Team:TU Munich/Project

From 2010.igem.org

Revision as of 16:19, 27 October 2010 by Hartlmueller (Talk | contribs)

Navigation:

Home →  Project

iGEM MainPage

Contents

bioLOGICS: Logical RNA-Devices Enabling BioBrick-Network Formation


Vision

Until today, 13.628 biobrick sequences[1] have been submitted to partsregistry, thereof 102 reporter units, 12 signaling bricks and xx sensing parts. Since there, people are trying to arrange these single biological building blocks in such a manner that allows producing special biotechnological products (metabolic engineering), developing biological sensory circuits (biosensors) and even giving microorganisms the ability to react on multiple environmental factors and serve both as disease indicator and drug. These examples and further promising ideas were implemented on previous iGEM-competitions.[2][3][4]

The idea of combining the outcome of several iGEM competitions to construct complex synthetic biological systems falls at the last hurdle - the fact, that each team uses a different principle how to access and functionally connect the respectively used biobricks. For example, it is a major challenge to create a system that uses several sensoring BioBricks from different iGEM-teams which in turn regulates reportering BioBricks from various teams. In order to combine and fully take advantage of these promising projects, our vision is to develop an adapter that allows interconnecting any arbitrary biobricks on a functional level. Such a system easily allows to setup sensor-reporter circuits and interconnect them to complete biological chips... A further step towards artificial cells.

Read more

Generally speaking, the above descirbed adapter has to meet the following requirements:

  • Universality
The adapter has to be compatible to as many BioBricks as possible. This objective will guarantee that a large number of BioBricks can be connected.
  • Scalability
Once the basic design of the system is established, the construction of the system is supposed to be automated in silico. This way it will be possible to create an adapter connecting a large amount of BioBricks.
  • Biological orthogonality
Interfering with cellular components has to be as low as possible in order to avoid unwanted and perturbing side effects.
  • Logic
The adapter is supposed to not only associate different BioBricks, but to functionally connect BioBricks in a precisely determined manner (including operations such as AND/OR/NOT).


Several biological logic units, devices and circuits have been developed so far[5], but to our knowledge, none was shown to meet all requirement listed above.

Close

Implementation

To functionally connect BioBricks, there are several possibilities including genetic switches, riboswitches and direct protein-protein interactions. We investigated several hypothetically principles, and decided to focus our practical work on the development of a RNA-RNA interaction-based switch. These switches are capable of changing between two states, a state of antitermination and termination, and make use of highly-specific RNA-RNA interaction. In principle such a switch can fulfill all requirements mentioned previously. The following text clarifies how these switches work in detail.

How to connect BioBricks

Our adapter is a system, that activates or disables BioBricks (output BioBricks) in response to the presence of other Biobricks (input Biobricks). Our approach uses a molecular network to put this into practice and consists of four major elements:

Read more


Networks.png

In order to connect different BioBricks, our network requires four major types of components:

  • Input elements
  • Transmitter molecules
  • Logic gates
  • Output elements

Computer vs. molecular network - and our approach
Logic gates in a molecular network are often compared to transistors used in a computer, where billions of transistors are incorporated[7]. The main advantage on a computer chip is, all transistors share the same functional principle, and only the way connecting them in a special sequence allows specific addressing of only a subset of other transistors by an input. However, spatial fixed connections of molecular logic gates are not possible in a living cell. The "wiring" within a cell relies on the specific interaction between transmitter molecule and their corresponding logic gates, for example implemented by protein-protein/ligand-protein interactions or specific ligand-riboswitch interactions.[8][9] As a result, in a cell, each occurring logic gate ("transistor") has to be different, at least in a special recognition site[10] - for example like different transcription factors, recognizing different DNA-sites. Thanks to evolution, nature easily can invent a new transistor for each task - science achieves this only on a limited scale, and producing synthetic molecular logic gates artificially by either rational or evolutionary protein or riboswitch engineering, is limited to small circuits so far[11]. Our project aims to establish a real molecular transistor, which shares the same functional principle for all logic gates. At the same time, we want to design a easily exchangeable recognition site, which can individually be designed by everyone!

