Aquaculture Research, 2011, 42,360^365 doi:10.1111/j.1365-2109.2010.02630.x

Geosmin causes off-flavour in arctic charr in recirculating aquaculture systems

Ste¤phanie Houle1, Kevin K Schrader2, Nathalie R Le Francois3,4,Yves Comeau5, Mourad Kharoune6, Steven T Summerfelt7, Arianne Savoie3 & Grant W Vandenberg1 1De¤partement des sciences animales, Universite¤Laval, Que¤bec, QC, Canada 2United States Department of Agriculture, Natural Products Utilization Research, Agricultural Research Service, University of Mississippi, University, MS, USA 3Universite¤du Que¤bec aØRimouski, Rimouski, QC, Canada 4BiodoŒ me de Montre¤al, Montre¤al, QC, Canada 5Department of Civil, Geological and Mining Engineering, EŁ cole Polytechnique de Montre¤al, Montre¤al, QC, Canada 6De¤partement de Ge¤nie de la Production Automatise¤e, EŁ cole deTechnologie Supe¤rieure, Montre¤al, QC, Canada 7The Conservation Fund Freshwater Institute, Shepherdstown,WV,USA

Correspondence: G W Vandenberg, De¤partement des sciences animales, Universite¤Laval,2425, rue de l’Agriculture, Que¤bec, QC, Canada G1V 0A6. E-mail: [email protected]

Abstract Keywords: geosmin, o¡-£avour, arctic charr, recirculating aquaculture systems The ‘earthy’ and ‘muddy’ o¡-£avours in pond-reared ¢sh are due to the presence of geosmin or 2-methyli- soborneol in the £esh of the ¢sh. Similar o¡-£avours Introduction have been reported in ¢sh raised in recirculating aquaculture systems (RAS); however, little informa- ‘O¡-£avour’ in aquacultured products is a problem tion is available regarding the cause of these o¡-£a- that can be costly for the industry and can also lead vours. Our hypothesis was that earthy and muddy to the loss of sales. The problem was ¢rst recognized o¡-£avour compounds, found previously in pond- in farmed ¢sh raised in outdoor ponds but it is now raised ¢sh, are also responsible for o¡-£avours in ¢sh also being observed in ¢sh raised in recirculating raised in RAS. In this preliminary study, we exam- and partial recirculating systems (Schrader & Dennis ined water, bio¢lms in RAS and ¢llets from cultured 2005; Robin, Cravedi, Hillenweck, Deshayes & Vallod arctic charr known to have o¡-£avours and requiring 2006). The o¡-£avours most frequently encountered depuration using instrumental [solid-phase microex- are ‘earthy’ and ‘muddy’. The earthy and muddy £a- traction procedure and gas chromatograph-mass vours in cultured ¢sh are due to the presence of geos- spectrometry (GC-MS)] and human sensory ana- min or 2-methylisoborneol (MIB) in the £esh of the lyses. Geosmin was present in the samples taken from ¢sh (Tucker 2000; Howgate 2004).These compounds the bio¢lter and on the side walls of the tanks. Two- are semivolatiles and, when present in the water, are methylisoborneol was only found in low levels in the absorbed by the ¢sh and stored in lipid-rich tissues. samples. The GC-MS results indicated the presence of Planktonic and benthic cyanobacteria and actinomy- geosmin in the ¢llets (705 ng kg À 1), but lower levels cetes are known producers of MIB and geosmin were found in the water (30.5 ng L À 1). Sensory ana- (Grimm, Lloyd, Zimba & Palmer 1999). The process lyses also detected an earthy £avour (i.e., geosmin of elimination of these compounds by the ¢sh is presence) in the ¢llets, and, therefore, it appears that much longer then the absorption. The elimination geosmin is the main compound responsible for the rate decreases with lower temperature water and is o¡-£avour in RAS. Further studies are being per- slower in ¢sh with more fatty tissues (Tucker 2000). formed to identify the microorganisms responsible In a commercial operation, a depuration time is re- for geosmin production in RAS. quired before the harvesting in order to gradually

