Introduction

Effective cooling prolongs the shelf life of food and fish in particular. Different attempts have been approached to delay the spoilage of fresh fish, using superchilling techniques [1], modified atmosphere packaging [2], a combination of both [3, 4] or even radiation [5]. Changes of environmental conditions will delay or even stop the typical spoilers from growing, but at the same time, a new niche is opened, allowing the onset of other bacteria. Because of the short shelf life of fish, traceability and information on quality parameters are of great importance during trading of fish. It has been shown before that the correlation between the quality as judged by sensory evaluation and total bacterial counts is not as accurate when compared to the levels of specific spoilage organisms (SSO) [6, 7]. Knowledge on how the microbiota will respond to different cooling methods can help to better understand the spoilage process and design cooling methods that can prolong shelf life, as well as to improve bacterial growth models and pinpoint the most relevant biosensors for evaluation of fish freshness.

Cooling technologies have been improving in recent years with the development and commercialisation of slurry ice machines of different kinds where various mixtures of water, ice, salt and gases have been used to increase cooling rate and optimise temperature during transportation. All these developments aim at reducing bacterial growth and thereby extending the shelf life and increasing the quality of the product [8]. Several different slurry ice production technologies and a number of slurry ice brands exist (e.g., Liquid Ice, Flow Ice, Optim-Ice, Fluid Ice and Bubble Slurry Ice). The thermal and rheological properties along with other characteristics of ice slurries have been thoroughly described [9]. Information on how these new techniques affect the succession of the microbiota present on the skin and flesh of the fish is limited at present.

Examples of well-documented fish spoilage organism are members of the Pseudomonas genus and Shewanella putrefaciens which are frequently used as bacterial indicators for spoilage and are present in different food types [10–12]. Photobacterium phosphoreum has not been investigated as intensively but there are scientific evidences that it is the main spoilage organism when environmental conditions are favourable such as in chilled modified atmosphere packed fish [4, 13, 14]. Other workers have reported its presence in aerobically spoiled fish [15, 16] and its importance in the spoilage of air-stored gadoid fish products was recently reported [1, 17].

Since the emergence of molecular biology and PCR-based strategies, a completely new vision on microbial diversity has been established. These methodologies have shown the presence of uncultivable populations and lateral gene transfer between species which has greatly complicated the picture for microbiological evolution and speciation of bacteria. These new insights are in large proportions from the ocean microbiota which is likely to influence the bacterial composition on any fish species living in the same surroundings [18, 19].

The aim of the present study was to investigate the practical application of four different cooling methods available to the fish industry for whole fish storage with regard to their cooling performance and effects on fish deterioration using conventional methods such as microbial enumeration by plate counting and physicochemical analysis. Furthermore, the treatments’ effect on bacterial community changes during storage was evaluated using a cultivation-independent method (16S rRNA clone analysis).

Materials and methods

Experimental setup

Four cooling treatments were evaluated during the 8-day storage trial of whole haddock in October 2008: (1) Plate ice (PI) with a top layer of plate ice (PI + PI) constituted 100 kg of fish layered with 45 kg of plate ice (with a top layer of ice) in a drained tub (310 L); (2) liquid ice A with a top layer of ice (LIA + PI) prepared with 100 kg of fish layered in 100 kg of liquid ice A (ice ratio = 34%); (3) liquid ice B with no top layer of ice (LIB), with 100 kg of fish layered in 100 kg of liquid ice (ice ratio 47%); (4) LIB + PI prepared as for LIA + PI but using liquid ice B. The tubs in treatments 2, 3 and 4 were chilled for 2 h, undrained and then drained followed by top icing (12 kg PI) for treatments 2 and 4. Haddock was caught in Hvalfjordur using a longline, gutted and rinsed upon landing and was received at the laboratory almost un-iced 12-h post-catch.

Temperature logging

In each of the four tubs comprising five layers of fish, temperature loggers were inserted into the fish flesh in four individuals weighing approximately 600–1,300 g, positioned at the bottom, middle and top layers, resulting in 12 loggers recording the fish flesh temperature every 5 min for each treatment. These loggers were iButton type DS1922L (Maxim Integrated Products, Sunnyvale, CA, USA) with an accuracy of ±0.5 °C, a resolution of 0.0625 °C and an operating range of −40–85 °C. In addition to those loggers, one temperature logger was used to measure ambient temperature outside each tub and another one to record the medium temperature, approximately 10 cm below surface. These were Onset temperature loggers, type UTBI-001 (Onset Computer Corporation, Pocasset, MA, USA), with an accuracy of ±0.2 °C, a resolution of 0.02 °C and an operating range of −20–70 °C.

