Assisted reproductive technologies in the common marmoset: an integral species for developing nonhuman primate models of human diseases

Abstract

Generation of nonhuman primate models of human disease conditions will foster the development of novel therapeutic strategies. Callithrix jacchus, or the common marmoset, is a New World, nonhuman primate species that exhibits great reproductive fitness in captivity with an ovarian cycle that can be easily managed with pharmacological agents. This characteristic, among others, provides an opportunity to employ assisted reproductive technologies to generate embryos that can be genetically manipulated to create a variety of nonhuman primate models for human disease. Here, we review methods to synchronize the marmoset ovarian cycle and stimulate oocyte donors, and compare various protocols for in vitro production of embryos. In light of advances in genomic editing, recent approaches used to generate transgenic or genetically edited embryos in the marmoset and also future perspective are reviewed.

Summary Sentence

This review presents a summary of assisted reproductive technologies utilized to generate in vitro produced embryos in the common marmoset and describes prospects for genomic editing and transgenesis to create nonhuman primate models of human disease.

Introduction

Nonhuman primates have become exceedingly important for human disease modeling to not only investigate disease progression at a mechanistic level, but also for developing treatments and therapies. The common marmoset (Callithrix jacchus) has novel attributes which position this nonhuman primate species as integral for research in neuroscience [1–3] and immunology [4] among other areas of biomedical research [5–7]. Marmosets are a unique primate model in that they are small in size (350–450 g), relatively easy to handle, and live within families. An advantage to utilizing marmosets over other nonhuman primate species is their robust reproductive capacity: females reach sexual maturity around 18 months of age, ovulate two or more oocytes per cycle, and have a reproductive cycle which can be readily managed with exogenous pharmacological agents. Although assisted reproductive techniques are well developed in other nonhuman primates, such as rhesus macaques [8, 9], the marmoset model is unique in that multiple offspring reach maturity sooner and can be generated in a relatively short time frame. Marmosets are therefore well suited for human disease modeling experiments.

Advances in genetic technology and genomic editing provide opportunities for the development of powerful new models of human disease in a spectrum of species, including nonhuman primates. To create an animal model by way of genetic modifications, in vitro production of embryos is a critical component for either the isolation of stem cells or manipulation of the embryonic genome to create genetically modified offspring. Understanding of the marmoset ovarian cycle and improvements to in vitro methods has enhanced embryonic development rates. Methodology varies greatly, however, across research groups and comparison of published protocols may reveal optimal approaches that can be tailored to specific applications. Herein, we review in vitro production of marmoset embryos including oocyte donor stimulation, in vitro fertilization (IVF), embryo culture, and embryo transfer, as well as highlight recent approaches to generating marmoset models of human disease.

Ovarian cycle synchronization and stimulation

A central feature of the marmoset ovarian cycle, which assigns it tremendous value for assisted reproductive technologies, is the ability to enforce ovarian cycle synchronization of oocyte donors and embryo recipients. The common marmoset can have an average ovarian cycle length of about 28 days, with a follicular phase lasting about 8 days and the luteal phase about 19 days [10], but luteal phase length can be highly variable if not controlled [11]. By evaluating serum progesterone concentrations, the phase of the cycle can be determined with levels of serum progesterone <10 ng/ml correspond to the follicular phase, while concentrations >10 ng/ml are indicative of the luteal phase [10]. The day preceding a rise in serum progesterone to levels >10 ng/ml is deemed the day of ovulation [10]. An advantage of the marmoset compared to other nonhuman primates is the ability to regulate the ovarian cycle by administering a prostaglandin F_2α analog, cloprostenol (Estrumate), to induce luteal regression [12]. The timing of ovulation can also be controlled in the marmoset, as Hodges et al. [13] demonstrated that intramuscular administration of 50 IU human chorionic gonadotropin (hCG), 7 days following Estrumate induced ovulated within days 9–10 post-Estrumate. The facile manipulation of the marmoset ovarian cycle with pharmacological agents offers a distinct advantage over other nonhuman primates, as fewer animals are needed to achieve synchronization between oocyte donors and embryo recipients.