r 2010 The Authors 360 Aquaculture Research r 2010 Blackwell Publishing Ltd Aquaculture Research, 2011, 42, 360^365 Geosmin is the dominant o¡-£avour compound in RAS SHouleet al. eliminate the earthy-muddy taste produced by geos- Sample collection of water and bio¢lms min and MIB. Most o¡-£avour-related research has Six water samples were taken in each tank; the been performed with channel cat¢sh raised in USA samples were taken from di¡erent places in the (Tucker 2000) and relatively little research has been tank. Six bio¢lm samples were also taken from performed with other types of ¢sh, such as charr, each tank. Individual water samples were placed raised in recirculating systems. This is the ¢rst report in 20 mL glass scintillation vials with foil-lined on the issue of o¡-£avour aimed at Arctic charr caps (Fisher Scienti¢c, Ottawa, ON, Canada).The vials (Salvelinus alpinus) largely raised in recirculating were ¢lled completely so that no air bubbles were ob- aquaculture systems (RAS) as it displays clear adapt- served when the vial was capped and inverted. Indi- ability to crowding, tolerance to disease and fetches a vidual bio¢lm samples were also placed in 20 mL high market price (Le Francois, Lemieux & Blier glass scintillation vials. Bio¢lm material was scraped 2002). Our hypothesis was that the compounds in- from the side of the tanks and the bio¢lm was then volved in the production o¡-£avours in pond-raised placed into the vials. The samples were refrigerated ¢sh are responsible for o¡-£avours encountered in (4 1C) until shipping to the United States Department RAS. In this primary study, we analysed ¢llets from of Agriculture, Agricultural Research Service, cultured Arctic charr that were o¡-£avour and sub- NPURU at theThad Cochran Research Center in Uni- sequently required depuration before they could be versity,Mississippi. sold to processors. Instrumental and human sensory analyses were used to identify the types of o¡-£avour and the compound(s) responsible for the earthy o¡- £avour in the ¢llets. Analysis of ¢llets by microwave distillation and solid-phase microextraction-gas chromatograph-mass spectrometer Materials and methods (SPME-GC-MS) Sample collection and ¢llet preparation The method of Schrader, Dhammika Nanayakkara, Tucker, Rimando, Ganzera and Schaneberg (2003) Arctic charr were from a commercial farm operated in as modi¢ed from the method of Lloyd, Lea, Zimba Man, WV, USA. The charr were cultured indoor in and Grimm (1998) was used to analyse geosmin and 3 dual-drain ¢breglass circular tanks (640 m ) under MIB from ¢llets. For each ¢llet of Arctic charr, dim light. The water temperature was maintained at 20 Æ 0.5 g of ¢sh were weighed into a glass distilla- 12.6^13.0 1C. The Arctic charr densities in the grow- tion £ask. The £ask was then heated in the micro- À 3 out tanks were between 120 and 140 kg m .The wave (General Electric, New York, NY, USA; model production system was a water-recirculation unit JES1036WF0001) for 5 min at power level ‘3’ while (95% water exchange) that included bio¢ltration and À 1 purging with 80 mL L of N2 gas. The collected dis- pure oxygen injection. Fish were fed industrial manu- tillate was cooled in a water and antifreeze bath factured pellets from Zeigler (Gardners, PA, USA), (Fisher Scienti¢c, model Isotemp 3006) and the vo- using an automatic feeder for16h a day at10^12-min lume was adjusted to 20 mL using nanopure water. intervals and contained 45% protein and 20% fat. Ten From each 20 mL of sample, aliquots of 600 mLwere ¢sh were collected from four tanks (10 ¢sh  four placed into 2 mL glass vials containing 0.3 g of NaCl. tanks): two grow-out tanks and two depuration tanks, Each vial was sealed with a crimp cap. These vials 3 and 7 days of depuration. The depuration system in- were stored at 4 1C until ready for analysis by SPME- cluded a sand ¢ltration and aeration. Collected ¢sh GC-MS using the same method as that described for were euthanized by farm personnel with a sharp blow the analysis of water samples. to the head immediately ¢lleted and the skin was also removed. Two ¢llets per ¢sh were obtained and placed in separate plastic (Ziploc, SC Johnson Canada, Brant- Sensory determination of ¢llet taste and ford, ON, Canada) bags. The ¢llets were then frozen aroma until they were shipped overnight for the o¡-£avour analysis to the United States Department of Agricul- The ¢llets (40 in total) were sent to the Garrison Sen- ture, Agricultural Research Service, Natural Pro- sory Evaluation Laboratory at Mississippi State Uni- ducts Utilization Research Unit (NPURU) at the Thad versity (Starkville, MS, USA) for sensory analysis by a Cochran Research Center in University,Mississippi. trained panel. The samples were removed from the

