# Model

**Computational Modelling of Biological Resonators**

Optical resonator is an essential component of a working laser system. To create a true biological laser, we are planning to turn cells into optical resonators. For this end, we are planning to cover the cell membrane with two different proteins, reflectin and silicatein to increase the reflectivity of the cell membrane.

In order to get lasing from cell lasers, our proposed optical cavity (the cell itself) should be showing some properties which can support whispering gallery modes (WGM). Since we are interested in structures which have spherical symmetry, we are implementing WGMs into our models.

**
A. Whispering Gallery Modes**

Light can be conned into an optical structure demonstrating a spherical symmetry by almost total internal reflections (TIR) at the interface of the resonator and the outside environment. Since the light undergoes specular reflection in reality, there is some loss of energy which is described by the quality factor of the resonator.

**
Total Internal Reflection**

Imagine a light wave in an environment with a high refractive index travelling through an environment with low refractive index. When the light wave hits the interface, it can be totally reflected if the angle of incidence is higher than the critical angle. This is called total internal reflection. WGM is a phenomenon where electromagnetic waves travel around the resonator by almost TIR since the light wave is only partially reflected. Figure 1 illustrates a WGM where the light wave undergoes TIR.

Figure 1: Light waves can undergo total internal reflection off an interface if the angle of incidence is larger than the critical angle.

In total internal reflection, both the angle of incidence and the refractive index difference is important because the critical angle for TIR is determined by the following equation:

**
Figure 2: A spherical cavity sustaining whispering gallery modes by total
internal reflection on the surfaces.**

**
Constructive and Destructive Interference**

In WGM, light waves need to constructively interfere with each other to form a closed path as shown in Figure 1. Constructive interference happens when waves are in phase with each other, so their amplitudes add up. To create WGM in a spherical cavity, we need constructive interference happening around the cavity.

The criteria for constructive interference is being in phase within light waves. So the optical path length needs to be integer multiple of the wavelength.

**
Figure 3: If two waves are in phase with each other, they constructively
interfere. However if they are out of phase by 90 degree, they
destructively interfere with each other and vanish.**

As described previously, light waves need to be undergoing total internal reflection to create WGM. As can be seen in Figure 10, a full wave should be fitted inside a polygon which can fit into the cavity, for a minimum sized cavity which can support whispering gallery modes. Since we know the refractive indices of the environment and the cavity surface, we can deduce the critical angle for TIR using Snell's law. Our initial guesses revealed that the polygon which fits into the minimum sized cavity should have nine-sides. A previous iGEM team, TU_Delft 2016 has performed these calculations as can be seen in their wiki and obtained some results which were weak to support for their models because it is a very rough estimation for such a complex model environment. We have performed time domain simulations of these models to propose stronger evidence that we can obtain lasing from biological resonators.

**
Figure 4: A full wave should be fitted inside a polygon which can fit into
the cavity, for a minimum sized cavity which can support whispering gallery
modes**

**
Figure 5: Our initial guesstimation for a minimum sized biological cavity.
One can find the minimum diameter of a spherical cavity using this very
basic model (we have found it around 3 micrometers) although a more
thorough model is developed using finite difference time domain method.**

**
B. Solving the Electromagnetic Wave Equation**

WGMs are described by electromagnetic wave equation:

Since the wave equation cannot be solved analytically, we need to turn into numerical methods because we need a time domain method so that we can solve the electromagnetic wave equation every second for every wave in the model. It is not an easy job so we need to use a software which can solve the wave equation in our model every time steps which we describe.

We have chosen the time difference time domain (FDTD) method to solve the wave equation for our model because we are dealing with structures with dimensions of micrometers, and also we need to solve the wave equation in time domain. For our simulations we have used Lumerical FDTD software 2020 R2 version.

We have designed our computational models with three different cell geometry in mind: spherical, toroidal and disk-shaped cells. For each type of geometry we have some cells in our experimental plan with spherical and disk-shaped models for generic biological cells but red blood cells. We have designed the toroidal model for red blood cells specifically.

