In vivo Intelligent Agents with CRISPR-Cas9 for Molecular Learning

Christina Baek1, Byoung-Tak Zhang2

1Interdisciplinary Program in Neuroscience, Seoul National University,

2Department of Computer Science & Engineering, Seoul National University dsbaek@bi.snu.ac.kr

In vivo 환경의 지능형 에이전트를 위한 CRISPR-Cas9기반의 분자

학습 방법

백다솜 1, 장병탁 2 1서울대학교 뇌과학 협동과정, 2서울대학교 컴퓨터공학부

dsbaek@bi.snu.ac.kr

Abstract In vivo computing is rapidly evolving how science today is understood and engineered to create

smarter agents, capable of co-existing in the complex and dynamic environments of biological systems. Here, we discuss the various breakthroughs in DNA nanotechnology and limitations of current in vivo methods where DNA nanodevices are used for drug delivery. We propose alternative approaches for in vivo agents, which avoid the limitations of unwanted immune responses or unclear location of nanodevices. We discuss the revolutionary CRISPR-Cas9 genome editing technique and how this could be used in conjunction with implementing molecular learning with mammalian cells and model organisms.

1. Introduction

Biologists and engineers alike have recently realized the

idea of biocomputers, capable of sensing, interacting and

responding to dynamic biological environments. Such

artificial biological systems are built for a deeper

understanding of the many natural biological phenomena

yet to be elucidated as well as for a range of personalized

medical applications in diagnostics, disease prevention and

therapeutics.

One key example is in the advancement of DNA

nanotechnology, where dynamic DNA molecules are used to

build cellular controllers with molecular circuitry [1].

Boolean logic circuits [2], finite-state machines [3], and

neural networks [4] are a few of many demonstrations of

DNA computing. A study on DNA origami nanorobots was

first published in 2012, where the nanorobot could

recognize cell surface proteins for targeted delivery of

molecular payloads to specific cell lines [5]. The biological

compatibility of DNA technology [6] has rapidly conjured

development of in vivo applications.

Here, in vivo refers to studies implemented “within the

living”, meaning in living cells or whole organisms. From cell

specific uptake in Caenorhabditis elegans (C. elegans) [7]

to DNA nanostructures targeting tumors in a living animal

in 2012 [8], to the one example of a DNA nanostructure

performing logical and dynamic operations in living

cockroaches [9] and now even a thought-controlled robot,

living in a host [10], in vivo methods of drug delivery are

rapidly revolutionizing existing methods of treating diseases

such as cancer.

The concept of DNA nanostructures targeting only

cancer cells with a lock and key mechanism, releasing

chemotherapy drugs at specific sites, leaving the healthy

cells untouched sounds very promising [5]. But, why are

Figure 1. Molecular learning starting from random DNA sequences transfected into cells, selection of specific genes

with disease-related gRNA sequences (insertion of DNA at these sites), addition of independent cell culture with

new set of random DNA transfected into them, second round of selection and so forth to finally produce a culture

of cells with inserts in between all disease-related genes.

823

2017년 한국컴퓨터종합학술대회 논문집

clinical trials still not underway? Although these drug

delivery techniques have been implemented in vitro, and

in organisms with the benefit of low systemic nuclease

activity and volumes such as C. elegans and cockroaches,

the immense complexity and obstacles associated with the

human body, biological barriers and the immune system

response cannot be represented through existing studies.

Several issues with molecular, cellular and organism level

pathways responding to the immunogenic DNA is a critical

part of investigating into in vivo applications [11]. There is

now one report of reduced tumor growth in mice from

administering DNA tetrahedrons, where the DNA structures

were found to accumulate in the liver, and kidneys as well

[12]. There is no study into the effects of foreign DNA

nanorobots to other highly correlated factors such as cell

apoptosis which raises many questions about the safety of

such nanostructures for clinical trials.

Here, we introduce alternative methods to specifically

target diseased cells and induce preventative or therapeutic

strategies from within the cell, avoiding the issues related

to inserting nanostructures into organisms. Although these

ideas require future work to fully implement experimentally,

the purpose of this paper is to propose a new experimental

paradigm, where in vivo intelligent agents could be realized

through the use of the new frontier of genome engineering,

CRISPR Cas-9 together with concepts of molecular learning.

