). Although these results are intriguing and hold promise for early embryo selection, they require further validation in clinical

6 Table 1 Study Payne et al. (1997) Lemmen et al. (2008) Mio and Maeda (2008) Pribenszky et al. (2010b) Wong et al. (2010) Summary of embryo TLM studies. Sample TLM duration Embryo and cell events assessed 50 2PN embryos h after ICSI at 1-min intervals 102 2PN oocytes h at 5-min intervals 286 oocytes and embryos (data of 5 embryos reported in paper) 5 2PN embryos from one patient 2 5 days at 2-min intervals 5 days at 10-min intervals 242 2PN embryos 2 6 days at 5-min intervals Details of the fertilization process (1) PN appearance and disappearance Outcomes NA Pregnancy rate Observations (1) Observed sequence and timing of the human fertilization process (2) Observed novel ooplasm wave (3) Observed novel cytoplasmic flare (1) Early PN disappearance and 1st cleavage correlates with higher cell number on day 2 (2) Timing of cell cleavage (2) Synchrony in nuclei appearance after 1st cleavage correlates with pregnancy success (3) Synchrony of nuclei appearance after the 1st cleavage (1) Details of the fertilization process to the 1st/2nd cleavage (2) Progression from the 2- cell stage to the hatching stage (3) Formation of monozygotic twins Clinical pregnancy rate, pregnancy rate and live births (1) Observed embryo development from sperm penetration to blastocyst formation (2) Observed formation of two inner cell mass in three blastocysts (3) TLM has no observable negative effects on embryonic morphology and clinical pregnancy rate (1) Embryo morphology Live birth Observed that a single blastocyst chosen for transfer based on lack of fragmentation and earlier cleavage timing resulted in a live birth (2) Timing of cell divisions (1) Embryo morphology Blastocyst formation as outcome (2) Timing and synchrony of cell divisions (3) Duration of 1st cytokinesis (1) Three parameters collectively predict blastocyst formation before day 3: (i) duration of 1st cytokinesis, (ii) time between 1st and 2nd mitosis, (iii) synchronicity of 2nd and 3rd mitosis (2) Single cell/embryo gene expression profiles correlate morphology phenotypes to aberrant gene expression patterns Time-lapse microscopy in early embryo development research 125

7 126 C Wong et al. Table 1 (continued) Three parameters collectively predict implantation: (i) time to 5-cell embryo, (ii) time between 1st and 2nd mitosis, (iii) synchronicity of 2nd and 3rd mitosis (1) Embryo morphology Implantation as outcome measurement To 72 h after insemination at 15-min intervals 247 2PN transferred embryos Meseguer et al. (2011) (2) Timing and synchrony of cell divisions Two parameters predict the formation of highscoring blastocysts: (i) shorter time to complete 2nd mitosis, (ii) shorter time to complete 3rd mitosis Timing of cell divisions Formation of high- versus lowscoring blastocysts To 5 days at 10-min intervals 80 2PN embryos Hashimoto et al. (2012) settings. In addition, non-trivial technical challenges in imaging cytoplasmic waves at the high frequencies required may limit the clinical adoption of the approach. In summary, a number of embryo TLM studies have contributed to our understanding of early embryo development and to the identification and validation of promising new markers for early embryo selection. Those utilizing human embryos are summarized in Table 1. Non-invasive TLM markers: the promise for embryo selection In order to improve assisted reproduction success, researchers have been searching for biomarkers that can predict the developmental and implantation potential of early preimplantation embryos. During a standard IVF cycle, multiple embryos are collected per couple, and the embryo(s) determined to be of the highest quality among the cohort will be transferred typically on day 3 or day 5 of development. Data show that the clinical pregnancy rate increases with the number of embryos transferred, yet transfer of multiple embryos leads to higher chance of multiple births and significantly increases health risks and costs during pregnancy and labour. Culturing embryos to the day-5 blastocyst stage has been shown to significantly improve the chance of pregnancy compared with the transfer of 8-cell embryos on day 3(Papanikolaou et al., 2006, 2005; Blake et al., 2007) and is gaining adoption as an embryo selection approach (Diamond et al., 2012). However, culturing embryos for 5 6 days can be risky for patients with only a few embryos, and many embryos arrest before reaching this critical stage. Additional data suggest that prolonged in-vitro culture may lead to imprinting errors and subsequently epigenetic disorders (Niemitz and Feinberg, 2004; Horsthemke and Ludwig, 2005; Manipalviratn et al., 2009), as well as higher risk of preterm birth or congenital malformations (Kallen et al., 2010a,b). Therefore, the ideal