Unlocking the microbial diversity and the chemical changes throughout the fermentation process of “hákarl”, Greenland shark

Abstract

Hákarl is a unique traditional Icelandic product and is obtained by fermenting and drying Greenland shark (Somniosus microcephalus). However, little is known about the chemical and microbial changes occurring during the process. In this small-scale industrial study, fresh and frozen shark meat was fermented for eight and seven weeks, respectively, and then dried for five weeks. During the fermentation, trimethylamine N-oxide levels decreased to below the limit of detection within five weeks and pH increased from about 6 to 9. Simultaneously, trimethylamine and dimethylamine levels increased significantly. Total viable plate counts, and specific spoilage organisms increased during the first weeks of the fermentation period but decreased during drying. Culture-independent analyses (16S rRNA) revealed gradual shifts in the bacterial community structure as fermentation progressed, dividing the fermentation process into three distinct phases but stayed rather similar throughout the drying process. During the first three weeks of fermentation, Photobacterium was dominant in the fresh group, compared to Pseudoalteromonas in the frozen group. However, as the fermentation progressed, the groups became more alike with Atopostipes, Pseudomonas and Tissierella being dominant. The PCA analysis done on the chemical variables and 16S rRNA analysis variables confirmed the correlation between high concentrations of TMAO and Pseudoalteromonas, and Photobacterium at the initial fermentation phase. During the final fermentation phase, correlation was detected between high concentrations of TMA/DMA and Atopostipes, Pseudomonas and Tissierella. The results indicate the possibility to shortening the fermentation period and it is suggested that the microbial community can potentially be standardized with starter cultures to gain an optimal fermentation procedure.

Highlights

• •Bacterial counts reached a maximum after 2–3 weeks of fermentation.
• •TMAO levels decreased early in the process while TMA and DMA increased.
• •TMAO reducing bacteria were dominant in the initial phase of fermentation.
• •Bacteria using TMA, urea and ammonia were abundant in the mid fermentation phase.
• •Ready-to-eat hákarl frozen before fermentation contained less TMA and DMA.

Abbreviations

ASV

Amplicon sequence variant

DMA

Dimethylamine

HPLC

High performance liquid chromatography

LC-MS/MS

Liquid chromatography triple quad mass spectrometry

LOD

Limit of detection

LOQ

Limit of quantification

MAP

Modified atmospheric packaging

PCA

Principal component analysis

RTE

Ready-to-eat

SSO

Specific spoilage organism

TMA

Trimethylamine

TMAO

Trimethylamine oxide

TVC

Total viable counts

VOC

Volatile organic compound

1. Introduction

Fermentation represents one of the most ancient techniques of food preservation and is mainly based on metabolic activities of microorganisms that are either naturally present in the raw material or artificially inoculated [[1], [2], [3]]. Fermentation of Greenland shark has a long tradition in Iceland, and the oldest documentation of processed shark in Iceland is from the year 1374 [4]. The ready-to-eat (RTE) final product is called hákarl, and was for centuries a valuable source of protein and energy for the Icelandic population [5]. Fresh or unprocessed meat of Greenland shark is considered toxic if consumed excessively due to high content of ammonia containing compounds such as urea and trimethylamine-N-oxide (TMAO) [6]. During the fermentation, urea is converted into ammonia and the TMAO into TMA and hence the fermentation process has been considered to be not only a preservation process but also a detoxification process [6]. TMA and TMAO are found in all prokaryote and eukaryote organisms as well as in the environment, and have diverse biochemical and biological functions [7]. TMAO is a natural osmolyte used by marine animals living at great depths to stabilize against pressure and urea denaturation [8]. Although the fermented shark is still eaten in Iceland, the consumption habits have changed as the fish is now consumed as a snack or delicacy, primarily as part of the food served during the “Thorri” festival. Since fermented Greenland shark has only been produced and consumed in Iceland, this unique food is also an item of curiosity for visiting tourists and is frequently referred to as the national dish of Icelanders. The traditional production of hákarl has been described previously [9,10] and consists of two steps, fermentation and drying. In the past, the cut-up shark was fermented in gravel pits often close to the sea in such a way that the seawater washed over it at high tide. The fish was “buried” by covering the pits with stones, seaweed, or turf and left for several weeks or months [5]. After the fermentation period, the shark cuts were dried outdoors in drying sheds. The process has remained largely unchanged to this day, although nowadays the shark cuts are mainly fermented indoors in perforated containers, allowing the liquid to drain off. Furthermore, the shark meat is also sometimes kept frozen for some time before the fermentation is initiated. The fermentation step may take between six and twelve weeks depending on the temperature and season. Similarly, the outdoor drying varies from several weeks to months.

Even though this process is traditional and has been practiced for centuries, little is still known about the chemical and microbial changes that occur during the whole production process, and how the raw material and different processing parameters affect the quality and characteristics of the final RTE product. Only one study has been conducted on the chemical and microbial changes that occur during the whole production period [11]. In that study, microbial analyses were carried out by cultivations and TMA, ammonia, fat, dry matter content and pH were measured. The study revealed that during the fermentation process (35 days) urea was degraded into ammonia by bacterial enzymes which raised the pH in the shark flesh from 6 to 9 in the final product. The total number of bacteria in the shark flesh increased significantly during the fermentation process where Moraxella/Acinetobacter predominated along with Lactobacillus and on iron agar, the total counts reached about 10⁸/g. Furthermore, at the end of the drying process (70 days), the number of bacteria and the amount of ammonia and TMA had decreased. The water content in the shark flesh decreased during the fermentation process and especially during the drying process from 75 % to 60 % and was about 45 % in the final product. Moreover, a few studies have been carried out on lab scale RTE hákarl where the microbiome, pH, nitrogen and ammonia containing compounds were examined [[12], [13], [14]]. These studies show that RTE hákarl produced from different production facilities vary in terms of their chemical composition [12,14]. PCR-DGGE analysis and 16S rRNA amplicon sequencing have also revealed that RTE hákarl harbors diverse microbial communities which are mostly dominated by alkaliphilic and halophilic bacteria such as Tissierella, Oceanobacillus and Abyssivirga. Additionally, diverse mycobiome occupy RTE hákarl where Debarymyces, Candida and Saccharomyces dominate [13].

