MicroRNAs fine-tune oncolytic viruses - Nature

© 2008 Nature Publishing Group http://www.nature.com/naturebiotechnology

news and views immobilized antibodies that capture analytes present in the plasma for subsequent detection using fluorescent secondary antibodies. Fan et al.3 show that the levels of prostatespecific antigen detected in patient serum by their chip correlate well with those measured by conventional techniques, and the overall sensitivity of the device is sufficient to quantify multiple proteins and cytokines at levels relevant to human disease. However, whether fluorescence will ultimately be the best way to detect biomarkers remains unclear; a wide variety of detection methods are under development, including several that are label-free7. Nearly a decade after widespread clinical use of microfluidics was first anticipated, the field remains plagued by technical problems. One issue concerns the sensitivity and durability of antibody arrays, on which most biochemical assays rely. The integrated blood barcode chip addresses this problem by using DNA-directed protein immobilization11, in which a surface coated with single-stranded DNA is hybridized to antibodies chemically conjugated to complementary DNA. This seemingly baroque scheme has important advantages over direct antibody conjugation, including reliance on robust and stable DNA-glass surfaces during device fabrication and storage, formation of delicate antibody-containing features just before use, and creation of capture features with densely packed antigen-combining domains, thereby increasing sensitivity12. Passivation and the prevention of fouling are two additional challenges in using microfluidic devices with biological samples. Passivation refers to the modification of surfaces so as to minimize nonspecific binding, and fouling to unwanted binding to device surfaces and to plugging of microchannels by components of the sample. Fouling leads to irreproducible measurements and device failure. Because most microfluidic devices based on soft lithography (such as the integrated blood barcode chip) are bonded to glass, and glass has a very high nonspecific binding capacity for proteins, lipids and other biomolecules, effective passivation is difficult to achieve. Perfect device passivation and prefiltering of samples, which would solve the problem of fouling, has never been achieved. Fortunately, the multilayer polyamine-DNA-antibody surfaces created by DNA-encoded antibody immobilization are anti-fouling and appear to be very effectively passivated. The remaining fouling problems relate to background that limits the ultimate sensitivity of the device. Fan et al.3 demonstrate impressive sensitivity with their chip, but largely using serum (plasma cleared of fibrinogen and other clotting factors) rather than plasma itself, which gives

1346

much higher nonspecific binding than that of serum13. Further experiments with fresh blood from healthy volunteers and patients, as outlined by the authors, should resolve whether the rapid processing of blood made possible by their chip will solve the sensitivity problems that have impaired less sophisticated devices. It is now clear that the micro total analysis system (µTAS or lab-on-a-chip) proposed over a decade ago14 will be substantially harder to achieve than was first envisioned. Indeed, several companies founded on the µTAS concept have already faltered or failed. Looking forward, we can nonetheless expect steady improvement in microfluidic technologies, antibodyand aptamer-based capture of biomolecules, label-free detection and on-chip integration of multiple preparative and analytical units with diverse functions. It seems highly likely that devices such as the integrated blood barcode chip presage commercial analytic devices

whose reliability, simplicity and low cost will revolutionize the use of protein biomarkers. 1. Braunwald, E. N. Engl. J. Med. 358, 2148–2159 (2008). 2. Morrow, D.A. & de Lemos, J.A. Circulation 115, 949– 952 (2007). 3. Fan, R. et al. Nat. Biotechnol. 26, 1373–1378 (2008). 4. Barry, M.J. N. Engl. J. Med. 344, 1373–1377 (2001). 5. Kling, J. Nat. Biotechnol. 24, 891–893 (2006). 6. Goluch, E.D. et al. Lab Chip 6, 1293–1299 (2006). 7. Burg, T.P. et al. Nature 446, 1066–1069 (2007). 8. El-Ali, J., Sorger, P.K. & Jensen, K.F. Nature 442, 403– 411 (2006). 9. Svanes, K. & Zweifach, B.W. Microvasc. Res. 1, 210– 220 (1968). 10. Yang, S., Undar, A. & Zahn, J.D. Lab Chip 6, 871–880 (2006). 11. Boozer, C., Ladd, J., Chen, S. & Jiang, S. Anal. Chem. 78, 1515–1519 (2006). 12. Bailey, R.C. et al. J. Am. Chem. Soc. 129, 1959–1967 (2007). 13. Vaisocherova, H. et al. Anal. Chem. 80, 7894–7901 (2008). 14. van den Berg, A. & Lammerink, T.S.J. in Microsystem Technology in Chemistry and Life Sciences (eds. Manz, A. & Becker, H.) pp. 21–50 (Springer, Berlin, 1999).

MicroRNAs fine-tune oncolytic viruses John C Bell & David Kirn Targeting by tissue-specific microRNAs enhances the efficacy and safety of tumor-killing viruses. Gene silencing by endogenous microRNAs (miRNAs) has recently been exploited to control the tropism of gene-therapy vectors. By including the target sequence of a tissue-specific miRNA in the genome of lentiviral vectors, Naldini and colleagues suppressed transgene expression specifically in selected cell types such as hematopoietic cells and hepatocytes1,2. This strategy has now been extended to control targeting of oncolytic viruses. The new studies, appearing in Molecular Therapy3, Journal of Virology4 and Nature Medicine5, mark the beginning of innovative efforts aimed at discovering combinations of viruses and miRNA targets that yield safer and more effective anticancer virotherapeutics. miRNAs are versatile, noncoding RNAs, ~22 nucleotides in length, that exert postJohn C. Bell is in the Ottawa Health Research Institute, Ottawa, Ontario K1G 0K8, Canada and David Kirn is at Jennerex Inc., One Market St., Spear Tower, Suite 2260, San Francisco, California 94105, USA. e-mail: [email protected] or [email protected]

transcriptional regulation through specific recognition of short sequences, often located in the 3′ untranslated region (UTR), in target mRNAs6. Depending on its degree of complementarity to the target, the miRNA can affect either the stability or translation of the mRNA7. miRNAs have complex expression profiles that reflect the important roles they play in the control of mammalian growth and development. The human genome is estimated to contain >500 miRNA genes. Some are expressed in a tissue-specific fashion, whereas others are constitutively expressed or turned on in response to endogenous cues or stress signals8. Another important function of miRNAs, clearly demonstrated in plants and invertebrates9, is to suppress viruses by binding to cognate sequences in viral mRNAs. Oncolytic viruses are vectors engineered or selected to infect and kill cancer cells while leaving normal cells relatively unscathed10,11. Clinical data show that this class of therapeutic is safe and cancer selective but also suggest that more potent viruses—targeted specifically to malignancies—would be desirable to

volume 26 number 12 DECEMBER 2008 nature biotechnology

improve efficacy10,11. This goal was achieved in the three new studies3–5 using miRNA for tumor targeting. Coxsackievirus A21 (CVA21) is a naturally tumoricidal virus that rapidly replicates in and kills a variety of cancer cell types12. However, CVA21 also induces pathology in mice and humans by replicating in normal muscle cells5,13. Kelly et al.5 showed that by incorporating four miRNA target sequences—two for each of two muscle-specific miRNAs—into the CVA21 genome, they could reduce viral replication in normal muscle tissue by >100,000-fold and thereby eliminate muscle pathology without compromising tumor-killing activity (Fig. 1). Using a very similar strategy, Ylösmäki et al.4 created an adenovirus regulated by the liver-specific miR122. Tagging an early gene product of adenovirus (E1A) with the target sequence for this miRNA, in combination with another targeting mutation, effectively eliminated the ability of the adenovirus to replicate in normal liver cells. This is important for the entire field of adenoviral therapeutics as liver toxicity associated with systemic administration is a significant limitation for this technology. Given the pleiotropic effects of miRNAs on normal cell growth, differentiation, metabolism and apoptosis, it is not surprising that aberrant miRNA expression contributes to the generation of malignancies14. Indeed, there is an ever-growing list of ‘tumor-suppressor’ miRNAs that are ubiquitously expressed in normal tissues but often reduced or absent in malignant cells15. Edge et al.3 demonstrated that incorporating the target sequence for the let-7 tumor-suppressor miRNA into the genome of vesicular stomatitis virus generated a virus that replicates preferentially in tumor cells. The let-7 miRNA is often expressed at lower levels in tumors than in normal tissues. Thus, this strategy may be more broadly applicable than the tissue-specific approaches of Kelly et al.5 and Ylösmäki et al.4. The oncolytic virus field has explored many creative ways of restricting viral replication to malignant tissues. These include engineering virus coat proteins to recognize unique tumor antigens on the cell surface, transcriptionally regulating viral genes and exploiting activated signaling pathways unique to tumor cells10. However, miRNA targeting has several features that may ultimately make it one of the preferred methods for engineering viral tropism, either alone or in combination with complementary approaches. For instance, the miRNA sequences required to regulate virus replication are short and por-

Tumor cells

Viral mRNA miRNA target

Muscle cells

Viral replication

Tumor cell lysis

Muscle-specific miRNA

Cleavage of viral RNA

No muscle toxicity

Katie Ris-Vicari


news and views

Figure 1 Control of transgene expression by tissue-specific miRNAs. Incorporation into a viral genome (blue) of target sequences (red) recognized by muscle-specific miRNAs (purple) ensures that the virus replicates in and kills only cancer cells, while leaving muscle cells unharmed. Although Kelly et al.5 inserted four miRNA targets in the 3′ UTR of coxsackievirus A21, a single copy is shown for simplicity. This strategy may enhance the safety of oncolytic viruses for cancer therapy and of attenuated viruses for vaccination against infectious diseases.

table, making them easy to add to compact viral genomes. In contrast, redirecting virus infectivity to the tumor cell surface (transductional targeting) requires sophisticated alterations to the viral genome that often attenuate virus growth, even in target cells. Another advantage shown by the miRNA targeting studies reported to date1–5 is that the strategy apparently works for both RNA and DNA viruses, whether they replicate in the cytoplasm or in the nucleus. In contrast, transcriptional targeting of oncolytic viruses can be used only for DNA viruses that replicate in the nucleus. As Kelly et al.5 point out, regulating oncolytic virus replication with preexisting, highly expressed cellular miRNAs may provide a more potent defense that does not need to be primed by virus infection. As with any promising new approach, several issues require further investigation. First, how does one select the optimal miRNAs? The use of tissue-specific miRNAs could, as in the case of CVA21 and adenovirus, eliminate the dominant acute organ toxicity. However, previously unrecognized secondary complications may be revealed5. For instance, Kelly et al.5 found that although their miRNA-targeted virus no longer caused acute muscle myositis, myelitis associated with tremors and paralysis was observed in a small number of animals; this may have been due to the extended viremia exhibited by their engineered virus. Perhaps the best strategy would be to incorporate a tissue-specific virus ‘silencer’ together with a more ubiquitous tumor-suppressor miRNA target.

nature biotechnology volume 26 number 12 DECEMBER 2008

A second issue concerns the number of miRNA targets to use and where to insert them. In the studies discussed here3–5, multiple copies of miRNA targets were placed in the 3′ UTRs of viral genes. However, other locations within the genome may be preferable and should be explored. The optimal site may well be specific to the virus under investigation. It is worth noting that Barnes et al.16 recently created an attenuated poliovirus vaccine strain by including one miRNA target sequence in the 5′ UTR and a second one between two coding regions. The attenuated strain was unable to replicate in the central nervous system, thus eliciting protective immunity without pathological side effects. Targeting multiple viral genes seems like an obvious next step for the oncolytic virus platform. Finally, what is the likelihood of generating escape mutants? Kelly et al.5 identified mutations in miRNA target sequences in some viruses collected from a minority of treated mice, and at least one of these animals had symptoms of myositis. As the authors suggest, incorporation of multiple miRNA targets throughout the genome would mitigate this problem. In addition, even when rare revertants arise, they might be eliminated by innate or adaptive immune responses3,5,16. As the complexities of miRNA expression and function are elucidated, many opportunities to improve the design of oncolytic viruses will emerge. Used in combination with miRNA profiling of a patient’s tumor17, such viruses could provide an unprecedented level of personalized medicine and enhance clinical outcomes for cancer patients.

1347

news and views

1. Brown, B.D. et al. Nat. Med. 12, 585–591 (2006). 2. Brown, B.D. et al. Nat. Biotechnol. 25, 1457–1467 (2007). 3. Edge, R.E. et al. Mol. Ther. 16, 1437–1443 (2008). 4. Ylösmäki, E. et al. J. Virol. 82, 11009–11015 (2008). 5. Kelly, E.J., Hadac, E.M., Greiner, S. & Russell, S.J. Nat. Med. 14, 1278–1283 (2008). 6. Ivanovska, I. & Cleary, M.A. Cell Cycle 7, 3137–3142 (2008). 7. Zeng, Y., Yi, R. & Cullen, B.R. Proc. Natl. Acad. Sci. USA 100, 9779–9784 (2003).

8. Xu, S. et al. J. Biol. Chem. 282, 25053–25066 (2007). 9. Obbard, D.J., Gordon, K.H., Buck, A.H. & Jiggins, F.M. Phil. Trans. R. Soc. Lond. B Biol. Sci. (2008). 10. Parato, K.A., Senger, D., Forsyth, P.A. & Bell, J.C. Nat. Rev. Cancer 5, 965–976 (2005). 11. Liu, T.C., Galanis, E. & Kirn, D. Nat. Clin. Pract. Oncol. 4, 101–117 (2007). 12. Berry, L.J., Au, G.G., Barry, R.D. & Shafren, D.R. Prostate 68, 577–587 (2008). 13. Dekel, B. et al. Acta Paediatr. 91, 357–359 (2002). 14. Deng, S. et al. Cell Cycle 7, 2643–2646 (2008). 15. He, L., He, X., Lowe, S.W. & Hannon, G.J. Nat. Rev. Cancer 7, 819–822 (2007). 16. Barnes, D. et al. Cell Host Microbe 4, 239–248 (2008). 17. Kong, W., Zhao, J.J., He, L. & Cheng, J.Q. J. Cell. Physiol. 218, 22–25 (2008).

