Computational genes: a tool for molecular diagnosis and therapy of aberrant mutational phenotype
© Martínez-Pérez et al; licensee BioMed Central Ltd. 2007
Received: 20 February 2007
Accepted: 28 September 2007
Published: 28 September 2007
A finite state machine manipulating information-carrying DNA strands can be used to perform autonomous molecular-scale computations at the cellular level.
We propose a new finite state machine able to detect and correct aberrant molecular phenotype given by mutated genetic transcripts. The aberrant mutations trigger a cascade reaction: specific molecular markers as input are released and induce a spontaneous self-assembly of a wild type protein or peptide, while the mutational disease phenotype is silenced. We experimentally demostrated in in vitro translation system that a viable protein can be autonomously assembled.
Our work demostrates the basic principles of computational genes and particularly, their potential to detect mutations, and as a response thereafter administer an output that suppresses the aberrant disease phenotype and/or restores the lost physiological function.
Finite state automata operating at molecular scale [1–6] can conceptually be used for applications in the living cell. Head et al.  proposed an in vivo biomolecular model in which computations manipulate a plasmid DNA. The solution of the problem is given by either the longest or shortest plasmid at the end of the computation. Henkel et al.  implemented a similar in vivo mechanism in which computations are conducted on a DNA sequence constituting of an open reading frame controlled through a strong promoter. Sequential plasmid manipulation finds a final plasmid construct containing the computational solution allowing the in vivo transcription and translation into a protein. The first autonomous finite state automaton was proposed in [3–5]. It is a 2-state 2-symbol automaton composed of DNA strands and enzymes. Benenson et al.  used this automaton to carry out in vitro molecular diagnosis and therapy. For this, an anti-sense drug (oligonucleotide – a short single stranded DNA molecule) is released if certain diagnostic conditions are true, i.e., low expression levels of certain mRNAs and high expression levels of others. The diagnostic conditions can be viewed as transitional steps in a finite state automaton. If the automaton reached a final state, indicating that all conditions are met, then the drug enclosed in the loop of a hairpin shaped oligonucleotide is released. This if-then mechanism is a new element of autonomous molecular computation. Unfortunately, the proposed mechanisms would not work in a living cell as unwanted side effects raised by supporting molecules (particularly the Fok I enzyme carrying out the transitions) would be a major problem [7, 8]. Moreover, this molecular mechanism is limited in its ability to logically control gene expression as it allows for administering as output only small molecules.
Nevertheless, this work is an important conceptual step forward to link molecular automata to molecular diagnosis and therapy. The first autonomous finite state machine working in the living cell of E. coli was proposed by Nakagawa et al. . This approach is based on a length encoding automaton model  whose input string is encoded by an mRNA molecule. The computation of this mRNA molecule is accomplished by the biosynthesis mechanism of the cell combined with four-base codon techniques [11, 12] yielding a protein of interest.
Results and discussion
Diagnosis and therapy
We study the hypothetical application of simple molecular computations to correct an aberrant mutation in a gene that can trigger a disease-phenotype. One of the most prominent examples is the tumor suppressor p53 gene, which is present in every cell, and acts as a guard to control the growth. Mutations in this gene can abolish its function, allowing uncontrolled growth that can lead to cancer [15, 16].
To process diagnostic rule (1), the molecular automaton must be able to detect point mutations. This task is based on a diagnostic complex, a dsDNA molecule consisting of a single-stranded mutation signal and a single-stranded diagnostic signal (Figure 2A, [see Additional file 2]). Both strands imperfectly pair in the region that resembles an aberrant mutation to be detected. An mRNA molecule bearing a mutation (e.g., oncogenic) will trigger the dissociation of the diagnostic complex and will pair to the mutation signal. The latter process is thermodynamically driven by the higher stability of the DNA/RNA duplex over the mismatched DNA/DNA diagnostic complex and favourable due to its increased complementarity [19, 20].