The aim of this study was to investigate the complex chemical and microbiome changes that occur during the traditional small-scale industrial fermentation and drying of Greenland shark. To evaluate whether the processing time might be decreased and what impact the initial handling of the raw material may have on the RTE final product, fresh and thawed meat cuts from the same shark were compared throughout 13 weeks of processing time. The microbial and chemical changes were examined with 16S rRNA amplicon sequencing, selective plate counts, pH, TMAO, DMA, TMA and water content measurements.

2. Materials and methods

2.1. Study design, shark fermentation and drying

The Greenland shark used for the study was a by-catch caught by a commercial fishing vessel off the west coast of Iceland in March 2021. On board, the shark was gutted, beheaded and kept on deck for two days before landing. After landing, the shark, a female of about 4–4.5 m length, was transported to the processing site. The shark was split into two identical halves and each half was divided into similar 4–5 kg meat cuts. Cuts from one side of the shark were placed immediately in a perforated wooden box (fresh group) according to the traditional fermenting method, where the flap cuts were layered at the bottom and the top with the skin facing out to cover the thicker back cuts which were stacked in a few layers in-between. The cuts from the other half of the shark were frozen at −21 °C for four days (frozen group), then defrosted at 6 °C for three days and stacked like the fresh group in a separate wooden box which delayed the first sampling point of a week for the frozen sample group. This setup resulted in the following four different groups: fresh back cuts, fresh flap cuts, frozen back cuts and frozen flap cuts (Fig. 1). The fermentation took place in a cold room and was simultaneously sampled for fresh and frozen cuts, resulting in a seven-weeks fermentation period for the frozen-groups and eight weeks for the fresh groups. After the cessation of the fermentation, the meat cuts were hung outdoors for drying in half open wooden sheds for 5 weeks, and therefore giving a total of thirteen weeks of processing. Approximately 100 g of samples were collected weekly during the fermentation and biweekly during the drying period for the different analyses (Fig. 1). Samples were sent in a cooling box from the processing site to the laboratory and treated within 20 h from sampling. For chemical analyses (2.2) samples were frozen at −20 °C until measurements were performed, but samples for microbial analyses (2.3) were measured right upon arrival. The ambient temperature and humidity during the fermentation were logged using iButton DS1923-F5 thermocron temperature and humidity loggers (Maxim Integrated, USA).

Fig. 1.

Diagram of the methodology. Triplicate samples (n = 3) were collected for each sampling group and sampling point (weekly during fermentation and biweekly during drying), where selective microbiological plate counts, 16S rRNA amplicon sequencing, TMAO, DMA, TMA, pH and water content were analyzed. X = RTE hákarl.

2.2. Chemical analysis

2.2.1. Chemicals, solvents and reagents

All chemicals, solvents and reagents used for the chemical analysis are listed in the supplementary material.

2.2.2. Sample preparation for methylamine analysis

Methylamine extraction and derivatization was based on the method by Baliño-Zuazo and Barranco [15] with modifications. Briefly, 2.5 g of thawed and homogenized muscle sample was weighed accurately and mixed with 20 mL of 10 mM acetic acid (pH 3.5). The solution was homogenized for 60 s with an IKA T25 digital ultraturrax (S25N-8G dispersing tool). The rod was rinsed with 5 mL of 10 mM acetic acid to obtain a total volume of 25 mL. The solution was centrifuged at 4617×g for 20 min at 4 °C (centrifuge TJ 25, rotor TS – 5.a-500, Beckman Coulter, USA), supernatant filtered through a Whatman n°4 filter paper (Maidstone, UK) and filtrate centrifuged at 1610×g for 5 min at 4 °C. An aliquot (20 μL) was mixed with acetonitrile/water (6:4, v/v) containing TEAC (3.2 μg/mL) (480 μL), 20 μL 0.5 M bicarbonate buffer and 1 mL TBBA (5 mg/mL in acetonitrile). The mixture was vortexed for 0.5–1 min, incubated at 60 °C for 60 min, put on ice and 10 μL 5 M formic acid added. The mixture was vortexed for 0.5–1 min, diluted with acetonitrile, filtered into an HPLC vial and subjected to LC-MS/MS analysis. Each sample was extracted and derivatized in duplicate and injected twice (n = 4). The multi-analyte calibration solution was derivatized by the same procedure.

2.2.3. LC-MS/MS analysis

Measurements were done by using a ThermoFisher Scientific (USA) Ultimate 3000 UHPLC Series liquid chromatograph coupled to a Thermo Scientific TSQ Endura Triple-Stage Quadrupole Mass Spectrometer. The column was an XBridge® BEH Amide column (3 × 100 mm, 2.5 μm, 130 Å, Waters, Ireland) with gradient solvent system consisting of solvents (A): 15 mM ammonium formate in MQ water and (B): acetonitrile. The flow rate was 0.4 mL/min and 6 μL of sample was injected. Elution conditions and MS parameters are detailed in the supplementary material.

2.2.4. Verification of the analytical method

Verification experiments were performed using fresh cod purchased from a local market. The linear range was determined both with standard solutions (7 concentration levels and a reagent blank) and spiked samples (3 concentration levels and a reagent blank). As fish free of DMA, TMA and TMAO was impossible to get, the LOD and LOQ were calculated from the calibration curves as the average of the area of six reagent blanks for each compound plus 3 (LOD) or 10 (LOQ) times the standard deviation. The final values were presented as μg/g of fish. Recovery and repeatability were determined by analyzing blank and spiked samples at three different spiking levels (10, 25 and 50 μg/g for DMA and TMA, and 25 times higher for TMAO). The spiking solution was added before homogenization. Then, samples were prepared following the same protocol as for the shark samples (2.2.2).