Original experiments that aimed to produce marmoset embryos by IVF did not attempt to stimulate the ovary to increase the pool of preovulatory follicles, but rather performed timed laparotomy following a synchronization protocol of Estrumate, and then stimulation of ovulation with hCG. Figure 1 presents a schematic summary of published protocols in which optimal timing of pharmacological agents was administered to oocyte donors for ovarian cycle synchronization and stimulation. A pioneering study by Lopata et al. [14] was the first to evaluate the timing of delivery of Estrumate and hCG for controlling the ovarian cycle prior to oocyte retrieval. It was found that hCG given on day 8 versus day 7 post-Estrumate led to ovulation prior to laparotomy. The optimal regimen consisted of an injection of hCG at 5 pm, 7 days after Estrumate followed by laparotomy 24 h after hCG, which yielded a mean of 2.2 oocytes per animal from follicles that were 2.5 mm in diameter. Follicles greater than 2.5 mm in diameter were deemed preovulatory, as the oocytes derived from these follicles had expanded cumulus cells and thus were more mature. A subsequent study by Wilton et al. [15] further optimized this protocol by modifying the time of hCG injection to 1 pm on day 7 post-Estrumate, followed by laparotomy to aspirate follicles 21–24 h post-hCG. The latter study was the first to establish a grading system for marmoset oocyte maturation following follicle aspiration. Consistent with the previous study by Lopata et al. [14], about 2.7 follicles of preovulatory size (>2 mm diameter) were aspirated per animal, and increased follicle size corresponded to a more mature oocyte. Overall, these studies provided a time course for the best recovery of oocytes within the follicular phase, and established that follicles of >2.5 mm were preovulatory follicles containing oocytes of greater maturity than smaller follicles.

Figure 1.

Cycle synchronization and ovarian stimulation protocols of the marmoset. Each timeline represents the time course of synchronization or stimulation prior to laparotomy (Lap) or removal of the ovaries to aspirate follicles. Preparations of FSH include ¹Fertinorm HP75, Serono; ²Gonal-f, Serono; ³Folyrmon-P, Fuji Pharma. Bracketed numbers denote reference citations.

The innate limited abundance of preovulatory follicles per cycle from these approaches necessitated the integration of alternative approaches to increase the oocyte population at the time of retrieval. In a study by Gilchrist et al. [16], ovaries were removed and antral follicles were excised to investigate the effects of follicle-stimulating hormone (FSH) priming on oocyte competence of in vitro matured oocytes. Interestingly, this was the first study to report administration of FSH in the marmoset, in which a dose of 1.5 IU human FSH (hFSH) was delivered twice daily from days 0 to 3 of the follicular phase. While the FSH dosage was equivalent to the dosage administered to rhesus monkeys, and twice that administered during human IVF stimulation, FSH priming had no effect on the proportion of cumulus oocyte complexes retrieved [16]. More oocytes could be obtained, however, by simply removing the ovary and dissecting individual follicles to obtain oocytes from follicles of all sizes.

Ovarian hyperstimulation protocols

To further increase the population of preovulatory follicles prior to retrieval without removal of the ovaries, subsequent studies integrated recombinant hFSH (r-hFSH) into ovarian synchronization protocols to stimulate follicular development. Superovulation was assessed by Marshall et al. [17] in which a dose of 1–50 IU r-hFSH per day was administered for 5–6 days followed by timed hCG injection and laparotomy. The response to FSH was not different between control oocyte donors and those administered twice daily doses of 1, 10, or 25 IU r-hFSH, as the mean number of preovulatory follicles plus ovulation points was 2.9. Females administered 50 IU r-hFSH, however, demonstrated a response to stimulation with a mean of 14.1 preovulatory follicles (>2 mm) per animal. Oocyte donors were stimulated up to three times with a 50 IU r-hFSH protocol and had increased follicle populations over control animals, although there was a trend for a decreased response to r-hFSH with each stimulation [17]. Similarly, a study by Grupen et al. [18] also utilized doses of 25 IU r-hFSH twice daily, but increased the dose of hCG to 500 IU to investigate in vivo maturation. Interestingly, the high dosage of hCG in comparison with animals who did not receive hCG resulted in an increased proportion of expanded cumulus–oocyte complexes to about 23.5 per animal from follicles of >1.5mm diameter. Moreover, extension of FSH administration to 9 days, and reduction to once daily injection of a 50 IU r-hFSH dose, does not dramatically reduce the ovarian response to FSH. Takahashi et al. [19] reported collection of 104 in vitro matured oocytes derived from 12 female marmosets, equating to a mean of 8.67 oocytes per donor. Taken together, these studies revealed that a relatively high dose of FSH compared to humans and macaques is required in the marmoset to facilitate ovarian stimulation, thereby increasing oocyte yield for subsequent in vitro embryo production.

Oocyte maturation

Oocyte maturity at the time of follicular aspiration is associated with follicle size, as more mature oocytes are obtained from larger follicles. Excision of antral follicles of all sizes and subsequent oocyte maturation revealed that those aspirated from follicles greater than 1 mm in diameter are often cumulus enclosed [16]. Interestingly, germinal vesicle break down and MII competency increased with follicle size regardless of the presence of cumulus cells [16, 20], where 68.9% of cumulus enclosed complexes and 63.2% of denuded oocytes completed meiotic maturation within 24 h of in vitro maturation [16]. Notably, oocytes derived from smaller follicles that fail to complete in vitro maturation have abnormal spindles.