Moreover, the resulting DNA/RNA hybrid complex will act as a substrate for the cellular RNase H, destroying the RNA component of the duplex . The released single-stranded diagnostic signal links the assembly of the functional gene (Figure 2B, [see Additional file 2]), whose structure is completed by cellular ligase present in both eukaryotic and prokaryotic cells. The transcription and translation machinery of the cell is then in charge of therapy and administers either a wild-type protein or an anti-drug. In both cases, the pathogenic phenotype is suppressed, either by a replacement with the wild-type protein or by a release of a small molecule that may stabilise the aberrant mutant protein, and thus providing the physiological functionality of the wild-type. This idea of favouring the full over the partial complementarity of the base pairing was implemented in robust DNA-based machines 'fuelled' by the DNA/DNA-cross-pairing [22–24]. Although mechanistically simple and quite robust on molecular level, several issues need to be addressed before an in vivo implementation of computational genes can be considered. First, the DNA material must be internalised into the cell, specifically into the nucleus. The transfer of DNA or RNA through biological membranes is a key step in the drug delivery . Nuclear localisation signals can be irreversibly linked to one end of oligonucleotides, forming an oligonucleotide-peptide conjugate that allows effective internalisation of DNA into the nucleus [26, 27]. In addition, the DNA complexes should have low immunogenicity to guarantee their integrity in the cell and their resistance to cellular nucleases. Current strategies to eliminate nuclease sensitivity include modifications of the oligonucleotide backbone such as methylphosphonate  and phosphorothioate (S-ODN) oligodeoxynucleotides , but along with their increased stability, modified oligonucleotides often have altered pharmacologic properties [30, 31]. Finally, similar to any other drug, DNA complexes could cause nonspecific and toxic side effects. In vivo applications of antisense oligonucleotides showed that toxicity is largely due to impurities in the oligonucleotide preparation and lack of specifity of the particular sequence used [30–32]. Undoubtedly, progress on antisense biotechnology will also result in a direct benefit to the model of computational genes.
To test the feasibility of the principle of computational genes, we developed an in vitro system making use of the following assumptions: (i) diagnostic and therapy reactions are simulated in one-single step process; (ii) diagnostic complexes contain three adjacent mismatches representing a whole codon replacement (single amino-acid point mutation); (iii) RNA molecules are replaced by ssDNA during simulations.
Rate-constants of the strand-exchange reactions in phosphate buffer
Ratio [Am/AB']: [Am']
1.258 ± 0.049
2.325 ± 0.029
3.325 ± 0.047
3.846 ± 0.038
4.476 ± 0.022
4.524 ± 0.054
In vitro translation
If the promotor-containing fragment starts transcription before the gene self-assembly (hence leading to an incomplete mRNA transcript), the reliability of the computation would not be affected. In eukaryotic cells, quality control mechanisms ensure that an mRNA molecule is complete before translation is initiated. These mechanisms include cleavage and polyadenylation of the 3' end of proper RNA transcripts protecting them from cellular degradation .
This rule can be realized by using an (n + 2)-state n-symbol automaton (Figure 1C). In the (i - 1)-th state, the input symbol given by the i-th mutation lets the automaton transit into the i-th state. The automaton can be implemented by two partially dsDNA molecules, ssDNA molecules related one-to-one with the mutations, and further (complementary) ssDNA molecules necessary for self-assembly. The functional gene will be self-assembled only if the n-th (final) state is reached, that is, if all n diagnosed mutations are present [see Additional file 6]. In this way, computational genes may be used for the diagnosis and therapy of diseases related to mutation of transcribed genetic material. The rule (2) may even be generalised to involve mutations from different proteins allowing a combined diagnosis and therapy. Furthermore, computational genes are extendable to prokaryotic genes evidencing the generality of the principle. A prokaryotic model could release several different output molecules in response to different environmental conditions [see Additional file 7]. Diagnostic rule (2) can be implemented by linear self-assembly that was successfully demonstrated in the Adleman's first experiment , and recently has been used to implement in vitro finite small-state automata .