These elements can be combined to build up a molecular network (see illustration). Each input molecule (such as a BioBrick) produces a unique transmitter molecule. All transmitters belong to the same type of molecule and share a common design. However each transmitter molecule can only interact and activate a certain subset of logic gates. In other words, logic gates have to recognize as well as bind the corresponding transmitter molecules and are capable of producing a new output transmitter molecule. Depending on the type of the logic gate (AND, OR or NOT[6]), an output transmitter is only created if both input transmitter molecules are present (AND), at least one of two input transmitters is present (OR) or if no input transmitter is present at all (NOT). Once a logic gates has produced a new output transmitter, these transmitters can in turn address another subset ("layer") of logic gates. In theory many layers of logic gates can be connected this way allowing the creation of large networks. Until this step, various transmitter molecules might have been produced. But in order to create a Biobrick output, the last layer logic gates finally generates transmitter molecules that will not active logic gates, but will rather interact with the cell metabolism to produce a BioBrick responds. In other words, the last layer of transmitter molecules is capable of regulating BioBrick formation.


Summarizing, the network established a connection between input BioBricks and output BioBricks in a functional manner. Having addressed the basic layout of the molecular network, the next step is to determine what type of molecules can perform the required functions. We decided to use RNA, both as transmitter molecules and for constructing logic gates. Several advantages result from the utilization of RNA as the central element:

  • During the last years, many Biobricks were designed that are sensitive to various chemicals and substances. These BioBricks often function as a transcription factor that binds to a specific DNA sequence and consequently would be capable to produce a specific transmitter RNA molecule. Thus, in principle each BioBrick which involves transcription can be integrated in our network.
  • Since all logic gates are capable of producing transmitter RNA, they can also produce functional mRNA encoding any protein. This means, each BioBrick consisting of protein or RNA can be produced as an output of our network.
  • If RNA forms both, the transmitter molecule and the logic gates, they can specifically interact by RNA-RNA interaction, which is highly predictable compared to protein interactions. This allows to generate a library of transmitters and gates in silico. Such a library is essential for the creation of large networks.
  • RNA production is fast and energy saving for a cell. Consequently, operating a network that only produces RNA rather than proteins will also be faster and more efficient for the host cell. Since our logic gates are based on transcription, translation and resource consuming protein production will only be required at the very last step.
  • As the half-time of RNA can be rather short, transmitter RNA will not accumulate within the cell and it is therefore less likely for the system to become saturated.

Close

Design and functional principle of logic gates

The concept introduced above provides a framework that can potentially serve as an universal adapter between different BioBricks. However, the logic gates have not been specified more precisely so far. This will be done in the following section.

Read more

Generally speaking, our logic gates are to posses the following characteristics:

  • Logic gates, such as AND, OR and NOT, have to be implemented by RNA-interaction based principles (see How to connect BioBricks).
  • All logic gates have to recognize their corresponding transmitter RNA´s and, in response, produce an output transmitter molecule.
  • Logic gates should follow a basic design rule, in such a way, that their creation can be automated in silico.
  • The response efficiency of a logic gate toward a transmitter molecule should be comparable for all logic gates to provide calculable robustness and sensitivity. This will ensure comparable molecular concentrations and functionality of large networks.
  • The system has to be designed for in vivo utilization at the first place. As a reference we always assumed a temperature of 37 °C and an E. coli environment.


In order to build a logic gates for our bioLOGICS system we will first create a simple toggle switch. A toggle switch can be activated by one transmitter RNA and produce an output transmitter RNA. In contrast to a logic gate, a toggle switch does not perform logic operations. However by combining toggle switches, logic gates can be created. The following text will first describe how the developed toggle switch works and secondly, how logic gates such as AND/OR/NOT can be created using these toggle switches.

Toggle Switch

Roughly speaking, a toggle switch can be regarded as an enhance transcriptional terminator. The enhancement can be described easier by dividing a toggle switch into its components:

