Table 4. Results of feeding unsaponifiables from "fatty acid by-product" to weanling rats by stomach tube (1.0 ml/rat/day) indicated in the urea filtrate was shown in later experiments to reside in the unsaponifiable residue that had not been completely extracted. It is clear that the edema disease-producing material is associated with the unsaponifiable portion of the fatty byproduct. In Table 5 are also shown the variations in severity of the disease produced by a series of graded doses. Although such a tabulation as shown here cannot give a complete picture of the range of observations encountered, it indicates how the pattern of symptoms seen in any one group depends upon the dose of active material fed. In these studies the dose was usually high enough to produce severe symptoms, and as we became more familiar with the disease we became more confident of the results obtainable with smaller groups. Whenever possible we now use five chicks per group; 10.03 11.8 0.035 0.162 0.234 0.16* 14.9 0.084 0.073 occasionally we have used as few as two, when the amount, of sample was limited. Concentration of Toxic Factor Alumina Chromatography-The first step in our attempts to isolate the factor that caused the "edema discase" from the unsaponifiable portion of the fatty by-pro-luet was chromatography on alumina (1:6). A chromatogram (Fig. 2) was obtained by plotting the weight of material eluted by increments of solvent. Petzoleum ether was used until the clution curve seemed to have reached a minimum, then was followed by 25% ethyl ether in petroleum ether. The elution was completed with 100% ethyl ether to give three distinct fractions of increasing polarity. In general, Fraction I could be characterized as hydrocarbon (HC), Fraction I as containing carbonyl compounds, and Fraction III as containing sterols. The ultraviolet absorpcon spectra show that Fraction I has characteristic absorption peaks at 223, 227, 234, and 243 mp, suggesting the presence of significant amounts of cholestadiene, either the 2,4 or 3,5 isomer or both. In anticipation of the possibility that the toxicity would reside in action I (83% of the unsaponifiable mater ) a sample of 3,5-cholestadiene was prepaed by W. L. Hall in time to be includ in the third feeding trial. The results of this experiment indicated clearly that there was no response to cholestadiene but that all the activity responsible for the chick a disease was in Fraction II, which coituted approximately 8, of the unsapor, fiable or about 0.85% of the original proba Purifcation of Fruction 11-Three differ Table 5. Effect of by-product fat and its fractions on chicks • Figures in parentheses indicate average survival time in days. 10 mg d-alpha tocopherol per chick per day by mouth. The toxicity score is computed by assigning a score of 5 for each chick that died of the edema disease, 4 for each chick that showed one or more of the characteristic symptoms, i.e., hydropericardium, hydroperitoneum,or subcutaneous edema, in severe degree, a score of 3 for moderate symptoms, and a score of 1 or 2 for mild symptoms. The average of the individual scores is tabulated as the toxicity score. ent approaches were made simultaneously to the problem of further purifying Fraction II: (1) counter current distribution (iso-octane: methanol solvent system with 500 transfers 1), (2) chemical separation of carbonyl containing compounds by the Girard Reagent T, and (8) re-chromatography on alumina with a higher ratio of adsorbent to sample and more gradual elution. For the counter-current distribution experiment 8.4 grams of Fraction II were loaded into the first 10 tubes of a 500 tube Craig all-glass counter-current apparatus. After 500 transfers, the contents of every 10 consecutive tubes were combined, and after evaporation of the solvents the weight of solids was determined in each of the 50 fractions obtained. The resulting distribution curve, i.e., solute vs. tube numbers, is shown in Fig. 3. The 50 fractions were combined according to the peaks indicated by the distribution curve into 10 fractions for biological testing (Fig. 3). Each fraction was added to the test diet at a level equivalent to 7% of the original fatty by-product. Figure 4 shows the separation procedure diagramnmatically, and indicates the amount of material recovered in each fraction. The activity seemed to be equally distributed between Fraction 5 (tubes 240-320 containing 2.239 grams) and Fraction 6 (tubes 321-3-13 containing 0.600 grams), as indicated by the toxicity scores. Fraction 6 was the most We gratefully acknowledge our debt to Procter & Gamble workers for suggesting the solvent system. potent, showing as much activity at a level of 4.2 mg per 100 grams of feed as had been observed with 5 grams of original product. It is of interest that the toxic fractions 5 and 6 did not react with digitonin and did not show the L.B. reaction. In preparing material for the chemical separation (Fig. 5) Fraction II was cut into two approximately equal portions; a characteristic yellow band was used as the landmark for the separation. Each of the fractions (IIA and IIB) was treated identically. First, after evaporation of the ether the material was heated with methanol, then cooled. When the hot methanol solution cooled, white crystals were obtained. The infrared spectrum for these crystals indicated a carbonyl compound with a long aliphatic chain. A synthetic compound, dipalmitone (prepared by Jonas Carol), had a very similar infrared spectrum. Approximately 16% of Fraction IIA was insoluble in methanol. To the methanolsoluble portion was added twice the weight of Girard Reagent T and to the mixture was added 10% its volume of glacial acetic acid. The mixture was allowed to stand for two hours at room temperature before the water-soluble derivatives were separated from the unreacted material ("non-ketonic"). The "non-ketonic" portion was dissolved in hot acetic acid and heated with two times its weight of Girard Reagent T on the steam bath for 10 minutes, and the unreacted material was separated from the water-soluble derivatives. Approximately 23% of sample Fig. 4-Partition of Fraction II with methyl alcohol:iso-octane counter-current system. (Num bers follnord by + sign are toxicity scores; see footnote e, Teble 6.) had reacted in the cold and 8% more with heat; 69% did not react under these conditions and was classified as "non-ketonic" (11N). The various fractions derived from Fraction IIA in Fig. 5, as well as a similar set derived from Fraction IIB, were also fed at a 7% equivalent level in the diet. The synthetic dipalmitone was fed at a level of 10 mg per 100 grams. As indicated, there was no disease-producing activity in any of the fractions obtained from IIB and only the "non-ketonic" Fraction 11N was very active. Surprisingly, however, the infrared spectrun of Fraction 11N definitely indicated the presence of carbonyl groups. This so-called "non-ketonic" material was again treated with Girard Reagent T, this time with five times the weight of reagent instead of two times, and was heated at 80° C for 15 min utes. After the water-soluble derivatives were separated from the unreacted nonketonic material, two-thirds of the 11N fraction proved to have reacted. The nonketonic residue was again treated with excess Girard reagent and heat, and this time 50% of the "non-ketonic" material reacted. When the original material from the Girard derivatives was regenerated and the reaction was repeated with excess Girard reagent, again only 67% reacted to form ketoneGirard derivatives. It was clear that our active "non-ketonic" Fraction 11N consisted largely of compounds containing "hindered" carbonyl groups that would react with Girard reagent with great reluctance. Of additional interest is the fact that these "hindered ketone" fractions had characteristic ultraviolet spectra with a major absorption max |