biomarker will minimize the chance of multiple births and enable earlier transfer without sacrificing the success rate. Study data suggest that markers of developmental timing might be better prognostic factors for embryo growth potential than morphology alone (Shoukir et al., 1997; Sakkas et al., 1998; Lundin et al., 2001; Hesters et al., 2008; Montag et al., 2011; Filho et al., 2010). Using TLM, new predictive morphokinetic parameters that correlate with good-quality embryos determined by either morphology, blastocyst formation or embryo implantation have been found, including: (i) time between fertilization and PN abuttal (Payne et al., 1997); (ii) appearance of nuclei after first cleavage (Lemmen et al., 2008); (iii) duration of first cytokinesis (Wong et al., 2010); (iv) time between the first and second mitosis (Meseguer et al., 2011; Wong et al., 2010); (v) synchronicity of the second and third mitosis (Meseguer et al., 2011; Wong et al., 2010) and (vi) time to reach 5-cell stage (Meseguer et al., 2011). Notably, all of these parameters occur before day 3 of development, suggesting that the quality of human preimplantation embryo is predetermined before embryonic gene activation. Practical use of a combination of these parameters, perhaps in addition to the knowledge obtained from traditional morphological assessment of embryos, will likely improve the embryo selection

8 Time-lapse microscopy in early embryo development research 127 and subsequently increase the success of assisted reproduction. The risk of multiple births will also be lowered as fewer embryos will need to be transferred to achieve successful pregnancy. Finally, the ability to predict the developmental potential of an embryo before embryonic genome activation may enable embryo transfer on day 3. Although there still may be a need for extended culture in certain situations, for example to assess paternal genomic contributions, embryo transfer on day 3 may minimize the risks potentially associated with prolonged in-vitro embryo culture (Niemitz and Feinberg, 2004; Horsthemke and Ludwig, 2005; Manipalviratn et al., 2009). TLM markers: practical clinical application for embryo selection While TLM may prove to be safe and effective at capturing critical embryo information, it is important to consider that the manual processing of abundant image data poses a huge hurdle for practical clinical translation. Manual scoring of embryos based on qualitative measurements and/or at discontinued time points has been shown to present substantial inter-observer as well as intra-observer variability which significantly impact the clinical success of morphology-based embryo selection (Baxter Bendus et al., 2006). Alternatively, computer vision software that can measure quantifiable parameters may provide an objective, standardized embryo quality assessment free of human biases. Such computer vision-based image analysis systems have shown enormous promise in aiding clinicians, for example by enabling quantitative and rapid assessment of sperm quality (Krause and Viethen, 1999) and by augmenting cervical cytology screens (Dziura et al., 2006; Lozano, 2007). To be clinically useful for human embryology, image analysis software should be accurate, automated, performed in real-time and scalable in order to accommodate many embryos at the same time. To date, such image analysis software has been applied to human embryos in only one published report (Wong et al., 2010), wherein the cell tracking software utilized a probabilistic model estimation technique to simulate images, which were then compared with the observed image data to infer cell membranes and cell division events. The observed image data consisted of a single image of each embryo captured at the centre (or equator), which allows the tracking of the cell membranes up to the 4-cell stage and the automated quantification of the three predictive cell division parameters discovered in their study. While this algorithm was a significant advance, it was limited by the fact that it could only track cell divisions from the 1-cell stage to the 4-cell stage. Automated tracking of cell divisions beyond 4-cells is a more difficult problem due to the increasing size, depth and complexity of the developing embryo and the broad spectrum of biological phenomena that can occur. At the current state of research, later events that have been used as embryo quality markers, such as compaction, blastulation, length of the 5 8-cell interval and timing of the fourth cleavage can only be assessed manually. Automated tracking of these complicated events will likely involve more sophisticated algorithms, a combination of illumination techniques, and/or additional focal planes captured throughout the embryo. Nonetheless, the development of any computerized algorithm will be faced with technical challenges and the possibility