OD calibration factors drift

Cells-per-OD is not a constant. Across a growth curve it moved 11x.

OD₆₀₀ is often used as an inexpensive, real-time proxy for cell concentrations. A conversion factor is needed to convert the arbitrary OD₆₀₀ units to cell concentrations. The factor is determined once by making a standard curve.

That only works if the factor stays put. It does not. OD₆₀₀ is turbidity, so the signal per cell scales with cell size, and cells change size as a culture grows. The conversion factor moves with them.

This study tracked one E. coli culture in LB medium with BactoBox®, plate counts (CFU), and OD₆₀₀ at the same time points It asks one question. Which ratio between two methods holds still?

Take-home messages

• Cells-per-OD moved 11.4x across the run, from 1.6×10⁸ to 1.8×10⁹ cells/mL per OD unit. That is a 11× change during the growth curve
• The BactoBox®-to-CFU ratio held at a geometric mean of 1.01 with a geometric CV of 21%. It varied 2.2x in total, five times less than cells-per-OD.
• The drift in cells-per-OD and cells-per-CFU is not noise. It rose steadily with time as cell size decreased and scaled with CIZE⁻²·⁵ .
• CFU-per-OD drifted the same way, 6.6x. The problem is OD₆₀₀, not the counting method it is paired with.

Study design

• A single E. coli ATCC 8739 LB medium shake-flask culture inoculated at 1×10⁶ cells/mL and sampled 28 times over 11.6 hours. Cultivated at 37 °C, 200 RPM.
• Every sample was measured with BactoBox® (cells/mL and CIZE), by compact dry TC plate counts (CFU/mL), and by OD₆₀₀ using an Inplen Diluphotometer and standard cuvettes.
• Samples were diluted into the linear range of each instrument before measurement. A single measurement was done for each method.
• We report ratios as geometric means, because a ratio is a multiplicative quantity. Span is max divided by min.

The first nine OD readings sit in the blank

The photometer was read at every time point, including the first nine. Those nine readings came back at 0.00 ± 0.01 AU, which is below the lower limit of detection. The detection limit, taken as the blank mean plus three standard deviations, is 0.03 AU.

The first reading above that limit arrived at 2.2 hours, at 0.07 AU. By then the culture had passed 1×10⁷ cells/mL. This means that the first decade of growth carries no usable OD₆₀₀ signal at all.

An E. coli shake flask culture measured by BactoBox® (lavender), plate counts (yellow) and OD₆₀₀ (red). CIZE (green) from BactoBox® measurements is also shown (secondary axis). The y-axis is symlog, i.e. linear below the limit of detection (grey box) and logarithmic beyond 0.027 AU.

Stability during different growth stages

Ratios between the three different measurands are presented below. Values before 2.2 hours are not available for the OD method as OD below the detection limit would be division by noise.

Ratio between methods as a function of incubation time. The BactoBox®-to-CFU ratio stays flat across four log decades. The per-OD ratios climb with the inverse square of CIZE.

OD₆₀₀ ratios are unstable

The geometric mean of OD-to-CFU and OD-to-cells is 8.6×10⁸ CFU/AU and 8.9×10⁹ cells/AU, respectively. A curve centered at 1× would represent a stable calibration factor.

This is clearly not the case. For OD-to-CFU comparisons the factor rose 6.6x between early exponential phase and stationary phase. For OD-to-cells the factor rose 11.4×.

The drift is one-directional. It is not scatter around a stable mean that more replicates would average away. Pick your calibration point early and every later number is too low. Pick it late and every earlier number is too high.

Phase	Cells-per-OD (cells/mL per OD)	CIZE (µm)
Early exponential, 2.2–3.35 h	2.5×108	1.74–2.08
Late exponential, 4.42–5.53 h	9.0×108	1.05–1.34
Stationary, 6.12–11.63 h	1.7×109	0.85–0.94

BactoBox to CFU ratios are stable

The BactoBox®-to-CFU ratio is stable, and it is stable at 1.01. On average, one BactoBox® cell equals one colony-forming unit here, with no fitted factor in between. The span is 2.2× which is expectable when single measurements are compared, especially for a method that is as imprecise as CFU determination. In this study, frequent sampling points were preferred over precision.

In any case the 2.2× span is much lower than the 6× and 11× spans observed in the OD-to-CFU and OD-to-CFU results.

Cell size explains the difference

CIZE peaked at 2.13 µm after 1.6 hours and fell to 0.85 µm by the end of the run.

Turbidity is scattering, and scattering follows the cross-sectional area of the cell. Halve the diameter and each cell contributes roughly a quarter of the signal. The number of cells behind one OD unit therefore rises as cells shrink.

The data match that. Over this curve, cells-per-OD scaled as CIZE⁻²·⁵ (log-log r = -0.98) and CFU-per-OD as CIZE⁻² (r = -0.98). Both exponents sit near -2, which is the inverse-square behaviour you expect from an area term.

BactoBox® counts cells one at a time. Its ratio to plate counts compares one count to another count, so no size term enters. That is why it holds still while the per-OD ratios do not.

Implications

A cells-per-OD factor is only valid for one growth phase, in one medium, with one strain. Carry it outside those bounds and it brings an error you cannot see in the OD₆₀₀ reading itself.

The error is largest exactly where people want the factor most: comparing a culture early against the same culture late, or comparing two conditions that grow at different rates. Cell size differs in both cases. This means that growth rates are often not reliable when derived from OD₆₀₀ data. OD600 also confounds growth medium screening, since one condition may have few, but large cells, while another condition has small but copious cells.

When you need a number that means cells, count cells. BactoBox® reports cells/mL directly, matched CFU/mL one to one across this entire curve, and needs no calibration factor to do it.