Bacteriology of Spoilage of Fish Muscle IV. Role of Protein

APPLIED MICROBIOLOGY, July 1967, p. 770-776 Copyright 1967 American Society for Microbiology Vol. 15, No. 4 Printed in U.S.A. Bacteriology of Spoilage of Fish Muscle IV. Role of Protein PETER LERKE, LIONEL FARBER, AND RALPH ADAMS Seafood and Nutrition Research Laboratory, The G. W. Hooper Foundation, University of California San Francisco Medical Center, San Francisco, California 94122 Received for publication 27 December 1966 Clarified muscle press juice from English sole (Parophrys vetulus) was fractionated by gel filtration into a protein and a protein-free fraction. Upon inoculation with spoilage bacteria, the protein fraction failed to show any signs of spoilage, but the protein-free fraction spoiled according to the usual organoleptic and chemical criteria. Despite its spoilage-resistant qualities, the protein fraction accelerated spoilage of the protein-free fraction when the two were combined. Protein breadkown due to bacterial action was greatest in the unfractionated juice and was least in the protein fraction. No significant proteolysis occurred until after spoilage became evident. During the early studies on the spoilage of flesh foods, such as meat, poultry, and fish, it was commonly thought that protein degradation was primarily involved. In the case of fish, it was gradually realized that proteolysis actually represented an advanced stage of spoilage (2, 10). More recently, considerable attention has been given to the presence in fresh fish muscle of lowmolecular-weight nitrogenous extractives such as trimethylamine oxide, simple peptides, amino acids, etc. (8). These substances can serve as growth substrates for spoilage bacteria and give rise to spoilage products (3). It used to be thought that these extractives, more specifically the amino acids and peptides, must be the end products of protein breakdown; this implied that proteolysis was a necessary first step in spoilage. Nevertheless, it has been shown repeatedly that the ability of certain bacteria to spoil fish muscle press juice is in no way connected with proteolytic activity. Thus, Castell and Greenough (3) have commented on the inability of Pseudomonasfragi to hydrolyze proteins; yet this organism was able to produce strong off-odors from fish muscle extractives. Similarly, while attempting a biochemical characterization of fish-spoilage bacteria (7), we noted that proteolytic ability, as determined by gelatin liquefaction and digestion of egg albumin, was not a constant characteristic of spoilers and could often be found among nonspoilers. As a result, a question arose concerning the importance of proteolysis in fish spoilage; namely, could precipitable protein supply the necessary lowmolecular-weight substrates from which spoilage 770 products could be formed through bacteria action? The present report attempts to furnish the answer. MATERIALS AND MEI HODS Bacteria. The two bacterial cultures used were pseudomonads isolated from spoiling fish. The spoiler was a nonpigmented Pseudomonas belonging to group III of Shewan's classification (9); the nonspoiler was a pigment-producing Pseudomonas, belonging to group I. Spoilage ability was originally determined by growing the organism for 5 days at 5 C in the sterile muscle juice described below. An organism was designated as a spoiler if, under these conditions, it was able to produce: (i) an off odor, (ii) more than 1 mg/100 ml of trimethylamine, and (iii) volatile reducing substances in an amount greater than 15 meq of reduction per 5 ml of sample. Organisms which grew but did not fulfill the above criteria were classified as nonspoilers. Substrates. The basic medium used, previously described by Lerke et al. (6), consisted of filtrationsterilized saline-diluted fish muscle press juice. The idea for a similar medium was advanced by Anderson and Fellers (1), and we wish to express our regret that, in earlier publications, we inadvertently omitted reference to this work. The fish used to prepare the substrate was local English sole (Parophyrs vetulus) obtainable in San Francisco. After preliminary filtration through Whatman no. 1 paper on a Buchner funnel, the juice was centrifuged at 4 C for 1 hr at 57,000 X g and sterilized by pressure filtration through 0.22-,u membrane filter (Millipore Corp., Bedford, Mass.). For purposes of fractionation, the juice was simply clarified by passing it through a Seitz pad, and a sample of the liquid was placed on a Sephadex G-25 column at 0 C. By using 0.8% NaCl solution as eluant, a clear separation could be achieved into a protein and