Regardless of stimulation regimen, the length of culture needed prior to fertilization is determined by oocyte maturity status at the time of aspiration. When a high dose of hCG was administered to assess whether in vivo maturation could be achieved, it was found that a higher proportion of expanded cumulus oocyte complexes were recovered. The proportion of oocytes to reach MII, however, was not different compared to those that did not receive an hCG injection [18]. It has been shown that oocytes with expanded cumulus cells require a shorter duration of culture prior to fertilization in comparison to less mature oocytes. For example, Wilton et al. [15] assigned a maturity grade to each oocyte and then assessed the duration of oocyte preincubation required as evidenced by fertilization rates. Oocytes with extensive but not fully expanded cumulus cells had the highest fertilization rate at 2–5 h of incubation, whereas oocytes that were less mature required a longer preincubation period to achieve higher fertilization rates. Indeed, Marshall et al. [17] recovered oocytes at the germinal vesicle stage and 85% of those oocytes extruded a polar body at 24–26 h of culture. This observation is also in concordance with the study by Takahashi et al. [19], in which polar body extrusion was similarly observed between 20 and 24 h of in vitro maturation.

In comparison to in vitro maturation duration, there is greater disparity in culture conditions across studies reporting in vitro oocyte maturation. Fertilization was reported following oocyte maturation in higher ambient oxygen levels [15–18] and also for oocytes matured in a low oxygen environment [19, 21], although the authors of the Wilton et al. [15] study claimed that there was no difference in oocyte maturity following incubation in either high or low oxygen conditions. Basal media and supplemented components also varied greatly, although all aforementioned studies had supplemented media with FBS. To support in vitro maturation, media have been supplemented with FSH [16, 18, 19] as well as luteinizing hormone (LH) [16], hCG [18], epidermal growth factor (EGF) [22], and estradiol [16, 21, 23]. Medium optimization has been a focus, where, for example, experimentation with a chemically defined medium, porcine oocyte medium compared to Waymouth medium resulted in fewer degraded oocytes during the maturation process [21]. Optimization and progression to a standardized maturation medium across research laboratories would allow direct comparison of stimulation and retrieval protocols.

Semen collection and fertilization

A limiting factor to IVF for producing marmoset embryos is the timely accessibility of a quality semen sample. Semen collection in the marmoset was initially performed in a similar manner as Old World primates, by rectal probe electrostimulation to obtain a sample for IVF [14] or semen analysis [24]. However, this method requires administration of anesthesia and is relatively invasive. Collection by electroejaculation was reported to negatively impact the ability of spermatozoa to separate from the coagulum following collection [15]. Alternatively, other noninvasive approaches can be used to obtain a semen sample for IVF, including isolation of epididymal sperm [16] and collection via vaginal washing after copulation [25]. Successful semen collection in squirrel monkeys by penile vibratory stimulation (PVS) led to the trial of this method in the marmoset, where 35% of attempts led to ejaculate collection [26]. Similar rates of collection were achieved by males housed singly or paired; however, there were significant increases in the total number of sperm and concentration in singly house males [26]. Further modification to the PVS collection technique improved the positive collection rate to 89.2%, where comparable electroejaculation success rates in the same study were 30.4% [27]. Collection by PVS in comparison to electroejaculation also yielded samples of greater concentration, and percentage of both viable and motile spermatozoa. Since the first success reports of PVS semen collection in the marmoset, this has become the preferred method for semen collection.

Following collection, various protocols have been utilized to isolate a viable, motile fraction of spermatozoa for IVF. Addition of media following collection facilitates dispersal of the coagulum from the spermatozoa within a 30-min incubation. Initial semen preparation protocols were modeled after macaques in which dibutryl cAMP and caffeine were added to the basal medium, minimal essential medium, or Tyrode's albumin lactate pyruvate, to promote capacitation [14–16, 28]. A protocol adapted from rhesus macaques [29] revealed that ejaculated spermatozoa compared to epididymal spermatozoa have a greater sensitivity to dbcAMP and caffeine, where increased hyperactivity was observed followed by a rapid decline in motility. Likewise, Marshall et al. [17] reported a fertilization rate as low as 57% when oocytes were fertilized with ejaculated spermatozoa prepared in a similar manner. The studies by Gilchrist et al. [16] and Marshall et al. [17] incubated the sperm for 2–3 h to allow for the viable, motile sperm to swim up into the upper phase of the media, which was the fraction used for fertilization. More recently, a fertilization rate of 82.2% was achieved in which oocytes were co-incubated with fresh ejaculated semen following a 30-min swim-up incubation without adding dbcAMP or caffeine to promote capacitation [19].

