Team:Peking/Results
Proof of Concept
Sequence-Specific Binding of dCas9 to DNA is Validated by Classical CRISPR-interference Method
To investigate CRISPR-dCas9 binding specificity and affinity with DNA, we started with arabinose-induced expression of dCas9 targeted to mRFP coding region, with assistance of constantly-expressed single guide RNA that is complementary to the corresponding sequence (Figure 1A). Expression level of dCas9 induced by arabinose is tested by fluorescence of fused dCas9-GFP (Figure 1B). Since normal mRNA elongation is interrupted by occurrence of dCas9, fluorescence is greatly decreased as the arabinose concentration increases (Figure 1C), comparing to a single guide RNA that has no binding specificity to DNA. This proves the basic concept that a dCas9 protein is able to bind to DNA with sequence specificity and interferes with the physiological process. All plasmids we use to interfere with DNA replication subsequently are derived from this.
Fig.1:Sequence-Specific Binding of dCas9 to DNA is Validated by Classical CRISPR-interference Method Fig.1A:CRISPR-interference inhibits mRFP expression by interrput the elongation of mRNA. Fig.1B:Measurement of dCas9 expression level in E. coli by test the fluorescence of dCas9-GFP. Fig.1C:Dose-dependent effect of CRISPRi on mRFP expression level. NT1 is on the coding region of mRFP. Ctrl group is PolyA.
An sgRNA Library Allows for a Rapid Scan for Essential DNA Boxes in vivo
In-vitro methods for identifying protein-nucleic-acid interactions are always controversial for significant distinctions between their experimental conditions and real in-vivo physiological environment. However, classical in-vivo method to search for essential DNA boxes involved in biological process is mainly by site-directed random mutagenesis, which is almost impossible for rapid and high-throughput scanning. Based on CRISPR-interference method for transcription inhibition, we develop a novel approach for prokaryotic genome replication interference (CRISPRri). Hence, a 20-bp sgRNA is designed to be complementary to OriC, the genome replication origin (Figure 2A). Instead of site-directed mutations one by one, CRISPRri allows for 20-bp scan each time. Although CRISPRri requires a PAM ("NGG") sequence to execute its function, we found a high occurrence frequency of PAM in the region of replication origin and all available sgRNAs can cover 76.2% (221 out of 290) of OriC.
Seven different targeting sites for dCas9 is designed to test the effect on cell growth (Figure 2B). Binding box nomenclature used here is the conventional name of binding box (e.g. "M") followed by a "+" or "-", which stands for the direction of sgRNA from 5' to 3'. For instance, M box with its bottom strand bound by dCas9 is thus called M+ box. TOP 10 strain is used as the chassis organism.
To precisely record the bacteria growth under stable conditions, a microfluidic chip is developed to adapt to observed features of bacteria (Figure 2C). All repeat groups are under flow of the same culture to ensure that the experiment results will not be affected by irrelevant external conditions. We have pointed out that interference of genome replication initiation would result in longer cell cycle and cell number doubling time. Here we take a 90 um * 90 um microscopic view each repeat group for cell counting every half an hour. It turns out that CRISPRri targeted to different boxes on OriC results in variant levels of cell doubling time extension, even though intervals between these boxes are only tens of base pairs (Figure 2D). This is consistent with our expectations based on literatures, that functions and essence of different DNA boxes on the OriC and their contributions to genome replication vary a lot. Combined with known mechanism in DNA replication initiation, it is found out that our results accord with the DnaA binding affinity reported previously. High DnaA affinity boxes, like R1 and R3, were shown to have severe inhibition effect when targeted by dCas9. For typical low affinity box, like M box, the effect of CRISPRri is much milder. The only exception is R4, which was reported to be a high-affinity box but shows slight effect on cell growth..
Other boxes that are not considered to be DnaA binding boxes, like MR13 and IHF—they are, in fact, binding boxes for proteins other than DnaA—have mild effects. Moreover, we found another box which shows no inhibition ability on cell growth and the box happens to locate at the linker sequence of two DnaA binding boxes (M and R2) and is never reported to have specific biological functions before. These phenomena conform to our knowledge about DNA replication initiation that cooperative binding of DnaA is the rate-determining step.
Development, Characterization and Optimization of CRISPR-Based DNA Replication Interference (CRISPRri)
To develop a programmable and highly-adjustable method to realize precise control of DNA replication, we finely tuned the CRISPRri on multiple aspects , including plasmid copy number, inducer, targeted boxes and other extension for wider and smarter use of the system(Figure 3A). Multiple methods are developed to characterize and measure the system in detail and completeness. Cell number doubling time, nucleo-cytoplasmic ratio, morphology and irrelated protein productivity are seen as the outputs of the system and are all well described and tuned. These four parameters of E. coli are taken into consideration because they stand for different features of evaluating the cell state. A rounded characterization system, including multiple measurement methods, is well developed as a full-scale quantitative description of E. coli general states..