Sampling of whole, gutted fish

At initial sampling (day 0), four fish were removed aseptically from the fish tubs, pooling two fish per sample. During subsequent sampling (d2-4-8), four fish were removed aseptically from the top layer of the differently treated tubs, pooling two fish per sample for each treatment examined. Fish were transferred to the laboratory in expanded polystyrene foam (EPS) boxes and analysed within 30–60 min.

Cultivation-dependent microbial analyses

Microbiological counts were evaluated for both the skin and the flesh of haddock. Two pieces of skin were aseptically cut from one side of each fish (two fish = one sample), giving a total area of 30 cm2, and mixed with 60 mL of cooled Maximum Recovery Diluent (MRD, Oxoid, Hampshire, UK) for 1 min in a stomacher (Stomacher Lab Blender 400, A.-J. Seward Laboratories, London, UK). Successive tenfold dilutions were done as required. The other side of each fish was aseptically skinned, pieces of flesh removed, minced and diluted tenfold in cooled MRD. Cultivation methods used are described by Olafsdottir et al. [1]. Presumptive Pseudomonas were cultivated on modified Cephaloridine Fucidin Cetrimide (CFC, 22 °C) agar (Oxoid) based on Stanbridge and Board [20], H2S-producing bacteria on Iron Agar (IA, 17 °C), Photobacterium phosphoreum with Malthus conductance method [21] and total viable psychrotrophic counts (TVC) on Long and Hammer’s medium (LH, 17 °C). All samples were analysed in duplicate and results presented as an average. Microbiological quality of freshly prepared ice media (PI, LIA, LIB) was evaluated on LH, IA and CFC following serial dilution of melted samples.

Chemical analyses

The minced flesh samples prepared for microbiological analyses were used. The method of Malle and Tao was used for total volatile bases (TVB-N) and trimethylamine (TMA) measurements. TVB-N was measured by steam distillation (STRUERS, Copenhagen, Denmark) and titration, after extracting the fish mince with 7.5% aqueous trichloroacetic acid solution. The distilled TVB-N was collected in boric acid solution and then titrated with sulphuric acid solution. TMA was measured in trichloroacetic acid (TCA) extract by adding 20 mL of 35% formaldehyde, an alkaline binding mono- and diamine, TMA being the only volatile and measurable amine. All chemical analyses were done in duplicate.

The pH measurements were performed in 5 g of mince moistened with 5 mL of deionised water. Salt (NaCl) content was measured with the Volhard Titrino method according to AOAC ed. 17 from 2000 (no. 976.18).

DNA extraction for cultivation-independent analysis

Upon sampling, tenfold diluted fish samples were collected and kept at −20 °C until DNA extraction. Prior to extraction, the duplicate samples of each treatment were pooled and 1 mL (0.1 g) of this preparation was used for DNA extraction. Raw material (day 0), four skin samples (day 8) and four flesh samples (day 8) were selected for 16S rRNA clone analysis. Template genomic DNA was isolated as described before [22].

Cultivation-independent bacterial analysis

The 16S rRNA analysis was performed in order to determine the dominating bacterial groups in early and late storage and to obtain a separate measure on bacterial succession than cultivation alone. PCR reaction was done by amplifying the 16S rRNA gene with universal primers, 9F and 805R (5′-GAGTTTGATCCTGGCTCAG-3′ and 5′-GACTACCAGGGTATCTAATCC-3′, respectively). PCR conditions, cloning and sequencing of PCR products obtained from the fish samples were performed essentially as described before [23]. From each pooled sample, a total of 48 clones were picked for sequencing resulting in 384 sequencing reactions in total. The results were analysed with Sequencer 4.8, built 3767 (Gene Codes Corporation, Ann Arbor, MI, USA) and identified with BLAST search through NCBI 16S rRNA database.

Statistical analysis

Statistical analysis of microbial and chemical data was carried out with NCSS 2000 software (UT, USA) using analysis of variance methods (Kruskal–Wallis one-way ANOVA on ranks for salt content and one-way ANOVA for all other variables). Comparison of data with respect to treatments was performed using the Kruskal–Wallis multiple comparison z-value test and Fisher’s LSD multiple comparison test, respectively. The threshold for significance was 0.05.