**
C. Computational Models**

**
Whispering Gallery Modes of A Spherical Cavity**

**
Details of Spherical Model**

Spherical models consist of an inner layer of varying diameters (0.5 µm, 4
µm, 1µm, 2 µm, 3 µm, 4 µm, 5 µm, 6 µm, 7 µm, 8 µm, 9 µm, 10 µm) and a
surface layer of 50 nm thickness. Inner layer of the spherical model is
built from a material with refractive index of 1.37 which represents the
cytoplasm of biological cells [1, 2] while the surface layer is built from
di

erent materials of varying refractive indices (between 1.5-2.0) to find out
the minimum refractive index for each model of varying diameters. Also, we
have integrated Gaussian distributed random particles of varying sizes
(diameters between 10-50 nm) into our models. Refractive index of these
particles is 1.5 which represents the inclusion bodies of silicatein and
reflectin proteins. These random particles also represent other structures
and molecules inside biological cells which have higher reflectivity than
cytoplasm. These random particles can be seen in Figure 2 in yellow.

In Figure 1, we can see the spherical model from the wide-angle view. The biological cell is represented in color turquoise; up-down arrows in the gure represent electric dipole point sources which model the fluorescent molecules; orange lines represent the simulation region whereas yellow lines represent the monitor region.

The time domain model we have designed is run on 2-dimensions (on the x-y plane) with uniaxial anisotropic PML boundaries. We have specified the background material as H2O(water)-Palik to represent biological cells inside water. Our time domain model runs at 320K temperature to make it close to the cell's real environment. We have meshed the FDTD with a maximum of 20 nm steps. As final settings of the FDTD, we have run the simulation more than 2000 fs to allow the simulation to have more time than needed to make it flawless whereas we have run the simulation with “early shutdown" function closed. This functionality allows us on some models to obtain resonance modes although that resonance modes may not be accessible on experimental settings because the power level of the waves are on the order of -10 to -13 arbitrary units when we start with 1 arbitrary unit power source.

We have used a “frequency domain field and power” monitor to find the resonance modes, then obtain the heatmaps on those resonance points using a “frequency domain field profile” monitor. We have applied a full apodization filter at 1000 fs center with 500 fs time-width to filter out the initial and final response of the cavity to the power sources. Position of the point spectrum monitor can be seen in Figure 2 as a small red dot on the right hand of the figure.

**
Figure 1: Wide angle view of the spherical model.**

**
Figure 2: Top view (x-y axis) of the spherical model.**

**
Details of Toroidal Model**

Toroidal models consist of an inner layer of varying radiuses (0.5 µm, 1 µm, 1.5 µm, 2 µm, 2.5 µm, 3 µm, 3.5µm, 4 µm) with 1µm cross section, and a surface layer of 50nm thickness. Inner layer of the toroidal model is built from a material with refractive index of 1.37 which represents the cytoplasm of biological cells [1, 2] while the surface layer is built from different materials of varying refractive indices (between 1.5 and 2.0) to find out the minimum refractive index for each models of varying diameters. Also, we have integrated Gaussian distributed random particles of varying sizes (diameters between 10-50 nm) into our models. Refractive index of these particles is 1.5 which represent structures and molecules inside biological cells which have higher reflectivity than cytoplasm. These random particles can be seen in Figure 4 in yellow.

In Figure 3, we can see the toroidal model from the wide-angle view. The biological cell is represented in color turquoise; up-down arrows in the figure represent electric dipole point sources which model the fluorescent molecules; orange lines represent the simulation region whereas yellow lines represent the monitor region.

The time domain model we have designed is run on 2-dimensions (on the x-y plane) with uniaxial anisotropic PML boundaries. We have specified the background material as H20(water)-Palik to represent biological cells inside water. Our time domain model runs at 320K temperature to make it close to the cell's real environment. We have meshed the FDTD with a maximum of 20 nm steps. As final settings of the FDTD, we have run the simulation more than 2000 fs to allow the simulation to have more time than needed to make it flawless whereas we have run the simulation with “early shutoff” function closed. This functionality allows us on some models to obtain resonance modes although that resonance modes may not be accessible on experimental settings because the power level of the waves are on the order of -10 to -13 arbitrary units when we start with 1 arbitrary unit power source.