An alternative approach to prepare semen is placement of the liquefied ejaculate, following the initial 30-min incubation, over a density gradient of Percoll or PureSperm, followed by centrifugation [30]. An advantage to this method is that the gradient will capture potential contaminants and debris, while pelleting the viable, motile spermatozoa. The study by Lopez et al. [30] who first reported the assessment of a density gradient to prepare sperm did not incorporate caffeine or dbcAMP into the medium, but rather referenced an abstract that suggests the addition of bovine serum albumin to Tyrode's lactate (TL) medium is sufficient for inducing capacitation in the marmoset. Optimal medium and agents for capacitation remain to be elucidated in the marmoset.

Fertilization by methods of gamete co-incubation and intracytoplasmic sperm injection (ICSI) have both proven successful in the marmoset. Across studies, the time point for IVF is largely dependent on the maturity of the oocytes. At the time of gamete co-incubation, addition of sperm to a final concentration ranges from 0.5 to 5.0 × 10⁶ sperm/ml, and the length of gamete co-incubation varies between 14 and 24 h. In an attempt to improve fertilization rates, ICSI has also been implemented, with the first report by Grupen et al. [18], and subsequently, by Takahashi et al. [19], which was the first study to report live offspring following ICSI as fertilization method. Time course experiments revealed that higher fertilization rates were achieved when ICSI was performed at least 2 h after first polar body extrusion, at about 20–24 h of in vitro maturation in this study [19]. A comparison between IVF and ICSI fertilization methods within the same study demonstrated that despite a significant increase in fertilization rates achieved with ICSI, the percent that developed to the blastocyst stage was no different [19].

Selection of males whose spermatozoa perform well within an in vitro system and long-term preservation of marmoset semen from those males remains a challenge. Several studies report a high degree of variation within and across males. Collection by PVS yielded ejaculates that ranged in volume from 10 to 79 μl, total sperm/ejaculate from 4.4 to 138.4 × 10⁶, and on average the viability was 74.6% whereas motility averaged 59.6% [26]. A study by Cui et al. [24] similarly reported high variability within and across ejaculates of males, in addition to finding three common head defects and eight tail defects across ejaculates. Considerable variation within and across ejaculates of males was similarly reported following computerized sperm analysis [31]. Aside from analysis of viability, sperm concentration, and motility for a specific ejaculate, selection of males for gamete co-incubation or ICSI has not been well developed. A method for staining the acrosome reaction has been developed in the marmoset [32], although methods for characterizing an individual males’ ability to capacitate in vitro have yet to be elucidated. Upon selection of males that perform well in vitro, a method for semen freezing in the marmoset would be of great utility. Freezing of marmoset semen has been reported [33]; however, motility is relatively low (40%–50%) post-thaw [34]. Overall, experiments are needed to better optimize the selection of males for in vitro production of embryos, in addition to developing an improved semen freezing method to ensure the availability of a semen sample.

Embryo culture and development

Embryonic development is comparatively slower in the marmoset relative to other nonhuman primates, including cynomolgus macaques [35], rhesus macaques [8, 29, 36, 37], African green monkeys [38], squirrel monkeys [39], and humans [40] (Table 1). Uterine flushings on days 7 and 8 of pregnancy recovered a greater proportion of morula and few early blastocysts, whereas on day 9 a mixed population of early and expanded blastocysts was recovered [41]. Collection of embryos from the oviduct and the uterus from days 4–11 of pregnancy suggested that the 10–16 cell-stage embryos pass from the oviduct to the uterus around day 5 after ovulation [41]. Likewise, the blastocyst stage is reached around days 9 and 10 of in vitro culture. Figure 2 shows the time course of each developmental stage in vivo and in vitro as well as representative images of each stage of marmoset embryonic development in vitro.

Table 1.

Comparison of in vitro embryonic development rates in humans and nonhuman primate species.

	Human [39]	Cynomolgus macaque [33]	Rhesus macaque [27, 34–36]	African Green Monkey [37]	Squirrel Monkey [38]	Marmoset [14, 15]
D1	2 pronuclei		2 cells	2 cells	2 cells	2 cells
D2	2–4 cells		4–8 cells	2–4 cells	4 cells	2–4 cells
D3	8–16 cells	>8 cells	8 cells	4 cells	8 cells	4–8 cells
D4	Compacted morula		16 cells	4–8 cells	Blastocyst	8–16 cells
D5	Blastocyst	Morula	Morula	8–16 cells	Blastocyst	16 cells Morula
D6	Hatching blastocyst	Morula blastocyst	Early blastocyst	Morula		Morula
D7	Implantation	Blastocyst	Blastocyst- expanded blastocyst	Early blastocyst- blastocyst		Morula blastocyst
D8		Blastocyst	Expanded blastocyst	Blastocyst		Blastocyst
D9		Blastocyst	Hatching blastocyst	Blastocyst- expanded blastocyst		Blastocyst
D10				Expanded blastocyst		Blastocyst
D11						Expanded blastocyst

*Bracketed numbers denote reference citations.