2.2.5. Water content analysis and pH measurement

Thawed shark samples were homogenized using a Waring Commercial® Pro Prep TM chopper grinder. The water content of the homogenized samples was measured one time per sample by determining the difference in weight before and after drying for 4 h at temperatures of 102–104 °C [16]. pH of the homogenized samples was measured in triplicate using a Portavo 902 pH meter (Knick, Germany), with pH probe S2104 N (Knick, Germany).

2.3. Microbiological analysis

2.3.1. Sample preparation for microbiological analysis

During the fermentation and drying process, microorganisms were detected and quantified from fresh samples using both culture-dependent plate counts and culture-independent 16S rRNA amplicon sequencing. Each sample was homogenized separately in a blender, 20 g weighed into a sterile stomacher plastic bag, diluted tenfold with chilled Maximum Recovery Diluent (MRD; Oxoid, Hampshire, UK) and stomached for 2 min. Stomacher solutions were then processed for culture-dependent microbiological analysis as described in section 2.3.2, followed by culture-independent microbiological analysis as described in section 2.3.3. For the culture-independent microbiological analysis, 1.5 mL of stomacher solution was pipetted into Eppendorf tubes, centrifuged at 16000×g for 10 min at 4 °C and the pellet collected. All pellets were stored at −80 °C until DNA extraction.

2.3.2. Culture-dependent microbiological analysis

Successive 10-fold serial dilutions with MRD were performed from the stomacher bags and spread-plated on nine different selective media for detection and quantification of food and fish associated microorganisms. For the detection of specific spoilage organisms (SSO), modified cephaloridine fucidin cetrimide (CFC) agar was used for the enumeration of presumptive Pseudomonas spp. and modified iron agar (IA) was used for both H₂S-producing bacteria and total viable psychrotrophic counts (TVC) [17]. The total count of heterotrophic bacteria was visualized by spread-plating ½ Marine Broth Agar (Difco, UK) both under aerobic and anaerobic conditions for five days at 17 °C. De Man, Rogosa and Sharpe agar (MRS, Oxoid, UK) was spread-plated to count the total Lactobacilli bacteria after three days at 30 °C under anaerobic conditions. Enumeration of food spoiling yeasts and molds was performed by spread-plating Dichloran-Rose Bengal Chloramphenicol (DRBC, Oxoid, UK) agar. The plates were incubated at 22 °C for five days. Blood agar (Oxoid, UK) was spread-plated to count the total of Bacillus cereus in the samples, and plates incubated at 30 °C for one day. Enumeration of Staphylococcus aureus was performed by spread-plating on Baird-Parker agar (BP, Oxoid, UK) by incubation at 37 °C for two days. The total count of sulphite-reducing Clostridia bacteria was done by spread-plating on Perfringens Agar Base (PAB, Oxoid, UK) and incubated for two days at 37 °C. Finally, Violet Red Bile Glucose Agar (VRBGA, Oxoid, UK) was spread-plated to count the total bacteria belonging to the group Enterobacteriaceae after one day at 37 °C. Microbiological data were expressed as log CFU/g.

2.3.3. Culture-independent microbiological analysis

DNA was extracted from defrosted cell pellets (see 2.3.1) using the QIAamp PowerFecal Pro DNA Kit (QIAGEN, Germany). A negative DNA extraction control was processed using only the kit's buffers to monitor contamination during the DNA extraction workflow. DNA quantification and purity were evaluated on a NanoDrop Spectrophotometer ND-1000 (Thermo Fisher Scientific, USA) and DNA diluted to 20 ng/μL and kept at −20 °C until polymerase chain reaction (PCR). All samples were PCR amplified in triplicate on MiniAmp Thermal Cycler (Applied Biosystems, USA) in 25 μL reaction volumes using the universal prokaryotic 16S rRNA gene primer pair 515F [18] and 806R [19], high fidelity Q5 polymerase (New England Biolabs, USA) and 20 ng DNA as described previously [17]. ZymoBIOMICS Gut Microbiome Standard D6331 (Zymo Research, USA) was used as a positive PCR and sequencing control. Pooled triplicates were purified using the HighPrep PCR Clean Up System (MagBio Genomics, USA) according to the manufacturer and when subjected to Miseq Illumina sequencing as previously described [20]. Bioinformatic analysis were done as described earlier [21] on Rstudio using R version 3.6.0 [22,23]. Sequence variants were inferred in R using DADA2 package version 1.18.0 [24].

Demultiplexed sequences were trimmed using the following trimming parameters: trimLeft = c(20, 20), maxN = 0, maxEE = c(2,5), truncQ = 2, and truncLen = c(280, 210). The trimmed and cleaned forward and reverse reads were then merged to remove chimeric variants, singletons and to obtain Amplicon Sequence Variants (ASVs). Reads that did not reach target length were removed from the dataset. ASV table was performed using the function “mergeSequenceTables”. The ASVs were compared against a training set of the SILVA SSU database release 138 [25] using the function “assignTaxonomy”. Subsequent microbial community analysis was performed in R implemented in RStudio [23] using the packages phyloseq (version 1.30.0) [26] and vegan (version 2.6.2) [27]. Contaminant ASVs were identified with the function “contamdf.prev” using the package decontam (version 1.6.0) [28]. Contamination removal was done by using the function “prune_taxa” using phyloseq package. Additionally, sequences that were assigned to the taxa Eukaryote at kingdom level, chloroplasts at the order level, mitochondria at the family level, Thermus at the genus level and unidentified taxa at kingdom level were treaded as contaminants and removed from the dataset by using the function “subset_taxa” with phyloseq before further processing. For subsequence analysis, all samples were normalized by transforming data to proportion using the function “transform_sample_counts” in phyloseq.