We have used a “frequency domain field and power” monitor to find the resonance modes, then obtain the heatmaps on those resonance points using a “frequency domain field profile” monitor. We have applied a full apodization filter at 1000 fs center with 500 fs time-width to filter out the initial and final response of the cavity to the power sources. Position of the point spectrum monitor can be seen in Figure 4 as a small red dot on the right hand of the figure.

**
Figure 3: Wide angle view of the toroidal model.**

**
Figure 4: Top view (x-y axis) of the toroidal model.**

**
Details of Disk-Shaped Model**

Disk-shaped models consist of an inner layer of varying diameters (0.5 µm, 1 µm, 2 µm, 3 µm, 4 µm, 5 µm, 6 µm, 7 µm, 8 µm, 9 µm, 10 µm) with 1 µm cross section, and a surface layer of 50 nm thickness. Inner layer of the toroidal model is built from a material with refractive index of 1.37 which represents the cytoplasm of biological cells [1, 2] while the surface layer is built from different materials of varying refractive indices (between 1.5 and 2.0) to find out the minimum refractive index for each models of varying diameters. Also, we have integrated Gaussian distributed random particles of varying sizes (diameters between 10-50 nm) into our models. Refractive index of these particles is 1.5 which represent structures and molecules inside biological cells which have higher reflectivity than cytoplasm. These random particles can be seen in Figure 6 in yellow.

In Figure 5, we can see the toroidal model from the wide-angle view. The biological cell is represented in color turquoise; up-down arrows in the figure represent electric dipole point sources which model the fluorescent molecules; orange lines represent the simulation region whereas yellow lines represent the monitor region.

The time domain model we have designed is run on 2-dimensions (on the x-y plane) with uniaxial anisotropic PML boundaries. We have specified the background material as H20(water)-Palik to represent biological cells inside water. Our time domain model runs at 320K temperature to make it close to the cell's real environment. We have meshed the FDTD with a maximum of 20 nm steps. As final settings of the FDTD, we have run the simulation more than 2000 fs to allow the simulation to have more time than needed to make it flawless whereas we have run the simulation with “early shutoff” function closed. This functionality allows us on some models to obtain resonance modes although that resonance modes may not be accessible on experimental settings because the power level of the waves are on the order of -10 to -13 arbitrary units when we start with 1 arbitrary unit power source.

We have used a “frequency domain field and power” monitor to find the resonance modes, then obtain the heatmaps on those resonance points using a “frequency domain field profile” monitor. We have applied a full apodization filter at 1000 fs center with 500 fs time-width to filter out the initial and final response of the cavity to the power sources. Position of the point spectrum monitor can be seen in Figure 6 as a small red dot on the right hand of the figure.

**
Figure 5: Wide angle view of the disk-shaped model.**

**
Figure 6: Top view (x-y axis) of the disk-shaped model.**

**
Figure 7: Default window on Lumerical FDTD to create and modify time domain
models. There is a toolbar upper side of the screen where one can add
different structures, power sources, monitors and many more functionalities
into the model. On the left side of the screen, one can see the added
elements into the current model. For example, in the model on the screen,
two spherical structures can be seen in turquoise, power sources can be
seen as two sided arrows in blue, and the simulation region is marked with
orange lines. On the four different screens on the center of the window,
one can see the model in different angles, a wide angle view (upper-right),
on the xy plane (upper-left), on the xz plane (lower left), and on the yz
plane (lower right).**

**
References**

[1] Q. Zhang et al. (2017) Quantitative refractive index distribution of single cell by combining phase-shifting interferometry and AFM imaging. Scientific Reports https://doi.org/10.1038/s41598-017-02797-8

[2] Z. A. Steelman (2017) Is the nuclear refractive index lower than cytoplasm? Validation of phase measurements and implications for light scattering technologies. Journal of Biophotonics https://doi.org/10.1002/jbio.201600314

