Multi-input Logic Characterization

Here we show an experimental result of an example of what we call our internal processing module, which is essentially a 2 input chemical AND gate. Taking the individual functional transcription factor systems characterized above, we integrated them into an AND gate with two inputs: IPTG and doxycycline levels. The system integrates the UAS/GV16, rtTA3/TRE, and lacI/lacO systems described above. As discussed earlier, IPTG can bind to lacI and relieve its repression of lacO. Thus in this particular circuit, IPTG relieves inhibition of rtTA3 expression. Then, only in the presence of doxycycline can rtTA3 activate expression of the TRE controlled reporter. Therefore, we require both IPTG and dox to achieve expression of our final Delta-mCherry reporter.

The data show the desired 2 input AND behavior. We see maximum Delta-mCherry expression at the highest levels of both dox and IPTG. We also observe significant activation at high dox levels even in the absence of IPTG. This suggests some leakiness in the lacI inhibition, allowing for some rtTA3 to be produced and activated by dox. This behavior can be addressed using a different inhibitor such as lacI-KRAB which exhibits a much stronger repression.

2-input AND Gate: Gal4-VP16→LacI¬rtTA3

The above graph displays flow cytometry measuring mCherry fluorescence of transfected Hek293 cells in the presence of different IPTG and doxycycline concentrations. We observe highest reporter levels at max IPTG and dox concentrations. Meanwhile, we see decreasing levels of reporter expression as we decrease either dox or IPTG levels. Note the high background, especially at high dox levels due to leakiness of lacI inhibition.

Patterning Modeling Results

All simulations were done using our modeling framework, mcell. For more information on mcell, please look in the Tools section. Some sample runs can be downloaded from here: here.

Most of our modeling was done on the following circuit:

MODERATE is a promoter that has moderate activity under the experimental conditions; for instance, TRE without any rtTA3 in the system would work. All cells contain this circuit; the values of all proteins are 0 at start.

We followed the model of Delta-Notch interaction proposed by Elowitz[1]. In it, Delta and Notch in the same cell cis-inhibit at a rate proportional to their amounts; Delta in one cell can bind to Notch in another cell (at a rate proportional to the available Notch and Delta), destroying both proteins and releasing the Notch Intracellular Domain (NICD), which then goes on bind to a promoter and stimulated production of genes.

Applying this model, along with standard mass action to model protein binding events and Hill equations to model promoter stimulation (along with the standard set of simplifying assumptions that go along with these methods), we arrive at the following series of differential equations:

d[NCis]/dt = n_prod - k_cis*[NCis]*[DCis] - k_trans*[NCis]*[DTrans]
d[DCis]/dt = d_min + d_prod/(1+[LacI]*b_L) - k_cis*[NCis]*[DCis] - k_trans*[NTrans]*[DCis]
d[NICD]/dt = k_trans*[NCis]*[DTrans]
d[Reporter]/dt = d[NCAD]/dt = d[LacI]/dt = V_max*(Gal4^h)/(k_d+Gal4^h)

DTrans is the amount of Delta in neighboring cells touching the current cell; likewise, NTrans is the amount of Notch in neighboring cells touching the current cell. We assumed that Delta and Notch are all located on the cell surface and are evenly distributed. NCis is the amount of Notch inside the given cell; DCis is the amount of Delta inside the given cell. The values for V_max, h, and k_d were taken from [1]; for the simulations shown on this page, n_prod = 5, d_min = 1, d_prod = 9 (so that the Delta production rate varies between 1 and 10); k_cis = 1, k_trans = 0.04, and b_L = 10.

Bimodality in HEK cells

Running this circuit in HEK cells (modeled by setting the energy of contact between cells in CompuCell3D to -10.0 and initializing the cells with no gaps) produces stunning bimodality.

Video

Analysis

What happens here is that, initially, all the cells develop a large amount of Delta. However, some Notch still remains in the cells, and is strongly activated; Delta production rates begin to drop because of the LacI. Some cells gain significant amounts of Notch before others, due to the random movements of the cells. Once a cell gains a significant amount of Notch, it's Delta level drops off sharply, and it ceases to activate cells in its vincinity. Thus, once a cell is fully "activated", other cells around it are much less likely to be. The stable state of such dynamics is a mosaic pattern, in which single cells are activated and are surrounded by a sea of inactivated cells.

But what if they are not sticky?

There is a very large discrepancy between the behavior of the circuit in HEK cells in CHO cells, which are modeled with no adhesion and are initialized with gaps in between them. In CHO cells, the following ensues:

Movie

Analysis

Bimodality almost completely drops off in this model. This is clear from the video; looking at the histograms, we can see that only an extremely small hump develops at the end in place of where the hump for the activated mode should be. The cells drift around randomly, most of the far away from others and thus inactivated; sometimes they bump in to one another, and a cell or two begins the road to activation, but they never stay together for long enough for the dominance of the activated cell to be established and for bimodality to be enforced.

Adding in NCAD

Now, we add in the NCad. Using the AdhesionFlex plugin, we set the adhesion strength of NCad to 1 (modifying this within reasonable bounds does not really change things much) and set the amount of NCad in CompuCell3D to equal the amount of Reporter in our BionetSolver circuit. This created the following results:

Movie

Note: this is a long video; the behavior of the cells never really settles down into a stable state.