2.4. Statistical analysis and data representation

Methylamine levels were calculated as an average of duplicates injected twice and the standard deviation was calculated from those four points. For the culture-dependent analysis, the bacterial growth numbers obtained from each culture medium was compared between fresh and frozen shark cuts at every timepoint in the process. One-way Analysis of variance (ANOVA) was used for this statistical measurement comparison on Rstudio using R version 4.1.3 [22,23]. For the culture-independent analyses, microbial community diversity (Shannon) was compared between samples throughout the process time. Two-way analysis of variance (ANOVA) was used to predict if treatment method and/or sample type had significant effect on the microbial community diversity of the samples by grouping each sample after either fermentation or drying. Bray-Curtis dissimilarity index was used to quantify the dissimilarity of the microbial communities between the samples. The dispersion of variances of the microbial communities was tested in the samples with the function “betadisper” from the vegan package [27]. To test if there was statistically significant difference between the microbial communities of the samples based on group and sample type, permutational multivariate ANOVA (PERMANOVA) was done using the functions “adonis” from the vegan package. For all statistical analyses, the significance level was set at 5 % unless otherwise noted. All the plots were created using R software (version 3.6.1) [22]. The packages ggplot2 [29] and gridExtra [30] were used to create the graphs. Principal component analysis (PCA) was performed using Unscrambler software (version 11 Client, CAMO ASA, Trondheim, Norway) to analyze the connection between the chemical and microbial communities' variables and the variation between the samples. The data was centered, and the chemical variables (TMAO, TMA and DMA) were weighted with the inverse of the standard deviation to correct for the difference of scales in the variables. The models were cross validated.

3. Results and discussion

3.1. Chemical analysis

3.1.1. Temperature and humidity

During the first week, the environmental humidity increased from 40 % to close to 90 % in the fermentation room and stayed stable afterwards. The temperature decreased from 20 °C to close to −3 °C during the two first weeks of fermentation and increased again to between 0 °C and 5 °C for the remaining five to six weeks of fermentation. During drying, the humidity stayed at 75 % and the temperature was around 5 °C. Controlling factors such as humidity and temperature can act to prevent or slow down the growth of microorganisms [31]. However, the effects of humidity and temperature on bacterial survival and growth is species dependent [32,33].

3.1.2. Water content and pH during fermentation and drying

The initial pH of the fresh shark samples was 6.0 ± 0.0 for both back and flap cuts, while the frozen thawed samples had a slightly higher initial pH of 6.1 ± 0.0 and 6.7 ± 0.0, respectively. Between the first and second week of fermentation, the pH increased for all four sample groups to reach a value between 8.9 ± 0.0 and 9.2 ± 0.1, where the pH of the flap cuts was lower compared to the back cuts (both for the fresh and frozen samples). The pH stayed at these values throughout the fermentation and decreased slightly during the drying process for all groups by 0.1–0.2 pH units. The difference between flap and back cuts remained similar with a final pH of 8.9 ± 0.0 for the back cuts and 8.8 ± 0.0 and 8.7 ± 0.0 for the flap cuts frozen and fresh respectively. These values are in accordance with earlier findings [11,13]. The increase in pH during the fermentation seems to be related to the increase in total viable counts and changes in the microbiome as shown in section 3.2. Such alkali fermentation, elevating the pH to 8–9 inhibits the growth of spoilage-associated microbes [34]. The water content in the meat cuts remained constant between 60 and 70 % until drying. After five weeks of drying, the water content had decreased to between 36 % and 41 % for all the samples. This correlates with previous results [11].

Water is an important constituent for bacterial growth and water activity is correlated with food stability deterioration due to the growth of microorganisms. Drying is one of the oldest methods of food preservation, to reduce water activity by separating out water [31,35].

3.1.3. Linearity, LOD, LOQ and recovery of LC-MS/MS method

The linearity, LOD, LOQ and recovery determined for the target methylamines in shark flesh samples are listed in Table 1. Recoveries were used as a correction factor during the quantitation of samples. To minimize matrix effects and increase recovery, samples were diluted at least three times and did not exceed thirty-fold dilution.

Table 1.

Linearity, recovery, and limits of detection (LOD) and quantification (LOQ) for the different analytes.

Compound	Linearity (ng/mL)	LOD (ng/mL)	LOQ (ng/mL)	Recovery (%)
DMA	0.5–40	1.7	8.6	74–107
TMA	5–600	6.7	18	68–113
TMAO	250–7000	213	351	82–136
TEAC				80–100

3.1.4. Methylamine levels throughout fermentation and drying

As shown in Fig. 2, TMAO decreased between 12-fold and 3-fold during the first five weeks of the fermentation period, i.e., going from 25852 ± 1591 μg/g of dry weight to 2141 ± 930 μg/g of dry weight for fresh flap cuts (Fig. 2A) and from 51510 ± 7910 μg/g dry weight (d.w.) to 6217 ± 1362 μg/g d.w. for the frozen flap cuts (Fig. 2B). These results are in concordance with the findings of Anthoni et al. [6] where it was also concluded that the concentration of TMAO in the shark flesh increases upon freezing.

Fig. 2.

Methylamine levels throughout fermentation and drying. TMAO, DMA, and TMA content in shark flesh during fermentation and drying in A) fresh sample groups and B) frozen sample groups. TMAO is represented by the blue line with the triangles, TMA is represented by the red line without points and DMA by the green line.

