TECHNOLOGY FEATURE

Microbiology: the return of culture Vivien Marx

Sequencing technology drives and gives researchers new reasons to draw on classic techniques.

The number might be 100, 1,000 or 1,500. Three scientists might offer three different numbers when asked how many bacterial species a healthy gut contains, says Trevor Lawley, a microbiologist at the Wellcome Trust Sanger Institute. Sequencing approaches have yet to reveal definitive answers. Sequencing, notably 16S ribosomal RNA (rRNA) gene amplification and sequencing, and metagenomics, in which genetic materi- al from samples chock full of microbial spe- cies are sequenced in one go, has galvanized microbiology. As microbiome research has grown, sequencing has become a standard assay to identify and count microbes and their genes. Ideally, sequencing provides a complete census of a sample, but often the M. Porcar census data have gaping holes. These meth- A culture revival is underway. Culture is being combined with methods such as omics, phenotyping, and ods generate data about genome fragments, mass spectrometry. not entire genomes, says Lawley. There are too few reference genomes with which to assign these data to a known species, and key to a newborn’s immune system or coral that are potent PCR inhibitors. A remedy many species remain unidentified because reef health, have pathogenic traits, or might is the mechanical lysis of samples plus deg- they do not grow in the lab. The microbi- be healthful in a probiotic drink. radation of these polysaccharides. But not ome field has not focused on its foundation, The good news is that many researchers every research team will have done this, which is microbiology, says Lawley. “It’s a have realized what they need to do to fill the which is why he advises caution for data major deficiency.” gaps in data. “We are living in a revival of comparison. For years, researchers have disregarded the culture,” says Raoult. “Culturing is going Often a bug is difficult to grow in the lab. the use of culture to establish the reper- to live a ‘second youth’, in parallel with the Raoult has long focused on growing diffi- toires of microbes associated with the gut, massive use of ‘omics’.” This revival com- cult bacteria, including many species of the says Didier Raoult, bines culture with other methods, such as pathogen Rickettsia. He and his team also a microbiologist phenotyping, mass spectrometry, in situ used PCR to identify Yersinia pestis–specif- at Aix-Marseille hybridization and omics. ic sequences in dental pulp extracted from University. Culture skeletons in mass graves dating back to the is the classic task of Vive la culture sixteenth and eighteenth centuries. Over coaxing microbes Metagenomics is unquestionably power- time, he expanded his approaches, combin- to grow on agar ful, says Raoult, but it needs to be com- ing culturing, 16S rRNA sequencing and plates. Labs want to plemented by what he calls ‘culturomics’, matrix-assisted laser desorption/ioniza- Aix-Marseille University Aix-Marseille identify and grow which combines culturing, sequencing and tion–time of flight (MALDI-TOF) to create Metagenomics is microbes to explore mass spectrometry to identify bacteria in a high-throughput way to analyze samples 1 powerful, says Didier how they function samples . Metagenomics results are hard to to find what might otherwise be missed. Raoult, but it needs to and discover, for compare between labs—they may use differ- Raoult says he likes the speed of MALDI- be complemented with example, which ing DNA extraction methods, for example. TOF and its role as a discovery tool. ‘culturomics’. bacterial species are Escherichia coli produces ­polysaccharides Species have unique protein signatures that

NATURE METHODS | VOL.14 NO.1 | JANUARY 2017 | 37 TECHNOLOGY FEATURE

­correspond to different mass spectra. He has samples and 30 samples from the small Bacteroides thetaiotaomicron grows in the eight MALDI-TOF instruments in his lab. intestine and colon of people in Europe, lower chamber, producing hydrogen that Each costs $100,000 to $150,000, and the rural and urban Africa, Polynesia and India. helps the methanogenic bacteria in the cost of reagents is negligible; “this is why we They tested nearly 1 million colonies and upper chamber to grow. can test so much,” he says. His lab has gener- discovered previously uncultured microbes Culturomics and metagenomics stand ated a total of 10 million mass spectra from from the human gut2. They identified more side by side, says Raoult; each shows scien- between 3 million and 5 million microbial than 1,525 bacterial species in the fecal sam- tists a different part of the microbial world. colonies tested. The team puts the spectra ples, including more than 140 human bac- “It’s not inferior, it’s not superior, it’s a differ- in the database that comes with the Bruker teria not previously known to be in the gut, ent part of the world,” he says. Culturomics instruments, and the lab also has a home- more than 180 bacteria and 1 archaeal spe- helps with finding and culturing bacterial built database. cies not previously isolated from humans, species. Labs can then use these discover- Raoult and his group also worked out 212 and almost 200 previously unknown bacte- ies to help annotate or re-annotate metage- distinct medium-based culture conditions, rial species. nomic data, creating a microbiological El such as preincubation in a blood culture bot- The scientists used magnifying glasses to Dorado of sorts, he says. Over the past 2 tle or the addition of sheep blood or rumen detect tiny ‘microcolonies’, some as small years, a PhD student in Raoult’s lab has dis- fluid, which is not commercially available as 100 μm in diameter, on culture plates. covered 50 bacterial species. Many of the and must be prepared in the lab. He and his They grew these colonies and performed lab’s 18 members discover 3–10 new micro- team have reduced the list to 18 and are cur- MALDI-TOF or 16S rRNA analysis. To bial species a week. rently whittling it down to 5, he says. culture methanogenic bacteria, which are Discovery matters for science and Applying culturomics, Raoult and his expensive and highly challenging, they potential applications (see Box 1). When team recently analyzed nearly 1,000 stool used a double-chambered assay in which microbiologist Ruth Ley, director of the