Our work demonstrated the basic principles of computational genes and particularly, their potential to detect mutations, and as a response thereafter to administer an output that suppresses the aberrant disease phenotype and/or restores the lost physiological function. In this way, computational genes might allow implementation in situ of a therapy as soon as the cell starts developing defective material. Computational genes combine the techniques of gene therapy which allows to insert a healthy gene into the genome replacing an aberrant gene, as well as antisense technology mediating gene silencing. In addition, a computational gene can theoretically implement general m-state n-symbol automata and could be designed to solve all types of finite state applications at the molecular level, thus offering tremendous advantage for broader applicability than previous approaches [3–6], whose complexity is limited to fewer states and symbols. Clearly, in vivo application of computational genes is the ultimate goal, but before this some hurdles need to be overcome, i.e., the internalisation of the computational gene into the cell, its longevity, and its integrity in the cell. The issue of integrating computational genes into cellular regulatory pathways is directly linked to other biomolecular computing approaches in which gene regulation is the basis for making computations . Computational genes provide a major step towards bringing molecular computers to work inside living organisms, to detect and correct physiological abnormalities, and to administer healthy solutions.
The hID1 gene (pBLAST49-hID1, Invivogen) was subcloned from into the pIVEX 2.3d vector (Roche-Applied-Science). The resulting pIVEX-hID1 plasmid was mutated (Quickchange, Stratagene) using conservative replacements (no change in the amino acid sequence) to introduce the necessary restriction sites for generating initial and final accepting states and to modify both termini with NotI and SacI restriction sites [Additional file 2].
In vitro translation
The plasmid pIVEX-hID1 was expressed in in vitro translation TNT T7 coupled reticulocyte lysate system (Promega) changing thereby the procedure provided by the manufacturer as follows: To the transcription/translation reaction mixture provided in the kit (total volume of 25μ l) 2μ l T7 RNA polymerase, 1μ l T4 DNA Ligase (New England Biolabs) and 2 μ l 35S-methionine (Amersham) were added. The self-assembly reaction contained 2μ l digested hID1, overhang A, overhang B, Am/AB' and Am', pre-mixed in equimolar concentration and added to the final concentration of 0.4μ M to the in vitro translation reaction. The reaction for testing the self-ligation of the computational gene contained all the components as described for the self-assembly reaction, except for the Am' strand. The positive control reaction was supplied with 2μ l intact pIVEX-hID1 plasmid, and the negative control reaction contained only 2μ l pIVEX vector. The reactions were carried out at 30°C for 3.5 h. The translation product was resolved on 15% SDS-PAGE and detected by autoradiography.
The full sequences of the oligonucleotides Am, Am', and AB' are included in Additional file 6. Equimolar amounts of Am and AB' dissolved in 150 m M sodium phosphate buffer pH 7.4 were denatured at 95°C for 30 min and slowly cooled down to 37°C with 1°C step to form stable dsDNA complexes. The displacement reaction was initiated by addition of the displacing strand Am' in different molar ratios to 1μ M Am/AB' duplex and was monitored at 260 nm . The rate constant was determined by fitting the experimental data to a first-order reaction equation as described elsewhere [40, 41]: (A t - A0) = (A∞ - A0)(1 - e-kt), where t is time, and A t , A0, and A∞ are the absorbance values at time t, zero and infinite time, respectively. The melting curves of the 1μ M dsDNA-duplexes Am/Am' and Am/AB' were monitored at 260 nm in phosphate buffer (see above) or in the presence of Ficoll 70 (150 g/L) mimicking the crowded environment in the cells .
This research was supported by grants from the DFG (IG 73/4-1) and Heisenberg Foundation (IG 73/1-1) to ZI, the CONACYT and DAAD fellowship to IMMP and the KAAD fellowship to GZ. The authors thank the anonymous referees for helpful comments.