r 2010 The Authors Aquaculture Research r 2010 Blackwell Publishing Ltd, Aquaculture Research, 42,360^365 361 Geosmin is the dominant o¡-£avour compound in RAS SHouleet al. Aquaculture Research, 2011, 42,360^365 freezer and immediately wrapped and sealed in an Isolation of microorganisms in bio¢lm aluminium foil. The samples were prepared and samples cooked using AOAC method 976.16 for baking in foil. The analyses were performed at the NPURU,Thad Co- They were speci¢cally baked at 190.56 1Cinapre- chran Research Center in University,Mississippi. Bio- heated oven to an internal temperature of 71.11 1C. ¢lm samples were inoculated into di¡erent media After the samples were cooked, a small amount of (ASM, BG-11, Allen, BG-11and J) contained in 6 well the ¢llet was placed in a Te£onTM (Dupont,Wilming- plates and incubated for 2 weeks at 28 1C under cool- ton, DE, USA) sni¡ bottle. The remainder of the ¢llet white £uorescent lights. After 2 weeks, a sample from was left in the aluminium foil. The samples were pre- each well was centrifuged to concentrate the algae at sented to a group of six panellists, and the aroma and the bottom of the centrifuge tube and observed under taste were recorded on a10cm hedonic scale. a phase-contrast microscope to determine the pre- sence of cyanobacteria.When present, the cyanobac- teria were isolated using the manual direct isolation Analysis of geosmin and MIB levels in water technique. For isolation, the sample was observed and bio¢lm samples using a dissecting stereoscope (Olympus SZH10, CenterValley,PA, USA) and one ¢lament of the cyano- The analyses were performed at the NPURU, Thad bacteria was removed with a capillary tube that had Cochran Research Center in University, Mississippi. been separated in half over a Bunsen burner £ame. The method used to quantify the levels of geosmin The ¢lament was then placed in a well containing and MIB utilized the SPME procedure adapted from BG-11at 28 1C for a week. The isolated ¢laments were Lloyd et al. (1998). Aliquots of 600 mL were pipetted then observed under the microscope to determine individually into 2 mL glass screw-top vials with so- whether isolation had been achieved. Isolated cyano- dium chloride at the bottom (0.3 g vial À 1). The vials bacteria were identi¢ed using the reference book were heated at 40 1C for 20 min before the volatile Fresh Algae of North America: Ecology and Classi¢ca- compounds were absorbed onto a 100 mm poly- tion (Wehr & Sheath 2003). The isolated cyanobacter- dimethyl siloxane SPME ¢bre (Supelco, Bellfonte, PA, ial cultures were also tested for the production of USA). The ¢bre assembly was then shaken for 10min geosmin and MIB using SPME-GC-MS. during the absorption period and desorbed for 2 min at 250 1C in the injection port of an HP 6890 GC-MS (Agilent, Palo Alto, CA, USA) operated in the selected ion monitoring mode. The conditions of the gas chro- Isolation of actinomycetes matograph were as follows: (1) the initial oven tem- Aliquots (0.1mL) of water and bio¢lm samples were perature was 60 1C for 0.5 min; (2) followed by a spread-plated on yeast^dextrose agar and also on ac- À 1 ramp rate of 30 1Cmin to 100 1C; (3) followed by a tinomycete isolation agar plates.The plates were then À 1 ramp rate of 20 1Cmin to 300 1C with an isotherm incubated at di¡erent temperatures:14, 20 and 28 1C. time of 2 min; and (4) the maintenance of £ow pres- The plates were inspected after 2 days and at 7 days À 2 sure was at 18lb in with helium used as a carrier of incubation for the presence of actinomycete colo- gas. The molecular ion base peaks were monitored at nies. Suspected colonies were transferred to fresh m/z 168, 95 and 135 for MIB and at m/z 182, 112 and agar plates and restreaked to con¢rm isolation.These 126 for geosmin. The capillary column used was a ‘isolation’plates were inspected after 2 and 7 days of DB-5 (5%-phenyl-methylsiloxane, 30 m, 0.25 mm in- incubation for earthy or musty odour production. side diameter,0.25 mm ¢lm thickness; J&W Scienti¢c, Folsom, CA, USA). The retention time was 6.8 min for geosmin and 5.2 min for MIB. Standards for MIB and Statistical analysis geosmin were prepared at 0.1,0.5,1.0 and 2.5 mgLÀ 1 in deionized water. These standards were obtained The tanks were used as the experimental units from Wako Chemicals USA (Richmond,VA, USA) and (n 5 4). Partially hierarchical analysis of variance were included at the beginning, middle and end of with three factors [tank, treatment (grow-out or de- each group of samples analysed using a CombiPal puration) and the location of measure (¢llet, water Autosampler (Leap Technologies, Carrboro, NC, or bio¢lm)] (split-plot with replication) was per- USA). Each sample was run in triplicate (Schrader formed to detect di¡erences between treatments, fol- et al. 2003). lowed by Tukey’s multiple comparison test. The tank