  • Target site
The target site is the functional core element of our switches, allowing a shift between an "on" and "off" state. Since we work on the level of RNA-production (transcription), a "switchable" transcriptional terminator is suitable for this purpose. By allowing or preventing formation of a transcriptional terminator, that is by switching between termination and antitermination it is possible to represent an "off" and an "on" state, respectively. Therefore, the target site is the 5' ending of the terminator and is required for a stable terminator formation. It should be noted that this principle was also observed in nature.
To highlight and illustrate the functional principle of our switches, only the part of the terminator which is involved in interacting with a transmitter molecule and which is responsible for shifting between "on" and "off" state is called target site. The remaining terminator sequence is called terminator in the following, even if both, target site and terminator build up the terminator structure occurring in nature.
The important aspect of our toggle switches is the fact that all switches will hold the same identical target site. Therefore having found one functional "switchable" terminator, will allow almost unlimited upscaling since this terminator can be used for a large library of switches. This is the main difference to previous works done on this field, which always required developing a new shifting principle for each switch.[12][13][14] Beside this scalability, this principle provides a comparable on/off shifting rate (responds function) for all switches, avoiding complex fine tuning of molecular networks.
To sum it up, the target site, allows to switch between an "on" and "off" state. But so far, the toggle switch is not capable of performing specific interaction with transmitter molecules. This is where the recognition site comes into play.
  • Recognition site
The recognition site defines which transmitter molecule can actually interact with the toggle switch. Therefore, a unique recognition site is generated for each switch and is positioned right upstream of the target site. In principle the recognition can be any random sequence as long as it remains unique within the molecular network.

Summing up, the recognition site allows a specific interaction between switches and transmitter molecules. Once this interaction is formed, an interaction between the transmitter and the target will actually switch the state of the terminator. This allows the specific arrange and interconnection of numerous of these switches by transmitter molecules, without changing the target site. Comparable to wires connecting many identical transistors, our target site remains the same.

Transmitter RNA´s, input and output of bioLOGICS toggle switches

Transmitters present the "trigger" to shift switches between the "on" and "off" state. They requires the ability to change the terminators secondary structure and cause antitermination, BUT only if a special recognition site is detected. Thus, each signal consists of a trigger site, interacting with the switches trigger site and an identity site, interacting with the toggles recognition site. Practically, the shift between antitermination and termination is induced by a complementary RNA-sequence, influencing the terminators secondary structure. But in contrast to previous approaches on this field [12], we introduced the described synthetic trigger site in such a manner it is not able to change the terminator´s state on its own, but only in combination with the identity site, which is complementary to the recognition site.

The challenge is to arrange and optimize these elementary building blocks thermodynamically, that a trigger site is only able to switch in combination with its respective identity site. This was done by in silico design using NUPACK, presented in section in silico design.


Putting it all together: the switching process


Rectangles present the composition of our functional units on the level of DNA. Fringed lines represent RNA produced by RNA polymerase. The stem loop structure depicts the switchable terminator. Terminator and target site are illustrated in blue and turquoise, respectively. Recognition sites are indicated in different colors, in this case red for the input transmitter and green for the output transmitter.Each toggle switch and or later logical unit has to be flanked by a promotor and another constitutive terminator, to allow RNA-production by RNA-polymerase in a proper way.
The functional principle of the designed toggle switches is illustrated in the figure above. The switch is implemented upstream of a desired output transmitter. So in the ground state, transcription will be canceled by the formation of a RNA stem loop in the nascent RNA-chain, which then causes the RNA polymerase to stop transcription and fall off the DNA and no output RNA will be produced. We rely on rho-independent termination

On the other hand, in the presence of a input transmitter, this small functional RNA inhibits the stem loop formation by complementary base-pairing and hence avoids termination of transcription. In detail, the identity site (red part on transmitter) binds the recognition site (red part on switch) and serves as toehold, which will thermodynamically allow the trigger site (turquoise part on transmitter) to perform a strand displacement and open the stem loop structure, allowing the polymerase to read through and form the output RNA.
Summing up, we use can use this concept to create a switch that can be toggled by a transmitter RNA molecule and in response, is able to produce another transmitter RNA.

Main challenges are

  • to find a suitable terminator construct and design a complementary trigger unit, which is only functional in combination with a specificity site - meaning an optimization of the thermodynamically parameters (see in silico design)
  • to investigate whether the signal/switch interaction reaction is really on a timescale

to be competitive to terminator formation - meaning an comparison of kinetic parameters (see Modeling page)


The bioLOGIC toggle switch with regard of logical operations

As described, each switch can be accessed by a specific RNA-transmitter molecule, illustrating the input. In turn, another RNA-transmitter molecule will be produced if the switch shifts its state. This output transmitter of one switch can serve as input transmitter for the next toggle switch by meaningful selection and design of the respective recognition sites. This easily allows arranging several switches in specific sequences and faulty wiring - the corner stone of a logical network.