Cancer and complement Suzanne Ostrand-Rosenberg A component of the complement system promotes tumor growth by activating myeloid-derived suppressor cells. The failure of cancer immunotherapies in clinical trials is due largely to tumor-induced immune suppression, mediated by cells such as myeloid-derived suppressor (MDS) cells and T regulatory cells. In a recent Nature Immunology article, Markiewski et al.1 elegantly demonstrate that MDS cells are recruited to tumors by the protein C5a, a component of the classical complement cascade. This study reveals yet another connection between inflammation and cancer and points to C5a as an attractive drug target for fighting cancer by suppressing the MDS cell population, possibly in conjunction with immunotherapeutics. Complement is a collection of serum proteins that are integral to inflammatory processes and to innate immune responses to infection. C5a, also known as anaphylatoxin, is generated from C5 by C5 convertase, which includes C3a, a cleavage product of C3 (Fig. 1). Both C5a and C3a are chemoattractants and pro-inflammatory mediators. Based on epidemiological and experimental data linking chronic inflammation with tumor onset and progression2, Markiewski et al.1 hypothesized that the inflammatory effects of complement may actually protect established tumors and enhance their growth. Using a murine model of cervical cancer and mice deficient in variSuzanne Ostrand-Rosenberg is in the Department of Biological Sciences, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, Maryland 21250, USA. e-mail: [email protected]

1348

ous complement components, they confirmed this hypothesis, showing that C5a deposited in the tumor vasculature attracts MDS cells, which express receptors for C5a, and boosts the potency of these cells by increasing their

Classical and lectin complement pathways C3

content of reactive oxygen and nitrogen species and of arginase—all of which are documented to contribute to MDS cell–mediated immunosuppression. Moreover, a peptide antagonist of the C5a receptor enhanced CD8+ T-cell antitumor responses and was as effective as the chemotherapeutic paclitaxel (Taxol) in retarding tumor growth. Although MDS cells were originally observed in cancer patients and experimental animals >30 years ago, their role as spoilers of anti-tumor immunity is only now being appreciated. A heterogenous population of normal myeloid cells trapped in intermediate stages of differentiation, MDS cells accumulate in the blood, lymph nodes and at tumor sites in virtually all cancer patients. In healthy individuals, these cells differentiate into macrophages, dendritic cells and neutrophils, but tumors secrete a range of factors that disrupt normal differentiation of immune progenitor cells. MDS cells promote tumor growth by preventing the activation of CD4+ and CD8+ T lymphocytes, inhibiting natural-killercell cytotoxicity, stimulating tumorigenic cytokine production and increasing angiogenesis. Because MDS cells promote tumor growth through so many mechanisms, their elimination or inactivation should remove numerous barriers that interfere with cancer immunotherapies.

COX2 VEGF

IL-6

IL-1β COX2

Inflammation

COX2 PGE2

C3a VEGFR

C3b

IL-6R

C5

C5aR

C5a

S100 A8/A9

S100 A8/A9

4

2/

EP

VEGF S100 A8/A9

MDS cell

PGE2

Tumor cell

MDS cells block anti-tumor immunity

Katie Ris-Vicari


COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/ naturebiotechnology/.

Figure 1 Multiple pro-inflammatory mediators in the tumor microenvironment increase the accumulation of MDS cells and enhance their immunosuppressive activity. MDS cell accumulation is driven by the proinflammatory factors prostaglandin E2 (PGE2), S100A8/A9 proteins, vascular endothelial growth factor (VEGF), cytokines IL-1β and IL-6, and complement component C5a (ref. 1), all of which are produced in the tumor microenvironment by host stromal cells. COX2 and PGE 2 can also be secreted by tumor cells. MDS cells express receptors for all of these molecules except IL-1β. Besides increasing MDS cell abundance and potency, C5a and S100A8/A9 proteins are chemotactic, attracting MDS cells to tumor sites. VEGF and S100A8/A9 proteins are produced by both MDS and other host cells, providing an autocrine regulatory pathway for maintaining MDS cell levels. By activating MDS cells, these pro-inflammatory mediators inhibit anti-tumor immunity and interfere with cell-mediated immunotherapies.

volume 26 number 12 DECEMBER 2008 nature biotechnology