r 2010 The Authors 362 Aquaculture Research r 2010 Blackwell Publishing Ltd, Aquaculture Research, 42,360^365 Aquaculture Research, 2011, 42, 360^365 Geosmin is the dominant o¡-£avour compound in RAS SHouleet al. was nested within treatment. The equality of var- from the grow-out tanks and the depurations tanks are iance of residuals was analysed using Levene’s test not signi¢cantly di¡erent, likely as a result that the ¢sh and normality of residuals using the Kolmogorov^ in the depuration tanks had only been depurated for 3 Smirnov test. To reach normality, MIB data were log and 7 days. Depuration requires more time for fatty ¢sh transformed and geosmin data were ranked follow- such as charr, and when held at a low temperature ing the general rank approach (Conover & Iman (Tucker 2000). Therefore, geosmin appeared to be the 1981). When no di¡erences were found between the main compound responsible for the ‘earthy’ o¡-£avour tanks and the treatment, linear regression analyses in the ¢llet. were used to determine the correlation between the Linear regression analysis con¢rmed a correlation geosmin levels in the ¢llets and in the water and also (correlation coe⁄cient r2 50.88) between geosmin between the geosmin levels in the ¢llets and the levels in the water and the geosmin levels in the ¢l- earthy-£avour intensity of the ¢sh. All statistical tests lets, although not signi¢cant (R 5 0.776, P 5 0.224; were performed using SAS version 9.1.3 (SAS Institute, Fig. 1). We acknowledge that the number of samples Cary,NC, USA) and SYSTAT version12 (Systat Software, used for the calculation of the linear regression is Chicago, IL, USA). low (n 5 4) but our approach involved working in a commercial operation environment (limited access to tanks) in order to provide the presence or absence Results and discussion of MIB and geosmin in conjunction with o¡-£avour The GC-MS results indicated the presence of geosmin expression in cultivated ¢sh in intensive recircula- in the ¢llets, water and bio¢lm samples (Table 1). tion systems, i.e. RAS. The fact that geosmin levels in Geosmin concentrations were higher in the samples the water and ¢sh are not highlycorrelated is not sur- taken from the bio¢lm on the side walls of the tanks prising. If cyanobacteria are the source of the geos- and in the Arctic charr ¢llets (bio¢lm:381 Æ 634 and min, the timing of cyanobacterial synthesis of 348 Æ 393 ng kg À 1 in grow-out and depuration geosmin, release of geosmin in the tank water from tanks respectively; ¢llet: 703 Æ 493 and 706 Æ senescing cyanobacterial cells, the uptake by the ¢sh 619 ng kg À 1 in grow-out and depuration tanks re- and the depuration time were probably not synchro- spectively; Table 1). Two-methylisoborneol was pre- nized (Grimm & Zimba 2003). The depuration of sent in ‘trace’ amounts in the ¢llets (8 Æ 8and 2 Æ 2ngkgÀ 1 in grow-out and depuration tanks re- 1100 spectively; Table 1) compared with geosmin levels. An ANOVA indicated that the values did not di¡er signi¢- 1000 cantly between the treatments (F1,2 50.123; P 5 0.759) 900 or the tanks (F1,4 50.377; P 5 0.708) for the geosmin le- 800 vels. The same results were obtained for MIB (treat- 700 ments: F 0.833; P 5 0.458; tanks: F 51.166; 1,2 1,4 fillet (ng/kg) P 5 0.399). The MIB does not appear to be the com- 600 pound responsible for the o¡-£avour in the charr, be- 500 Geosmin level in Arctic charr in Arctic Geosmin level cause the levels found in this study are well bellow the 400 detectable limit in rainbow trout (600 ng kg À 1)(Tucker 10 20 30 40 50 2000). The detectable limit for the geosmin in rainbow Geosmin level in tank water (ng/L) 1 trout ¢llet is 900 ng kg À (Robertson, Jauncey, Bever- Figure 1 Linear regression comparing geosmin levels in idge & Lawton 2005).The geosmin levels in the ¢sh ¢llet water and Arctic charr, r 5 0.77.