To ease the building of logical networks, applying mathematical logics, e.g. Boolean logics like in computational science would be worthwhile. It is possible to establish general Boolean operators with our switches and thus build "logical modules". Since AND/OR/NOT are the most simple logic operations which can be implemented with the presented toggle switches, and all remaining operations can be expressed by these three operators[ZITAAAT Wiki oder so)we exemplary designed them.

*AND consists of a parallel circuit of two switches
AND2.png
*OR is implemented by connecting two switches in series
OR2.png
*NOT is more complex to explain. In principle, it consists only of one toggle switch which contains its respective signal molecule intrinsic, so via intramolecular interaction, antitermination is the initial state. The signal is composed of the same components as usual to allow interconnection with other logic gates.
NOT2.png

Network construction

Designing complex biological networks based on either traditional protein engineering or our new bioLOGICS is still a complex task. We developed a software which allows the fast construction of a bioLOGICS based networks.
To read more about this, look at our Software page

Close


Our Objective

Putting the implementation described above into pratise, will be a major challenge. For this year's iGEM competition our goal is to do the first step: design and build a switch that can be toggled by a RNA molecule. To be precise, we want to modify a transcription terminator, in such a way, that it interacts with a second RNA molecule and as a result is no longer capable of forming a stem loop. Once the objective mentioned above is accomplished, these basic RNA/RNA-interactions have to be modified in such a manner that the described identity/trigger site pattern for the transmitter and the complementary recognition/target site switch composition has to be established. The most important requirement is to is to optimize these modules that the transmitter is only able to toggle switches specifically, meaning only in the presence of both, identity AND trigger site.
Once the objective mentioned above is accomplished, the creation of an OR gate will be rather simple since it only requires two switches. However the creation of an AND or NOT gate and optimizing the logic gates to improve their responds function will remain the goal of future work. Also the creation of small networks and the correct integration of BioBricks as input and output molecules will be future challenges. Furthermore, we wanted to rather focus on the development and the testing of our structural design of the switches, rather than developing a variety of new BioBricks.

In silico design

As desribed above, our toggle switches are based on certain design rules. However, there still are different strucutral paramters that need to be tested and optimized (length of recognition site and target site, choice of terminator, etc.). We used in silico design and modeling ) to test different parameters. Furthermore we tried to use the antitermination principle observed in nature, such as attenuation in E. coli or tiny abortive RNA´s of T7-phage.

Evaluation and Measurements

To evaluate the functionality of our molecular switches, we first had to establish several assays. Therefore, we improved an existing in vivo assay and developed an in vitro assay for this purpose. For more information please refer to the lab section.

Results

Every network starts with a basic unit. While our declared aim is to enable networks allowing fine-tuning of gene expression beyond the regular on/off, exploring such an on/off switch/signal pair is the first step towards a functional network. We constructed several units and tested their efficiency, robustness and reproducibility in vivo, in vitro and in silico. Furthermore we developed a software which allows easy constructions of networks based on our designed logic gates. Conclusive elaboration of a few first RNA-based logic units is the major contribution of our iGEM team.

in silico Design of Switching and Trigger Unit

Read more

attenuation principle

A random sequence was derived from xxx. A complementary sequence, reaching within the terminator´s stem loop was stepwise shortened to find the length, where the formation of the terminator is thermodynamically favored compared to the strand displacement by the signal. The trigger sequence was defined by selecting the shortest unit which still is able to "destroy" the stem loop. Subsequently, the trigger unit was tested in regard of not being able to resolve the stem loop on its on. As table xxx illustrates, the terminator is thermodynamically favorite toward the trigger unit, but in combination with the specificity site, binding becomes possible.


element free energy [kcal/mol] ratio of pairing [%]
toggle switch (terminator secondary structure) -30.10 -
transmitter + toggle switch -72.88 100
trigger site + target site -24.21 0
identity site + recognition site 0
Toggle switch:Comparison of free energy for secondary structure formation at 37°C, simulated by using NUPACK analysis.