Results

Temperature monitoring

The treatments including liquid ice B provided slightly faster cooling than the other two media tested, but they did not maintain the fish temperature as well throughout the 8-day storage period. In those LIB groups, 0 °C was reached in 21–28 min, but 57 min with liquid ice A. When plate ice was used solely, it took 262 min to reach 0 °C (Fig. 1). The average fish temperature in the different cooling treatments ranged from −0.09 to −0.64 °C (Table 1). The temperature in the PI + PI and LIA + PI increased by approximately 0.2 °C during the storage period and by 0.8 and 0.9 °C with LIB and LIB + PI, respectively (Fig. 1). No significant difference was found in the mean ambient temperature between groups (data not shown). During the 8-day storage, the temperature fluctuated between 2 and 5 °C for all the groups with a mean temperature of 3.4 ± 0.8 °C.

Fig. 1
figure 1

Maintenance of temperature—mean temperature of all layers (initial fish temperature = 7.5 °C) over the 8-day storage period in the differently treated fish. Samples were stored in plate ice with a top ice layer (PI + PI), liquid ice A with a top ice layer (LIA + PI), liquid ice B with (LIB + PI) or without (LIB) a top ice layer

Full size image
Table 1 Measured parameters of each cooling treatment during the storage experiment
Full size table

Effects of cooling treatments on physicochemical changes in haddock flesh

Measurements of pH in haddock flesh throughout the storage time showed little variation and were constantly in the range of pH 6.5–6.7 (data not shown). Initial NaCl content of fish flesh was found to be 0.15 ± 0.07%, while the salt content of the cooling media ranged from 0 to 2.9% (Table 1). After 2 and 4 days of storage, a slight salt uptake was noticed, but after 8 days, a significant (p < 0.05) increase was found in all liquid ice treatments compared to the NaCl content of the raw material (Fig. 2). Initial TVB-N content in the raw material amounted to 11.9 ± 2.8 mg N/100 g flesh. After 8 days of storage, a significant (p < 0.05) increase in TVB-N and TMA was found for liquid ice treatments compared to traditional icing (PI), reaching the highest levels when an ice top layer was not applied (LIB) (Fig. 2).

Fig. 2
figure 2

Production of volatile basic compounds and P ratio (%, TMA/TVB-N) in haddock flesh after the 8-day storage period as influenced by the cooling treatments. Fish stored in plate ice with a top ice layer (PI + PI), liquid ice A with a top ice layer (LIA + PI), liquid ice B with (LIB + PI) or without (LIB) a top ice layer. Different letters indicate significant differences among cooling treatments (p < 0.05)

Full size image

Microbial changes during storage using cultivation-dependent methods

Analysis of freshly prepared ice media indicated the higher microbial contamination of LIB and plate ice compared to LIA (Table 1). Skin counts were assessed initially on the raw material used, resulting in an overall psychrotrophic load (TVC) of log 3.1 ± 0.1 colony-forming units (cfu)/cm2. P. phosphoreum (Pp, log 2.9 ± 0.3/cm2) was found to dominate among the SSO assessed (H2S-producers, log 2.2 ± 0.3 cfu/cm2; pseudomonads, log 0.8 ± 0.3 cfu/cm2). The growth behaviour of these SSO was influenced by the cooling treatments, as shown in Fig. 3. The least effect was observed during the first 4 days of storage for pseudomonads which grew steadily, independently of the cooling methods used. However, after 8 days of storage, pseudomonad counts were significantly (p < 0.05) lower on fish cooled with LIB + PI than LIB alone or LIA + PI. Ice storage performed as well as LIB + PI in slowing down pseudomonads. Pp growth was delayed for at least 4 days by all cooling methods, after which a significant growth increase was seen on fish stored in LIB compared to other treatments (p < 0.05). A similar but shorter delay was seen for H2S producers, followed by significant growth in LIB-cooled fish. Overall, the psychrotrophic skin microbiota developed significantly faster on LIB-stored fish than in other treatments.

Fig. 3
figure 3

Development of microbiota on haddock skin as influenced by the cooling treatments. Fish stored in plate ice with a top ice layer (PI + PI), liquid ice A with a top ice layer (LIA + PI), liquid ice B with (LIB + PI) or without (LIB) a top ice layer. Different letters indicate significant count differences among cooling treatments on respective sampling day (p < 0.05)

Full size image

Since the flesh of a newly caught fish is generally known to be sterile, the flesh at day 0 was not analysed. Psychrotrophic flesh counts were slightly influenced by the cooling treatments on day 2, but no significant differences were seen for TVC and pseudomonad counts among the four treatments as time progressed (Fig. 4). P. phosphoreum growth in fish flesh was delayed by all cooling treatments for at least 4 days, coinciding with the results for skin counts (Figs. 3, 4). After 8 days of storage, slowest Pp growth was seen in LIB + PI and PI + PI stored fish but fastest growth was observed in LIB-stored fish with no ice on top. At that time point, H2S producers had reached lowest levels (log 2.4 cfu/g) in LIA + PI stored fish, but highest (log 3.3–3.9 cfu/g) in the other groups (Fig. 4).