The Greenland shark is known for having one of the highest levels of TMAO in the body [36] and after its death, TMAO is degraded by bacterial enzymes present in muscle tissue to form TMA and DMA [[37], [38], [39]]. Spoilage bacteria of genera such as Alteromonas, Photobacteria, and Vibrio express TMAO-reductase and utilize TMAO as a terminal electron acceptor during anaerobic growth which reduces TMAO into TMA. DMA is formed in the absence of oxygen by removal of the methyl group of TMAO by TMAO-demethylase enzymes. Freezing and frozen storage do not deactivate these TMAO converting enzymes but may alter the catalyzed reaction [38]. The reactions are exponential, starting slowly and increase rapidly after a few days of chilled storage [40], which is demonstrated by the results presented in Fig. 2. For the first two to three weeks of fermentation, the levels of DMA and TMA were very low. Between three to six weeks of fermentation, DMA and TMA levels increased rapidly, oppositely correlated to the TMAO degradation (Fig. 2) and reached levels of DMA between 466 ± 30 μg/g (d.w.) and 672 ± 38 μg/g (d.w.) in the fresh back and flap cuts (Figs. 2A) and 354 ± 29 μg/g (d.w.) to 405 ± 28 μg/g (d.w.) for the frozen flap and back cuts (Fig. 2B). The TMA levels were respectively 21259 ± 2433 μg/g (d.w.) and 26078 ± 844 μg/g (d.w.) for the fresh back and flap cuts (Figs. 2A) and 17441 ± 1355 μg/g (d.w.) and 17994 ± 2367 μg/g (d.w.) for the frozen back and flap cuts (Fig. 2B) at the end of the drying period. The ability of bacteria to utilize TMAO is species-dependent at certain level [41], suggesting that TMAO reduction could be dependent on the bacterial composition when initiating fermentation. The difference in microbial composition when initiating fermentation of the frozen and fresh shark cuts is demonstrated in section 3.2. Additionally, it has been demonstrated in hake that there is a relationship between the quantity of TMA produced and the number of bacteria in the muscle [42].

After six weeks of fermentation, the level of TMAO in the shark cuts was below the limit of detection (LOD) of the method (213 μg/g d.w.) (results therefore not presented after five weeks processing time) and the DMA and TMA contents had stabilized in correlation with the breakdown of TMAO. This suggested that stopping the fermentation after 5 weeks instead of continuing to 8 weeks could be an option to reduce the fermentation time but still obtain a similar end product.

3.2. Microbiological content

3.2.1. Culture-dependent microbiological plate counts

Results of microbial plate counts are shown in Fig. 3. During the first three to four weeks of fermentation, there was a steady increase of total viable counts (TVC) of psychrotrophic bacteria in all sample groups (Fig. 3A and B) which persisted until drying, when it dropped to values similar to the initial sampling. At the initial sampling point, the TVC bacterial numbers were significantly lower (p < 0.01) in the fresh back cuts (3.5 ± 0.5 log CFU/g) compared to the frozen back cuts (5.0 ± 0.2 log CFU/g). However no significant difference was detected between fresh and frozen flap cuts at this sampling time. During the first and third weeks of fermentation, there was a significant difference between fresh back cuts (first week 6.3 ± 0.2 log CFU/g; third week 8.0 ± 0.1 log CFU/g) and frozen back cuts (first week 7.3 ± 0.4 log CFU/g; third week 7.5 ± 0.5 log CFU/g). Throughout the fermentation process no significant difference in TVC bacterial numbers was detected between the fresh and frozen flap cuts. A statistically significant difference in the TVC load was found between fresh back cuts (6.9 ± 0.2 log CFU/g) and frozen back cuts (4.3 ± 1.2 log CFU/g) during the first week of drying. However, no significant difference was found in TVC load between fresh and frozen flap cuts for the whole drying period.

Fig. 3.

Microbial growth during fermentation and drying of fresh and frozen cuts of Greenland shark. Total viable psychotrophic counts on IA and H₂S producer levels in A) fresh sample groups and B) frozen sample groups. Aerobic heterotrophic bacteria and anaerobic heterotrophic bacteria counts on ½ Marine Broth Agar in C) fresh sample groups and D) frozen sample groups. Pseudomonas spp. counts on CFC in E) fresh sample groups and F) frozen sample groups.

H₂S producing bacteria like Shewanella putrefaciens that grow on iron agar are defined as a specific spoilage organisms (SSO) in fish and are characterized for producing metabolites which contribute to the development of the fishy sulfite odors and flavors associated with fish spoilage [43]. The number of H₂S producing bacteria was similar in all sample groups at all sampling times and followed similar patterns, reaching a maximum within the first two weeks of fermentation (fresh back cuts 5.4 ± 0.8 log CFU/g; fresh flap cuts 5.6 ± 0.1 log CFU/g; frozen back cuts 5.5 ± 0.2 log CFU/g; frozen flap cuts 6.1 ± 0.3 log CFU/g) (Fig. 3A and B). After that, the H₂S producing bacteria numbers began to decrease and were not detected in any of the shark cuts at the end of the drying process. Similarly, as for TVC, there was only a statistically significant difference in H₂S producing bacteria load between fresh and frozen back cuts at the initial sampling point but not during fermentation and drying. The results of TVC and H₂S producing bacteria followed a similar growth pattern in the shark cuts as previously observed [11].