- Head T, Chen X, Nichols MJ, Yamamura M, Gal S: Aqueuos solutions of algorithmic problems emphasizing knights on a 3 × 3. In DNA7, LNCS 2340 Edited by: Jonoska N, Seeman N. 2002, 191–202.Google Scholar
- Henkel CV, Bladergroen RS, Balog CI, Deelder AM, Head T: Protein output for DNA computing. Natural Computing 2005, 4: 1–10.View ArticleGoogle Scholar
- Adar R, Benenson Y, Linshiz G, Rozner A, Tishby N, Shapiro E: Stochastic computing with biomolecular automata. Proc Natl Acad Sci USA 2004, 101: 9960–9965.PubMed CentralView ArticlePubMedGoogle Scholar
- Benenson Y, Paz-Elizur T, Adar R, Keinan E, Livneh Z, Shapiro E: Programmable and autonomous computing machine made of biomolecules. Nature 2001, 414: 430–434.View ArticlePubMedGoogle Scholar
- Benenson Y, Adar R, Paz-Elizur T, Livneh Z, Shapiro E: DNA molecule provides a computing machine with both data and fuel. Proc Natl Acad Sci USA 2003, 100: 2191–2196.PubMed CentralView ArticlePubMedGoogle Scholar
- Benenson Y, Gil B, Ben-Dor U, Adar R, Shapiro E: An autonomous molecular computer for logical control of gene expression. Nature 2004, 429: 423–429.View ArticlePubMedGoogle Scholar
- Condon A: Automata make antisense. Nature 2004, 429: 351–352.View ArticlePubMedGoogle Scholar
- Stein CA: The experimental use of antisense oligonucleotides: a guide for the perplexed. J Clin Invest 2001, 108: 641–644.PubMed CentralView ArticlePubMedGoogle Scholar
- Nakagawa H, Sakamoto K, Sakakibara Y: Development of an in vitro computer based on escherichia coli . In DNA11, LNCS 3892 Edited by: Carbone A, Pierce N. 2006, 203–212.Google Scholar
- Yokomori T, Sakakibara Y, Kobayashi S: A magic pot: Self-assembly computation revisited. In Formal and Natural Computing, LNCS 2300 Edited by: Brauer W, Ehrig H, Karhumaki J, Salomaa AK. 2002, 418–429.Google Scholar
- Sakakibara Y, Hohsaka T: In vitro translation-based computations. In DNA computing, LNCS 2943 Edited by: Chen J, Reif J. 2004, 197–202.Google Scholar
- Anderson JC, Magliery TJ, Schultz PG: Exploring the limits of codon and anticodon size. Chemistry & Biology 2002, 9: 237–244.View ArticleGoogle Scholar
- Sidransky D: Emerging molecular markers of cancer. Nat Rev Cancer 2002, 2: 210–219.View ArticlePubMedGoogle Scholar
- Winfree E: Algorithmic self-assembly of DNA. PhD thesis. California Institute of Technology, Computer Science Department; 1998.Google Scholar
- Montesano R, Hainaut P, Wild CP: Hepatocellular carcinoma: from gene to public health. J Natl Cancer Inst 1997, 89: 1844–1851.View ArticlePubMedGoogle Scholar
- Jackson PE, Kuang SY, Wang JB, Strickland PT, Munoz A, Kensler TV, Qian GV, Groopman JD: Prospective detection of codon 249 mutations in plasma of hepatocellular carcinoma patients. Carcinogenesis 2003, 24: 1657–1263.View ArticlePubMedGoogle Scholar
- Friedler A, Hansson LO, Veprintsev DB, Freund SMV, Rippin TM, Nikolova PV, Proctor MR, Rudiger S, Fersht AR: A peptide that binds and stabilises p53 core domain: chaperone strategy for rescue of oncogenic mutants. Proc Natl Acad Sci USA 2002, 99: 937–942.PubMed CentralView ArticlePubMedGoogle Scholar
- Bullock AN, Fersht AR: Rescuing the function of mutant p53. Nat Rev Cancer 2001, 1: 68–76.View ArticlePubMedGoogle Scholar
- Sugimoto N, Nakano S, Katoh M, Matsumura A, Nakamuta H, Ohmichi T, Yoneyama M: Thermodynamic parameters to predict stability of RNA/DNA hybrid duplexes. Biochemistry 1995, 34: 11211–11216.View ArticlePubMedGoogle Scholar
- Breslauer K, Frank R, Blocker H, Marky LA: Predicting DNA duplex stability from the base sequence. Proc Natl Acad Sci USA 1986, 83: 3746–3750.PubMed CentralView ArticlePubMedGoogle Scholar
