news and views

DNA sequencing with nanopores

Grégory F Schneider & Cees Dekker

Major hurdles in the quest to sequence DNA with biological nanopores have now been overcome.

Sequencing DNA with nanopores offers excit- The idea of using nanopores for DNA fast (on the order of a microsecond per base) to ing potential advantages over other sequenc- sequencing was proposed more than 20 years identify individual bases6. ing technologies, but thus far the reading of the ago4. The first experimental proof of trans- Several strategies have been tried to address bases from a DNA molecule in a nanopore has location of DNA molecules was reported the fast translocation speeds. Single-stranded been hampered by the fast translocation speed in 1996 using the biological protein pore DNA molecules were statically captured in the of DNA together with the fact that several α-hemolysin5 (Fig. 1b), albeit without α-hemolysin nanopore using a DNA hairpin nucleotides contribute to the recorded signal. sequence information. It has taken sig- or a protein attached to the DNA end, either In ­theory, the basic concept is straightforward: nificantly more time to establish that single of which are too large to enter the pore. This pass a DNA molecule through a nanoscale pore nucleotides can be discriminated within the allowed the detection of single-nucleotide in a membrane from head to tail, and read off traversing DNA, most prominently because mutations in the immobilized DNA7–9. In other each base when it is located at the ­narrowest the speed of translocating DNA has been too work, α-hemolysin was modified to include a constriction of the pore, using the ion current passing through the pore to probe the identity of the base. ab In this issue, two groups, Cherf et al.1 led -hemolysin by Mark Akeson and Manrao et al.2 led by α Jens Gundlach, independently provide the first reports that sequence information can be obtained upon traversing a DNA molecule through a protein nanopore. Both groups use a polymerase (Fig. 1a) to slow DNA trans­ © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature location through a protein nanopore to a speed that is conducive to reading nucleotide- c specific ­current levels, which paves the way for MspA npg sequencing. Together with UK-based Oxford Nanopore Technologies’ recent announcement of the first commercial nanopore sequencing devices, these studies represent a clear step toward sequencing of single DNA molecules using nanopores. DNA sequencing is a rapidly growing industry, d Graphene making inroads in basic research, clinical medi- cine, pathogen epidemiology, forensic science and agricultural and industrial biotechnology. A variety of ‘conventional’ sequencing techniques Figure 1 Nanopores for DNA sequencing. (a) DNA inserted in a nanopore, with speed control provided by a phi29 DNA polymerase (brown). An -hemolysin nanopore (gray) is embedded in a lipid bilayer exist, each with their own pros and cons3. Most α that separates two chambers containing buffered potassium chloride solution. The DNA substrate (red techniques currently on the market require backbone) is inserted into the pore by an applied electric field. Its motion into or out of the pore (arrows) extensive biochemical labeling and sample can be controlled by the applied electric field and the polymerase activity. DNA sequence information is preparation, and importantly do not allow long obtained by changes in the ionic current running along the DNA through the nanopore, which occur as the read lengths to be achieved. DNA is ratcheted through the pore by the polymerase. This image has been extracted from Supplementary Video 1 from Cherf et al.1; the full movie shows the blocking strategy in better detail. (b) Structure of an -hemolysin nanopore inserted into a lipid bilayer. (c) Structure of a MspA nanopore inserted into a lipid Grégory F. Schneider and Cees Dekker are in the α bilayer. (d) A graphene nanopore drawn on the same scale as the α-hemolysin and MspA protein pores in b Department of Bionanoscience, Kavli Institute and c; inset above represents a three-dimensional rendering of the graphene nanopore. Green spheres of Nanoscience, Delft University of Technology, (b–d) represent individual carbon atoms in the supporting membrane. Purple color represents atoms Delft, The Netherlands. in the close vicinity of the nanopore. Pink displays the cross-section of the nanopore at its center. e-mail: [email protected] Credit: Iemedia solutions (a) and Aleksei Aksimentiev at the University of Illinois (b–d).

