Michael E. Dawson, Ph.D., RAC
Director of Regulatory Affairs
Associates of Cape Cod, Inc.
The question is often asked, how many bacteria does it take to make one Endotoxin Unit (EU)? Although it sounds like the beginning of a joke, unfortunately, there is neither a punch line nor a simple answer. In the case of sterile solutions, which may contain high concentrations of endotoxin, the answer is none, regardless of how much endotoxin is present. In the case of non-sterile solutions, endotoxin concentration may or may not correlate with bacterial number. When there is a correlation, the relationship between the two parameters can vary substantially. Great caution should be exercised in any attempt to relate endotoxin concentrations to numbers of organisms because of the variability of this relationship. With those caveats in mind, this article reviews some aspects of this relationship and the measurements involved.
As endotoxin is derived from the outer membrane of gram negative bacteria, it is not unreasonable to suppose there might be a relationship between bacterial number and endotoxin concentration. Watson and coworkers  demonstrated strong positive correlations between the two parameters both in the laboratory for an Escherichia coli culture in log phase of growth and in seawater samples from the open ocean. For E. coli in log phase the authors reported 49.4 fg of bound lipopolysaccharide (LPS) per cell; this declined to 28.9 fg/cell (i.e. 2.89 10-16 g/cell) in stationary phase. By contrast the average concentration of LPS for the seawater samples was 2.78 fg/cell. Thus, the relationships were quite different for the two data sets. For natural freshwaters, Evans et al.  have reported strong correlations between endotoxin concentration and both coliform and heterotrophic bacterial counts.
A significant correlation between endotoxin concentration and bacterial number has also been demonstrated in a pure water system . In this case, the concentration of LPS per cell varied throughout the system from a high of 15.7 fg/cell to a low of 2.1 fg/cell. The high value was recorded early in the treatment train and the low at the beginning of the distribution loop, which was the cleanest (highly oligotrophic) part of the system. In this study, as in the study by Watson et al. , bacterial number was determined by epifluorescence direct counting (EDC), not by culture. It should be noted that the numbers obtained by EDC can be an order of magnitude, or more, greater than those obtained by culture (for example, see Armisen and Servais ), particularly in the cases of natural and oligotrophic systems. Consequently, the amounts of endotoxin per cell that are determined using EDC values are lower than if culture techniques are used.
Another major caveat surrounding these determinations is that the values for mass of LPS reported are not based on an absolute determination of the mass of LPS. The amounts of LPS per cell are obtained in bacterial endotoxin tests using Limulus amebocyte lysate (LAL) reagent. The results represent the mass of standard endotoxin that has the same activity as that in the sample, not the amount of endotoxin present in the sample. Thus, a value of 2.1 fg/cell means that the activity of the endotoxin from a single cell is equal to that of 2.1 fg of the standard endotoxin preparation; we do not know the actual mass of endotoxin in the cell which gives that activity.
In the work of Watson et al.  and Dawson et al. , LPS preparations from of E. coli were used as endotoxin standards and results are expressed in units of mass of that standard. The problem is that different endotoxins can differ markedly in activity (or potency) per unit mass. This is true regardless of whether activity is determined in pyrogen tests or in LAL tests (,,, and ). Consequently, results expressed in units of mass of one endotoxin cannot be readily compared with another study in which a different standard endotoxin was used. The results within any one such study can be meaningfully compared with other results in the same study. The standard endotoxins used in such studies are usually derived from a single organism (commonly an E. coli strain) and consist of a purified preparation of lipopolysaccharide. In contrast, the endotoxin being measured may come from a variety of bacteria, which may not include the organism from which the standard endotoxin was obtained. In the two studies discussed at the beginning of this paragraph the bacteria in waters studied probably did not include E. coli. This problem of the different relative potencies of different endotoxin standards has now largely been addressed by the development of a reference standard endotoxin (RSE) and the establishment of the Endotoxin Unit (EU).