BOX 1 MICROBIOLOGY-BASED PROSPECTING

Manuel Porcar and Cristina Vilanova, of the University of Valencia, and their colleagues are self-professed ‘bioprospectors.’ They hunt for organisms, especially microbes that have potential for many applications. Porcar cofounded the company Darwin Bioprospecting Excellence in early 2016. Vilanova joined later that year after completing her PhD. The company offers bioprospecting services for academic labs in Europe and North America as well as for small and medium- sized companies in the biotech and agro-food industry, says Vilanova. The team combines sequencing, culturing and

other techniques. They are also characterizing strains they K. Tanner have collected to explore their potential in areas such as Manuel Porcar and Cristina Vilanova hunt for microbes they can bioremediation, biomass conversion and food processing. ‘domesticate’ in the lab. “There’s always a bunch of microbes naturally trained to do whatever you need,” says Vilanova. the fragmentary state of much metagenomic data will not In both their lab and their company, the researchers suffice for discovering this . Annotation methods bioprospect with a view to industrial applications. Using omics and databases are improving, says Vilanova, but labs cannot techniques and culturing, they have, for example, found a currently identify the total diversity of in their diversity of microorganisms on solar panels. And the diversity environmental samples. And the complex data sets still have to increases the chances of finding a microbe they can isolate be analyzed for the core information they hold, especially for and ‘domesticate’ in the lab through culturing to test these applied research. microbes’ potential, says Vilanova. For example, they hunt Many extreme and unexplored habitats are likely to harbor for “super-producers” of carotinoids, which are used in the novel species or gene sequences, but omics techniques pharmaceutical and cosmetics industries. alone will not be enough to find them. The genome is “just Although microbial genes or pathways can be chosen and the shadow of an organism; it’s not it,” says Porcar. “If you cloned into E. coli or , Porcar says the team is focusing can culture a , it’s literally on your hands.” on culturable organisms from the environment as potential Metagenomics can help to discover industrial enzymes, for biofactories to make metabolites of interest. These microbes example, but strain-level information is hard to obtain, promise to be more robust for many applications, and their especially when the microbes cannot be cultured. That is why, use avoids regulatory issues related to genetically modified he says, “culture-dependent and -independent techniques organisms. need each other.” Many microbiome-focused companies have The estimated 1 trillion bacterial species on Earth emerged, and in this competitive environment, the details represent huge biodiversity, says Porcar. Even if the estimate of Darwin’s approach are proprietary. What differentiates is off, there are also subspecies and strains to consider. A the approach from others, he says, is a focus on improving metagenomics-centered comprehension of the biosphere and culturing methods to “isolate the unisolatable.”