Analysis

This circuit never really settles down into any clear behavior. Looking at the histograms, we can see that there is a fairly large gap between most "activated" and "non-activated" cells, but the "activated" cells have a much wider range of Reporter values and the range drifts over time. Looking at the video, one can see certain cells that end up always being unactivated, certain cells that are always activated, and certain cells with a moderate Reporter level that oscillates in a complicated fashion. What seems to be happening is that as a cell randomly becomes activated, it causes other cells to stick to it; this creates enough cells in close proximity to one another to cause more cells to activate, which weakens the aggregate structure, causing the cells to start drifting apart. Analysis of this model is ongoing. We tried clustering the cells, but standard algorithms did not do a very good job in clustering cells into groups that share the same behavior. This may be because cells seem to change behaviors -- a cell that looks like it is always off can, in the middle of the simulation, become an always on cell.

---

[1] Sprinzak, et al. Cis-interactions between Notch and Delta generate mutually exclusive signalling states. Nature 465:86-90

Patterning Experimental Results

One of the key modules in our self-organizing cell project is intercellular communication. We achieved this through the use of the Notch/Delta signalling system. As described under the Project page, Notch/Delta is a cell-cell contact based signalling system with Delta acting as the sending ligand and Notch as the receiver. The Notch we used was a fusion of the Notch extracellular domain with an intracellular Gal4 domain. Upon Delta binding, the intracellular domain is cleaved and Gal4 can migrate to the nucleus and activate UAS. In the following experiments, we tested out this interaction ----

TO-DMC (Sender) cells

CHO T-REx stably integrated with: pcDNA5/TO‐Deltamcherry

Purpose: These cells normally inhibit expression of Delta-mCherry. However, doxycyline can relieve this repression and start expression of Delta. These cells can only send using their Delta ligand, so we termed them "senders".

hN1G4esn‐No‐Delta (Receiver Type)

CHO T-REx stably integrated with: CMV-Cerulean, UAS-H2B-Citrine, pcDNA3‐hNECD‐Gal4esn

Purpose: These cells express Notch constitutively and have a Histone-Citrine fusion under UAS control as the readout. We termed these cells as receivers.

hN1G4esn (Cis-inhibitor Type)

CHO T-REx stably integrated with: CMV-Cerulean, UAS-H2B-Citrine, pcDNA3‐hNECD‐Gal4esn, pcDNA5/TO‐Deltamcherry

Purpose: These cells express Notch constitutively and inhibit Delta expression in the absence of doxycycline. In this state, they are effectively receiver type. However, as dox levels increase, we see more and more Delta expression and due to cis-inhibition, we see a switch to sender state. Because these cells allow us to observe and modulate cis-inhibition, we termed them "cis-inhibitor type".

Explanation:

We decided to test our Notch/Delta constructs by performing a co-culture experiment. We transfected our Hef1a:Delta-mCherrry construct into Hek293 cells. We mixed these Hek sender cells with receiver CHO cells in the specified ratios. The results of flow cytometry are displayed above.

From the microscopy images above, we can see clear expression of Delta-mCherry by our HEK cells. In addition, we can see activation of Notch on adjacent CHO receiver cells as seen by the presence of Citrine (yellow signal). This spatial correlation is strong evidence showing Notch activation by Delta.

The above bar chart displays average FITC signal of our cells as determined by flow cytometry. We can see an increase in fluorescence in the test samples compared to the controls. This suggests that our Delta construct is also capable of signaling to Notch.

Mammalian Cell Adhesion Results

Cadherins are a medically important calcium-dependent surface protein that can bind to same-type cadherins on other cells. The existence of different types of cadherins allows for the development of highly organized tissues, as the differential expression of different types of cadherins in various cells drives the development of complex three-dimensional structures.

We incorporated cadherins into our existing characterized signaling work, with the goal of achieving control of cadherin expression in mammalian cells. By integrating cadherins into our pattern-formation mechanism, we will be able to create patterns based on differential adhesion strengths between cells.

Constitutive expression of NCad-eGFP.

The construct shown above uses a Ncadherin eGFP fusion protein that allows us to visualize localization of Ncadherin. Using confocal microscopy, we were able to observe membrane localization of our Ncadherin protein as expected. Additionally, we can also observe very bright signal at the cell interfaces, showing clustering of Ncad at these interfaces and thus adhesion.

Constitutive expression of eGFP.

Meanwhile, in our control, we transfected a constitutive GFP construct. We can see diffuse cytoplasmic localization versus membrane localization.

This series of experiments show that our Ncad construct does display expected behavior and provides a promising start to incorporating adhesion into our designs. We observe membrane localization and sticking which is what we were looking for.

Attributions

Our instructors were very helpful not only in giving feedback on our designs, cloning strategies, and data, but also in training us for lab work. The training for tissue culture work was conducted by Linda Stockdale in the Griffiths lab. Gibson assembly techniques and FACS training from Deepak Mishra, one of our instructors. Jonathan Babb, another one of our instructors helped us with setting up the robot.

Other than the initial training, all programming, DNA, and transfection work was done by our undergraduate team.

Special credit belongs to Semon Rezchikov for simulations and modeling, Jenny Cheng for animations, and Tiffany Huang for wiki design.

Special thanks to Michael Elowitz for generously providing not only the CMV:hN1ECD-Gal4esn, CMV-TO:Delta-mCherry, and UAS:H2B-Citrine constructs, but also the stably integrated cell lines. We could not have implemented the Notch-Delta system without these valuable contributions.