**
Biological Modelling**

Reference Literature: A computational Model Approach for calculating Fluorescence Intensity in Arbitrary Units according to Gel Densitometry, Flow Cytometry and Microscopy [Enrique Balleza et al., 2018, Nature Methods]

**
Why Do We Use This Model?**

We want to observe the lasing changes during the growth of bacteria. So we decided to use this model to calculate the fluorescent intensity in arbitrary units during growth of bacteria and maturation of fluorescent proteins to compare the fluorescence intensity of different fluorophores. We also thought that future iGEM teams can use this model to calculate the fluorescence intensity(relatively) in growing cells.

**
1) Two regulatory parameters as inputs to the function, quantum yield and
Molar extinction coefficient**

First, we used 2 regulatory factor parameters for fluorescent proteins with different ex / em spectra. These were the quantum yield for expression and the molar extinction coefficient for emission. It had to be an experimental setup to obtain these two data, but we couldn't do this because we are an online team. However, we have already decided to use a ready-made dataset prepared according to certain concentration, temperature and ph values.

When we get the experimental results of gel densitometry which is just a Picture. We need to get it analyzed by a software tool specialized on this. After that by using a standard curve the next step is to estimate the total protein in every well. The molar concentration was calculated by assuming protein weights in kD’s. After that by using Beer’s law we can find the molar extinction coefficient.

When we move to the calculation of quantum yield, a function exists to calculate quantum yields by using parameters integral photon flux, absorption factor at the excitation wavelength, and the refractive index of dye reference and the sample. To sum up the general relative quantum yield formula is:

In the formula above F denotes the integral photon flux, f is the absorption factor at the excitation wavelength which can be calculated as (1-exp(-abs(λex)), and last n denotes the refractive index. Since it is relative, x is used to denote the sample and st denotes the dye reference.

We used the equation below to calculate the correction factor of Excitation. fex shows how many times more FP absorbs the light with a broader excitation window.

Note: Absorption max is:

Excitation Factor calculation:

Ex1 and Ex2=limits of excitation filter

Abs=absorption spectrum

λmax=wavelength of maximum absorption

Note: the numerator is just a part of the light absorbed between the filters because:

T=Transmittance

The emission factor was equal to the fraction of the emission spectrum covered by the emission filter.

**
2) Relative Protein Expression**

The method for calculating the protein expression in this study is the BSA method.

In the BSA method they used BSA concentrations of 1.0, 0.666, 0.500, 0.333, 0.250, 0.167, 0.125, 0.083, 0.062, 0.041, 0.031 ug/ul.

However, the authors didn’t share the calculations they derived from the BSA but they only put the image of BSA SDS-Page with unknown FPs. So we thought that by referencing these BSA concentrations and calculating their densities from GelAnalyzer we can estimate the concentration of their FPs by using their relative expression rates even if it is not so accurate. However, if iGEM teams in the future can do this assay in their labs (we didn’t have access to the lab because of COVID19 to do the assays) they can calculate the expression levels and use this method to calculate the estimation of fluorescence intensity by using its density!

Relative Protein Expression can be derived from Gel Densitometry Method as well. Two orders made to reach the relative solution, inside-out Ordering and regular ordering to get rid of the systematic gel distortions, then the values obtained by using these two methods will be summarized by median operation. First the protein amounts are calculated on three reference bands 100kDa and 40kDa and FP reference band (the last band has chosen with respect to the each proteins’ molar weight) by using a software. (GelAnalyzer 2013a in original work).

First, you need to do the gel density calculations on no fluorescent protein bacteria extract proteins in order to understand the background proteins in other FP band ranges. FP band range(25 kDa range)

Reference Estimation Parameter for 100 kDa = (FP Band Protein Amount) / ( 100 kDa Band Protein Amount)

Reference Estimation Parameter for 40 kDa = (FP Band Protein Amount) / ( 40 kDa Band Protein Amount)

You can calculate the background amount of protein by using the parameters found above. The operation is just a multiplication of the protein amount on each reference band with the parameter that have been found above for each band.