38 | VOL.14 NO.1 | JANUARY 2017 | NATURE METHODS TECHNOLOGY FEATURE

But sequencing is not the first step in his are transmitted between individuals, such as workflow, nicknamed CRAP, for ‘culture, Clostridium difficile, which causes diarrhea. re-streaking, archiving and phenotyping’. Somehow this anaerobic bacterium survives “We flipped it around,” says Lawley of the exposure to air, a feat that is perhaps enabled approach. He starts with microbial pheno- by sporulation. type, applies large-scale culturing, and then The researchers found spore-forming follows with sequencing. bacteria in the microbiota to be much more diverse than non-spore-forming bacte- Start with phenotype ria. Exploring spores will help to annotate To test this workflow, Lawley and his team metagenomic data that indicate that the gut studied a distinct, metabolically dormant microbiome har- state of bacterial life: sporulation3. Spores are bors many sporu- C. Vilanova a cellular phenotype of a bacterial species or lation-associated Culture methods coax microbes to grow on agar strain; a thick shell envelops the bacterium genes. To date, these plates, enabling study of their functions. and it is resistant to oxygen and ethanol. bacteria have been The researchers did targeted phenotyping poorly character- Max Planck Institute for Developmental by testing for ethanol-resistant spores in the ized, says Lawley. Biology, was at Cornell University, she and human gut microbiome. They isolated 137 His workflow tar- her colleagues isolated a new anaerobic bacterial species, 45 of which were potential- gets a phenotype in bacterium from the gut, Christensenella ly novel. They also cultured 90 species from Gary Borisy is a fecal sample, but minuta. Ley and her team showed weight the Human Microbiome Project’s ‘most- scaling up FISH to labs can apply it to loss and less adiposity in mice with this wanted list’ of uncultured and unsequenced explore microbial other phenotypes, bacterium in their gut. And some research- microbes. communities. he says, as part of ers think C. minuta could become part of The team plated the spores, coaxing them studying the func- a probiotic with which to study and pos- to germinate and return to active growth tional differences between the microbiomes sibly treat obesity. Raoult’s lab uncovered using YCFA agar, to which glucose, malt- of individuals, for example. two more Christensenella species, C. mas- ose and cellobiose were added. The plates Sequencing is not a ‘functional tell-all’. siliensis and C. timonensis. These microbial were subjected to anaerobic conditions, and For example, says Lawley, microbes might species are difficult to isolate and grow. intestinal bile salts were added. Next came have a gene known to confer resistance to a A gene sequence can indicate a physiologi- sequencing to find the genotype linked to particular antibiotic. But the gene’s presence cal role, as in the case with C. minuta, says the sporulation phenotype. does not mean the organism is resistant, he Raoult, but to establish the scientific basis Lawley and his team call sporulation an says. Combining metagenomics with phe- for a microbe’s metabolic effect, labs need to “unappreciated” phenotype that might, for notyping and culturing helps to discern grow, sequence and characterize it. His group example, help researchers to better under- which microbes are present and offers is discovering too many species to be able to stand bacterial transmission. The human functional insight. When scientists use characterize all of their biology, he says. But gut contains many anaerobic bacteria that only metagenomic approaches, they make others can use the spectra, genomic data and strains. Ultimately, he says, labs need the bug. And they need microbiologists. Some microbiome researchers hail from computational biology, and others focus on phylogeny, but microbiologists have a valu- able biological perspective for researching the gut, says Raoult. For example, many gut disorders are connected to anaerobic bac- teria, so labs must tend to the differences between aerobes and anaerobes. “Microbiota need microbiologists,” he says. At the Sanger Institute, Lawley and his team considered mass spectrometry–based methods, and he agrees that they are fast and inexpensive. Mass spectrometry also offers “a shorter-term gain with identification,” he says. But a mass spec signature from a pre- viously uncultured microbe means that labs still have to sequence its genome, so Lawley prefers sequencing over mass spectrometry G. Borisy/Forsyth Institute Laboratory, Biological Welch/Marine J. Mark to accompany culturing methods. “I’m at a The spatial organization of microbial community in human dental plaque, rendered visible at the genome center; that’s what we do,” he says. micrometer scale by FISH.

NATURE METHODS | VOL.14 NO.1 | JANUARY 2017 | 39 TECHNOLOGY FEATURE

Laboratory have adapted FISH to target Each computer-designed probe still has to rRNA and identify a wealth of microbes be validated for efficacy and specificity in distinctly, in a single experiment5. Their experiments run with a range of taxa. approach is called combinatorial labeling and Borisy and his colleagues have used spectral imaging (CLASI)-FISH. CLASI-FISH to map microbes at the genus Metagenomics is powerful, says Borisy, but level, and they reached species level for a its weakness is that the experiment begins few of the genera. “We are pushing harder with grinding a sample to extract DNA. “In in this direction now,” says Borisy. They col- the process, you lose all spatial information lected 3D data, which is important because at micron-to-millimeter scales,” he says. The biofilms are not monolayers but 3D com- dynamics of individual cells in the sample munities. CLASI-FISH lets scientists “look are also lost. Metagenomics offers data more holistically” at a microbial community