326 volume 30 number 4 april 2012 nature biotechnology news and views

cyclodextrin ring that binds free mononucleo­ each experiment. Using this method, they were technological advantage, as the conventional tides, which therefore reside in the pore for able to match current steps to the nucleotide short read lengths call for an elaborate process long enough (up to 10 ms) to be distinguished sequence of the known DNA molecule with an to computationally assemble the separate reads by different ionic current levels for each of the error of 10–25%. It remains to be seen whether to deduce a genome sequence. Long reads also four bases10. this approach can be applied to DNA having an directly provide information on insertions, More recently, the Akeson group has a priori unknown sequence. deletions and gene duplicates that are some- advanced the use of DNA polymerases that Manrao et al.2 pursued the same strategy times difficult to determine from assembling drive a DNA template through a nanopore in of combining a polymerase and a nanopore. short reads. Multiple DNA repeats are found single-nucleotide steps as DNA is synthesized. However, instead of α-hemolysin they used in regulatory DNA sequences, and long read They found that a DNA polymerase from the a mutated version of the protein pore MspA, lengths will thus help resolve haplotypes. phage phi29 was potentially suitable to this which has a slightly different geometry (Fig. 1c). In principle, the head-to-tail sequential approach because it remained bound to DNA, Whereas many (~10) DNA bases within the feed-through of DNA molecules allows for even against the force of an applied voltage long beta barrel of α-hemolysin contribute to ­unlimited read length, although with the needed to insert the DNA into the pore11. The the magnitude of the current level, MspA locally ­current approach, the processivity of the DNA polymerase, which processes DNA at a rate of confines the DNA merely over a few bases at its polymerase will likely become the limiting about one nucleotide every ten milliseconds smallest constriction (cf. Fig. 1b,c). ­factor. Other advantages of nanopore sequenc- or slower, lowered the translocation speed of Even with this advantage, the ­current sig- ing include the single-molecule read-out, the DNA by four orders of magnitude compared natures obtained by Manrao et al.2 were more fact that multiple molecules can be processed with freely translocating DNA. This study, complicated than anticipated. If the ionic current by a single pore, the absence of expensive however, relied on the transient chemical pro- were strictly influenced by the single base located fluorophore labeling and the lack of need for tection of the DNA primer to prevent elonga- at the narrowest point of the constriction, four amplification steps. tion and excision in bulk phase, yielding only current levels corres­ponding to the four bases Looking ahead, several potential improve- a short timeframe to capture and process DNA should be observed. Instead, Manrao et al.2 ments are apparent. The error rate must be molecules from bulk (~20 minutes). observed a wide variety of distinct current levels. decreased (a 4% error rate was reported by Now, Cherf et al.1 describe a ‘blocking This likely can be attributed to the influence of Oxford Nanopore), the read length can likely be oligomer’ strategy in which up to 500 DNA nucleotides neighboring the base of interest, that increased further and the stability of the protein molecules can be sequentially translocated is, a small series of adjacent bases, rather than a pores over time may need to be improved. through an individual nanopore over a period single base, determining the current signal. An alternative approach to nanopore of many hours. To achieve this, they bound Thus, although both groups have achieved sequencing may involve solid-state nano­ phi29 DNA polymerase to a single-stranded the translocation of DNA through a nanopore pores12. Solid-state nanopores have tunable DNA template hybridized to a primer, and in single-­nucleotide steps at an appropriate pore size, are more stable than biological threaded the single-strand end through an rate, determining the identity of the nucleo­ membranes, can be reused after cleaning and α-hemolysin nanopore using an applied volt- tides from the current levels calls for further allow for scaling and device integration. So age. As the polymerase extends the primer to advances in data analysis for base calling. far, however, DNA sequencing has not been synthesize double-stranded DNA, it acts like a Interestingly, both groups also show that the demon­strated with these devices. Conventional motor that ratchets the DNA through the pore sequence information is present in the ionic silicon-based nanopore membranes are rela- © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature in single-nucleotide steps. current traces recorded during both the unzip- tively thick, typically ~30 nm, which corres­ To protect the 3′ end of the primer in solu- ping and synthesis modes of passing DNA ponds to ~60 bases along a single-stranded tion and allow the template to be positioned through the pore, thereby providing double- DNA molecule. Therefore, they are not directly npg in the pore, Cherf et al.1 used a special oligo­ pass readings of one molecule. applicable in DNA sequencing applications, nucleotide that binds downstream of the These two scientific reports coincided with although solid-state nanopores are excellent primer. The resulting complex is thermo- an announcement by Oxford Nanopore at new tools for biophysical studies12. dynamically stable in bulk, but when the February’s Advances in Genome Biology and The recently introduced solid-state nano- single-stranded end of the template is electro­ Technology (AGBT) conference in Marco pores made of graphene, on the other hand, phoretically driven through the pore by an Island, Florida, that the company will launch the present­ an intriguing alternative13–15. applied voltage, the force resulting from the first commercial nanopore sequencing devices Graphene is the ultimate nanopore membrane applied voltage can mechanically dissociate later this year (http://www.nanoporetech. as it is a carbon sheet with a thickness of only a the blocking oligomer in a process referred to com/news/press-releases/view/39). These single carbon atom (Fig. 1d). It is also electri- as ‘unzipping,’ thereby allowing DNA synthesis nanopore-based DNA sequencing devices cally conductive, which opens up new modali- to proceed (Fig. 1a). presumably use a layout similar to the one dis- ties for measuring the traversing nucleotides, Although Cherf et al.1 succeeded in slowing cussed above. The achievements reported are for example by running a tunneling current the DNA translocation to a speed that should quite impressive, but no data have been released through the DNA molecule that is passing be compatible with sequencing, they were not publicly yet. Most importantly, the 48,490-bp through a graphene gap to directly probe the able to directly associate current levels with genome of lambda DNA was sequenced as a chemical nature of the bases16. DNA sequenc- individual nucleotides. They measured a vari- complete fragment—in one read. This read ing using graphene nanopores is currently still ety of current levels and manufactured a ref- length is three orders of magnitude larger than science fiction, but so was sequencing with erence map to correlate current amplitudes to the typical length of DNA fragments that are biological pores two decades ago. known nucleotide sequences by carrying out read in current sequencing approaches using Even with the recent rapid progress with DNA four independent experiments using the same fluorescent nucleotide labeling3. sequencing using protein nanopores, the future DNA template but with one of the four dNTP The long read lengths potentially achievable commercial success of nanopore sequencing is substrates present at lower concentrations in with nanopore sequencing present an important hard to estimate, given the many competitors