Figure 1 shows a simple overview of this.

2. Methods and Design

Cell culture and transfection protocols typically used with

mammalian cell lines will be used. Human embryonic kidney

cells 293 (HEK 293) are commonly used in cell biology for

healthy growth in culture and propensity for transfection

[13]. Dulbecco's Modified Eagle Medium (DMEM) and fetal

bovine serum (FBS) containing cell culture media is used,

Dulbecco’s Phosphate Buffered Saline (DPBS) for washing

cells and 0.05% Trypsin-EDTA for detaching cells for

subcultures. Cells should be incubated at 37 °C in 5% CO2

and subcultured every 48 hours. Transfection of DNA into

cells is a relatively simple procedure for inserting DNA to be

expressed into eukaryotic cells. We will use a form of

chemical transfection using Lipofectamine [14]. DNA is

transfected into cells by this agent, which envelopes the

DNA of interest into a liposome which then merges with the

cell membrane of the cell, allowing the uptake of the

liposomal contents (Figure 2).

CRISPR-Cas9 system is the new era of genome editing

technology. It has transformed genetic engineering, the

next big break since polymerase chain reaction (PCR),

whole-genome sequencing and cloning came about.

CRISPR-Cas9 offers a faster, cheaper and highly precise

technique of editing DNA. CRISPR is short for clustered

regularly interspaced palindromic repeats and contains the

Figure 3 CRISPR-Cas9 system used to cleave and

insert a desired DNA sequence (green) at specifically

the site in the original DNA (targed DNA sequence)

using the sequenced guide RNA (gRNA)

Figure 2. Transfection of desired DNA sequence into

cells can induce targeted protein expression

824

2017년 한국컴퓨터종합학술대회 논문집

molecular scissors, the nuclease, to cut the DNA. Protein

Cas9 is associated with the guide RNA (gRNA) provided to

the system, to cut precisely where the gRNA binds with

perfect complementarity to the DNA inside the cell. When

and only when the gRNA binds to the target DNA, the DNA

separates allowing for the ribonucleoprotein complex to

cleave and inactivate this gene [15]. The modification

technique of inserting a DNA at this very site will be used to

inactive and add a DNA sequence of our interest (Figure 3).

Molecular learning is intertwined with these molecular

biology techniques to design an in vivo intelligent agent

which selects for specific disease-related genes with CRISPR

Cas-9 and adds in a specific gene of our choice in that place.

First, random DNA sequences of varying lengths are

transfected into the cells. Next, disease-related genes are

cleaved according to the type of gRNA we design. For

example, targeting the γ-secretase gene may be beneficial

for the reducing the well-known hallmark of Alzheimer’s

disease, the amyloid-β (Aβ) plaque aggregation from

incorrectly spliced precursor protein, APP [16]. Inactivation

of this gene may lower the Aβ plaque buildup. Furthermore,

an insertion of the α-secretase gene in its place could

promote correctly cleaved APP. Repeated iterations of this

process with new batch of cells containing different DNA

transfections may lead to a convergence on the types of

DNA relevant for targeting Aβ plaque buildup. It is a way of

unsupervised learning as cells will eventually have targeted

out unnecessary or pathogenic DNA sequences and leave in

beneficial DNA in its place (Figure 1).

Finally, a method of studying the molecular agent’s

behavior is thought of as an exciting avenue to link these

regimes of molecular learning with actual movable,

autonomous organisms where decision making can be

observed. A suitable method is the PDMS maze for C.

elegans, designed inside an agar plate with a decision

making area in the center and various directions for the C.

elegans to move to inject bacteria containing different DNA

sequences (Figure 4) [17]. Similar to that explained in

Figure 1, here, transfection will be replaced by the ingestion

of bacteria containing DNA, and through multiple iterations,

including training with volatile substances with gRNA for

gene inactivation or pheromones for chemotaxis with gRNA

for gene insertion, at the end of training, C. elegans with

the best set of DNA combinations will survive.