J. Peretó about the genetic potential of a community, than is possible with one or a few probes, but “it doesn’t say whether that potential is he says. In this sense, the method helps labs Using omics techniques and culture, researchers expressed.” That requires meta-transcrip- understand complete communities as well at the University of Valencia and the startup tomics and genome assembly, which are chal- as their range of variability, which he hopes Darwin Bioprospecting Excellence have found diverse microorganisms on solar panels. lenging. Beyond knowing about genes and will make it easier to compare and contrast gene expression, scientists want to connect results from different labs. assumptions about phenotypes. Yet they this information to the molecular In Lawley’s view, a microbiome’s spa- still must do experiments to see whether of individual taxa and assess how microbes tial organization is “pretty key.” He has an assumption is true, he says, if they want, interact in a community. seen CLASI-FISH data and likes the pat- for example, to understand disease-causing On its own, says Borisy, metagenomics terns it reveals. Lawley sees his approach anaerobes in the human gut. cannot address all of a microbiome’s dimen- as complementary to phenotyping accom- Conceptually, says Lawley, his approach sions, and this realization led him to explore plished with CLASI-FISH. Once bugs are addresses the ‘great plate count anomaly’, a scaling up FISH. “It can’t be scaled up like cultured, labs can use antibodies and other longstanding frustration in microbiology DNA sequencing, but we think it can be reagents, as well as FISH-based approaches, in which many microbes can be detected pushed further,” he says. He and his team to explore how species interact in a microbial and phenotyped but not grown in the lab have begun using binary labels as a practical community. for further analysis. In his view, the CRAP way to obtain a large number of fluorescent An estimated 1011 to 1012 microbial spe- workflow avoids this situation by rendering signatures. The team did a “proof-of-imag- cies inhabit Earth, according to Indiana many more, and perhaps all, microbes cul- ing test” with E. coli and showed that with 16 University researchers Jay Lennon and turable and opens up the human microbiome individual fluors they could distinguish 120 Kenneth Locey6. Of these microbes, around for phenotypic and functional analysis. “We binary combinations of labels. In principle, 104 species have been cultured and fewer 5 can culture almost everything,” he says. “The he says, it’s possible to use higher-order com- than 10 are represented in annotated dogma of the unculturable doesn’t really binations and expand the number of label genome sequence information. A number exist for the human microbiome.” It has been types. Interpreting the resulting emission of microbiologists believe that character- replaced by trial and error to find the ideal spectra involves linear unmixing, and the izing these riches will require both culture- growth media, nutrients and conditions linear equations to do this are usually built independent sequencing and classic micro- microbes need. into platforms such as MATLAB. “We use biology techniques2,3,7. This combination in-house scripts drawing upon MATLAB has potential for applications, too. Together, In situ approaches functions,” he says. Most manufacturers of the methods might let microbiologists bid At the California Institute of Technology, spectral imaging systems build unmixing adieu to the great plate count anomaly and Victoria Orphan and her team characterize operations into their software. turn uncultivable microbes into ones that microbiomes in extreme environments, such Borisy and his team are now working on can be studied in the lab. as methane-producing bacteria in deep-sea how to address variable brightness and sig- locations. She applies a combination of nal strength of CLASI-FISH fluorophores. Vivien Marx is technology editor for techniques, mass spec- Probes also differ in hybridization effi- Nature Methods ([email protected]). trometry and methods such as fluorescence ciency, he says, and it can be a challenge to in situ hybridization (FISH) to functionally get the probes across the bacterial envelope. 1. Lagier, J.-C. et al. Clin. Microbiol. Rev. 28, analyze the microbes. For example, she and Ultimately, he says, the method can deliver 237–264 (2015). 2. Lagier, J.-C. et al. Nat. Microbiol. 1, 16203 her team use bio-orthogonal noncanonical species-level information because probes (2016). amino acid tagging (BONCAT-FISH) to can be designed at almost any phylogenetic 3. Browne, H.P. et al. Nature 533, 543–546 label proteins that a microbe is actively pro- level. “Of course, probe design becomes (2016). 4 4. Hatzenpichler, R. et al. Environ. Microbiol. 16, ducing to assess metabolic activity . more difficult when one is trying to differ- 2568–2590 (2014). Another emerging in situ method uses entiate closely related taxa that might differ 5. Mark Welch, J.L. et al. Proc. Natl. Acad. Sci. USA fluorescence to capture the spatial qualities in only one or two bases,” he says. For probe 113, E791–E800 (2016). 6. Locey, K.J. & Lennon, J.T. Proc. Natl. Acad. Sci. of a microbiome, such as in oral biofilms. design, he draws on sequence databases, USA 113, 5970–5975 (2016). Gary Borisy at the Forsyth Institute and hybridization know-how and tools such as 7. Vilanova, C. & Porcar, M. Nat. Microbiol. 1, Jessica Mark Welch at the Marine Biological the web-based probe design tool mathFISH. 16101 (2016).

40 | VOL.14 NO.1 | JANUARY 2017 | NATURE METHODS