The total number of heterotrophic bacteria on ½ MA media was lower under anaerobic conditions compared to aerobic conditions in all sample groups at all sampling times (Fig. 3C and D), which is consistent with previous research on bacterial growth during the processing of Greenland shark [11]. A similar pattern was seen in all sample groups as for TVC and H₂S producing bacteria, reaching a maximum after four weeks of fermentation with a difference varying from log 1.5–2.3 CFU/g between aerobic and anaerobic culture conditions for the same sample groups (fresh back cuts 8.9 ± 0.3 log CFU/g vs. 6.6 ± 0.1 log CFU/g; fresh flap cuts 9.2 ± 0.4 log CFU/g vs. 7.5 ± 0.1 log CFU/g; frozen back cuts 8.5 ± 0.2 log CFU/g vs. 6.8 ± 0.1 log CFU/g; frozen flap cuts 9.0 ± 0.5 log CFU/g vs. 7.5 ± 0.9 log CFU/g), followed by a decrease in numbers which continued throughout the drying process. There was a significant difference in heterotrophic bacteria load at the initial sample time between fresh (3.8 ± 0.6 log CFU/g) and frozen (5.2 ± 0.3 log CFU/g) back cuts and between fresh (7.0 ± 0.5 log CFU/g) and frozen (6.1 ± 0.3 log CFU/g) flap cuts in the aerobic group. There was again a statistically significant difference (p < 0.01) in aerobic heterotrophic bacteria between fresh (7.7 ± 0.1 log CFU/g) and frozen (8.5 ± 0.2 log CFU/g) back cuts after seven weeks. During the drying process no significant difference was detected in heterotrophic bacterial load between the fresh and frozen shark cuts in the aerobic group. A significant difference (p < 0.01) was found in anaerobic heterotrophic bacteria load between fresh (5.5 ± 0.1 log CFU/g) and frozen (6.7 ± 0.0 log CFU/g) back cuts at week seven. However, significant difference was detected in anaerobic heterotrophic bacteria load between fresh and frozen flap cuts at the initial sampling (fresh flap cuts 6.7 ± 0.7 log CFU/g; frozen flap cuts 4.7 ± 0.6 log CFU/g) and at week one (fresh flap cuts 4.9 ± 0.1 log CFU/g; frozen flap cuts 5.9 ± 0.2 log CFU/g) (p < 0.01). During the drying process no significant difference was detected in the heterotrophic bacterial load between the fresh and frozen shark cuts in the anaerobic group.

Psychrophilic Pseudomonas spp. are one of the major SSO in chilled aerobically stored foods [44] and show high spoilage potential in proteinaceous foods such as fish [45]. Again, in all sample groups, Pseudomonas spp. reached a maximum load within the first two and three weeks of fermentation for frozen and fresh cuts respectively (fresh back cuts 4.0 ± 0.1 log CFU/g; fresh flap cuts 4.3 ± 0.3 log CFU/g; frozen back cuts 5.2 ± 0.4 log CFU/g; and frozen flap cuts 5.3 ± 0.7 log CFU/g) (Fig. 3E and F). The counts then declined and were below the detection limit for all sample groups at the end of the drying process. No significant difference was found in Pseudomonas spp. load between fresh and frozen back cuts during the whole process, but a significant difference was detected between fresh flap cuts (4.3 ± 0.3 log CFU/g) and frozen flap cuts (2.0 ± 0.0 log CFU/g) at the third week (p < 0.01).

The bacterial growth results are in line with earlier microbial cultivation findings [11], which may point to a rather uniform growth pattern during the production of fermented shark. During the whole processing time, only a few samples had low counts of food-spoiling yeasts (one sample), molds (two samples), Lactobacilli (one sample) and sulphite-reducing Clostridium (one sample). The counts of Staphylococcus aureus, Bacillus cereus, and Enterobacteriaceae were below the detection limit in all samples.

3.2.2. Microbial community succession

The total number of the 16S rRNA amplicon sequencing reached about 13 million raw reads for all samples. Raw 16S rRNA gene sequences are deposited in the Sequence Read Archive of the National Center for Biotechnology Information (“NCBI: PRJNA990677”). Positive, negative and replicate controls that were included in the workflow during the molecular analysis workflow confirmed scientific consistency across the research. Strains from the mock community with ≥0.1 % abundances were identified in the positive controls. The negative controls had a relatively low combined prevalence of contamination and taxa were not found in the samples. The replicate controls confirmed that the microbial composition and sequencing depth in each sample showed no statistical difference which indicates that the PCR and sequencing was successful. After assigning reads to taxa and removing contaminants and control samples from the dataset, the average number of 86 amplicon sequence variants (ASV) was obtained in each sample assigned to total of 17 phyla and 198 genera. In this study significant difference was detected in the microbial community diversity (Shannon) between the fresh and frozen sample groups during fermentation (p < 0.05) and drying (p < 0.01). Additionally, microbial community diversity was significantly different between the flap cuts and back cuts during fermentation and drying.

The results of the culture-independent microbial analysis added to the knowledge obtained from the cultivations and the chemical analyses. Characterized by the activity and 16S rRNA microbial community assembly, the fermentation process could be divided into three separate phases; initial phase (first two weeks), intermediate phase (next two to three weeks) and final phase (until end of fermentation) (Fig. 4, Fig. 5) which then could be correlated to the changes in concentration of TMAO, TMA, DMA (Fig. 2A and B) and pH and the culture-depended analyses (Fig. 3). During the initial phase, the microbial communities varied between the frozen and fresh sample groups. However, as the fermentation progressed, the microbial community variability decreased in all samples and remained similar throughout the drying process (Fig. 4). The whole drying process showed similar abundance as the final fermentation phase in all sample groups. These results indicate that as the shark fermentation process advances, the microbial community gets more stable and the number of ASV decreases. The average number of observed ASV was measured and compared between fresh (back and flap cuts) and frozen (back and flap cuts) sample groups throughout the process. During the initial week, the ASV were lower in the fresh group (105) than in the frozen group (153). After one week of fermentation, the number of ASVs increased slightly for the fresh group (117) and decreased for the frozen group (130). For the rest of the fermentation process the ASV fluctuated in both groups but at the end of the fermentation period the number of ASV was similar between groups in the fresh group (90) and in the frozen group (88). During the drying process (weeks 9–13) the number of ASV stayed relatively same in the fresh group (ranging between 82 and 93) but decreased in the frozen group (ranging between 64 and 74).

Fig. 4.

Non-metric dimensional scaling plot based on Bray-Curtis dissimilarity, illustrating the microbial community dissimilarity between fresh and frozen shark cut samples during the fermentation and drying process.