VOL. 15, 1967 SPOILAGE OF FISH MUSCLE 771 a nonprotein fraction. The approximate exclusion limit for Sephadex G-25 is molecular weight 5,000. The fractions were quite distinct, with no ninhydrinpositive material coming through in between. Any dilution developing during separation was taken into account, and all substrates were diluted to match the most highly diluted fraction (1:5). Thus, essentially three different media were used: whole juice, the protein fraction, and the nonprotein fraction. After separation, the fractions were also filtration-sterilized. As different batches of medium sometimes had to be used in the same experiment, different original levels of protein resulted (Table 5, for example). Assessment of spoilage. At each sampling, the material was examined sensorily, bacteriologically, and chemically. Sensory evaluation involved general appearance and odor. Bacteriological examinations consisted of total aerobic counts on Difco Penassay Base Agar: a sample of the appropriate dilution was spread on the surface of the medium with a sterile bent glass rod and the plate was incubated 48 hr at 22 C. Chemical tests included the estimation of volatile reducing substances (VRS), total volatile nitrogen (TVN), and trimethylamine nitrogen (TMN). The method used to determine VRS was the one described by Farber and Ferro (5), modified to allow colorimetric rather than titrimetric estimation of the extent of MNnO4 reduction. A description of this modification is being prepared for publication. Determinations of TVN and TMN were made by the microdiffusion procedure of Conway (4). Chemical tests sometimes not run after samples had spoiled. Estimation ofprotein and nonprotein nitrogen. Total nitrogen and protein nitrogen, after precipitation with 10% trichloroacetic acid, were determined by the micro-kjeldahl method. Nonprotein nitrogen was obtained by difference. RESuLTS Protein as a spoilage substrate. It was shown in several ways that isolated fish protein cannot spoil as whole juice does. Thus, if protein is heatprecipitated out of fish juice, washed, resuspended in saline, and inoculated with spoilage bacteria, no detectable spoilage results although bacterial TABLE 1. Mixed culture spoilage offish preparations incubated 24 hr at 22 C Medium Odor VRSa TMNb 1. Raw (uncooked) fish juice... Spoiled >36.0 5.9 2. Cooked fish juice... Spoiled >36.0 5.9 3. Supernatant fluid from 2... Spoiled >36.0 5.9 4. Resuspended precipitate from 2... Bland 2.8 0.0 a Microequivalents of reduction per 5 ml. bmilligrams trimethylamine nitrogen per 100 ml. TABLE 2. Effect of spoilage bacteria on substances of high and low molecular weight from fish muscle press juice, incubated 48 hr at 22 C Fraction Odor VRSa TMNb Dialyzable... Spoiled >36.0 4.8 Nondialyzable... Bland 3.8 0.0 a In microequivalents of reduction per 5-ml sample. 6 In milligrams of nitrogen per 100-ml sample. growth is abundant. The figures in Table 1 were obtained after 24 hr of incubation at 22 C. Similar results were obtained with pure cultures of spoilers. To avoid heat denaturation, some freshly obtained press juice was dialyzed against distilled water at 0 C for 48 hr. The first change of water was concentrated by forced evaporation, Seitzfiltered, and tubed. Dialysis was then continued for 7 more days with frequent changes of water. At the end of that period, the contents of the bag were concentrated to the original volume. The dialyzable and nondialyzable fractions were then inoculated with a spoilage organism and incubated for 48 hr at 22 C. The results of the chemical tests are shown in Table 2. As a result of these findings, the procedure described in Materials and Methods was