Figure 2.

Time course of in vitro embryonic development in the marmoset. In vitro day 0 is the day of insemination, whereas in vivo day 0 is the day of ovulation. Data are summarized from in vitro studies including Gilchrist et al. [16] and Marshall et al. [17] and in vivo studies including Harlow et al. [10] and Summers et al. [41].

Overall, in vitro developmental rates have improved as stimulation and fertilization methods were optimized. Table 2 summarizes developmental rates reported across studies in which IVF was performed with ejaculated semen, and in one study where ICSI was performed. Across studies, the percentage of fertilized oocytes that cleaved was relatively high with a degree of developmental arrest at each stage. Indeed, the authors of the study by Grupen et al. [18] observed that the highest percentage of embryos arrested during the third and fourth division, and suggest arrest could be explained by failure to undergo embryonic genome activation. As marmoset embryos require an extended length of culture, improvements in developmental rates may be achieved by a better understanding of how to support development by optimizing long-term embryo culture conditions.

Table 2.

Summary of in vitro embryonic development rates.

Study	Oocyte (n)	% Fertilized	% Cleaved	% 8 cell	% 16 cell	% Morula	% Blastocyst
Gilchrist et al. 1997[20]	33	33	30	24	12	9	0
Marshall et al. 2003[17]	21	57	84	70	–	58	47
Grupen et al. 2007[18]	57	66.9	44.7	–	–	–	56.7
Tomioka et al. 2012[21]	75	60	95.6	–	–	17.8	13.3
Takahashi et al. 2014[19]	IVF 90	82.2	93.2	83.8	64.9	48.6	39.2
	ICSI 88	93.2	97.6	76.8	52.4	39	35.4

*All studies reported here utilized ejaculated semen for fertilization with the exception of the ICSI example. Bracketed numbers denote reference citations.

Similar to both oocyte maturation and the fertilization methods, embryo culture conditions vary greatly across studies. The methodology is similar in that several studies utilize a sequential or two-step medium system; however, the medium employed for the initial culture period and transition to the later culture period is not consistent across studies. These medium transitions include TL to CMRL-1066 [16], G1.2 to G2.2 (Vitrolife, Göteburg, Sweden) [17, 18], and ISM1 to ISM2 (Origio, Måløv, Denmark) [19, 21]. Embryos were cultured in the first medium for about 48 h (days 1–3 of culture) and then transferred to the second medium with replacement about every 2 days. Embryo culture in high oxygen as well as low oxygen has been reported; however, both conditions have not been evaluated in a single study to determine the impact of oxygen on marmoset embryonic development. Interestingly, a 47% blastocyst rate was reported for embryos cultured in higher ambient air conditions [17], whereas a 39.2% blastocyst rate was obtained for embryos cultured in a low oxygen condition [19]. Future studies should characterize optimal culture media and gas conditions for marmoset embryo culture, so that a standard system may be utilized allowing for comparisons of embryonic development outcomes across studies.

Embryo transfer methods

Transfer of marmoset embryos to recipient females has been achieved by both surgical and nonsurgical methods. Earliest attempts at embryo transfer by surgical delivery of fresh morula and blastocysts to the uterine lumen at the fundus resulted in a 66.6% pregnancy rate by day 40, with 10 young born from 11 transferred embryos; however, nonsurgical methods were deemed unsuccessful [41]. Interestingly, pregnancy rates were higher following surgical transfer (to the uterine lumen) of 4–10 cell-stage embryos with 100% of the recipients becoming pregnant while 50% of the recipients became pregnant following transfer of morula. In the same study, transfer of embryos that were frozen in DMSO was successful; however, pregnancy rates were reduced to 50% for 4–10 cell-stage embryos compared to a 100% for unfrozen controls.

Alternatively, embryo transfer by nonsurgical methods is less invasive. A study by Marshall et al. [42] was the first report of successful embryo transfer by a nonsurgical method, in which the cervix was cannulated and a needle containing the embryo(s) is then passed through the cannula. Pregnancy and birth rates were higher for embryos that were transferred to recipients that were asynchronous (44% pregnancy rate and 35% birth rate) to the oocyte donor, compared with those that were synchronous (9% and 0%). Asynchronous recipients were those females 2–4 days postovulation, whereas transfers to recipients synchronous to the oocyte donors were performed 5–8 days postovulation. A study by Ishibashi et al. [43] sought to assess the effect of transfer volume, cryopreservation, and embryo stage on transfer success following nonsurgical embryo transfer. Despite transfer of relatively few embryos (n = 23), the authors concluded that there was no effect of embryo stage, cryostorage, or recipient timing on conception rates. When the transfer volume was 1 μl or less, however, the conception rate was improved by 30% to a pregnancy rate of 80%. Although a volume of less than 1 μl of was suggested to improve conception rates, a later study by Takahashi et al. [19] reported successful birth rates when embryos were transferred in a volume of 2 μl. Moreover, birth rates were significantly higher when six to eight earlier stage cell embryos (28.6%) were transferred compared to blastocysts (2.7%) [19]. Collectively, observations from these studies suggested that transfer of one to three early-stage embryos in a volume of up to 2 μl to a recipient that is asynchronous to the oocyte donor may improve pregnancy and birth rates following embryo transfer.

Nonhuman primate models of human disease through transgenesis and genomic editing

Development of nonhuman primate models of human disease is currently receiving significant attention, particularly following recent breakthroughs in embryonic genomic editing. Techniques to create genetically modified nonhuman primates, by way of viral vectors or through genomic editing, have been piloted in macaques and the marmoset. Transgene and editing components are delivered to either the oocyte or zygote during the in vitro process, followed by transfer of embryos to a surrogate dam, and offspring from successful pregnancies are then screened for incorporation of the genetic alteration of interest. Here, we review approaches to creating nonhuman primate models for human diseases in macaques, and also pioneering studies in the marmoset. The reader is also referred to other recent reviews on this topic [6, 44, 45].

The generation of transgenic nonhuman primate models was first achieved utilizing viral vectors delivered to macaque oocytes and embryos. Microinjection of a human immunodeficiency virus-based lentiviral vector expressing enhanced green fluorescent protein (EGFP) into blastocoel cavity of rhesus macaque embryos followed by embryo transfer resulted in EGFP transgene expression restricted to placental tissues, suggesting that viral infection was limited to the trophectoderm [46]. Alternatively, microinjection of a Moloney murine leukemia virus retroviral vector expressing EGFP into rhesus macaque oocytes followed by ICSI and transfer of four-cell-stage embryos resulted in the birth of three live offspring, one of which exhibited transgene expression in all tissues analyzed [47]. In another study, delivery of a simian immunodeficiency virus-based vector expressing EGFP at early stages of embryo development achieved transgene expression in somatic tissues, whereas injection into four to eight cell embryos demonstrated that offspring were chimeric for transgene expression [48]. Similarly, injection of a lentiviral vector expressing EGFP into mature cynomolgus monkey oocytes prior to fertilization resulted in homongeneous transgene expression throughout tissues of the offspring, whereas partial EGFP expression was observed in a fetus derived from an embryo injected postfertilization as a zygote [49]. These pioneering studies in macaques demonstrate that transgene expression can be achieved through delivery of viral vectors to macaques IVF embryos.

Furthermore, studies utilizing viral vector delivery of a mutated form of a human gene produced transgenic offspring with phenotypes resembling the predicted human disease conditions. Yang et al. [50] leveraged lentiviral transgene technology to introduce a vector expressing the human Huntington (HTT, previously known as HD) gene containing expanded (84) CAG repeats, associated with development of the disease, into mature oocytes followed by in vitro production of embryos. Interestingly, a range of 27–88 repeats were observed in the five offspring born from the injected oocytes, and individuals with higher copy number of mutant HTT with greater expanded repeats were associated with early death, or demonstrated clinical symptoms of the disease in comparison to an apparently normal animal with 29 CAG repeats (within the normal range of non-Huntington humans). Another example of transgenesis in macaques was reported by Niu et al. [51] in which a lentiviral vector expressing mutant alpha-synuclein was introduced into matured oocytes to generate transgenic offspring as a potential nonhuman primate model of Parkinson disease. The six transgenic live offspring generated in this study expressed the mutated gene in the brain, and exhibited anxiety and cognitive defects, which the authors suggested resemble early symptoms of Parkinson disease in humans. In another study, transgene integration of the human methyl-CpG binding protein 2 (MECP2) gene in cynomolgus macaques varied in copy number from 1.0 to 7.3 [52]. Duplication of the gene in humans is associated with symptoms of autism spectrum disorder [52]. The transgenic MeCP2 monkeys exhibited higher levels of stress and less social interaction; however, the authors suggest that there is no evidence of correlation between the degree of social behavior and copy number. Collectively, these studies established that transgenesis can be achieved in the nonhuman primate by injection of viral vectors in the oocyte or early embryo and that the disease-associated transgene can impact physiology of the resulting offspring; however, the degree of mosaicism and variation in copy number are limitations that need to be addressed in the future.