Background 100kDa estimation = (Reference Estimation parameter for 100 kDa)*(Reference band 100kDa Protein Amount)

Background 40kDa estimation = (Reference Estimation parameter for 40 kDa)*(Reference band 40kDa Protein Amount)

So In order to find the corrected value in each bands we need to get rid of the background values that we have obtained from the formulas above

FP100 Background Corrected = FP Band Protein Amount - Background 100 kDa estimation

FP40 Background Corrected = FP Band Protein Amount - Background 40 kDa estimation

To make relative protein expression calculations, we need to take 1 FP band as a reference. To achieve that, we need to normalize other FPs bands based on 1 reference FP’s band values.

100 Normalized = (Reference Band 100 kDa Protein Amount) / (Reference Protein Selected in order to Normalize)

40 Normalized = (Reference Band 40 kDa Protein Amount) / (Reference Protein Selected in order to Normalize)

*** In the last two calculations the reference protein must be the same

FP Norm 100 = FP100 Background Corrected/ 100 Normalized

FP Norm 40 = FP40 Background Corrected/ 40 Normalized

%FP100 = FP Norm 100 / Normalization Protein (third)

%FP40 = FP Norm 40 / Normalization Protein (third)

When we finish this process, we will obtain 4 different values, two for inside-out ordering, two for regular ordering, they exactly refer to the same information but as we have mentioned this process’ main aim is to get rid of the gel distortion errors, in other words make it more accurate. When we get the median of these 4 values we will get the exact normalized protein expression that we can use for our general fluorescence formula.

**
3) In-vitro Brightness**

This data is a constant and unique data for each fluorescent protein, although it can be validated experimentally, we will use the reference data since we cannot establish an experimental system.

In-Vitro Brightness = QY(quantum yield) x molar extinction coefficient

molar extinction coefficient(ε)= A/c x l

A=Absorbance

c= Molar concentration

l= optical pathlength

Relative Quantum Yield:

In the formula above F denotes the integral photon flux, f is the absorption factor at the excitation wavelength which can be calculated as (1-exp(-abs(λex)), and last n denotes the refractive index. Since it is relative, x is used to denote the sample and st denotes the dye reference.

**
4) Maturation Factor**

Fluorescent proteins must mature in order to radiate. Calculations made without considering this factor will contain errors. So to overcome this problem, we need to add a maturation parameter to our equation. For the formation of this parameter, the maturation factor was created with 1 / (1 + t50)/tGrowth by using the maturation time data, t50, that is, the time when 50 percent of fluorescent proteins are mature. tGrowth is the doubling time of bacteria

Calculation of Fluorescence Intensity:

Excitation Correction Factor X Emission Correction Factor X Normalized Protein Expression(In density) X Maturation Factor= Fluorescence Intensity in Growing cells

Implementation of These Results into Our Experiment Design

With the understanding gained from the article, we have decided to include maturation time as a parameter for our pumping cycles during lasing experiments. In order to achieve the most intense radiation, we must determine the time periods in which the proteins have the most intense maturation. To achieve this, maturation times were noted for each protein, leaving a tolerance range instead of pumping a laser at a single time. These gaps for each protein will be detailed below.

**
1. Gamillus (On)**

**
Figure 1: The mean is chosen the denoted maturation time for Gamillus
protein. The gaussian distribution has prepared with the standard deviation
0.25 and planned pumping time tolerance has chosen as +0.25/-0.25 range.**

**
2. GFPmut3 **

**
Figure 2: The mean is chosen the denoted maturation time for GFPmut3
protein. The gaussian distribution has prepared with the standard deviation
0.25 and planned pumping time tolerance has chosen as +0.25/-0.25 range.
**

**
3. mScarlet-I **

**
Figure 3: The mean is chosen the denoted maturation time for GFPmut3
protein. The gaussian distribution has prepared with the standard deviation
0.25 and planned pumping time tolerance has chosen as +0.20/-0.20 range.
**

**
4. mVenus **

**
Figure 4: The mean is chosen the denoted maturation time for GFPmut3
protein. The gaussian distribution has prepared with the standard deviation
0.25 and planned pumping time tolerance has chosen as +0.20/-0.20 range.
**

**
5. SuperfolderGFP **

**
Figure 5: The mean is chosen the denoted maturation time for GFPmut3
protein. The gaussian distribution has prepared with the standard deviation
0.25 and planned pumping time tolerance has chosen as +0.20/-0.20 range.
**

This protocol is taken from the reference literature [Balleza et al. 2018].