3. Conclusion

With the use of the cutting edge technology of CRISPR Cas-

9, which has already reached human clinical trials to treat

aggressive lung cancer [18, 19], cunning design and

collaboration with molecular learning concepts may help

pave the way

to revolutionize the applications for in vivo intelligent

agents to diagnose or treat human diseases.

Acknowledgement

This work was supported by Samsung Research Funding Center

of Samsung Electronics under Project Number SRFC-IT1401-12

References 1. Zhang, David Yu, and Georg Seelig. "Dynamic DNA nanotechnology using

strand-displacement reactions." Nature chemistry 3.2 (2011): 103-113. 2. Qian, Lulu, and Erik Winfree. "Scaling up digital circuit computation with DNA

strand displacement cascades." Science 332.6034 (2011): 1196-1201. 3. Benenson, Yaakov, et al. "Programmable and autonomous computing machine

made of biomolecules." Nature 414.6862 (2001): 430-434. 4. Qian, Lulu, Erik Winfree, and Jehoshua Bruck. "Neural network computation with

DNA strand displacement cascades." Nature 475.7356 (2011): 368-372. 5. Douglas, Shawn M., Ido Bachelet, and George M. Church. "A logic-gated

nanorobot for targeted transport of molecular payloads." Science 335.6070 (2012): 831-834.

6. Chen, Yuan-Jyue, et al. "DNA nanotechnology from the test tube to the cell." Nature nanotechnology 10.9 (2015): 748-760.

7. Bhatia, Dhiraj, et al. "A synthetic icosahedral DNA-based host–cargo complex for functional in vivo imaging." Nature communications 2 (2011): 339.

8. Lee, Hyukjin, et al. "Molecularly self-assembled nucleic acid nanoparticles for targeted in vivo siRNA delivery." Nature nanotechnology 7.6 (2012): 389-393.

9. Amir, Yaniv, et al. "Universal computing by DNA origami robots in a living animal." Nature nanotechnology 9.5 (2014): 353-357.

10. Arnon, Shachar, et al. "Thought-controlled nanoscale robots in a living host." PloS one 11.8 (2016): e0161227.

11. Surana, Sunaina, Avinash R. Shenoy, and Yamuna Krishnan. "Designing DNA nanodevices for compatibility with the immune system of higher organisms." Nature nanotechnology 10.9 (2015): 741-747.

12. Huang, Yanyu, et al. "A multifunctional DNA origami as carrier of metal complexes to achieve enhanced tumoral delivery and nullified systemic toxicity." Biomaterials 103 (2016): 183-196.

13. Lin, Yao-Cheng, et al. "Genome dynamics of the human embryonic kidney 293 lineage in response to cell biology manipulations." Nature communications 5 (2014): 4767.

14. Felgner, Philip L., et al. "Lipofection: a highly efficient, lipid-mediated DNA-transfection procedure." Proceedings of the National Academy of Sciences 84.21 (1987): 7413-7417.

15. Doudna, Jennifer A., and Emmanuelle Charpentier. "The new frontier of genome engineering with CRISPR-Cas9." Science 346.6213 (2014): 1258096.

16. LaFerla, Frank M., Kim N. Green, and Salvatore Oddo. "Intracellular amyloid-β in Alzheimer's disease." Nature Reviews Neuroscience 8.7 (2007): 499-509.

17. Zhang, Yun, Hang Lu, and Cornelia I. Bargmann. "Pathogenic bacteria induce aversive olfactory learning in Caenorhabditis elegans." Nature 438.7065 (2005): 179-184.

18. Cyranoski, David. "CRISPR gene-editing tested in a person for the first time." Nature News 539.7630 (2016): 479.

19. Castillo, Andres. "Gene editing using CRISPR-Cas9 for the treatment of lung cancer." Colombia Médica 47.4 (2016): 178-180.

Figure 4. Maze assay for C. elegans allowing decision

making behavior to be observed.

825

2017년 한국컴퓨터종합학술대회 논문집