Table 1 Mean values of geosmin and MIB levels present in the ¢llets, water and bio¢lms from four tanks; two grow-out tanks and two depuration tanks,3 and 7 days of depuration

Fillet Water Biofilm

Ta n k s Geosmin (ng kg À 1)MIB(ngkgÀ 1)Geosmin(ngkgÀ 1)MIB(ngkgÀ 1)Geosmin(ngkgÀ 1)MIB(ngkgÀ 1)

Grow-out 703 Æ 493 8 Æ 837Æ 17 1 Æ 0.2 381 Æ 634 4 Æ 4 Depuration 706 Æ 619 2 Æ 224Æ 17 2 Æ 3 348 Æ 393 3 Æ 4

MIB, 2-methylisoborneol.

r 2010 The Authors Aquaculture Research r 2010 Blackwell Publishing Ltd, Aquaculture Research, 42,360^365 363 Geosmin is the dominant o¡-£avour compound in RAS SHouleet al. Aquaculture Research, 2011, 42,360^365 earthy and musty o¡-£avour compounds by ¢sh is 1100 slow compared with their uptake (Tucker 2000). If 950 the main route for geosmin accumulation in the 800 charr £esh is via passive di¡usion across the gills from the water, bioconcentration of geosmin in the 650 £esh is occurring because the geosmin levels in the