developed for the fractionation of fish muscle juice on Sephadex G-25. The nonprotein fraction, when inoculated with spoilage bacteria, showed most of the chemical criteria of spoilage (VRS, TVN, TMN); a similarly inoculated protein fraction did not. Bacteria grew as well in both fractions as in whole, unfractionated substrate; yet, the chemical results were entirely different; i.e., the protein fraction did not give rise to VRS, TVN, or TMN. The values for all three chemical indices rose in the nonprotein fraction. Table 3 presents these findings and shows that the nonprotein fraction contains certain substances of low molecular weight that provide a substrate for the formation of chemical products indicative of spoilage. The protein fraction is lacking in such substrates. On the other hand, both fractions are sufficient to support a bacterial population almost as large as that found in whole muscle juice. As pure cultures of bacteria were used, there could be no question of a selective growth of different organisms in the two fractions. Spoilage patterns of the various juice fractions. Figure 1 shows the changes in different substrates inoculated with a mixed spoilage flora. Although

772 LERKE, FARBER, AND ADAMS APPL. MICROBIOL. TABLE 3. Spoilage of various fractions offish muscle press juice at 15 C by a pure culture of a spoiler Determination Substrate Non Day Day WIole Protein juice protein fraction Volatile reducing 0 7.5 5.0 5.0 substances (micro- 0.5 7.5 7.5 9.5 equivalents of 1.0 7.5 4.5 5.0 reduction per S-ml 1.5 36.0 5.5 6.5 sample) 2.0 a 16.0 4.0 2.5-36.0 4.0 Total volatile nitro- 0 0.0 0.0 0 gen (mg per 100-mI 0.5 0.0 0.0 0 sample) 1.0 0.0 0.0 0 1.5 4.5 2.0 0 2.0 5.0 2.75 0 2.5 6.0 5.5 0 Trimethylamine 0 0.0 0.0 0 nitrogen (mg per 0.5 0.0 0.0 0 100-ml sample) 1.0 0.0 0.0 0 1.5 3.0 0.0 0 2.0 3.3 2.5 0 2.5 3.2 4.7 0 Log bacterial num- 0 3.30 3.30 3.30 bers per ml 0.5 6.10 5.60 5.20 1.0 7.35 6.80 6.30 1.5 8.0 8.20 7.20 2.0 8.5 8.60 7.80 2.5 8.9 8.60 8.10 a No sample taken. it can be said that the nonprotein fraction spoiled and the protein fraction did not, the nonprotein fraction did not spoil as quickly as the unfractionated substrate. Upon recombination of the protein and nonprotein fractions, however, the rate of spoilage was restored almost to the level of the unfractionated juice. This reconstituted substrate spoiled faster than either of its components alone. The bacterial growth curves in all four media were essentially identical. Similar results are shown in Tables 4-7. Moreover, the difference between a spoiler and a nonspoiler can clearly be seen in Table 4, showing that one organism caused complete spoilage of the whole juice in 2 days, whereas the same medium inoculated with another culture was still perfectly good on the 5th day. The difference between the two organisms was completely obliterated when they were grown in the protein fraction alone (Table 5). Here, although bacterial growth was ample, no trace of organoleptic or z 0 - v 0 M a. NL. LU 0 ea 14 14 14 * =unfractionated juice o protein fraction A=nonprotein fraction A=reconstituted juice 6 8 10 12 14 DAYS AT 50C FIG. 1. Effect of a mixed spoilage flora on fish muscle press juice and oni various fractions derived therefrom. The reconstituted juice is a 1:1 mixture of the protein and noniprotein fractions. chemical deterioration could be detected by the tests employed (VRS, TMN). When the two organisms were grown on the nonprotein fraction (Table 6), the spoiler again became distinguishable from the nonspoiler, although the rate of spoilage was considerably reduced. Upon recombination of the two fractions, the characteristics of the spoiler reappeared, this time with only a slightly reduced intensity when compared with unfractionated juice. Changes in the nitrogenous components of fish juice during spoilage. Several observations can be made from Tables 4-7. Autolysis seems rather insignificant. Of all the sterile substrates containing protein, only the whole juice decreased noticeably in protein nitrogen on storage (Table 4). Moreover, no apparent decrease took place until after the 5th day, although the medium had spoiled completely 3 days before. The breakdown of fish protein due to bacterial