Fig. 5.

Microbial community succession in A) fresh back cut samples, B) frozen back cut samples, C) fresh flap cut samples and, D) frozen flap cut samples.

Although the bacterial composition became more similar as the fermentation process progressed, there was a significant difference detected between fresh and frozen sample groups and back and flap sample groups throughout the process. During the initial phase (first two weeks), there was a clear difference in the microbial community structure between the fresh and frozen sample groups. The fresh sample groups (Fig. 5A and C) had relatively few but dominant genera where Photobacterium (fresh back cuts 37–52 %; fresh flap cuts 15–80 %), Pseudoalteromonas (fresh back cuts 2–35 %; fresh flap cuts 16–50 %) and Carnobacterium (fresh back cuts 0.5–23 %; fresh flap cuts 2–8 %) were the most common. However, during the same phase, the frozen sample groups (Fig. 5B and D) had a higher number of genera where Pseudoalteromonas (frozen back cuts 40–52 %; frozen flap cuts 39–67 %), Carnobacterium (frozen back cuts 0.5–20 %; frozen flap cuts 4–28 %), Jeotgalicoccus (frozen back cuts 3–6 %; frozen flap cuts 0.2–8 %) and Yaniella (frozen back cuts 6–9 %; frozen flap cuts 1–9 %) were the most dominant.

Photobacterium are commonly found in marine environments and the species, P. phosphoreum, is believed to be a key player in fish spoilage, with the ability to reduce TMAO into TMA [46]. There was a clear difference in the relative abundance of Photobacterium between fresh and frozen sample groups (Fig. 5) and the niche left behind by Photobacterium in the frozen shark cuts was replaced by other genera. Previous research has shown that storage under super chilled conditions and thawing, delays the growth of P. phosphoreum in super chilled MAP cod fillets [47]. Photobacterium abundance in the fresh shark cuts decreased as the fermentation progressed and was only detected in low abundance during the drying process (0–6 %). These results are consistent with a previous study of RTE shark [13] where Photobacterium was a small taxonomic group (<1 %). Similarly, Photobacterium has also been detected as one of the key bacteria members in the fermentation process of skate [[48], [49], [50]].

In the frozen sample groups, Pseudolateromonas was the main genus found at the expense of Photobacterium. Pseudoalteromonas is commonly found in marine environments and contains members which show high spoilage potential and produce many volatile organic compounds (VOC) which are associated with seafood spoilage [51]. Pseudoalteromonas has e.g., been shown to be the main spoilage organism during aerobic storage of brown shrimp (Crangon crangon) and to be responsible for reducing TMAO into TMA [51]. Pseudoalteromonas has also been documented as one of the key players in the hydrolyzation of urea into ammonia during the early fermentation process of skate [50]. Herein, Pseudoltaromonas seems to be an important TMAO reducer in the fermentation process of shark meat and its decrease as the fermentation progresses, correlates with the concurrent decrease in TMAO levels throughout the shark fermentation process. Furthermore, during the drying process, neither Pseudoalteromonas (Fig. 5) nor TMAO could be detected (Fig. 2). The pH of the shark cuts increased significantly from 6 to 9 during the initial fermentation phase which could be linked to the reduction of TMAO into TMA and DMA (Fig. 2). Hence, the dominant bacteria during the initial phase reduced TMAO into alkaline metabolites which contributed to the increased pH and changed the nutrient availability in the shark.

During the final phase of fermentation and drying, the microbial community in the sample groups began to stabilize. At this point, fresh and frozen sample groups had fairly similar microbial community assemblies. However, during the intermediate phase, many of the earlier colonists had been replaced by new genera, i.e., Atopostipes, Pseudomonas and Tissierella which became more dominant as the fermentation progressed. Atopostipes became one of the predominant genera in both fresh and frozen sample groups after two to three weeks of fermentation (fresh back cuts 8–27 %; fresh flap cuts 37–43 %), (frozen back cuts 4–36 %; frozen flap cuts 8–44 %) and held that position throughout the drying process. Atopostipes is a Gram-positive anaerobic bacteria genus comprising only the species A. suicloacalis [52]. The bacterium has regularly been detected in alkaline environments rich in urea and ammonia such as in urine derived fertilizers [53]. It is therefore likely that the Atopostipes detected in the shark cuts utilizes urea and/or ammonia as a nitrogen source. These findings are consistent with the abundance of Atopostipes found in fermented skate that is also rich TMAO and ammonia [48,50]. It has also be suggested that Atopostipes might be an important starter bacteria for culture selection, which could improve the quality of fermented skate [48]. As shown in Fig. 7. TMAO and Atopostipes are oppositely related as indicated by decreasing TMAO levels and increasing TMA and DMA, Atopostipes abundance increased.

Fig. 7.

Correlation loadings from the PC1 and PC2 of the PCA of the shark samples. The chemical variables are identified in red and the 16S rRNA microbial communities' variables are identified in blue.

Pseudomonas is a large genus which contains members with diverse metabolic activities [54,55]. Similarly as for Atopostipes, it has been documented that Pseudomonas can hydrolyze urea into ammonia in the early fermentation process of skate [56] and some Pseudomonas spp. are known to reduce TMAO into TMA while others can dehydrogenate TMA into DMA and formaldehyde [57]. In the study of RTE hákarl, Osimani et al. [13] hypothesized that Pseudomonas could be important in the detoxification of fermented shark by metabolizing TMAO and TMA. However, the results from this study show that Pseudomonas became abundant in the fresh and frozen sample groups after three to four weeks of fermentation (fresh back cuts 19–28 %; fresh flap cuts 20–28 %), (frozen back cuts 23–28 %; frozen flap cuts 1–27 %) when almost all TMAO had been reduced to TMA and therefore indicates the utilization of TMA rather than TMAO as also shown in the PCA analysis (Fig. 6, Fig. 7).