Pairing probabilities of the His-Terminator based toggle switch simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based toggle switch secondary structure, generated by NUPACK utility tools
Pairing probabilities of the His-Terminator based toggle switch in presence of the respective transmitter RNA, simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based toggle switch secondary structure in presence of the respective transmitter RNA, generated by NUPACK utility tools
Pairing probabilities of the His-Terminator based toggle switch in presence of the transmitter RNA´s trigger site only, simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based toggle switch secondary structure in presence of the transmitter RNA´s trigger site only, generated by NUPACK utility tools
Pairing probabilities of the His-Terminator based NOT-gate in presence of the transmitter RNA´s identity site only, simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based NOT-gate secondary structure in presence of the transmitter RNA´s identity site only, generated by NUPACK utility tools



NOT Gate

element free energy [kcal/mol] ratio of paaring [%]
toggle switch (NOT-gate secondary structure) -68.18 -
transmitter + switch -98,18 100
trigger unit + target site - 0
identity site + recognition site 0
NOT-gate:Comparison of free energy for secondary structure formation at 37°C, simulated by using NUPACK analysis.


Pairing probabilities of the His-Terminator based NOT-gate simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based NOT-gate secondary structure, generated by NUPACK utility tools
Pairing probabilities of the His-Terminator based NOT-gate in presence of the respective transmitter RNA, simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based NOT-gate secondary structure in presence of the respective transmitter RNA, generated by NUPACK utility tools
Pairing probabilities of the His-Terminator based NOT-gate in presence of the transmitter RNA´s trigger site only, simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based NOT-gate secondary structure in presence of the transmitter RNA´s trigger site only, generated by NUPACK utility tools
Pairing probabilities of the His-Terminator based NOT-gate in presence of the transmitter RNA´s identity site only, simulated by NUPACK secondary structure analysis tool
Illustration of the His-Terminator based NOT-gate secondary structure in presence of the transmitter RNA´s identity site only, generated by NUPACK utility tools

tiny abortive principle

ubiquitous terminators

Close

Modeling

We modeled our bioLOGICS toggle switches theoretically to prove that our they are able to interrupt termination. We also optimized the switches and the corresponding signals and furthermore took a look a combination of toggle switches that result in logic gate. See our Modeling page for further details.

in vivo Functionality Screening

Since our logic gates are intended to function in living cells, in vivo measurements were essential. In a set of experiments we concentrated on two different toggle switches based on known [[1]] from nature: the HisTerm and TrpTerm. Focusing on fluorescent proteins for quantifiable input and output we designed a functional and robust screening system. For greater detail see Experimental Design. Unfortunately, setting up a working screening system failed twice. Only in redesigning and improving the screening plasmid pSB1A10 we succeeded, but lost precious time.

Ultimately, the two switches displayed remarkable differences in their terminator efficiency, but neither of them responded to their corresponding signal. However, screening one transmitter signal cannot disprove our principle system. Limited by time, we hope for future teams to take up our work and to use our improved test system that we submitted to the parts registry.

Read more

Considering the high complexity of in vivo measurements compared to other experimental challenges, a robust and easy to handle test system for PoPS-based devices is desirable. As described in Experimental design, we used fluorescent proteins: eGFP to normalize the input and a RFP or mCherry to measure the output. Our first t attempt, using the screening plasmid pSB1A10 yielded no interpretable results. Switching the fluorescent protein to mCherry did not work either, but after several experimental setups we determined a transcriptional problem causing no reporter protein expression regardless of the inserted part. Thereby we demonstrated the screening plasmid pSB1A10 to be malfunctioning. Finally a new design based on pSB1A10 lead to a functional and robust screening system. A second promoter with identical induction properties inside the BioBrick cloning site enforces transcription of the PoPS-based device and the mCherry output.

Exemplary, the graph below on the right shows the positive control, induced and uninduced at OD600=0.7 followed by 16 h incubation at 25 °C. Clearly visible are eGFP and mCherry fluorescence in the induced samples. The uninduced control showed no fluorescence at all, demonstrating the PBad promoter to be tight and providing no basal transcription. A major advantage for the screening system. This newly designed screening approach renders the characterization of PoPS-based devices in general and toggle switches in particular easy and robust.