**
Protocol Title**

Measurement of maturation kinetics in living Escherichia coli cells

**
Authors:**

Enrique Balleza, J. Mark Kim & Philippe Cluzel*

Department of Molecular and Cellular Biology, Harvard University, Cambridge, Massachusetts, USA

**
Abstract**

This protocol describes how to measure the maturation kinetics of fluorescent proteins (FPs) expressed in Escherichia coli using microfluidics and microscopy. This protocol is a guide to produce maturation kinetics curves with high temporal resolution and low experimental noise.

**
Introduction**

While the maturation kinetics of many FPs has been characterized by in vitro refolding assays, much less is known about the maturation process in vivo. We describe how to measure maturation kinetics in live E. coli cells using chloramphenicol. This protein synthesis inhibitor has been widely used to give a broad estimate of maturation time from batch culture measurements. The usage of microfluidics and a microscope greatly increases the temporal resolution and lowers the experimental noise. The refined measurements unveil a wide range of maturation kinetics associated with different FPs.

To measure fast maturing FPs, it is essential to control the cell environment and to track at relatively high temporal resolution the fluorescence signal in living cells. However, to measure very slow maturing FPs, this protocol is not necessary and a batch culture experiment should suffice. The extension of this protocol to other organisms is straightforward as long as there exists a powerful protein synthesis inhibitor.

**
Reagents and Equipment**

• Inverted microscope with phase contrast and epifluorescence capabilities.

• Software for microscopy automation, e.g. Micro-Manager or MetaMorph.

• Software or hardware based autofocus.

• A documentation camera with a quantum yield >50%. The optical system (camera + objective lens) should have a projected size of 0.2 um/pixel or smaller.

• A stable source of excitation light (laser or LED lamp).

• Microfluidics device that allows for the balance growth of E. coli and quick buffer exchange, e.g. the “single-cell chemostat”[1] or the “mother machine” [2].

• An incubation chamber or an objective heater to control sample temperature.

• Two peristaltic pumps.

• M9 rich media: M9 salts, casamino acids 0.1%, glucose 0.5%, thiamine 1ug/ml, MgSO4 2mM, CaCl2 0.1 mM; adjust pH to 7.1 with NaOH.

• Chloramphenicol 125 mg/ml.

• An Escherichia coli strain that expresses constitutively the FP of interest.

• Cell tracking software, e.g. the Schnitzcell program3, software developed for the “mother machine” [2, 4], or software developed for the single-cell chemostat (see associated publication).

**
Procedure**

**
A. Preparation of the microscopy setup**

1. Turn on the microscopy body and the temperature controlling device. Let the setup equilibrate to the desired temperature for about 3 hours before starting data acquisition.

2. Prepare M9 rich media with and without chloramphenicol. Use a final chloramphenicol concentration in the range 100-140 ug/ml.

3. Load cells inside the microfluidics device and incubate inside the microscope with a steady flow of M9 rich media. Incubate until cells reach balanced growth (the waiting time is about 7 cell divisions at 37C).

**
B. Data acquisition and buffer exchange**

1. Start time-lapse acquisition with a temporal resolution of 1 minute and a total acquisition time greater than 4hrs.

2. After 30 minutes, stop the M9 rich media flow from the first pump and immediately start the flow of M9 rich media + chloramphenicol from the second pump.

**
C. Data processing**

1. Using a cell tracking program, select ~100 cells that remain in field of view (FOV) from the start to the end of the experiment. Avoid selecting cells that develop obvious cell-wall defects towards the end of the experiment.