Relatively early in the application of LAL tests for the detection of endotoxin, FDA recognized the problems cause by the use of different endotoxin standards in different studies and the difficulty in comparing results. To address these problems, FDA decided to establish a standard endotoxin preparation expressed in units of activity as opposed to mass. The first significant batch RSE was designated EC-2. It comprised vials containing 1.0 µg of LPS derived from E. coli O113:H10 K negative lyophilized with 0.1% human serum albumin. A potency (a measure of biological activity) of 5,000 Endotoxin Units (EU) per vial was assigned to the standard. Given the mass of 1.0 µg of LPS per vial, this equates to 5 EU/ng. Batch EC-2 was followed by other reference preparations, including those provided by the current World Health Organization, the United States Pharmacopeia (USP) and European Pharmacopoeia, all of which are essentially the same material but with different labeling (and expressed in different Units in the case of the USP endotoxin reference standard).
For standard E. coli LPS preparations, it is common to think of a potency of approximately 10 EU/ng, which is a common value reported on certificates of analysis for control standard endotoxins (CSEs). CSEs are secondary standard endotoxin preparations (such as those provided by LAL reagent manufacturers) that have been standardized against RSE. Sometimes values of 20 EU/ng or higher are obtained for CSEs; in other cases values down to about 5 EU/ng are obtained. The potency of a given endotoxin preparation relative to RSE can vary depending upon the lot of lysate reagent used to make the determination. Consequently, it is important to use the potency stated on the appropriate certificate of analysis for the specific lots of CSE and of lysate reagent used.
Bearing in mind the caveats given above, the reported amounts of endotoxin per cell range between about 2 and 50 fg/cell. If we assume that 1 EU unit represents about 0.1 ng of endotoxin (i.e. 105 fg), then 1 EU would be equivalent to between 2,000 and 50,000 cells. However, given all the assumptions and caveats (including the fact that sterile solutions may contain high endotoxin concentrations), any predictions about the numbers of viable organisms that are derived from endotoxin concentrations should be treated with a healthy dose of skepticism.
 Watson, S. W., T. J. Novitsky, H. L. Quinby, and F. W. Valois. 1977. Determination of bacterial number and biomass in the marine environment. Appl.Environ.Microbiol. 33: 940-946.
 Evans, T. M., J. E. Schillinger, and D. G. Stuart. 1978. Rapid determination of bacteriological water quality by using Limulus lysate. Appl.Environ.Microbiol. 35: 376-382.
 Dawson, M. E., T. J. Novitsky, and M. J. Gould. 1988. Microbes, endotoxins and water. Pharm.Engineering. 8(2): 9-12.
 Armisen ,T.G. and P. Servais. 2004. Combining direct viable count (DVC) and fluorescent in situ hybridisation (FISH) to enumerate viable E. coil in rivers and wastewaters. Water Sci Technol. 50(1): 271-275.
 Greisman, S.E. and R. B. Hornick, R.B. 1969. Comparative pyrogenic reactivity of rabbit and man to bacterial endotoxin. Proc. Soc. Exp. Biol. Med. 131: 1154-1158.
 Pearson, F.C., M. E. Weary, J. Bohon. 1982. Relative potency of “environmental” endotoxin as measured by the Limulus amebocyte lysate test and the USP rabbit pyrogen test. In Watson S. W., J. Levin, T. J. Novitsky, eds. Endotoxins and their detection with the Limulus amebocyte lysate test. Alan R. Liss, Inc. New York. pp. 65-77.
 Pearson, F. C. III, M. E. Weary, H. E. Sargent, T. J. Novitsky, M. P. Winegar, H. Lin, G. Lindsay, R. N. Berzofsky, A. L. Lane, J. D. Wilson, J. F. Cooper, E. J. Helme, C. W. Twohy, H. I. Basch, M. Rech, and J. W. Slade. 1985. Comparison of several control standard endotoxins to the national reference standard endotoxin – an HIMA collaborative study. Appl. Environ. Microbiol. 50: 91-93.
 Weary, M. E., G. Donohue, F. C. Pearson, and K Story. 1980. Relative potencies of four reference endotoxin standards as measured by the Limulus amebocyte lysate and USP rabbit pyrogen tests. Appl. Environ. Microbiol. 40: 1148-1151.
 Weary, M. E., F. C. Pearson, III, J. Bohon, and G. Donohue. 1982. The activity of various endotoxins in the USP rabbit test and in three different LAL tests. In S. W. Watson, J. Levin. and T. J. Novitsky (eds.), Endotoxins and their Detection with the Limulus Amebocyte Lysate Test. Alan R. Liss, New York, 1982, pp. 365-379.