The first thing I will say is that measuring cell density by OD is not a completely accurate method to begin with. That being said, the first step would be to produce a standard curve of OD vs cfu/ml. This would be started by inoculating a culture with a low density of cells. Samples would be taken at roughly every 30mins-1hr and split into 2 subsamples. The OD would be taken with 1 subsample and direct plating would be done with the second sample. Care must be taken with the sample for OD to ensure it's dilution is within the linear range of the spectrophotometer (anything reading above 2 should be diluted 1:10 to obtain a reading within the linear range). Care must also be taken with plated samples to ensure that only between 30-300 colonies are on any one plate (tables exist for most bacterial species which states an approp. dilution series to achieve this). This process is repeated until no further significant increase in OD is observed. After the experiment has run its course, a graph can be plotted with cfu/ml(as calculated from the plating samples) on one axis and OD600 on the other axis. The OD required to produce a 1ml sample containing your desired number of cells can then simply be read from the graph. To ensure high accuracy this procedure should be run in triplicate to produce (hopefully quite small) error bars.
The first thing I will say is that measuring cell density by OD is not a completely accurate method to begin with. That being said, the first step would be to produce a standard curve of OD vs cfu/ml. This would be started by inoculating a culture with a low density of cells. Samples would be taken at roughly every 30mins-1hr and split into 2 subsamples. The OD would be taken with 1 subsample and direct plating would be done with the second sample. Care must be taken with the sample for OD to ensure it's dilution is within the linear range of the spectrophotometer (anything reading above 2 should be diluted 1:10 to obtain a reading within the linear range). Care must also be taken with plated samples to ensure that only between 30-300 colonies are on any one plate (tables exist for most bacterial species which states an approp. dilution series to achieve this). This process is repeated until no further significant increase in OD is observed. After the experiment has run its course, a graph can be plotted with cfu/ml(as calculated from the plating samples) on one axis and OD600 on the other axis. The OD required to produce a 1ml sample containing your desired number of cells can then simply be read from the graph. To ensure high accuracy this procedure should be run in triplicate to produce (hopefully quite small) error bars.
Sorry, I may have slightly misread your question but a standard curve would still be required for you to accurately achieve your desired result. If you are growing your bacteria to a predefined OD then the standard curve will tell you how many bacteria are contained within 1ml of that culture. You would simply have to calculate the volume of culture required to extract that number of cells. Eg if your OD produces 2.5x10^7 cells then you would require 2ml of culture or if it produced 1x10^8 you would require 0.5ml of culture.
In addition to Michael's comments, beware of the following:
- What you are measuring is dispersion of light rather than optical density. This has the very important practical implication that OD readings of the same culture will vary between spectrometers due to differences in detector geometry. So, whatever OD-cfu/mL equivalence you determine, applies only to your specific spectrometer.
- OD changes (sometimes quite dramatically) with cell shape, which in turn varies with culture medium and growth stage.
- Doing bacterial counts by plating has the disadvantage that it only detects viable bacteria, and can underestimate the number of bacterial cells in species where two or more cells associate (diplococci, etc). Depending on your purpose, you might want to try a counting chamber, or simply a FACS machine (there are kits on the market for this purpose).