TABLE 4. Actioni of a spoiler and a nonspoiler on unfractionated fish muscle juice Inoculum Days at 10 C VRSa TMNb Bacteriac PNd NPN6 Spoiler 0 2.0 0.0 5.30 212 (47) 230 (53) 1 2.0 0.0 6.38 212 (47) 230 (53) 2 36.0 2.8 7.4 190 (38) 254 (57) 3-8.23 167 (38) 274 (62) 4-147 (32) 301 (68) 5 _ 8 _ 75 (17) 370 (83) Nonspoiler 0 2.0 0.0 5.56 205 (46) 237 (54) 1 2.0 0.0 6.36 203 (46) 239 (54) 2 2.0 0.0 7.43 206 (50) 210 (50) 3 2.8 0.0 8.26 204 (46) 230 (54) 4 5.0 0.0 8.30 203 (48) 219 (52) 5 5.0 0.0 192 (45) 240 (55) 8 171 (41) 249 (59) None 0 2.8 0.0 200 (44) 256 (56) 1 2.8 0.0-234 (51) 220 (49) 2 2.8 0.0 193 (44) 247 (56) 3 3.8 0.0 210 (47) 240 (53) 4 5.8 0.0 185 (42) 255 (58) 5 8.0 0.0-195 (44) 251 (56) 8 148 (34) 282 (66) Log bacterial numbers per milliliter of medium. e Nonprotein nitrogen in micrograms per milliliter (percentage of total nitrogen). TABLE 5. Action of a spoiler and a nonspoiler on the protein fraction offish muscle press juice Inoculum Days at 10 C VRSa TMNb Bactera" PNd NPN6 Spoiler 0 2.8 0.0 3.67 635 (99) 10 (I) 2 2.8 0.0 7.23 629 (97) 21 (3) 3 3.1 0.0 7.54 613 (96) 25 (4) 6 4.0 0.0 7.85 571 (90) 66 (10) 8 5.0 0.0 8.0 501 (81) 116 (19) 10 4.3 0.0 7.96 505 (81) 117 (19) 13 0.7 0.0 6.85 417 (66) 211 (33) Nonspoiler 0 1.8 0.0 3.48 416 (97) 14 (3) 2 1.0 0.0 7.08 414 (97) 14 (3) 3 4.3 0.0 7.52 419 (98) 10 (3) 6 3.5 0.0 7.70 420 (100) 0 (0) 8 6.2 0.0 7.75 386 (100) 0 (0) 10 2.9 0.0 7.89 408 (97) 13 (3) 13 2.6 0.0 8.04 383 (96) 15 (4) None 0 1.0 0.0-415 (97) 15 (3) 2 3.8 0.0 430 (100) 0 (0) 3 6.7 0.0 415 (100) 0 (0) 6 2.0 0.0 413 (100) 0 (0) 8 1.7 0.0-378 (95) 19 (5) 10 3.3 0.0 405 (97) 12 (3) 13 2.5 0.0 383 (98) 7 (2) c Log bacterial numbers per milliliter of medium. e Nonprotein nitrogen in micrograms per milliliter (percentage of total nitrogen). 773

774 LERKE, FARBER, AND ADAMS APPL. MICROBIOL. TABLE 6. Actiont of a spoiler and a nonspoiler on the nonprotein fraction offish muscle press juice Inoculum Days at 10 C VRSa TMb Bacteriac PNd NPNe Spoiler 0 3.2 0.0 4.86 0 (0) 480 (100) 1 3.4 0.0 5.75 0 (0) 478 (100) 2 3.8 0.5 7.49 18 (4) 458 (96) 4 12.6 6.1 7.81 38 (8) 446 (92) 6 >36.0 5.4 8.0 58 (12) 422 (88) 7 >36.0 5.4 7.90 50 (10) 438 (90) 9 >36.0 6.9 7.85 46 (9) 444 (91) Nonspoiler 0 3.2 0.0 4.22 0 (0) 480 (100) 1 6.2 0.0 5.42 0 (0) 475 (100) 2 8.8 0.0 7.53 20 (4) 450 (96) 4 3.8 0.0 7.85 38 (8) 446 (92) 6 1.8 0.0 8.15 60 (12) 430 (88) 7 2.0 0.0 8.25 50 (10) 435 (90) 9 1.8 0.0 8.15 46 (9) 444 (91) None 0 3.2 0.0-0 (0) 480 (100) 1 4.1 0.0 0(0) 480 (100) 2 4.8 0.0 0(0) 484 (100) 4 2.8 0.0-0(0) 486 (100) 6 3.7 0.0-0(0) 480 (100) 7 4.0 0.0 0 (0) 482 (100) 9 4.2 0.0 0(0) 480 (100) c Log bacterial numbers per milliliter of medium. 6 Nonprotein nitrogen in micrograms per milliliter (percentage of total nitrogen). action seemed to occur to a significant degree in all protein-containing media. The spoiling organism was more active than the nonspoiler, although both were strong gelatin liquefiers. It should be kept in mind that the protein nitrogen figures, as shown in Table 4, 5, and 7, include protein of bacterial origin. If we took this into account and assumed that bacteria contribute protein to the extent shown in Table 6, where no fish protein is present to begin with, protein nitrogen figures can then be adjusted to represent only fish protein. Thus, Fig. 2 shows the percentage of fish protein remaining in the three substrates tested when subjected to the action of a spoilage organism. The greatest breakdown of protein was in the whole juice and the least in the protein fraction alone, with the values for the reconstituted juice falling in between. Again, the bacteria grew equally well in all media. The vertical marks indicate time of onset of spoilage of the substrate, but by that time proteolysis had barely begun. The curve for the protein fraction has no mark, of course, for this substrate never spoiled, either organoleptically or according to our chemical criteria. DIscussIoN The results presented above show that substances present in the muscle press juice of English sole that have a molecular weight