Fig. 6.

Principal component analysis (PCA) scores plot (PC1 vs. PC2) of all shark samples based on chemical variables (TMAO, TMA, DMA) and 16S rRNA microbial communities' variables. A) Scores overview of the PCA and B) zoom in on the left part of the scores plot. Fre = Fresh; Fro = Frozen; B = Back; F = Flap.

Tissierella became more dominant as the fermentation progressed and after approximately five to six weeks (fresh back cuts 5–19 %; fresh flap cuts 4–16 %) (frozen back cuts 7–10 %; frozen flap cuts 5–11 %) of fermentation and prevailed throughout the drying process in all sample groups. The high abundance of Tissierella during the drying process (fresh back cuts 15–16 %; fresh flap cuts 17–19 %) (frozen back cuts 16–18 %; frozen flap cuts 11–18 %) is consistent with findings in previous studies of RTE hákarl Osimani et al. [13]. Tissierella are Gram-positive protein degrading anaerobic bacteria which can produce volatile fatty acids such as acetic, butyric and propionic acids [58]. Some Tissierella species are also capable of degrading creatinine into creatine, sarcosine and glycine into methylamine, ammonia, CO₂ and acetyl phosphate [58]. Moreover, Tissierella has been found in urine derived fertilizers [53] which suggests that the bacterium utilizes urea and/or ammonia as a nitrogen source in the fermentation process of shark. Tissierella was highly abundant in fermented skate [48] and was suggested, along with Atopostipes, as important for selection of possible starter bacteria which could improve the quality of fermented skate.

Unidentified genera in the microbial community of fresh and frozen sample groups (grey columns in Fig. 5) constituted nine different genera which occupied the sample groups at different abundance throughout the fermentation and drying process, from below 5 % at the initial sampling and up to 28–38 % at the end of drying indicating that this is a unique ecosystem with many unexplored bacteria.

A principal component analysis (PCA) of the total data set was done by using the chemical variables (TMAO, TMA and DMA) and the 16S rRNA microbial community variables for all samples. Overall, the two principal components (PCs) describe 85 % of the variation of the samples (Fig. 6A and B and Fig. 7). PC1 and 2 indicate a difference in the sample composition between the three first weeks of fermentation and the other weeks. Those three first weeks of fermentation are correlated to high levels of TMAO, Pseudoalteromonas and Photobacterium. The presence of Psychrilyobacter is specifically correlated to the first two weeks of fermentation only for the fresh back cuts. Psychilyobacter has previously been found in Chinese fermented mandarin fish (Siniperca chuatsi) products where it was positively correlated to its characteristic flavor substances [59]. After the fifth week of fermentation, no significant difference could be found between the fresh and frozen groups, where both groups had high levels of TMA and DMA and Tissierella, Atopostices and Pseudomonas and unknown (NA) microbial communities were predominant.

4. Conclusion

Overall, the combination of chemical and bacterial analysis allowed detailed insights into the transformation process from fresh Greenland shark to RTE hákarl. The analysis of TMAO, TMA and DMA levels revealed depletion of TMAO early in the fermentation process and after five weeks, the levels of TMAO were under the limit of quantification (LOQ). Simultaneously, TMA and DMA levels increased and reached a maximum after seven weeks of fermentation and stayed at that level throughout the drying. From the second week of fermentation, the pH increased from 6 to about 9 and at that point the culture-dependent analysis generally showed maximum levels for all the bacteria counts.

The culture-independent analysis showed three distinct gradual shift phases in the bacterial community composition through the fermentation period that further can be correlated to the chemical changes and microbial plate counts as shown by PCA analysis. During the first two weeks of fermentation (initial phase), a clear difference was found between samples of fresh and frozen shark cuts, where the bacterial community of fresh samples were dominated by Photobacterium while Pseudoalteromonas dominated in the frozen samples. Both genera are known to reduce TMAO to TMA and DMA which was correlated with the measured TMAO, TMA and DMA levels. However, TMAO levels were found to be higher in the frozen samples than in the fresh. During the intermediate phase, both fresh and frozen samples were dominated by the genera Atopostipes and Pseudomonas along with a high abundance of Tissierella, until the end of drying. These genera have been detected in urea and ammonia rich environments indicating that they can utilize those compounds as a carbon or nitrogen source and could therefore utilize the high levels of DMA and TMA found in the RTE hákarl.

Since fermentation of shark is an artisan industry where each producer has developed their own specific processing method, there is no standardized protocol for the traditional production of RTE hákarl. Therefore, depending on the raw material and the producer's production process, the final product could be somewhat variable both in microbiological, chemical and nutritional composition. However, the results from this study showed a depletion of TMAO in the shark samples after the first five weeks of fermentation, and a rather stable bacterial community structure from that time and until the processing end in both fresh and frozen cuts, may indicate that reducing the duration of the fermentation period might be possible. To gain an optimal fermentation procedure, this would need to be researched further, along with the possibility of standardizing the microbial community of the raw material with starting cultures.

Funding

This work was funded by the Icelandic Government through Matvælasjóður fund.

Data availability statement

Sequence Read Archive of the National Center for Biotechnology Information “NCBI: PRJNA990677”.

CRediT authorship contribution statement

Sophie Jensen: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing. Snorri Páll Ólason: Data curation, Formal analysis, Investigation, Software, Visualization, Writing – original draft, Writing – review & editing. Sigurlaug Skírnisdóttir: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Software, Supervision, Validation, Writing – original draft, Writing – review & editing. Guðmundur Stefánsson: Conceptualization, Funding acquisition, Project administration, Resources, Writing – original draft, Writing – review & editing. Cecile Dargentolle: Data curation, Formal analysis, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. Viggó Thór Marteinsson: Conceptualization, Funding acquisition, Methodology, Resources, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: