Generation of germ-free common marmosets


 Recent studies using germ-free mice have demonstrated that microbiota have functional roles in host homeostasis1,2. However, the phylogenetic distance between rodents and humans translates into differences in their metabolism, immune response, neural function and microbiota colonisation abilities. Hence, translational research using nonhuman primates (NHPs) is important for bridging the gap between rodent studies and human medicine3. Although several attempts to produce germ-free NHPs were made more than 50 years ago4,5, currently none are available. Here, we generated germ-free common marmosets suitable for rearing and handling under sterile conditions and maintained them with no culturable bacteria/fungi, for 22 months. The faecal microbiota composition and metabolome in conventional marmosets are more similar to those in humans than to those in mice. The transplantation of a bacterial consortium isolated from humans6 into marmosets and mice resulted in a significantly steadier bacterial colonisation in the former than that in the latter. Germ-free marmosets exhibited low levels of faecal short-chain fatty acids, bile acid metabolites, plasma and faecal immunoglobulins, and enlarged caecum in contrast-enhanced X-ray. These stable germ-free marmosets can serve as novel models that enable the development of therapeutics that target gut microbiota and elucidation of their interaction with higher-order brain function.


66
modulating the host's immune system and influencing host development and physiology, 67 including that of the central nervous system 1,2,7 ). However, extrapolating the findings of these 68 mouse studies to human medicine is hampered by the evolutionary distance between rodents 69 and primates, and the differences in terms of metabolism, immune responses, neural functions, 70 and colonisation ability of microbiota 8 .

71
Non-human primates (NHPs) are critical for translational research to bridge such gaps and 72 are sometimes the only relevant animal models because of their close genetic, physiological, 73 and behavioural similarities with humans 3,9 . Recent research involving NHPs has enhanced our 74 understanding of the host-microbiota interactions in humans 10,11 . However, currently, no GF 75 NHPs are available. More than 50 years ago, a few reports catalogued the acquiring and rearing 76 of GF NHPs, i.e., rhesus (Macaca mulatta) 4,12 and cynomolgus (Macaca fascicularis) 5 monkeys.

77
These preliminary records mention a few individuals who remained GF for short spans of time 78 but do not mention whether the GF status was monitored continuously; thus, it remains 79 controversial whether these primates were indeed born and raised under GF conditions.

80
The common marmoset (Callithrix jacchus) is an important NHP used in biomedical research 81 and in preclinical studies aimed at drug development [14][15][16] . This New World monkey species has 82 some advantages as laboratory models, such as small body size, high fecundity and relatively 83 short life cycle. These advantages make the species relatively easy to breed and handle in a 84 sterile isolator. Furthermore, the marmoset has been developed as a model animal for research 85 in the field of neuroscience 17 and has been used as a system to model various human disease, 86 5 especially those reported to be associated with imbalances in intestinal bacterial flora 18,19 such 87 as Parkinson's 20 , multiple sclerosis 21 , obesity 22 and age-related diseases 23 . Recent progress in 88 transgenic and genome editing technology has further expanded the use of this NHP in 89 research [24][25][26] . Thus, if GF marmosets are available, they can serve as novel NHP models in 90 microbiota research, i.e., investigations of brain-gut microbiota interactions and preclinical 91 studies on microbiota-based therapeutics.

92
The purpose of this study was (i) to investigate the characteristics of the marmoset as an NHP

103
Actinobacteria, although present in marmosets and humans, were rarely detected in mice, e.g.

105
Principal coordinate analysis (PCoA) using operational taxonomic unit (OTU) data sets showed 106 that the composition of the faecal flora in marmosets was similar to that in humans than to that 107 in mice (Fig. 1a).

108
Next, metabolome analysis for primary metabolites, short chain fatty acids (SCFAs) and to human and marmoset metabolomes were separated from those of the mouse groups along the 113 PC1 axis, although those of the faecal metabolome were in proximity for marmosets and mice 114 fed the marmoset diet (Fig. 1b). Among the faecal metabolites, the bile acid composition 115 differed between mice and marmosets/humans, with muricholic acids (MCAs) being primarily 116 detected in mice (Supplementary Table 1). With respect to plasma metabolites, marmosets and 117 humans shared higher levels of amino acids, such as proline, histidine and alanine, and lower 118 levels of polyunsaturated fatty acids compared to those in mice (Supplementary Table 2).

