Contents

Assessing the accuracy of microsphere calibration of cell count estimates from OD₆₀₀ measurements

Currently, the most common method for quantitatively assessing cell count in a sample is measurement of optical density at 600 nm (OD₆₀₀), with the corresponding background measurement for a blank sample subtracted from it ^[1]. Optical density itself is a measurement of the attenuation of light through a sample, derived both from absorption and scattering of light by the sample ^[2]. However, such methods are limited by the contributions of components in culture samples, other than live cells. Moreover, estimating cell count via measurement of OD₆₀₀, rather than direct measurement of cell count itself, introduces errors associated with differences in detector setup between labs, further reducing the reproducibility of cell count estimates between labs. Consequently, to compare OD₆₀₀ measurements between labs, calibration curves must be generated which can be interpolated to estimate cell count ^[3].

The iGEM 2018 Interlab study, as well as the Beal et al. paper ^[3] summarising its results, proposed the use of a suspension of silica microspheres for the generation of such a calibration curve. In this study, the OD₆₀₀ of serially diluted suspensions of silica microspheres, with approximately the same size and optical density of individual E. coli cells, was measured. This data was used to generate calibration curves across the range over which the relationship between OD₆₀₀ and microsphere concentration is linear. This linear curve could then be interpolated with any OD₆₀₀ measurement from an E. coli cell culture to identify the corresponding microsphere concentration, which in turn approximates cell count. Their work identified that relative to other calibration methods, e.g. measuring CFUs (colony forming units - number of viable colonies obtained when a sample is plated out), microsphere calibration of OD₆₀₀ measurements is equally inexpensive and much more precise for estimating cell count ^[3].

Flow cytometry is a technique in which single cells are probed with one or more laser light sources to measure light scattering, as well as fluorescence at one or more wavelengths ^[4]. Conventional analysis of data obtained in flow cytometry involves ‘gating’ off a population in a scatter plot of 2 parameters to demarcate signals associated with a particular subset of cells of the overall sample. Often, with high throughput analysis of many samples, a single gating template is drawn and then applied to all subsequent samples in batch analysis ^[4]. Such analysis can be used to determine cell density in a sample (cells/ml). The ImageStream system, along with other imaging flow cytometers, integrate fluorescence microscopy imaging with flow cytometry. This allows direct observation of sample cells, allowing for more accurate gating in analysis, based on the number of cells present in each image and their form ^[5].

In order to assess the accuracy of methods described in ^[3], we compared estimates of cell count of samples obtained from our characterisation of CycA, whereby 2 strains of E. coli (control containing only an empty pX1900 plasmid backbone and CycA containing the gene cycA embedded in the pX1900 plasmid) were grown in blank LB (with an alanine concentration of 5.8 mM) and LB doped with alanine up to a final concentration of 100 mM. For 3 replicates of each condition, cell count was approximated every hour both via measurement of OD₆₀₀ and estimation of cell count using microsphere calibration and by direct measurement of cells/ml using an ImageStream. A microsphere calibration curve was generated using data obtained by our PI Dr Chloe Singleton, following the procedure detailed in ^[3].

The percentage difference between microsphere calibrated estimated cell counts and ImageStream cells/ml measurements against time after inoculation for all 4 conditions is plotted below in Figure 1. This data shows that microsphere based estimations consistently overestimate cell density by a minimum of 246%. It is logical that OD-based estimates overestimate cell density because the overall optical density of sample is contributed to not only by viable cells, but additionally by many other components including accumulated waste in the growth medium and debris from lysed cells. In contrast, Imagestream measurements filter out debris by excluding all signals which fall outside of the region gated off in the flow cytometry output data which defines E. coli cells. Percentage inaccuracy peaks both at the beginning of log phase and in the final 2 hours of the incubation period.

A plausible explanation for this is that given both highly distressed cells, with altered morphologies resulting from unfavourable environmental conditions, and aggregates of cell debris from lysed cells, contribute to the overall OD₆₀₀ of a sample, and in turn its estimated microsphere calibrated cell count. Therefore, such estimates of cell count do not take into account population ‘health’. The effect of this on the accuracy of estimations of cell count becomes more pronounced in the latter stages of incubation as the likelihood of E. coli cells being exposed to suboptimal growth conditions increases, in turn increasing the proportion of distressed and lysed cells. In contrast, post processing of ImageStream data can filter out aggregates of lysed cells and distressed cells likely to lyse, via gating of the collected data, based upon examination of the images captured for objects which fall on the borderline of being considered a viable cell to be counted. This illustrates the utility of imaging flow cytometry for more precisely capturing the density of viable cells in a culture.

