OD/F Device

On this page we present the technical details of our OD/F device. You can skip to specific chapters by clicking on the panels below:

Building the OD/F device has been an interesting task. On the one hand, this device has been developed mainly by the IT division of our team. On the other hand, we got assistance from biologists suffering from color-blindness, yet eager to help selecting the best color filters for the LEDs.

General Considerations

The measuring principle and guidelines for this project have already been presented in the project section. Here, details about selecting filters, code and a construction manual are presented.

Cuvette Holder

The essential part of this device is the cuvette holder which has also been the most tricky thing to design. In short, we had to overcome a dilemma created by the need for an optimal height for the sensor:

• A too low sensor position bears problems with sedimentation as well as light diffraction from the bottom of the cuvette.
• The sensor has to be as close as possible to the bottom so that enough light shines through for the fluorescence measurement.

As a compromise, we place the sensor at a height of 0.75 cm. It is comparable to one of the standard heights (0.2 cm, 0.8 cm, 1.2 cm) of OD meters. It is important to note that our device works with filling volumens of just 1 mL, which in fact comes close to reality in the lab.

The final cuvette holder design is rendered from a stl-file shown below:

Cuvette Holder devloped for our OD/F device.

Light Filters

Finding the right and optimal filters is a tough challenge. The main goal throughout our project has been to choose easily available parts which are also inexpensive. Thus choosing Schott glasses as filters could not be considered. Alternatively, filters used for illumination of theaters were found to be an ideal solution.

We tested serveral filter and the optimal configuration of filter is shown in the below table.

Mode Fluorescence Protein Filter Name Filter Peak Excitation Peak Emission
Fluorescence GFPmut3b Twickenham Green 501nm 511nm
Optical Density -- Fire 600nm 600nm

The fluorescence protein GFPmut3b has a peak excitation at 501 nm and a peak emission at 511 nm - too close together for our low-cost filters to block the excitation light but transmit the emitted light. Thus, we chose to excite at around 485 nm reduce false positive results below 500 nm. However, no adequate filter for these settings could be found. Eventually, using the dark greenish Twickenham Green filter only little amounts of light shorter than 500 nm gets through, reducing any bias from excitation illumination significantly. Unfortunately, the transmission rate of this filter is quite bad, 20% only, for the target emission wavelength of 511 nm.

Linearity

It is crucial that the selected hardware is mapping reality into the digital world of our $\mu$-Controller. In order to sense reality our setup uses a light to frequency sensor, TSL235R-LF. The light to frequency sensor resembles the most to a photo transistor and thus is less sensible to temperature than a light dependant resistor. Additionally counting a frequency using interrupts seems to be easier and more accurate than using the analog to digital converter.

Using a dilution series of purified iLOV we could determine the characteristic curve for the light sensor. Finally we can conclude that the sensor is linear as expected and shown in the datasheet.

OD Device

We are measuring optical density using the presented cuvette holder. Particularly for optical density measurement the amount of light shining through the sample is crucial. If there is too few light, there will be not enough light registered at the sensor, and the resolution of the measurement shrinks. This should be prevented. The chosen light to frequency sensor is reported to be very sensitive on the amount of light shining on it. There are reports of the sensor breaking when put into sunlight on a nice day, and not being sensitive at both high light or low light conditions.

Commercial systems usually use a laser beam to shine through the sample. For reasons of available components, we omitted this in our plans. Finding a suitable orange laser is not easy. Instead we chose an LED, which unfortunately scatters light, both in the spectrum as well as in all directions. Using a filter we reduce the effect of scattering. Finally the Fire 019 filters solves our problems, but is far away from ideal.

Evaluation

From Transmittance to True Optical Density

At very low levels, uncorrected photometric determinations of cell densities show a decreasing proportionaility to actual cell density.

This can also be observed using our OD measurement device.

In general, photometric determination of bacterial concentrations depends primarily on light scattering, rather than light absorption. Therefore. often not absorption is measured, but transmittance. For this, the relationship between optical density (OD) and transmitted light $\frac{I_0}{I}$ exists as:

$$OD = \frac{I_0}{I} = \kappa \cdot c$$

However, this equation is linear only in a certain range. While one can tackle this non-linearity by using dilutions of the culture, correcting the error systematically is another way to overcome this limitation.

For our OD device we needed to correlate the transmittance measured by our sensor to an optical density anyway. Our team members from the deterministic sciences emphasized on the correction method, which was conducted according to Lawrence and Maier [1]:

• The relative density ($RD$) of each sample in a dilution series is calculated using $\frac{min(dilution)}{dilution}$.
• The uncorrected optical density is derived from the transmission T [%]: $OD = 2 - \log T$
• Finally, the unit optical density is calculated as $\frac{OD}{TD}$.
• The average of the stable unit optical densities is used to calculate the true optical density $OD_{unit} \cdot RD$.