nature biotechnology volume 30 number 4 april 2012 327 news and views

and rapid pace of development in the field. It will 6. Branton, D. et al. Nat. Biotechnol. 26, 1146–1153 Both studies2,3 show that a broad ‘whole-cell (2008). be interesting to see how things progress in the 7. Ashkenasy, N., Sanchez-Quesada, J., Bayley, H. & library of antigens’ approach overcomes coming years, both in sequencing applications Ghadiri, M.R. Angew. Chem. Int. Ed. 44, 1401–1404 immuno­logical tolerance to cancer antigens. and in fundamental science, where nano­pores (2005). In addition, regardless of whether the cDNA 8. Stoddart, D., Heron, A.J., Mikhailova, E., Maglia, G. provide versatile single-molecule sensors. & Bayley, H. Proc. Natl. Acad. Sci. 106, 7702–7707 library is derived from cancer cells or normal (2009). tissue, the anti-tumor effect is greater when COMPETING FINANCIAL INTERESTS 9. Vercoutere, W. et al. Nat. Biotechnol. 19, 248–252 the cDNA originates from a different species The authors declare no competing financial interests. (2001). 10. Clarke, J. et al. Nat. Nanotechnol. 4, 265–270 (2009). ­(xenogeneic). This supports the hypothesis 1. Cherf et al. Nat. Biotechnol. 30, 344–348 (2012). 11. Lieberman, K.R. et al. J. Am. Chem. Soc. 132, that xenogeneic antigens are more immuno- 2. Manrao et al. Nat. Biotechnol. 30, 349–353 17961–17972 (2010). genic owing to small differences in the protein (2012). 12. Dekker, C. Nat. Nanotechnol. 2, 209–215 (2007). 3. Ansorge, W.J. New Biotechnol. 25, 195–203 13. Schneider, G.F. et al. Nano Lett. 10, 3163–3167 sequence that are capable of generating a cross- (2009). (2010). reactive immune response4. Interestingly, 4. Deamer, D.W. & Akeson, M. Trends Biotechnol. 18, 14. Merchant, C.A. et al. Nano Lett. 10, 2915–2921 147–151 (2000). (2010). although the VSV melanoma cDNA library 5. Kasianowicz, J.J., Brandin, E., Branton, D. & Deamer, D.W. 15. Garaj, S. et al. Nature 467, 190–193 (2010). contained antigens that are known to elicit Proc. Natl. Acad. Sci. 93, 13770–13773 (1996). 16. Postma, H.W.C. Nano Lett. 10, 420–425 (2010). CD8+ T-cell immunity, such as the melanoma differentiation antigen gp100, most of the anti- cancer effect did not involve cytotoxic CD8+ T cells but instead involved CD4+ T cells. This Selecting antigens for cancer finding is in contrast to the generally accepted view that CD8+ T cells have a central role in tumor immunity, but is in line with more recent vaccines data pointing to the ­importance of other arms of the immune response, including CD4+ T cells Francesca Avogadri & Jedd D Wolchok and antibodies5,6. The main novelty of the current study3 is the A cancer vaccine is streamlined by identifying a small set of potent use of an innovative screening strategy that is immunogens in a tumor cDNA library. based on the authors’ finding that immunity generated by the VSV library involves produc- Vaccination as a therapeutic approach to ­cancer potential to improve cancer vaccines because tion of interleukin (IL)-17. The cytokine IL-17 differs in important ways from classic vaccina- it avoids the delivery of whole-cell or cDNA- is secreted by T-helper cell 17 (TH17) cells tion against infectious agents. A cancer ­vaccine library antigens that could stimulate auto­ (a subset of CD4+ T cells), which have recently must induce an immune response to ‘self’ immunity or irrelevant immune responses. emerged as having a major role in autoimmu- rather than to ‘non-self’ antigens without pro- Different types of whole-cell vaccines have nity, inflammation, host defense and tumor voking harmful autoimmunity; it must work been proposed and are being investigated for the