Our work establishes the degree of accuracy of microsphere calibrated estimation of cell count, whilst illustrating the utility of imaging flow cytometry, for those with access to the necessary equipment, . As the instrumentation becomes cheaper and more widespread, we hope such techniques will provide a platform for the generation of reproducible cell density measurements without the need for additional calibration. Given the expense of the equipment, it would be useful to create of a public repository with the contact information for iGEM teams with image flow cytometers available, to increase access and encourage collaboration between teams.

Future iGEM teams could investigate the reproducibility of our results, as well as searching for an explanation as to why percentage inaccuracy of bead based measurements peaks not only in the final phase of growth, but also at the beginning of log phase. Moreover, similar experiments could be run with appropriately sized beads to benchmark the accuracy of microsphere-calibrated OD measurements of cell count for other species of microorganism such asSaccharomyces cerevisiae.

Estimating cell density from OD₆₀₀

As aforementioned, estimation of cell density from OD₆₀₀ measurements, calibrated using standard curves produced using known concentrations of silica microspheres, is limited in accuracy. However, substituting such methods with image flow cytometry, whilst yielding more reliable measurements, is not feasible for the majority of the iGEM and wider synbio research community, given the prohibitive expense of the instruments required. A more appealing alternative is the use of Imagestream data to plot standard curves of OD₆₀₀ against known cell density values, which can then be interpolated by other researchers to obtain more reliable estimates of cell density from OD₆₀₀ measurements.

We plotted such standard curves using subsets of our alanine assay 7 hour incubation datasets for 4 different strains of E. coli BL21 DE3 pLysS, ranging from up to and including t=2 hours after inoculation to the full 7 hour incubation period. We took OD₆₀₀ measurements using a Tecan Infinite M PLEX plate reader and corresponding cells/ml measurements using an Imagestream image flow cytometer. Linear regression of OD₆₀₀ as a function of log₁₀(cells/ml) (given OD₆₀₀ is a logarithmic metric of light attenuation) identifies the former is directly proportional to the latter across an OD₆₀₀ range from 0 to 0.9. As shown below in Figure 2, for all datasets, R² values above 0.93 are obtained, and a maximum value is obtained when data up until t=5 hours after inoculation is included in the statistical analysis. Therefore, across the OD₆₀₀ range studied, a high linear correlation strength between OD₆₀₀ and log₁₀(cells/ml) is observed, suggesting log₁₀(cells/ml) and thus cells/ml can be interpolated from OD₆₀₀ measurements with a reasonable degree of accuracy.

As indicated in Figure 3 below, the goodness of fit for a linear correlation is clearly superior for the up to t=5 hours dataset than the up to t=7 dataset. However, the latter dataset covers OD₆₀₀ values up to approximately 0.9, versus 0.5 for the t=5 dataset, significantly increasing the range over which cells/ml can be interpolated from the standard curve, rather than being extrapolated. Therefore, when looking to estimate cells/ml from OD₆₀₀ using the standard curves obtained below, a trade-off must be struck between the superior fit of the correlation for the t=5 dataset, versus the practicality of having to produce samples with OD₆₀₀ values which lie within (or in close proximity to) the narrower range measured over.

Overall, the results obtained above indicate that OD₆₀₀ can provide a good basis for estimating cells/ml of a sample, for OD₆₀₀ values ranging from 0.05 to 0.85. Using the data gathered, as well as the equations for the linear correlations of the 2 standard curves plotted, future iGEM teams working with BL21 DE3 and its derivatives will be able to rapidly estimate cells/ml for samples from easily obtained OD600 values recorded using the Tecan M PLEX plate reader (specifically using black clear bottom Corning 96 well plates without the lid on). This will provide a rapid and cheap way of estimating cells/ml for a variety of wet lab applications (e.g. normalising uptake of a substrate to assess uptake per cell).