Generation of transgenic marmosets has also been accomplished by introduction of lentiviral vectors expressing the EGFP gene into embryos, which resulted in birth of offspring with transgene expression in epidermal cells of the ear and stromal cells of the placenta [53]. This study also demonstrated the significant advance of germline transmission in offspring following IVF with semen of a transgenic male, in addition to transgene expression in a naturally produced embryo produced by a transgenic female [53]. The lentiviral transgenesis approach is advantageous in that lentiviral vectors can be injected into the perivitelline space of an embryo at any stage of development with good posttransduction embryo viability reported in several species [54]. This approach could also be extended to flushed embryos, as nonsurgical embryo collection has been developed for the marmoset [55, 56] as well as the macaque [57]. For instance, Sasaki et al. [53] obtained transgenic offspring from both in vitro and naturally produced embryos injected with a lentiviral vector expressing EGFP from the pronuclear to the morula stage of development.

There are limitations to lentiviral delivery of transgene. The lentiviral vector will randomly integrate the gene of interest into the host, and multiple copies may integrate, giving rise to unpredictable expression patterns [54]. Delivery of lentiviral vectors with tissue-specific promoters has been demonstrated in a study by Park et al. [58], where vectors expressing the calcium-binding protein calmodulin fused with GFP (GCaMP) under either a ubiquitous (cytomegalovirus, CMV) or a neuron-specific (human synapsin I) promoter were injected into marmoset embryos. Eight transgenic offspring were generated in which the offspring that had transgene integration of the GCaMP under the neuron-specific promoter expressed GCaMP only in the spinal cord [58]. Although transgene expression can thus be restricted to specific tissues, random transgene integration would still be a disadvantage if the insertion disrupts the function of neighboring genes. Disruption of gene function was suspected in a human gene therapy clinical trial, where, following lentiviral vector delivery to treat a form of Severe Combined Immunodeficiency (SCID) proliferation of a specific T-cell population led to leukemia [59, 60]. Hence, it is important to consider the effect of random insertion when utilizing this approach.

Alternative approaches to create animal models by genomic editing strategies, such as zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENS), and the clustered regularly interspaced short palindromic repeats (CRISPR)-associated protein 9 (Cas9), or CRISPR-Cas9 system, have been leveraged to create knockout and site-specific mutations. All of these technologies create a double-stranded DNA (dsDNA) break, whereby the cell's intrinsic DNA repair mechanisms will be recruited to repair the break by either nonhomologous end-joining or homology-directed repair. The site at which the dsDNA break is created is directed by the intrinsic properties of each technology. For example, cleavage by ZFNs relies on two zinc finger proteins in conjunction with a FokI restriction enzyme, which upon recognition and dimer formation cleaves the target DNA [61, 62]. Similarly, TALENs utilize a FokI nuclease; however, the restriction enzyme is fused to transcription activator-like effector proteins containing a DNA- binding domain comprised of 33–35 amino acid repeats [62–64]. In comparison to ZFNs and TALENs, the CRISPR-Cas9 is comprised of a Cas9 protein complexed with a guide RNA that recognizes a target DNA sequence adjacent to the protospacer adjacent motif, which directs the sequence-specific cleavage of the Cas9 nuclease domains. Although each genomic editing strategy is sufficient to target specific sites of the genome, the relative ease of design and rapid generation of the guide RNA of the CRISPR-Cas9 system makes this approach the preferable option for genomic editing [62, 65].

Recent reports of genomic editing in macaques have illustrated the feasibility of modeling human disease. A mutation in the MECP2 gene associated with an autism spectrum disorder (Rett syndrome), mediated by delivery of three TALEN DNA plasmids into one-cell zygotes of rhesus and cynomolgus macaques, resulted in miscarriage of male fetuses and the birth of a live female offspring, where MECP2 mutations were observed in these fetuses [66]. Rett-linked male embryonic lethality is also observed in humans, indicating a similar phenotype with MECP2 mutations in both the nonhuman primate model and humans.

Reports of TALEN-mediated genomic editing in macaques are limited, whereas CRISPR-Cas9 technology has been more intensively explored to create nonhuman primate models. Niu et al. [67] microinjected zygotes with Cas9 mRNA and either one or two sgRNAs for the genes Ppar-, Rag1, and NrOb1, and transfer of these embryos generated live offspring with various mutations in Ppar- and Rag1. This study demonstrated that multiple genes can be targeted in a single injection of CRISPR-Cas9 components to the early nonhuman primate zygote. Follow-up assessment of aborted fetuses from this study revealed that an aborted male fetus carried a nonsense deletion in the NrOb1 gene, also known as Dax1, in a wide range of somatic tissue [68]. Dax1 mutations are associated with hypogonadotropic gonadism and adrenal hypoplasia congenita in humans. Interestingly, the aborted macaque fetus with extensive mutations in the Dax1 gene displayed lack of Dax1 protein in the testis, and altered testicular and adrenal gland development, revealing the role of Dax1 in development [68]. In another study, CRISPR-Cas9 was utilized to disrupt the macaque homolog of the human dystrophin gene to model human muscular dystrophy, where two stillborn and nine live monkeys carried mutations in the dystrophin gene for a targeting mutation rate of 61.1% [69]. Tissue analysis of the stillborn monkeys revealed a high degree of mosaicism, although the protein levels of dystrophin were greatly reduced compared to controls and the muscle cells appeared similar to young human Duchenne muscular dystrophy patients. Collectively, these studies demonstrate that the CRISPR-Cas9 technology has proven successful for creating gene modifications that underlie phenotypes similarly observed in human disease conditions.