This way, the correlation between transmission and true optical density can be computed. The derived function allows the conversion from transmission to optical density on our device and therefore calibrates our device.

Lawrence and Maier could show that correcting transmittance this way, the corrected optical density shows a linear relationship between true optical density and dry weight in cell suspensions.

In our experiments, we find in accordance to [1] that the correction majorly depends on the technical equipment used, especially the LED, sensor and cuvettes. While this at first sight looks disappointing, it is also expected: Transmittance is the fraction of light not absorbed by some medium relative to the cell-free and clear medium. However, the transmittance is not only dependent on the amount of cells in the way of the light's beam, but also how much light shines through the cuvette in which fashion, and in which fraction is received by the sensor in which angles.

Using the above formula we performed this experiment for Pseudomonas putida and Saccharomyces cerevisiae.

Experiments

We performed several experiments during the development of the OD/F device. Finally we can relate the measured transmittance to the true Optical Density, and further, we can relate that true OD to the one of the photospectrometer in our lab. By doing to we can calibrate our device to meaningful values.

We have done this according to the previous section for Pseudomonas putida and Saccharomyces cerevisiae.

The final function for calculating the OD from the transmission calculated by our device can be calculated as

$$OD = f(T) \circ g(device)$$

where $f$ transforms transforms transmittance to true optical density for our device, and $g$ transforms true optical density of our device into the true optical density of the photospectrometer. This way our device is calibrated according to the photospectrometer.

<center>

Saccharomyces cerevisiae

From these plots it can first be seen that our device delivers robust and reproducible results for both procaryotes and eucaryotes. Also the function from transmittance to true od can be expressed as a lower polynomial function, making its calculation easily possible on a low-end device like a microcontroller.

Most encouraging is that the function for relating the true OD of our device to the photospectrometer is, as seen by the regression coefficient, close together for both P. putida and S. cerivisae. In fact, 3.416 and 3.461 are such close together, that the minor deviation could be just measuring inaccuracy. Therefore we fix the regression coefficient for converting true OD of our device to true OD of the photospectrometer to 3.432 .

It is interesting to note, that also the function $f$ for the conversion of transmittance to true optical density fit nicely together, as can be seen in the following figure.

By this evaluation we have shown that our self-build optical density measurement device can compete with commercial systems, and moreover, is easy to calibrate by just calculating the true optical density. Therefore we present a device which measures accurately and is made of easily available parts at a low cost.

F device

Similarly to the OD measurement, the fluorescence is measured using the same cuvette holder. In fact, if one does not build a combined device, the only thing one is supposed to change is the cuvette holder. However, as for optical density measurement, a filter needs to be placed between led, sample and the light sensor. Selecting the filter has been troublesome. Either the tried filters had a good transmittance but did not screen for the correct wavelength, or they screened for the correct wavelength but showed bad transmittance. Finally we chose the [ Twickenham green] filter with bad transmittance, and raised the sampling interval from 1 s to 4 s to allow a distinct signal. This is by far not optimal, but delivers stable and reliable results.

For fluorescence measurement we luckily are not that much relying on the optical density of the cell culture to measure (if the sample contains cells at all). We compared the values of our device against the platereader.

Evaluation

Figure 1 shows the absolute measurements for both the platereader and our OD/F device. The abrupt jump at 50% concentration can be explained by a second dilution step and is prevalent in both devices. It can be seen that the platereader show a much higher difference between the GFP and non-GFP cell culture at a higher standard deviation. Another interesting metric is the difference between the GFP and non-GFP, which can be seen as the normalized fluorescence measure.

If one compares the results there, as in Figure 2, interesting observations can be made. First, both platereader and OD/F device show very similar results. The regression curves differ only in a linear factor. Most interestingly the general fit of the OD/F device to a linear function seems to be better than the platereader. Overall the linearity which has been observed earlier (in testing the general setup) could be verified. Therefore our do-it-yourself OD/F device can be used to determine fluorescence. At higher concentrations our OD/F device struggles regarding accuracy. However, this is also true for the platereader, but at a lower rate.

DIY: How to Build Your Own Device

Technical Components

While the casing and the cuvette holder are custom made, most of the parts are pre-made and only need to be bought. The previous section lists all needed parts. To get all these parts for creating your own OD/F device is easy by using the internet. A lot of companies all over the world are specialized in selling electronically equipment not only in the internet but also in local shops. However potential customers have a market access connected to the parts of building.

Please find our custom parts for download below. Despite being custom parts, these are quite inenxpensive - so feel free to give our OD/F device a test :) !

You will need a special library for the display, which can not be uploaded for legal reasons.

All needed components their quantities and prices for creating your own OD/F device

References

Lawrence, J. V., & Maier, S. (1977). Correction for the inherent error in optical density readings. Applied and Environmental Microbiology, 33(2), 482–484. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=170707&tool=pmcentrez&rendertype=abstract.