- Nowotny M, Gaidamakov SA, Crouch RJ, Yang W: Crystal structures of RNase H bound to an RNA: substrate specificity and metal-dependent catalysis. Cell 2005, 121: 1005–1016.View ArticlePubMedGoogle Scholar
- Yurke B, Mills AP: Using DNA to power nanostructures. Genetic Programming and Evolvable Machines 2003, 4: 111–122.View ArticleGoogle Scholar
- Yurke B, Turberfield AJ, Mills AP, Simmel FC, Neumann JL: A DNA-fuelled molecular machine made of DNA. Nature 2000, 406: 605–608.View ArticlePubMedGoogle Scholar
- Yan H, Zhang X, Shen Z, Seeman NC: A robust DNA mechanical device controlled by hybridization topology. Nature 2002, 415: 62–65.View ArticlePubMedGoogle Scholar
- Lambert G, Fattal E, Couvreur P: Nanoparticulate systems for the delivery of antisense oligonucleotides. Adv Drug Deliv Rev 2001, 47: 99–112.View ArticlePubMedGoogle Scholar
- Aanta MA, Belguise-Valladier P, Behr J: Gene delivery: a single nuclear localization signal peptide is suffcient to carry DNA to the cell nucleus. Proc Natl Acad Sci USA 1999, 96: 91–96.View ArticleGoogle Scholar
- Liang L, Liu DP, Liang CC: Optimizing the delivery systems of chimeric RNA-DNA oligonucleotides. Beyond general oligonucleotide transfer. Eur J Biochem 2002, 269: 5753–5758.View ArticlePubMedGoogle Scholar
- Miller P, Tso PO: A new approach to chemotherapy based on molecular biology and nucleic acid chemistry: Matagen (masking tape for gene expression). Anticancer Drug Des 1987, 2: 117–128.PubMedGoogle Scholar
- Stec WJ, Zon G, Egan W, Stec B: Automated solid-phase synthesis, separation, and stereochemistry of phosphorothioate analogues of oligodeoxyribonucleotides. J Am Chem Soc 1984, 106: 6077–6080.View ArticleGoogle Scholar
- Brysch W, Schlingensiepen KH: Design and application of antisense oligonucleotides in cel culture, in vivo , and as therapeutic agents. Cellular and Molecular Neurobiology 1994, 14: 557–568.View ArticlePubMedGoogle Scholar
- Lebedva I, Stein CA: Antisense oligonucleotides: Promise and reality. Annu Rev Pharmacol Toxicol 2001, 41: 403–419.View ArticleGoogle Scholar
- Yadava PK: Nucleic acid therapeutics: Current targets for antisense oligonucleotides and ribozymes. Molecular Biology Today 2000, 1: 1–16.Google Scholar
- Van den Berg B, Ellis RJ, Dobson CM: Effects of macromolecular crowding on protein folding and aggregation. EMBO J 1999, 18: 6927–6933.PubMed CentralView ArticlePubMedGoogle Scholar
- Sun XH, Copeland NG, Jenkins NA, Baltimore D: Id proteins Id1 and Id2 selectively inhibit DNA binding by one class of helix-loop-helix proteins. Mol Cell Biol 1991, 11: 5603–5611.PubMed CentralView ArticlePubMedGoogle Scholar
- Proudfoot NJ, Furger A, Dye M: Integrating mRNA processing with transcription. Cell 2002, 108: 501–512.View ArticlePubMedGoogle Scholar
- Adleman L: Molecular computation of solutions to combinatorial problems. Science 1994, 266: 1021–1024.View ArticlePubMedGoogle Scholar
- Kuramochi J, Sakakibara Y: Intensive in vitro experiments of implementing and executing finite state automata in test tube. In DNA11, LNCS 3892 Edited by: A C, NA P. 2006, 93–202.Google Scholar
- Abelson H, Allen D, Coore D, Hanson C, Homsy G, Knight TF, Nagpal R, Rauch E, Sussman GS, Weiss R: Amorphous computing. Comm ACM 2000, 43: 74–82.View ArticleGoogle Scholar
- Applequist J: Estimation of base pairing of nucleic acid from hypochromism. J Am Chem Soc 1961, 83: 3158–3159.View ArticleGoogle Scholar
- Kankia B: Optical absorption assay for strand-exchange reactions in unlabeled nucleic acids. Nucleic Acids Res 2004, 32: e154.PubMed CentralView ArticlePubMedGoogle Scholar
- Reynaldo LP, Vologodskii AV, Lyamichev VI: The kinetic of oligonucleotide replacements. J Mol Biol 2000, 297: 511–520.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.