fillet (ng/kg) 500 water were signi¢cantly lower than those detected in the ¢sh £esh. However, until the source of geosmin 350 Geosmin level in Arctic Arctic charr Geosmin level in in the RAS is identi¢ed, geosmin tank-¢sh dynamics 200 cannot be further discussed here. 45678 The ¢llets were also analysed by a sensory panel to Taste test detect and rate the intensity of all types of o¡-£a- Figure 2 Linear regression comparing geosmin levels in vours. Each of the six panellists rated the ¢llets on a Arctic charr with the sensory analysis (rating), R 5 0.96. hedonic scale of 1.2^9.4 (no o¡-£avour taste 51.2; 9.4 5 strongest o¡-£avour), and the scores were aver- aged (Table 2).The o¡-£avour that the panellists were Planktonic cyanobacteria species are commonly expected to detect was earthy because the presence attributed as the causes for the musty and earthy of geosmin had previously been veri¢ed instrumen- o¡-£avours in commercial channel cat¢sh (Ictalurus tally. The panellists also detected other o¡-£avours punctatus) cultured in ponds in the southeastern in the charr that could be in£uencing the taste and USA, and examples include Oscillatoria perornata,a aroma values (Table 3). The sensory data related to producer of MIB, and Anabaena spp., producers of earthy-muddy intensity were analysed by linear re- geosmin (Tucker 2000). In our study,these genera of gression to determine whether it showed a relation- cyanobacteria were not observed in the water and ship with the instrumental analytical results. The bio¢lm samples. In addition, several other cyanobac- coe⁄cient of determination (r) was 0.96 and indi- teria species isolated from the bio¢lm samples were cated a signi¢cant relation (P 5 0.036) between the determined not to produce geosmin. However, a cya- sensory results and the instrumental results for the nobacterium identi¢ed to be a species of Pseudana- presence of geosmin (Fig. 2). Conclusively, the sen- baena (isolated from a bio¢lm sample) was con¢rmed sory analysis helped con¢rm that geosmin was re- to be a MIB-producer. Because the main o¡-£avour in sponsible for the earthyo¡-£avour in theArctic charr. the charr was earthy and identi¢ed to be due to the presence of geosmin in the ¢sh £esh, this Pseudana- baena sp. is not considered to be a contributor to the Table 2 Mean values of the sensory results for earthy o¡- earthy o¡-£avour problem. However, the isolation of £avour in the Arctic charr ¢llets an MIB-producing species of cyanobacteria from the recirculating system provides evidence for the poten- Earthy tial existence of other odour-producing cyanobacter- Ta n k s (sensory rating) ia as in situ sources for other types of o¡-£avours in

Tank 1 (grow-out) 6.916 similar types of indoor culture systems. No odour- Tank 2 (grow-out) 4.956 producing actinomycetes were detected in any of the D1 (depuration 3 days) 7.976 water or the bio¢lm samples, and so they are also not D2 (depuration 7 days) 5.096 believed to be the primary cause of the o¡-£avour problem. At this time, the main producer of geosmin Table 3 Other o¡-£avours detected by the trained sensory in these tanks is still unknown, but we believe cer- panellists tain species of cyanobacteria that were abundant in the bio¢lm layers on the inside of the tanks to be the Off-flavour Number of fillets likely source.

Metallic 12 Our attempts to isolate a geosmin-producing spe- Bitter 3 cies microorganism have relied on conventional cul- Amoniac 2 tivation techniques so far. The inability to isolate a Cat box 1 geosmin-producing bacterial species can be attribu- Umami 1 ted to one of the following: (1) the culture media use Mud 2 was inappropriate for culturing the particular geos-