This method is preferable to microsphere based calibration because it relies on real-world image flow cytometry data of E. coli cells rather than silica beads approximating the optical properties of these cells. Moreover, by interpolating standard curves of OD against cells/ml data, extraneous factors which reduce the validity of using OD itself as a measure of cells/ml, such as the contribution of aggregating cell debris to optical density, are accounted for. This method is also a far cheaper and more accessible alternative to image flow cytometry. We hope future iGEM teams, with access to the necessary flow cytometry equipment, will expand the applicability of this technique, by producing similar standard curves using other strains of E. coli and species of model organism (e.g. Saccharomyces cerevisiae), as well as recording OD600 using other 96 well plates and plate readers, and recording corresponding cells/ml data using other flow cytometers.

Accessibility and STEM

Our work investigating the relationship between medical research, synthetic biology, and cures has contributed to the nature of human practices, as our research has not only confirmed the continued need for human practices, but has also provided suggestions for future ways teams could integrate human practices into their project. Throughout the interviews our participants emphasised how important it was to them that the opinions of specific patient populations were taken into account when designing new equipment and treatment options for them. Currently, medical research seems to be significantly lacking on this front, as many studies seem to either gather patient input into the testing stages, or not at all, rather than integrating their input in from the beginning. Rather, our research suggests that patient input should be used to guide the research from the outset of the project, which would likely ensure that the goals of the research are meaningful and beneficial to the group affected by its results. While these responses were from a patient’s perspective, our research also investigated the benefits of patient advisory groups from a researcher’s perspective as well.

When discussing the presence of ableism in research during our “Barriers to STEM” interviews, our participants, who were students and professionals in the STEM field, discussed the benefits that such groups could provide for scientific research. The responses from our participants similarly emphasised the importance of integrating patient input from the beginning, which they felt would help address many of the biases present in research against disabled people and help guide the focus of the research, and suggested the formation of advisory boards to help assist these studies or through the increased presence of researchers who have their own relevant lived experience.

However, our research into Barriers to STEM did not just stop with invesigating the benefits of advisory groups to address biases found in the field, as we continued on to do a comprehesive investigation into the barriers that prevent disabled people from engaging with STEM. Currently, many attempts to make the STEM field more accessible focus on adaptations for individual people that tend to focus on physical barriers such as laboratory equipment. This not only tends to place the responsibility for accessibility on the disabled person as they become responsible for making the adaptations, but fails to address many of the larger systemic barriers that prevent a person from accessing the field. Based on the responses from our participants, these systemic issues tend to be some of the most signficant causes as to why they struggle to participate in the field, and in the cases of some particpants, the reason they are considering withdrawing from the field. Addressing these barriers requires effort from the entire field to address, however doing so will not only benefit the entire disability community, but likely other disadvantaged groups as well, as many of the barriers tend to overlap.

Ultimately the results of our research have not only confirmed the need for human practices and its importance in producing meaningful and accurate research, but also suggest that for projects and research that focus on people, that the opinions and involvement of the group of focus should be integrated into the research from the earliest moment, through the use of advisory boards, the presence of researchers who have lived experience as a member of that group, and an increased understanding of the significance and importance of lived experience when conducting research.

To find out more, please visit our Inclusivity Page.

References

1. Myers J et al. Improving accuracy of cell and chromophore concentration measurements using optical density. BMC biophysics. 2013;6(1): 1-16. doi: doi.org/10.1186/2046-1682-6-4
2. Sciencing. What is Turbidity & What Does It Indicate in Microbiology?. Available from: https://sciencing.com/difference-between-optical-density-absorbance-7842652.html [Accessed 12/09/22]

3. Beal J et al. Robust estimation of bacterial cell count from optical density. Communications biology. 2020;3(1): 1-29. doi: 10.1038/s42003-020-01127-5
4. McKinnon K. Flow Cytometry: An Overview. Curr Protoc Immunol. 2018;120: 5.1.1-5.1.11. doi: 10.1002/cpim.4
5. Zuba-Surma E. The ImageStream System: a key step to a new era in imaging. Folia histochemica et cytobiologica. 2007;45(4): 279-290. Available from: https://pubmed.ncbi.nlm.nih.gov/18165167/ [Accessed 12/10/22]