119
In addition, we investigated how intestinal bacterial strains from humans colonised in the 120 marmoset intestines. The 11-strains mixture, which was isolated from healthy human faeces and 121 found to induce interferon-γ-producing CD8 T cells in the intestine 6 , was orally inoculated into 122 marmosets and mice. Quantitative PCR (qPCR) analysis of the relative amounts of bacteria in 123 faecal samples detected all 11 strains steadily for 3 weeks post-inoculation in marmosets, 124 whereas some strains decreased over time in mice (Fig. 1c). These results showed the usability 125 of marmosets as NHPs for microbiota research. Based on this knowledge, we attempted to 126 generate GF marmosets.

128
Obtaining GF marmosets 129 A total of 18 impregnated females were operated upon for obtaining GF newborns (Fig. 2a).

130
First, as with acquiring GF mice or rats, on day 142 post determined ovulation, we performed 131 hysterectomy on a marmoset impregnated with embryo transfers (ET) and delivered a newborn 132 inside a sterile flexible film isolator. Next, we prepared a dedicated isolator to establish a 133 surgical procedure for aseptically obtaining marmoset neonates through caesarean section, 134 which was minimally invasive to dams (Fig. 2b, c, Extended Data Fig. 1). A total of 17 135 caesarean sections were performed on 12 ET-impregnated and 5 naturally impregnated females.

136
Operations were conducted between the 136th and 142nd day post expected ovulation in ET 137 cases. The foetal biparietal diameter, measured using ultrasonography, was found to be 17.4-7 20.3 mm 10 days prior to surgery (Extended Data Table 1). Of the 25 newborn marmosets 139 obtained from 18 operations, we successfully resuscitated 23. The time from the induction of 140 anaesthesia to delivery ranged from 13 to 30 min in the resuscitated cases, whereas it was 35 141 min in the non-resuscitated cases. All operated females recovered without apparent problems 142 after surgery.

143
Of the 23 resuscitated marmosets, 7 neonates were weaned by hand-rearing in sterile 144 isolators (Fig. 2a); all weaned animals were singletons and the four newborns from the latest 145 operation cases were weaned by improving hand-rearing procedures such as modifying milk 146 formula and careful maintenance of body temperature (Supplementary Table 3

152
and Staphylococcus warneri was detected in one dead individual (Extended Data Table 2).

153
Thus, two GF (culture negative) females, three GF males, and two mono-colonised (MC) 154 males were obtained and maintained in sterile isolators (Fig. 3a). Monthly culture tests of their 155 faecal and isolator swabs were negative for up to 22 months (Fig. 3b); Staphylococcus aureus 156 was detected in the animal (881M) that had the longest GF state at 22 months old when pinholes 157 were observed on the gloves of the isolator. A male-female pair (939M and 795F) were 158 maintained culture-negative for more than 15 months prior to manuscript submission (Fig. 3b).

159
Furthermore, daily administration of a sensitive antibiotic, kanamycin, caused MC animals 160 (905M, 926M, 792F and 947M) to be negative for culture tests (Fig. 3b). All the marmosets 161 reared under sterile conditions grew without apparent problems and gained body weights similar 162 to those in conventionally reared animals (Fig. 3c).

165
Faecal metabolome analysis revealed that GF and MC marmosets had as few metabolites 166 involved in intestinal bacteria as did GF mice; SCFA concentrations in GF/MC marmoset 167 faeces were significantly lower (P < 0.001 or P < 0.05) than those in conventional marmosets, 168 similar to those in GF mice when compared to those in SPF mice (Fig. 4a). Only conjugated and 169 no deconjugated primary/secondary bile acids were detected in GF/MC marmoset faecal 170 samples, similar to those in GF mice (Fig. 4a). However, muricholic acid, which is considered 171 to be rodent-specific, was not detected in marmosets.

183
To verify whether the caecal enlargement in GF rodents was also observed in the obtained 184 GF marmosets, contrast-enhanced radiography acceptable for live animals was performed on 185 the animals in sterile isolators and showed enlargement of the caecum in a GF marmoset 186 compared to that in conventional marmosets (Fig. 4d). The caecal diameter measured from the 187 X-ray images of GF state marmosets, including those from culture-negative individuals after 188 antibiotic administration (906M, 926M and 939M), was significantly larger than that measured 189 in conventional marmosets (P < 0.05).