r 2010 The Authors 364 Aquaculture Research r 2010 Blackwell Publishing Ltd, Aquaculture Research, 42,360^365 Aquaculture Research, 2011, 42, 360^365 Geosmin is the dominant o¡-£avour compound in RAS SHouleet al. min-producing microorganism or (2) the presence of water aquaculture. In: O¡-Flavors in Aquaculture (ed. by a geosmin-producing bacterial species in the bio¢lm A.M. Rimando & K.K. Schrader), pp. 209^222. American and possibly the bio¢lter that entered a viable but Chemical Society,Washington, DC, USA. nonculturable state. Future studies may include the Grimm C.C., Lloyd S.K., Zimba P.V. & Palmer M. (1999) Ana- use of speci¢c oligonucleotide probes (e.g.,16S rRNA) lysis of 2-methylisoborneol and geosmin in cat¢sh by mi- to help characterize and identify potential geosmin- crowave distillation-solid-phase microextraction. Journal of Agricultural and Food Chemistry 47,164^169. producing bacteria in the bio¢lter and RAS tank-wall Howgate P.(2004) Tainting of farmed ¢sh by geosmin and 2- bio¢lms (Kalmbach, Manz & Szewzyk1997). methyl-iso-bomeol: a review of sensory aspects and of up- take/depuration. Aquaculture 234,155^181. Kalmbach S., Manz W. & Szewzyk U. (1997) Isolation of new Conclusions bacterial species from drinking water bio¢lms and proof of their in situ dominance with highly speci¢c 16S rRNA This research is the ¢rst report identifying an earthy probes. Applied and Environmental Microbiology 63,4164^ o¡-£avour in Arctic charr cultured in RAS. In our 4170. study, we con¢rmed the presence of geosmin in the Le Francois N.R., Lemieux H. & Blier P.U. (2002) Biological ¢sh £esh as the cause of the earthy o¡-£avour in the and technical evaluation of the potential of marine and charr. Although the sources of geosmin in the RAS anadromous ¢sh species for cold-water mariculture. were not identi¢ed, preliminary evidence indicates Aquaculture Research 33,95^108. microorganisms found in the bio¢lm layers on the Lloyd S.W., Lea J.M., Zimba P.V. & Grimm C.C. (1998) Rapid tank walls to be one potential source. analysis of geosmin and 2-methylisoborneol in water using solid phase micro extraction procedures.Water Re- search 32, 2140^2146. Robertson R.F., Jauncey K., Beveridge M.C.M. & Lawton L.A. Acknowledgments (2005) Depuration rates and the sensory threshold con- We would like to thank West Virginia Aqua for their centration of geosmin responsible for earthy-muddy taint help in the sampling, Ramona Pace of the Thad Co- in rainbow trout, Onchorhynchus mykiss. Aquaculture chran Research Center in Oxford, Mississippi, for the 245,89^99. technical help and support and the Garrison Sensory Robin J., Cravedi J.P., Hillenweck A., Deshayes C. & Vallod D. (2006) O¡ £avor characterization and origin in French Evaluation Laboratory at Mississippi State University, trout farming. Aquaculture 260,128^138. Starkville, MS. The technical assistance of the Labor- Schrader K.K. & Dennis M.E. (2005) Cyanobacteria and atoire de recherche des sciences aquatiques, Univer- earthy/musty compounds found in commercial cat¢sh site¤Laval and the help of Alain Caron of the (Ictalurus punctatus) ponds in the Mississippi Delta and Universite¤du Que¤bec aØRimouski for statistical analy- Mississippi^Alabama Blackland Prairie. Water Research sis is gratefully acknowledged. This work was sup- 39,2807^2814. ported by the le Fonds que¤be¤cois de la recherche sur Schrader K.K., Dhammika Nanayakkara N.P., Tucker C.S., la nature et les technologies. Rimando A.M., Ganzera M. & Schaneberg B.T. (2003) Novel derivatives of 9,10-anthraquinone are selective algicides against the musty-odor cyanobacterium Oscilla- References toria perornata. Applied and Environmental Microbiology 69, 5319^ 5327. Conover W.J. & Iman R.L. (1981) Rank transformations as a Tucker C.S. (2000) O¡-£avor problems in aquaculture. bridge between parametric and nonparametric statistics. Reviews in Fisheries Science 8, 45^88. TheAmerican Statistician 35,124^128. Wehr J.D. & Sheath R.G. (eds) (2003) Freshwater Algae of Grimm C.C. & Zimba P.V. (2003) Applications of an instru- North America: Ecology and Classi¢cation. Academic Press, mental method for the analysis of o¡-£avors in fresh San Diego, CA, USA,950pp.

r 2010 The Authors Aquaculture Research r 2010 Blackwell Publishing Ltd, Aquaculture Research, 42,360^365 365