for missed events, incorrect classification or false detections. Therefore, the introduction of any automated algorithm must be accompanied with full validation studies to characterize accuracy, safety and efficacy. Conclusion TLM is a powerful technology for the study of early embryo development. The growing appreciation of TLM in both basic embryology research and clinical assisted reproduction is reflected in the increasing number of TLM studies published in recent years. When practiced with precautions, TLM does not appear to cause any observable, detrimental effect in embryo development, and thus can be used safely in the clinic as a tool to select the best embryo for transfer. Using TLM, a range of morphological and dynamic parameters can be extracted from individual embryos and potentially used as predictive markers for healthy embryo development. Prospective and randomized studies that will test the predictive power of cell division parameters for embryo selection are underway. These studies will require multicentre validation in order to prove the improvement of embryo selection of traditional grading methods. These same studies will also need to test parameter robustness in the face of variable fertilization methods, culture media and environmental conditions. In parallel, emerging technologies that can automate the tracking and quantitative measurements of TLM data are being validated to enable practical clinical utility. These technologies have the potential to augment the utility of TLM by eliminating human operator variation from manual scoring of complex image sequences and reducing a significant amount of time and labor. Leveraging a validated computer program to extract validated parameters may also enable real-time embryo viability predictions that, used in conjunction with conventional grading protocols, may assist embryologists in identifying the best embryo for transfer. References Ajduk, A., Ilozue, T., Windsor, S., Yu, Y., Seres, K.B., Bomphrey, R.J., Tom, B.D., Swann, K., Thomas, A., Graham, C., Zernicka-Goetz, M., Rhythmic actomyosin-driven contractions induced by sperm entry predict mammalian embryo viability. Nat. Commun. 2, 417. Alpha, E., The Istanbul consensus workshop on embryo assessment: proceedings of an expert meeting. Hum. Reprod. 26, Barlow, P., Puissant, F., Van der Zwalmen, P., Vandromme, J., Trigaux, P., Leroy, F., In vitro fertilization, development, and implantation after exposure of mature mouse oocytes to visible light. Mol. Reprod. Dev. 33, Baxter Bendus, A.E., Mayer, J.F., Shipley, S.K., Catherino, W.H., Interobserver and intraobserver variation in day 3 embryo grading. Fertil. Steril. 86, Blake, D., Farquhar, C., Johnson, N., Proctor, M., Cleavage stage versus blastocyst stage embryo transfer in assisted reproductive technology. Cochrane Database Syst. Rev. Art. No.: CD Cruz, M., Gadea, B., Garrido, N., Pedersen, K.S., Martinez, M., Perez-Cano, I., Munoz, M., Meseguer, M., Embryo quality,

9 128 C Wong et al. blastocyst and ongoing pregnancy rates in oocyte donation patients whose embryos were monitored by time-lapse imaging. J. Assist. Reprod. Genet. 28, Danuser, G., Computer vision in cell biology. Cell 147, De Placido, G., Wilding, M., Strina, I., Alviggi, E., Alviggi, C., Mollo, A., Varicchio, M.T., Tolino, A., Schiattarella, C., Dale, B., High outcome predictability after IVF using a combined score for zygote and embryo morphology and growth rate. Hum. Reprod. 17, Diamond, M.P., Willman, S., Chenette, P., Cedars, M.I., The clinical need for a method of identification of embryos destined to become a blastocyst in assisted reproductive technology cycles. J. Assist. Reprod. Genet. 29, Dziura, B., Quinn, S., Richard, K., Performance of an imaging system vs. manual screening in the detection of squamous intraepithelial lesions of the uterine cervix. Acta Cytol. 50, Filho, E.S., Noble, J.A., Wells, D., A review on automatic analysis of human embryo microscope images. Open Biomed. Eng. J. 4, Frigault, M.M., Lacoste, J., Swift, J.L., Brown, C.M., Live-cell microscopy tips and tools. J. Cell Sci. 122, Fujimori, T., Kurotaki, Y., Miyazaki, J., Nabeshima, Y., Analysis of cell lineage in two- and four-cell mouse embryos. Development 130, Gardner, R.L., The early blastocyst is bilaterally symmetrical and its axis of symmetry is aligned with the animal vegetal axis of the zygote in the mouse. Development 124, Gardner, R.L., Specification of embryonic axes begins before cleavage in normal mouse development. Development 128, Gonzales, D.S., Pinheiro, J.C., Bavister, B.D., Prediction of the developmental potential of hamster embryos in vitro by precise timing of the third cell cycle. J. Reprod. Fertil. 105, 1 8. Grisart, B., Massip, A., Dessy, F., Cinematographic analysis of bovine embryo development in serum-free oviduct-conditioned medium. J. Reprod. Fertil. 