Fig. 4
figure 4

Development of microbiota in haddock flesh as influenced by the cooling treatments. Fish stored in plate ice with a top ice layer (PI + PI), liquid ice A with a top ice layer (LIA + PI), liquid ice B with (LIB + PI) or without (LIB) a top ice layer. Different letters indicate significant count differences among cooling treatments on respective sampling day (p < 0.05); ns not significant

Full size image

Bacterial community changes during storage using a cultivation-independent method

From all the samples analysed, a total of 64 different species were detected in 348 successful sequencing reactions (Table 2). The skin bacteria of newly caught, gutted and rinsed haddock showed a highly diverse population with 26 species and 15 genera with no single dominating species (Fig. 5). Vibrio, Acidobacteria and Mycoplasma sp. showed somewhat higher relative abundance than others. The largest group is a collection of sequences that corresponded to uncultured species and could therefore not be identified. After 8 days of storage in different cooling treatments, large shifts in bacterial dominance were observed with reduced diversity when compared to the skin sample on day 0. Generally, a good agreement was observed between the skin and flesh microbiota composition for each treatment. Storage in plate ice (PI + PI) resulted in the lowest bacterial diversity compared to the other groups where Psychrobacter was the dominating species along with Flavobacterium to a lesser extent. The proportions of these bacteria were inverted in the LIB + PI group, where Flavobacterium became dominant at the expense of Psychrobacter, while the total bacterial numbers remained similar between the groups (Figs. 2, 5). A different scenario was observed in both LIA + PI and LIB groups where Photobacterium was most apparent but several other genera were also detected (Fig. 5). The LIB group, however, contained highest number of bacteria according to the plate count enumeration and highest relative abundance of P. phoshphoreum (Fig. 2). No results are presented for the skin sample from the LIA + PI group since PCR amplification was not successful.

Table 2 Predominant species identified in the pooled haddock samples by 16S rRNA gene analysis
Full size table
Fig. 5
figure 5

Relative distribution of bacterial populations in whole, gutted haddock sampled as a raw material prior to storage and after 8 days of storage in tubs using different cooling media. Skin bacteria (d0 for initial sample and d8 for 3 treatments) is shown to the left and flesh bacteria (d8) to the right for each treatment. Fish stored in plate ice with a top ice layer (PI + PI), liquid ice A with a top ice layer (LIA + PI), liquid ice B with (LIB + PI) or without (LIB) a top ice layer

Full size image

Discussion

The present study shows the effects of different cooling media on selected physicochemical and microbiological parameters of whole, gutted haddock over an 8-day period. It reveals the sensitivity and dynamics of bacterial populations that develop in fish during storage at subzero temperatures and how minor changes in storage conditions can impact the microbiota in this environment.

The results show that liquid ice provides faster cooling than plate ice alone. However, a top layer of plate ice was found necessary to maintain better the superchilled condition established in liquid ice treatments during extended storage. The average fish temperature in the LIB + PI tub raised more rapidly (0.11 °C day−1) between days 1 and 8 compared to the LIA + PI treatment (0.04 °C day−1) which could be explained by the larger ice crystals of LIA than LIB, hence maintaining better cooling and delaying the melting process. Despite the temperature difference observed between plate ice and liquid ice treatments, no significant difference was observed in skin microbial load among these treatments after 2 days of storage. However, significantly faster SSO growth was observed on LIB fish after 4 and 8 days of storage, coinciding with the ongoing average temperature increase seen in the fish. After 8 days, no or little difference was seen among the other treatments.

As evidenced in our study, SSO growth was not only temperature dependent, but was apparently influenced by the environmental conditions established in fish stored using liquid ice treatments. On day 8, the salt content had doubled to tripled in LI-stored fish compared to iced fish and may explain the more rapid development of P. phosphoreum under these conditions. This increased salt uptake might be explained by the 2-h undrained cooling period during initial storage of LI treatments, causing salt deposition on the fish skin. Sodium chloride lowers the freezing point of flesh waterphase, hence protecting bacteria to some extent, and because P. phosphoreum is a sodium-requiring bacterium [10], it can be expected that increasing salt concentration in haddock flesh stimulated its growth. A recent study, reporting on the effect of a 2-day brining immersion treatment of cod fillets prior to air-packaging, showed that shelf life of superchilled brined products was lesser than that of unbrined fish, explained by slightly faster SSO growth and enhanced production of basic volatile compounds [4, 24]. Indeed, TVB-N and TMA increase in flesh of LIB fish was highest, coinciding with the SSO proliferation described. No increase in TVB-N was observed in iced fish after 8 days of storage, in agreement with the undetected level of TMA, indicating a much slower spoilage process in conventionally iced haddock.