2. Quantify the length, l(t), and the total fluorescence, f(t), of individual cells as a function of time.

3. For background fluorescence correction, select a region without cells and quantify the background fluorescence signal, b(t), as a function of time.

4. If throughout the experiment the background fluorescence is not constant, subtract its value from the total fluorescence value of single cells, i.e. calculate fcorrected(t) = f(t) - b(t).

5. Using the fluorescence data of the complete population as a reference, identify cells that abnormally loose fluorescence and discard them from the analysis.

**
D. Estimation of maturation time**

1. To obtain a mean single-cell fluorescence curve, F(t), add the fluorescence of individual cells, fcorrected(t), and divide by the sample size, i.e. number of tracked cells. Similarly, obtain a mean single-cell length curve, L(t).

2. In a semi-log plot (or regular plot), use the mean length curve, L(t), to identify the moment at which the curve deviates from a straight line (or an exponential curve). This is the moment at which chloramphenicol starts acting.

3. To obtain the maturation kinetics curve (or fraction of immature protein), apply to the mean fluorescence, F(t), the formula 1-(F(t)-Ft=0)/(F∞-Ft=0) where Ft=0 is the mean fluorescence value at the moment of chloramphenicol arrival and F∞ is the saturation value of the mean fluorescence.

4. To calculate t50, find the moment at which the fraction of immature protein reaches a value of 0.5. Find t90 in an analogous way.

**
Troubleshooting**

A.1. For the maturation time estimation, it is crucial to determine as accurately as possible the sample temperature. We recommend to run a mock experiment with a thermocouple inside the microfluidics device and to contrast the target temperature with the actual thermocouple reading at different flow rates.

A.2. It is also crucial to use chloramphenicol at the recommended concentration. If the concentration is too low, cells will keep producing fluorescent protein; if it is too high, almost all cells will present cell-wall damage after the first hour of treatment.

B.2 and D.2. If cell length does not increase exponentially or if cell growth rate is too slow, check for unintended pretreatment with chloramphenicol. Chloramphenicol is a potent drug and even low concentrations have a major impact in growth rate. Run a mock experiment with no chloramphenicol to discard other sources not related to the drug that may impact growth rate. Also, check for unintended diffusion of the M9 rich media + chloramphenicol into the M9 rich media anywhere in the tubing that goes from the pumps to the microfluidics device.

C.5 Assuming the excitation intensity is low, it is very probable that the loss of fluorescence reflects loss of FPs from the cytoplasm due to cell-wall damage. Another cause for the loss of fluorescence is photobleaching affecting only cells within a sector of the FOV. Excitation intensity should be homogeneous (+10%). Check excitation homogeneity by analyzing a snapshot of an autofluorescent slide (Chroma; 92001).

D.3. If the mean fluorescence curve does not reach a steady saturation but instead decays then excitation intensity is too high. Reduce excitation intensity or exposure time to avoid photobleaching. If the FP does not exhibit complex photobleaching (non-exponential fluorescence decay), another possibility is to correct the mean fluorescence curve following the method described in the associated publication.

**
Time Taken**

A typical maturation time experiment takes two days: one day to set up the experiment and perform the data acquisition process and another day to analyze the acquired data.

**
References**

[1] Moffitt, J. R., Lee, J. B. & Cluzel, P. The single-cell chemostat: an agarose-based, microfluidic device for high-throughput, single-cell studies of bacteria and bacterial communities. Lab Chip 12, 1487-1494, doi:10.1039/c2lc00009a (2012).

[2] Wang, P. et al. Robust growth of Escherichia coli. Curr Biol 20, 1099-1103, doi:10.1016/j.cub.2010.04.045 (2010).

[3] Rosenfeld, N., Young, J. W., Alon, U., Swain, P. S. & Elowitz, M. B. Gene regulation at the single-cell level. Science 307, 1962-1965, doi:10.1126/science.1106914 (2005).

[4] Norman, T. M., Lord, N. D., Paulsson, J. & Losick, R. Memory and modularity in cell-fate decision making. Nature 503, 481-486, doi:10.1038/nature12804 (2013).