101, Guerif, F., Lemseffer, M., Leger, J., Bidault, R., Cadoret, V., Chavez, C., Gasnier, O., Saussereau, M.H., Royere, D., Does early morphology provide additional selection power to blastocyst selection for transfer? Reprod. Biomed. Online 21, Hardarson, T., Lofman, C., Coull, G., Sjogren, A., Hamberger, L., Edwards, R.G., Internalization of cellular fragments in a human embryo: time-lapse recordings. Reprod. Biomed. Online 5, Hashimoto, S., Kato, N., Saeki, K., Morimoto, Y., Selection of high-potential embryos by culture in poly(dimethylsiloxane) microwells and time-lapse imaging. Fertil. Steril. 97, Hesters, L., Prisant, N., Fanchin, R., Mendez Lozano, D.H., Feyereisen, E., Frydman, R., Tachdjian, G., Frydman, N., Impact of early cleaved zygote morphology on embryo development and in vitro fertilization embryo transfer outcome: a prospective study. Fertil. Steril. 89, Hiiragi, T., Solter, D., First cleavage plane of the mouse egg is not predetermined but defined by the topology of the two apposing pronuclei. Nature 430, Horsthemke, B., Ludwig, M., Assisted reproduction: the epigenetic perspective. Hum. Reprod. Update 11, Kallen, B., Finnstrom, O., Lindam, A., Nilsson, E., Nygren, K.G., Olausson, P.O., 2010a. Blastocyst versus cleavage stage transfer in in vitro fertilization: differences in neonatal outcome? Fertil. Steril. 94, Kallen, B., Finnstrom, O., Lindam, A., Nilsson, E., Nygren, K.G., Otterblad, P.O., 2010b. Congenital malformations in infants born after in vitro fertilization in Sweden. Birth Defects Res. A Clin. Mol. Teratol. 88, Kirkegaard, K., Agerholm, I.E., Ingerslev, H.J., 2012a. Time-lapse monitoring as a tool for clinical embryo assessment. Hum. Reprod. 27, Kirkegaard, K., Hindkjaer, J.J., Grondahl, M.L., Kesmodel, U.S., Ingerslev, H.J., 2012b. A randomized clinical trial comparing embryo culture in a conventional incubator with a time-lapse incubator. J. Assist. Reprod. Genet. 29, Kirkegaard, K., Hindkjaer, J.J., Ingerslev, H.J., 2012c. Human embryonic development after blastomere removal: a time-lapse analysis. Hum. Reprod. 27, Krause, W., Viethen, G., Quality assessment of computer-assisted semen analysis (CASA) in the andrology laboratory. Andrologia 31, Lemmen, J.G., Agerholm, I., Ziebe, S., Kinetic markers of human embryo quality using time-lapse recordings of IVF/ICSI-fertilized oocytes. Reprod. Biomed. Online 17, Lozano, R., Comparison of computer-assisted and manual screening of cervical cytology. Gynecol. Oncol. 104, Lundin, K., Bergh, C., Hardarson, T., Early embryo cleavage is a strong indicator of embryo quality in human IVF. Hum. Reprod. 16, Manipalviratn, S., DeCherney, A., Segars, J., Imprinting disorders and assisted reproductive technology. Fertil. Steril. 91, Meseguer, M., Herrero, J., Tejera, A., Hilligsoe, K.M., Ramsing, N.B., Remohi, J., The use of morphokinetics as a predictor of embryo implantation. Hum. Reprod. 26, Meseguer, M., Kruhne, U., Laursen, S., Full in vitro fertilization laboratory mechanization: toward robotic assisted reproduction? Fertil. Steril. 97, Mio, Y., Maeda, K., Time-lapse cinematography of dynamic changes occurring during in vitro development of human embryos. Am. J. Obstet. Gynecol. 199, 660.e1 660.e5. Montag, M., Liebenthron, J., Koster, M., Which morphological scoring system is relevant in human embryo development? Placenta 32, S252 S256. Motosugi, N., Bauer, T., Polanski, Z., Solter, D., Hiiragi, T., Polarity of the mouse embryo is established at blastocyst and is not prepatterned. Genes Dev. 19, Nakahara, T., Iwase, A., Goto, M., Harata, T., Suzuki, M., Ienaga, M., Kobayashi, H., Takikawa, S., Manabe, S., Kikkawa, F., Ando, H., Evaluation of the safety of time-lapse observations for human embryos. J. Assist. Reprod. Genet. 27, Niemitz, E.L., Feinberg, A.P., Epigenetics and assisted reproductive technology: a call for investigation. Am. J. Hum. Genet. 74, Papanikolaou, E.G., Camus, M., Kolibianakis, E.M., Van Landuyt, L., Van Steirteghem, A., Devroey, P., In vitro fertilization with single blastocyst-stage versus single cleavage-stage embryos. N. Engl. J. Med. 354, Papanikolaou, E.G., D 0 Haeseleer, E., Verheyen, G., Van de Velde, H., Camus, M., Van Steirteghem, A., Devroey, P., Tournaye, H., Live birth rate is significantly higher after blastocyst transfer than after cleavage-stage embryo transfer when at least four embryos are available on day 3 of embryo culture. A randomized prospective study. Hum. Reprod. 20, Payne, D., Flaherty, S.P., Barry, M.F., Matthews, C.D., Preliminary observations on polar body extrusion and pronuclear formation in human oocytes using time-lapse video cinematography. Hum. Reprod. 12, Piotrowska, K., Zernicka-Goetz, M., Role for sperm in spatial patterning of the early mouse embryo. Nature 409, Piotrowska, K., Wianny, F., Pedersen, R.A., Zernicka-Goetz, M., Blastomeres arising from the first cleavage division have