The impact of the cooling treatments on the quality preservation of whole, gutted haddock, as studied by the evaluation of microbial and chemical spoilage indicators, was well reflected by the evolving skin and flesh bacteria in the differently treated fish. The 16S rRNA clone analysis showed considerable differences in the developed bacterial population after 8-day storage among the different cooling treatments. The liquid ice treatments maintained some of the heterogeneity of the skin bacteria, dominating in Photobacterium, Shewanella, Psychrobacter and Flavobacterium among others, while iced fish (PI + PI) microbiota was restricted to the three latter genera. Photobacterium was not recovered from the skin of iced fish. Generally, the transition of skin bacteria into the flesh of iced fish was mostly proportional where microbial composition of both skin and flesh was comparable. At end of storage, the PI + PI fish flesh was dominated by Psychrobacter (76%) and Flavobacter (11%) to a lesser extent. These proportions were interestingly inverted in the LIB + PI group where Flavobacter showed 61% dominance and Psychrobacter 34%. The bacterial flora of LIA + PI and LIB was dominated by Photobacterium (45 and 75%, respectively) which is in agreement with cultivation and TMA formation. Whether these differences in bacterial community structure are strictly the consequence of different ice properties or contamination of ice media cannot be ascertained. The ice media used in the present study had different bacterial loads (log 2.6–5.8 cfu/mL) which can clearly affect microbial developments and reduce shelf life. Previous investigations have shown the presence of Salinibacter in high dominance in liquid ice as well as incidence of different bacteria such as Acinetobacter, Psychrobacter, Arcobacter and Flavobacter in other samples (unpublished results). In our study, the different ice media were prepared from the same water source, so chemical analysis of the media was restricted to their salt content, being the most influential factor for bacterial growth behaviour in the fish. The capability of liquid ice to rapidly lower fish temperature to a subzero temperature justifies its use for fish cooling. The ultimate aim is to reduce bacterial growth as much as possible to delay fish deterioration and therefore the salubrity of the ice medium is essential to minimise risks of bacterial contamination. It has been shown that SSO growth in air-stored gadoid fish fillets is influenced by temperature, especially P. phosphoreum being enhanced under temperature abuse but delayed at superchilled conditions (−1 °C). H2S-producing bacteria behave similarly but to a lesser extent, while pseudomonads are less influenced [1, 17]. Other studies have shown the superior performance of slurry ice storage (−1 to −1.5 °C) when compared to ice alone (0–1 °C) during storage of various fish species resulting in lower microbial loads, higher score in sensory analysis and shelf life extension at superchilled conditions [25–27]. Furthermore, the use of ozonated slurry ice increased the storage performance of farmed turbot and sardine even further [28, 29]. Salt uptake was not evaluated in these studies and could not be detected by the sensory method used, but was reported during slurry ice storage of horse mackerel [30]. The bacterial developments occurring in fish stored in slurry ice are likely to be species specific where gadoid and other marine fishes containing TMAO might be more susceptible for P. phosphoreum growth and related deteriorative changes. In the present study, it is suggested that the more salty and wet environment created by the liquid ice may have contributed to the transition of P. phosphoreum from the skin to the flesh and influenced its development despite this superchilled condition, which is supported both by cultivation-dependent and cultivation-independent analysis. It has been shown by another research group that ozonated slurry ice did not extend the shelf life of MA-packed ozonated cod fillets due to prevailing levels of P. phosphoreum [31].

In conclusion, this study has shown the effects of different commercial cooling media on quality parameters on whole, gutted haddock. Furthermore, the influence of the cooling media on the bacterial species composition was monitored and showed a significant divergence of bacterial developments explained by the environmental conditions caused by the cooling treatments applied to the fish. The use of liquid ice is in many cases the most suitable cooling medium for whole fish to achieve faster cooling. One drawback of liquid ice is its stimulating growth effect on P. phosphoreum during extended storage of marine coldwater fish as a consequence of increased salt concentrations although draining the tubs during initial cooling might minimise this effect.