192
In this study, we succeeded in obtaining GF marmosets in which no viable bacteria and fungi 193 were detected in culture and maintained them for a long period of up to 22 months (Fig. 3b).
194 Reyniers and Trexler (1943), in the first record of obtaining GF primates, reared GF-rhesus

203
Here, we showed that primates can grow under sterile conditions, as monitored by monthly 204 culture tests, for up to 22 months when reaching sexual maturity.

205
The comparison of faecal bacterial composition in this study indicated that intestinal bacterial 206 flora in conventional marmosets is similar to that in humans, such as Bifidobacterium 207 colonisation, which did not occur in SPF mice. Spontaneous Bifidobacterium colonisation was 208 also consistently observed in marmosets from other facilities 30,31 . Furthermore, the inoculation 209 of the intestinal bacterial strains obtained from humans showed that marmosets underwent 210 significantly steady colonisation by all these strains, suggesting similar colonisation milieu for 211 intestinal bacteria with humans. Faecal and plasma metabolome analyses also showed that the 212 metabolic profile of marmosets was more similar to that in humans than that in mice fed 213 marmoset diet. Phylogenetic relationships in metabolome profiles of plasma reported 214 previously 32 were consistent with those detected in the present study. These results demonstrate 10 that marmosets can serve as suitable NHP models to bridge the gap between mouse studies and 216 human medicine in microbiota research.

217
In this study, we developed a series of techniques for GF-marmoset production including 218 preparation of full-term pregnant animals, aseptic caesarean section that was safe for dams and 219 neonates ( Fig. 2, Extended data Fig. 1), hand-rearing of neonates and long-term maintenance of 220 animals under sterile conditions (Fig. 3). Hobbs et al. (1977) reported that aseptic caesarean 221 section for producing SPF marmosets resulted in resuscitation of 40% of the neonates 33 . Our 222 results showed that the resuscitation rate of neonates in caesarean section was more than 90%, 223 indicating the more accurate timing and method of our operations (Fig. 2a). However, hand-224 reared neonates often died of diarrhoea, suffered body weight loss within 2 weeks of birth and 225 had 30% weaning rates. Hand-rearing under conventional conditions was previously performed 226 to rear triplets or more marmoset neonates 15 whose resuscitation rate was reported to be 80% or 227 more 34,35 . The lower survival of neonates reared under sterile conditions may be related to 228 bacterial colonisation that contributes to postnatal gut development 1,36 . Nevertheless, arranged 229 rearing protocols such as feeding for preventing diarrhoea and maintaining the body 230 temperature in the latter four neonates led to successful weaning in the present study (Extended 231 Data Table 2, Supplementary Table 3). These results demonstrate that GF marmosets can be 232 reproducibly produced.

233
In the marmosets reared under sterile conditions, phenomena associated with the absence of 234 microbiota such as lack of the metabolites derived from gut microbiota, inactivation of the 235 immune system and enlargement of the caecum appeared ( Fig. 4), thus demonstrating their 236 characteristics as GF animals. Furthermore, our results from stable biomaterial sampling and 237 live imaging that could be performed while maintaining sterile conditions indicate that 238 gnotobiotic experiments using GF marmosets are practicable even in NHPs. Future gnotobiotic 239 marmoset studies will have a great potential to explore unknown fields in microbiota research.

240
Their physiological similarities to humans, including susceptibility to colonisation by 11 microbiota and metabolome profiles as observed in this study, can contribute toward high 242 predictability in preclinical research for therapeutics targeting gut microbiota. Moreover, 243 marmosets and humans share core features of brain architecture and function, and the complex 244 social and cognitive behaviours typical of primates 37 . Recent progress in research using GF 245 mice has been revealing the mechanisms by which the gut microbiota and its metabolites 246 influence the host central nervous system, including neurogenesis, neuronal activity, 247 neuroinflammation and host behaviour 38-40 . The GF marmosets generated in this study can 248 become a powerful resource to clarify the unknown phenomena of interactions between 249 microbiota and its primate host including the microbiota-gut-brain axis.