greater than 5,000 cannot serve as a substrate for spoilage in the usual sense. These substances of high molecular weight appear to consist mostly of protein; they can serve as an adequate growth substrate for bacteria, but cannot, under the action of known spoilers, give rise to spoilage products. Here, some qualifying statements are necessary regarding the nature of spoilage. We know that bacteria grow in the protein fraction, presumably utilizing in some way the large molecules present. We can measure any proteolysis that may occur by estimating the amount of protein remaining. But, in our view, whatever happens in the protein fraction does not constitute spoilage because, (i) no off odors commonly associated with spoiled fish appear, and (ii) there are no VRS and no trimethylamine, products that we have repeatedly found to be associated with spoiling sole muscle, or press juice therefrom. Therefore, if we define spoilage as the production of off odors, VRS, and

VOL. 15, 1967 SPOILAGE OF FISH MUSCLE 775 TABLE 7. Action of a spoiler and a nonspoiler on a 1:1 mixture of the protein and nionprotein fractions of fish muscle press juice Inoculum Days at 10 C VRSa TMNb Bacteriac PNd NPNe Spoiler 0 5.3 0.0 4.59 414 (61) 264 (39) 1 6.0 0.0 6.48 416 (62) 259 (38) 2 30.0 0.5 8.0 400 (59) 273 (41) 3 >36.0 0.85 8.10 374 (55) 289 (45) 5 _5 8.16 330 (48) 353 (52) 9 8.25 278 (40) 412 (60) 12 - _ - 248 (37) 422 (63) 19-207 (30) 463 (70) Nonspoiler 0 2.5 0.0 4.67 414 (61) 264 (39) 1 3.0 0.0 6.0 408 (59) 278 (41) 2 1.4 0.0 7.71 393 (58) 282 (42) 3 1.9 0.0 7.95 371 (57) 279 (43) 5 3.1 0.0 8.59 344 (52) 316 (48) 9 2.8 0.0 8.74 319 (47) 362 (53) 12 3.1 0.0 8.56 310 (46) 368 (54) 19 3.1 0.0 8.49 270 (40) 398 (60) None 0 2.0 0.0-414 (61) 264 (39) 1 3.8 0.0-415 (60) 275 (40) 2 1.7 0.0 413 (61) 267 (39) 3 2.5 0.0-414 (61) 265 (39) 5 3.8 0.0-394 (59) 273 (41) 9 3.2 0.0-400 (61) 249 (39) 12 3.2 0.0-393 (61) 256 (39) 19 4.8 0.0-388 (57) 291 (43) C Log bacterial numbers per milliliter of medium. e Nonprotein nitrogen in micrograms per milliliter (percentage of total nitrogen). trimethylamine, then we can state that the nonprotein fraction spoils and the protein fraction does not. Our results showed that a known spodler, capable of breaking down the protein, was still unable to utilize the fragments to produce off odors. This indicates that none of the breakdown products from the protein fraction can serve as a spoilage substrate. Unpublished experiments in which we subjected the protein fraction to the action of a strongly proteolytic Flavobacterium (a nonspoiler) corroborate these findings: after the native protein had almost entirely disappeared, the medium was filtration-sterilized and inoculated with a mixture of spoilage bacteriano spoilage resulted, although the bacteria grew. If the breakdown products of fish muscle protein cannot function as spoilage substrates, then a spoiler should not have to be proteolytic. There are numerous spoilage bacteria which exhibit no proteolytic capabilities whatever (7). Recently, J. M. Jay (Bacteriol. Proc., p. 13, 1966) was able to show that in beef there is no significant decrease in protein with the onset of definite spoilage. Although our spoiler happened to be more actively proteolytic than the nonspoiler, apparently linking spoilage to proteolytic ability, it was still unable to spoil (produce off odors from) fish muscle protein. It can be argued that the protein fraction does not spoil when inoculated with a mixed flora because the spoilers are selectively repressed in that medium. To circumvent this objection, we used a single spoiler. If fish muscle protein is not a spoilage substrate by itself, does it play any role in spoilage? Our results indicate that it does, for when it is added to the nonprotein fraction it hastens the spoilage. The reason for this phenomenon is still unknown. We have only tried combining the two fractions in a 1:1 ratio. The use of different ratios might furnish some clues about the nature of the substances having this enhancing effect. Furthermore, what we call our protein fraction contains substances having molecular weights as low as 5,000.