Genomic editing by both ZFN and TALEN strategies has recently been initiated in the marmoset, where microinjection of relevant components into marmoset pronuclear stage embryos by either strategy successfully knocked out the interleukin 2 receptor subunit gamma IL2RG (previously known as CIDX) gene [70]. To date, this study is the only report of genomic editing in the marmoset. It can be predicted that genomic editing to create nonhuman primate models by way of ZFN, TALEN, or CRISPR technology in the future will heavily rely on the availability of embryos, whether they are produced by in vitro methods, or naturally produced and recovered from the reproductive tract. Figure 3 summarizes how IVF or naturally produced marmoset embryos can be manipulated by either microinjection of components of a genomic editing technology or by lentiviral vectors. These embryos can then be transferred to recipient females, from which live offspring carrying the genetic modification may be born. To assess disease therapies, embryonic stem cells or iPS cell lines derived from macaques [71–74] and the marmoset [75–80] can be used in regenerative medicine paradigms. Pluripotent cells derived from the edited offspring can be modified to correct the edited mutation, treated with specific differentiation protocols, and delivered back to the animal, establishing a powerful experimental therapeutic strategy.

Figure 3.

A working model for the manipulation of marmoset embryos and cells. Embryos can be generated in vitro or naturally produced and flushed from the uterus. Microinjection of zygotes or embryos with either lentiviral vectors or gene editing components will facilitate modification to the genome. Transfer of embryos and subsequent birth of the edited offspring may then yield a nonhuman primate model as a result of the specific genetic modification. From the offspring, experimental therapeutic strategies may be implemented to assess treatment of human disease conditions in the nonhuman primate model including iPS cell generation and regenerative medicine approaches. Red arrow denotes genetically modified cells.

Although genomic editing is feasible in the nonhuman primate, these technologies need to be further optimized. Before initial injection of the genomic editing components into the early embryo, multiple constructs specific to the gene of interest should be assessed for targeting efficiency and specificity in a cell culture system. This allows for selection of an optimal construct to be delivered into the early embryo to achieve efficient and specific targeting. Despite injection into the oocyte or one-cell stage, a high degree of mosaicism has been widely reported in tissues of stillborn and live macaque offspring. Additionally, relatively few animals are born from injected embryos that are transferred in examples of macaque genomic editing. Hence, improvement of targeting efficiency is warranted to not only minimize the degree of mosaicism, but also to realistically increase the number of targeted offspring generated. Recently, biallelic mutations in the p53 gene were observed in eight tissues of a cynomolgus macaque offspring that was injected with CRISPR-Cas9 [81], and thus offering promise that both alleles can be targeted at a high rate and that homozygous nonhuman primates for a specific mutation may be derived to model human disease conditions.

Conclusions and future directions

The marmoset presents a unique species in terms of its reproductive patterns that can be exploited to create nonhuman primate models of human disease. The procurement of embryos by in vitro production methods has been a cornerstone for developing nonhuman primate models of disease and these technologies will be integral to future genomic editing in the marmoset. Progress to improve the efficiency of marmoset embryo production has resulted in higher development rates; however, variability in terms of ovarian stimulation protocols and culture conditions makes it difficult to ascertain optimal conditions for marmoset in vitro development. Hence, assessment of multiple variables within a study to optimize conditions and subsequent movement to a standardized methodology would be of great benefit. A severe limitation to the in vitro process is the variability between males and availability of a quality semen sample for fertilization. Timing of the in vitro process negates delay in processing of semen and in the event an ejaculate cannot be obtained, it would be advantageous to have frozen semen; however, this aspect warrants additional exploration. Altogether, future experiments should focus on improvement of fertilization rates and culture conditions to increase the number of embryos that develop and potentially result in delivery of offspring. Recent advances in genome editing, in which editing components are introduced to the one-cell stage embryo, accelerate the need to develop more efficient means of producing marmoset embryos for future development of nonhuman primate models of human diseases.