We picked up M+ box as the target site of dCas9 due to its relatively milder effect on cell growth, which avoids over-inhibition under low inducer concentration. Another important reason for choosing M+ box instead of other mild boxes like IHF box, is that its relation with DnaA binding shows a clear picture on how DNA replication initiation is delayed and might have higher predictability. For a same reason, we chose the vector with medium copy number to carry the dCas9 gene and sgRNA. We found an arabinose-dose-dependent increase in cell number doubling time with a considerable dynamic range (Figure 3B). This realizes preliminary adjustment of bacteria division time.
Other factors, such as expression level of sgRNA and length of sgRNA are also tested. Single guide RNA with T7 promoters is transformed into bacteria in which T7 polymerase is induced by IPTG. The feature of the promoter is tested by expression of GFP and also CRISPRi (Figure 3C). Similarly, we found a decrease in cell growth rate as we develop the IPTG concentration in T7-CRISPRri system. It has been pointed out that longer cell cycle is mainly caused by a longer time to initiate the DNA replication. Since that there is still normal biochemical synthesis and metabolic reactions occurring in the cell, temporary blocking of genome replication would result in a bigger mass per cell unit. Nucleic acid staining enables us to observe the distributions of nucleoids in single cell under laser scanning confocal microscope. As before, we use poly-adenine as the sgRNA control group. We found a decrease in average nucleo-cytoplasmic ratio when treated with CRISPRri targeted to OriC (Figure 3D).
In order to extend this system to other boxes which are shown to have over-inhibition on cell growth and small dynamic range, we improve the performance of the system by weakening its effect by adding a degradation signal peptide ssrA to dCas9. This largely accelerates the degradation rate of dCas9 and thus weaken its effect. Again, the CRISPRi system provides solid evidence for retention of dCas9 binding ability and degradation-promoting effect of ssrA. As a matter of fact, CRISPRi system with sgRNA targeted to mRFP coding region shows a gentler decrease in fluorescence when dCas9 is fused with ssrA tag, while non-binding dCas9 with or without ssrA has no influence on mRFP expression (Figure 3E). We tested the improved CRISPRri-ssrA system with target site to boxes which are shown to have excessive inhibition on cell growth, and found that the degradation tag make inhibition effect much milder, which allows for a wider adjusting range (Figure 3F).
Traditional methods to inhibit bacterial growth, including antibiotics, self-killing switches and endogenous expression of toxic proteins, all cannot enable affected bacteria to restore its normal functionality and morphology. We examined the reversibility of CRISPRri-ssrA system by repeated addition and elimination of the inducer. After we treat bacteria with high inducer concentration and cultivate for 8 hours, we dilute the bacteria solution and transfer a portion of it into non-inducer medium, and again cultivate for 8 hours. The morphology of almost all cells in the non-inducer medium restore to normal (Figure 3G). This supports good reversibility of the CRISPRri system, and also implies that our bacteria are kept viable (that is, they are still able to form a colony) throughout the process. To further validate this conclusion, again bacteria are transferred into high-inducer-concentration medium and the growth was found to be inhibited. This process can be repeated for over three times. Although we notice the reduction in effect of the system as the times of repetition increases, it is thought be attributed to instability of plasmid-based expression and might be overcome if CRISPRri system is knocked in to the genome.
CRISPRri Enhances Cell Adhesiveness to Surface Through a Harmless Lengthening in Morphology
Measuring the 600-nm light absorbance is a traditional approach for rapid and real-time measurement of bacterial concentration. We found no remarkable decrease in OD600 for bacteria solution treated with CRISPRri (Figure 4A), which is attributed to decrease in cell number but lengthening in cell morphology. As a proof of this concept, we calculated the total covering area of bacteria in microfluidic system, and found the same results (Figure 4B). CRISPRri-implanted E. coli is able to reach a length of 100μm on average and over 500μm at most.
There are numerous ways of turning E. coli morphology into abnormality, and many of them are harmful to cell. E. coli under high stress, like antibiotics environment, is often observed to have abnormal morphology. To distinguish CRISPRri effect on replication initiation from potential dCas9 protein toxicity from the system, we stained the nucleoids of cell and compared the CRISPRri-implanted long cells with cells in abnormal morphology in control group (very rare, but can be found) (Figure 4C). An apparent distinction in nucleo-cytoplasmic ratio can be observed, which suggests that morphology change is not from dCas9 toxicity. Moreover, lengthening of the cell seems to be a persistent process as many of them keep growing even at the endpoint of our recording (Figure 4D). The concept of harmless lengthening is further proved by irrelative protein productivity, which will be stated in next session.
Meanwhile, we found morphology of bacteria to be an adjustable and dose-dependent outputs of the system (Figure 4E). In other word, a given input including sgRNA sequence and arabinose would result in a predictable morphology of the cell, which is almost impossible to achieve in traditional gene-knockout method.
One of the biggest advantages for long-shape-type cell is that its membrane area per cell increases. This allows a stronger interaction between a cell and other interface, like surface of human intestinal tract or genital tract. Longer bacteria might be more stable and easy-to-plant compared to normal-sized cell. As a simplified model, we test the adhesiveness of E. coli with a uniquely designed microfluidic chip (Figure 4F). Flow velocity is adjustable by the injection pump. Long cells and normal cells are mixed in the same culture and are placed into the chip. We found ratio of long cells to normal cells greatly increases as the flow gets faster (Figure 4G).