776 LERKE, FARBER, AND ADAMS APPL. MICROBIOL. 100 go0 80 70 60 50s 40 30j 201 10 * =unfractionated juice o =protein fraction - =reconstituted juice (1:1 mixture of the protein and nonprotein fractions) 0 1 2 3 4 5 DAYS AT 100C 6 7 8 9 FIG. 2. Action of a proteolytic spoilage bacterium onz the protein of various preparations from fish muscle press juice. Proteolysis is expressed as percentage of the protein originally present. Although we have shown that such relatively small molecules are not spoilage substrates in the usually accepted sense, the possibility remains that these substances, being readily utilizable as a growth (as opposed to a spoilage) substrate, could contribute significantly to the overall spoilage process. It is not known why the addition of the nonprotein fraction to the protein fraction speeds up proteolysis of the latter. The rates of both spoilage and proteolysis are highest in the unfractionated juice and lowest in the separate fractions (proteolysis, of course, does not occur in the nonprotein fraction). These observations are not the result of different degrees of dilution. The importance of autolysis in the spoilage process has been questioned many times. Its possible effect would presumably be due to an increased supply of spoilage substrates. If we define autolysis as the hydrolytic breakdown of proteins and not of relatively small peptides, then our results show that (i) autolysis is very slight at the time spoilage occurs, and (ii) even if it were a prominent feature of spoilage, the fragments it would liberate would probably be useless as spoilage substrates. In this report, we have tried to show that protein plays a minor role in fish muscle spoilage. The major role apparently belongs to the substances of low molecular weight. Progress is being made in the identification of these substances, but, aside from a few instances, such as the reduction of trimethylamine oxide to trimethylamine, the identity of the spoilage products which arise from various substrates remains largely unknown. The protein fraction, however, may be a tool to provide some answers. This fraction or a subfraction thereof can serve as a basic medium that supports bacterial growth yet does not spoil. It has the added advantage of being closely related to raw fish muscle. We are now attempting to identify specific spoilage products by adding potential spoilage substrates of known composition to this basal medium and inoculating with spoilage bacteria. ACKNOWLEDGMENT This investigation was supported by Public Health Service research grant EF-00192. LITERATURE CITED 1. ANDERSON, D. W., AND C. R. FELLERS. 1949. Preparation of sterile fish muscle press juice. Food Technol. 3:274. 2. BEATTY, S. A., AND V. K. COLLINS. 1940. Studies of fish spoilage. VI. The breakdown of carbohydrates, proteins and amino acids during spoilage of cod muscle press juice. J. Fisheries Res. Board Can. 4:412-423. 3. CASTELL, C. H., AND M. F. GREENOUGH. 1959. The action of Pseudomonas on fish muscle. IV. Relation between substrate composition and the development of odours by Pseudomonas fragi. J. Fisheries Res. Board Can. 16:21-31. 4. CONWAY, E. J. 1947. Microdiffusion analysis and volumetric error, revised ed., p. 161. D. Van Nostrand Co., New York. 5. FARBER, L., AND M. FERRO. 1956. Volatile reducing substances (VRS) and volatile nitrogen compounds in relation to spoilage in canned fish. Food Technol. 10:303-304. 6. LERKE, P., R. ADAMS, AND L. FARBER. 1963. Bacteriology of spoilage of fish muscle. I. Sterile press juice as a suitable experimental medium. Appl. Microbiol. 11:458-462. 7. LERKE, P., R. ADAMS, AND L. FARBER. 1965. Bacteriology of spoilage of fish muscle. III. Characterization of spoilers. Appl. Microbiol. 13:625-630. 8. SHEWAN, J. M. 1961. The microbiology of seawater fish, p. 487-560. In. G. Bergstrom [ed.], Fish as food, vol. 1. Academic Press, Inc., New York. 9. SHEWAN, J. M., G. HOBBS, AND W. HODGKISS. 1960. The Pseudomonas and Achromobacter groups of bacteria in the spoilage of marine white fish. J. Appl. Bacteriol. 23:463-468. 10. TARR, H. L. A. 1953. Symposium on cured and frozen fish technology. Swedish Institute for Food Preservation Research, Goteborg, No. 100, Chapter 5.