Bacteria containing positive control
Emission spectra of induced (green/red) and uninduced(black) positive control BBa_K494002 ; green: eGFP fluorescence ex: 501 nm, red: mCherry fluorescence ex: 587 nm


Due to the time limitations of the iGEM completion we had to focus our efforts on few toggle switches after designing the screening system. Since nobody does it better than nature, we choose the HisTerm as well as the TrpTerm. Both switches are based on known natural [[2]]. Delorme et al. reported the His-Terminator to be a remarkable effective Terminator with more than 99% termination efficancy.[12] The exemplary measurement below on the right confirms the high terminator efficiency. In fact, we could not detect any mCherry fluorescence in any cells containing the HisTerm. Even induction of the corresponding signal transmitter RNA via IPTG did not alter the Terminator efficiency. Again time was the limiting factor and prevented us from testing more than one corresponding signal. Thus, the results are insufficient to disprove the functionality of the HisTerm or our concept in general.

Bacteria containing HisTerm
Emission spectra of induced and uninduced screening plasmid BBa_K494002 containing HisTerm ; green: eGFP fluorescence ex: 501 nm, red: mCherry fluorescence ex: 587 nm


Attaining only 90% terminator effiecency, the natural Trp [[3]] is known be less effective than the HisTerm.[13] The graph on the right depicts the to our designed TrpTerm characteristic efficiency of about 40 %, notably below the natural standard. Allowing 60% transcription in the “off” state excludes the TrpTerm from possible candidates for a scalable network of logic gates. Thus the TrpTerm is inoperative as intended, but may still be useful in other contexts. Similar to the HisTerm, the TrpTerm also did not react to the induction of the corresponding signal.

Bacteria containing TrpTerm
Emission spectra of induced and uninduced screening plasmid BBa_K494002 containing TrpTerm ; green: eGFP fluorescence ex: 501 nm, red: mCherry fluorescence ex: 587 nm


Making use of our improved screening system we also carried out some in vivo kinetic measurements in addition to the end-point measurements above. In contrast to the in vitro experiments we did not obtain significant results for the chracterization of our toggle switches. As the switching process is many times faster than protein synthesis our in vivo kinetics include the synthesis of mCherry as well as its maturation. Therefore we centered our attention on end-point experiments. For more information browse the lab book.
Considering our in vivo measurements, neither of the tested switches showed any effect regarding the signal induction. But due to the small number of tested switches and signals this can hardly be regarded as disprove of concept. In particular in light of the recent findings by Sooncheol proving antitermination in principle using a T7 system.[14]

Close

in vitro Screening

Read more

in vitro translation

in vitro transcription

E.coli RPO

T7 RPO

Close

Software

Although we could not show the full functionality of bioLOGICS in the lab we still want to demonstrate the potential of our approach. Hence we implemented the idea behind our logic gates in a program which illustrates how bioLOGCIS theoretically would allow the construction of complex information processing networks interconnecting BioBricks. For further details take a look at our Software page.


Outlook

... Future plans will also work with Synthetic Terminators, which will retrieve additional informations what drives the process of Termination. ...

References

[1] http://partsregistry.org/cgi/partsdb/Statistics.cgi [2] https://2009.igem.org/Team:Imperial_College_London/M1 encapsulation [3] https://2009.igem.org/Team:TUDelft [4] https://2008.igem.org/Team:Heidelberg [5] Smolke and so on.... [6] http://en.wikipedia.org/wiki/Logic_gate#Symbols [7] http://en.wikipedia.org/wiki/Moore's_law [8] http://en.wikipedia.org/wiki/Protein_interaction [9] http://en.wikipedia.org/wiki/Riboswitch [10] http://en.wikipedia.org/wiki/Binding_sites + http://en.wikipedia.org/wiki/Recognition_site [11] irgend ein damn review über directed evolution and so on [12] Delorme, Ehrlich and Renault, Regulation of Expression of the Lactococcus lactis Histidine Operon. Journal of Bacteriology, Apr. 1999, p. 2026–2037 [13] Trun and Trempy(2003): Fundamental Bacterial Genetics, Wiley-Blackwell, Chapter 12 [14]Sooncheol Lee, Huong Minh Nguyen and Changwon Kang, Tiny abortive initiation transcripts exert antitermination activity on an RNA hairpin-dependent intrinsic terminator. Nucleic Acids Research, 2010, 1–9 [15] [16]