immunity. Tumor-specific TH17 cells from when administered after a tumor has formed treatment of various cancers, including prostate immunized mice were detected after only two and is exerting immunosuppressive effects; and cancer, colon cancer, leukemia and melanoma1. rounds of stimulation with the viral library and © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature it should be insensitive to heterogeneous tumor Early efforts involved co-administration of a final pulse with tumor-cell lysates. However, antigen expression among patients. Promising mycobacterial adjuvants with irradiated or the laborious culture conditions and the large results have been achieved in animal models by lysed cancer cells derived either from a patient’s amount of virus needed for stimulation could npg vaccination with whole cells from autologous own surgically removed tumor (autologous) present a limitation when screening hundreds or allogeneic tumors1 or with cDNA libraries2, or from patients with the same type of cancer of pools of antigens. This could pose a real but it is debatable whether using a broad anti- (allogeneic). A second generation of whole-cell challenge if a similar strategy was used to treat gen repertoire is the safest and most effective vaccines comprised genetically modified cells individual patients. Addition of heat shock strategy. In this issue, Vile and colleagues3 that expressed immunostimulatory cytokines protein (hsp)70 to the culture medium allowed show that melanoma in mice can be treated by or surface receptors to fine-tune immuno­ a reduction in the multiplicity of infection that vaccination with a human melanoma cDNA genicity. These vaccines have not induced was necessary to induce detectable amounts library delivered by a vesicular stomatitis virus robust tumor regression, perhaps owing to the of IL-17 and represents a crucial technical (VSV) vector. Using a novel screening strategy, inherently low immunogenicity of mammalian advance for the scalability of this screening the authors identify a set of three antigens in cells. More recently, the Vile laboratory showed approach (Fig. 1). the library that, when used in combination to that a VSV-based cDNA library expressing The IL-17–based screening step led to the vaccinate mice bearing established melanoma ‘self’ epitopes from normal human prostate identification of three immunogenic proteins: tumors, were as effective as the complete tissue induces rejection of prostate tumors TYRP1, NRAS and cytochrome C1. TYRP1 library. Selection of optimal antigen sets has the in mice2. is a melanoma rejection antigen that acti- In the present report, Vile and colleagues3 vates CD4+ and CD8+ T-cell responses and Francesca Avogadri and Jedd D. Wolchok are extend the VSV-expression library approach to ­antibodies. Recent studies in an adoptive T-cell at Swim Across America Laboratory, Ludwig melanoma. In the new study, the cDNA is derived transfer mouse model have also shown that + Center for Cancer Immunotherapy and from tumor cell lines, which, in contrast to TH17-like CD4 T cells specific for TYRP1 have 5 Ludwig Institute for Cancer Research, Memorial normal tissue, contain numerous genetic muta- higher anti-tumor activity than TH1-like cells . Sloan-Kettering Cancer Center, New York, tions that are potentially immunogenic (a list Notably, only mild autoimmunity was induced New York, USA. of mutated epitopes is available at: http://cancer by the VSV-based cDNA library relative to that e-mail: [email protected] immunity.org/peptidedatabase/mutation.htm). of other vaccines targeting TYRP1. It would be

328 volume 30 number 4 april 2012 nature biotechnology