Irrelative Protein Productivity can be Largely Enhanced in CRISPRri System
As we have mentioned before, CRISPRri system inhibits the cell growth but not in the way of reducing its total biomass (Figure 4A and 4B). This typical feature means its potential industrial application will not be limited. Furthermore, since cell energy consumption in DNA replication is reduced but nutrition uptake rate might be a constant, production of proteins that has no direct relation with DNA replication would increase. This was proved by measuring the production of GFP per cell mass, which is calculated by GFP fluorescence divided by OD600. We found more than three-fold increase in GFP productivity, and a dose-dependent phenomenon suggests it is caused by CRISPRri system (Figure 5).
Quorum Sensing CRISPRri System Enables Spatial-level Growth Control and Ultrasensitive Autoregulation of Growth
We have revealed that multi-input CRISPRri system with rational modification can realize high-adjustable control of cell growth, morphology and protein productivity. However, fully-artificial interference system might not be feasible enough for medical use, as the state of bacteria cannot be supervised all the time to be adjusted immediately. In our expectation, we hope that our system has the ability to adjust in situ. When applied in therapeutic scenario, lengthening in morphology and enhancement of drug production should initiate when cell density reaches a certain threshold, namely, when a large number of cells gather and colonize near the tumor.
Hence, based on the well-characterized growth control system, we constructed a quorum sensing CRISPRri system (qs-CRISPRri) to realize smart regulation of bacterial overall states. A classical quorum sensing system based on cell secretion of AHL synthesized by LuxI, which in turn activates the transcription factor LuxR, is combined with CRISPRri system to realize cell state control which itself senses the population density (Figure 6A).
Instead of combination with CRISPRri, we started with GFP to test the performance of quorum sensing system. A complete quorum sensing system is transformed into the cell, in which LuxI and LuxR is endogenously expressed while GFP expression is activated by LuxR. A positive correlation between GFP production per cell mass (FI/OD600) and population density (OD600) is remarkable (Figure 6B).
A time-scale quorum sensing system can be transformed to spatial-level through a donor/receiver system (Figure 6C). The donor cells, which merely express and release AHL, would activate the GFP expression of receiver cells through AHL diffusion. This is validated on the agar plate, by dropping the donor cells in the center and receiver cells around them with different distances. We found a progressive decrease in fluorescence as the receiver cells locate farther from central AHL donor.
GFP gene is exchanged for dCas9 with companion of constantly-expressed single guide RNA to establish the qs-CRISPRri system. Through the donor-receiver system, we realized spatial-level control of cell growth (Figure 6D). Receiver colonies located nearer to the donor grow much slower than farther ones, which is observed in either dropping or smearing plate.
Furthermore, a complete quorum sensing system-implanted strain is tested with microfluidic system. Distinct from arabinose-induced CRISPRri system, decrease in cell division rate shows an ultrasensitive behavior (Figure 6E). The cells were found to almost stop division and only lengthen its own shape all of a sudden, which we believe to be attributed to accumulation of AHL. Since cell growth is exponential at early stage, AHL concentration would increase rapidly. As AHL concentration reaches a certain threshold, cell genome replication is interrupted by binding of dCas9 and cell cycle extends largely. This enables E. coli to rapidly autoregulates its own growth state, and provides a potential concept of self-tuning medical-used bacteria without external interference.
We hope to construct a self-regulating system, but in our experiment, the expression of LuxI is still artificially induced by IPTG. The only reason why we do not use a constant promoter is that we try to avoid over-inhibition of cell growth during molecular cloning. In fact, it can be replaced by any constant promoters with proper strength to realize full-automatic regulation of growth in real application.
Discussion
We have developed a novel toolbox to bridge overall states of bacteria with synthetic gene circuits that combine the inner mechanism of DNA replication and inherent programmability of CRISPR system. By exploiting the different target sites on genome replication origin, we establish a convenient approach for rapid in-vivo scan for those DNA boxes with potentially important physiological process. This approach is able to extend to other functional region of genome, especially for those sequences with unclear biological essence.
Based on the scanning results, we offer multiple strategies to realize control over complex behaviors and properties of bacteria, including cell cycle, morphology and protein production. Through this multi-input CRISPRri system, bacteria with any wanted growth states can be reached with user-defined adjustment that enables the system to potentially applied in laboratory and industry. Considering the difficulty of too much artificial interference in human body, we modified the system by combining it with quorum sensing system and realized spatial and automatic regulation, which offers possibility in its medical use.
This portability of the system is further proved by joining with other functional system in subsequent demonstration. In this process of developing and optimizing the system, we constructed a measurement platform of precisely recording and analyzing important parameters relative to overall states of bacteria and any of the methods can be supported by another. High robustness and precision of our measurement system has helped us build a quantitative insight into the inner link among different general state parameters of E. coli, which provides a potential characterization and evaluation framework to extend to any artificial bacteria system.