DYNAMICS OF LACTIC ACID BACTERIA DURING PASTIRMA PRODUCTION

and LAB counts increased during the production stages reaching 7.15 and 6.64 log cfu/g, respectively. The most dominant LAB for all stages was Latilactobacillus sakei group with a relative abundance (RA) of 52-73% RA . Weissella species W. viridescens and W. halotolerans followed the L. sakei group. Phylogenetic analysis of 16S rRNA gene indicated that all L. sakei isolates were of subsp. carnosus , however, (GTG)5 fingerprinting demonstrated a high degree of intraspecies variation. Moreover, fingerprinting analysis showed that L. sakei isolates of specific fingerprinting groups were selected towards the final production stages. The present study elucidates how the LAB diversity changes both at the species and intraspecies level during pastirma production.

agar plates containing the same isolation media. Single colonies were restreaked consecutively twice for purification. Isolates were grown in liquid media containing 20% glycerol and stored at -80°C (Dincer & Kivanc, 2012). Gram reaction and catalase test were performed on the isolates as described previously (Chester, 1979;Powers, 1995).

DNA extraction and polymerase chain reaction (PCR)
DNA extraction was performed using a salting-out method as described previously (Martín-Platero et al., 2007). The isolates were grouped by (GTG)5 fingerprinting using the PCR conditions described previously (Versalovic et al., 1994). PCR reaction mixture was prepared as described previously (Seri & Metin, 2021). PCR reactions were conducted using T100 thermal cycler (Bio-Rad Laboratories, Hercules, CA, USA). PCR products were run in 1.5% (w/v) agarose gels containing Red-Safe nucleic acid staining solution (Intron Biotechnology Inc., Korea) in 1 X TAE buffer (Bio-Rad) using Wide Mini-Sub Cell GT electrophoresis system (Bio-Rad) and visualized by Gel Doc EZ Imager (Bio-Rad). Selected isolates according to the grouping patterns were subjected to 16S rRNA PCR using the universal primers 27F (5'-AGAGTTTGATCCTGGCTCAG) and 1492R (5'-GGTTACCTTGTTACGACTT) (Lane, 1991). The PCR mixture was prepared as described for (GTG)5 PCR reaction except that 2 L of both forward and reverse primer was used. PCR was carried out as described previously (Seri & Metin, 2021). PCR reaction products were analyzed on 0.8% (w/v) agarose gels, purified using a GeneJet PCR purification kit (Thermo Fisher Scientific) according to manufacturer's instructions and sequenced with the primers used for PCR.

Statistical analyses
Data were analyzed using JMP 14.1 software (SAS Institute Inc., Cary, NC, USA). A comparison among different stages was performed using a one-way analysis of variance (ANOVA) at a confidence level of 95% (p < 0.05).

(GTG)5 fingerprinting analysis
(GTG)5 fingerprinting patterns were analyzed using temporary Bionumerics (ver 8, Applied Maths, Sint-Martens-Latem, Belgium) evaluation licence that we have received permission to publish. A cluster analysis was performed using Ochiai similarity coefficient with 1% optimization and 1% band matching tolerance. Dendograms were generated using unweighted pair grouping by mathematical averaging (UPGMA).

Phylogenetic analyses of 16S rRNA gene
Phylogenetic analyses were conducted using MEGA X (Kumar et al., 2018). The model describing the substitution pattern the best was determined to be Kimura 2parameter with Gamma distribution, which resulted in the lowest Bayesian information criterion score. Evolutionary analyses were conducted using maximum Likelihood method and Kimura 2-parameter model (Kimura, 1980) with Gamma distribution (5 categories, parameter = 0.1275).

RESULTS AND DISCUSSION
Chemical properties of pastirma during the production process Change of pH and aw was monitored during the production process because of their importance in microbiota development. Results indicated an increase of pH from 5.63 (0.02) at stage 1 (after curing) to 5.81 (0.01) at stage 3 (before çemen coating) (p˂0.05), after which no significant change was observed (Tab 1). Previous studies on the pastirma production process indicated a similar trend of pH increase especially after the first drying step Kaban, 2009;Ozturk, 2015). For instance, in the study conducted by Kaban (2009), pH value first decreased during the curing stage ( pH 5.5) compared to the fresh meat, and then increased significantly during the later stages until the final product ( pH 5.9) is obtained. Similar to our study, çemen addition had little effect in pH change. Although there are reports of a relatively constant pH level during the production of certain dry-cured meat products around the world ( . For example, in Spanish Celta dry-cured loin, the pH value increases from 5.6 to 5.8 during dry-ripening (Pateiro et al., 2015). A similar pH increase from 5.7 to 5.9 was observed in Italian dry-cured ham during ageing (Virgili et al., 2007). The increase of pH during the ripening period is suggested to be due to proteolytic activity taking place in the muscle (Virgili et al., 2007). The origin of proteolysis in meat products has been suggested to be mainly due to meat-originated proteolytic enzymes; however, microbial enzymes have also been reported to be involved (  During pastirma production, aw significantly decreased from 0.923 (0.003) at stage 1 in the salted meat to 0.859 (0.003) at stage 2 after first pressing (p˂0.05) resulting in about 7% reduction, after which, it remained relatively constant until çemen coating. Osmotic pressure created by salt and pressing operation accelerated the water loss and drying resulting in a significant decrease in aw. Çemen addition slightly increased the aw of the final product to 0.889 (0.002) (p˂0.05), likely because of diffusion of water in çemen into the meat. Similar aw decreases were reported during the pastırma production process (Inat, 2008;Kaban, 2009;Ozturk, 2015); however, the stages where the most significant reduction takes place differ among the studies. Kaban (2009) reported a significant decrease after the end of first drying (0.96) to the final product (0.87). According to their results, aw decrease continued at a similar rate during the first and the second drying periods until the final product was obtained. However, our results indicated that the most significant reduction occurred during the first drying and the first pressing stages. In the study conducted by Ozturk (2015), a decrease of aw from 0.97 to 0.90 occurred during salting to the first drying stage, while pressing and the second drying stage had little effect. The differences observed between the studies might be due to the differences in the process conditions such as temperature and relative humidity during drying, or the pressing force applied. Aw is an important parameter determining the shelf life of the product. Reduction of aw during the drying steps makes pastirma an intermediate-moisture food, increasing its microbial stability (Leistner, 1985).

Bacterial counts of pastirma during the production process
During the production process, TMAB and LAB counts changed in a similar manner: they increased during the production process until the final step, çemen coating, which did not change the counts significantly (p˂0.05  , 1998). Differences observed in the LAB loads of different pastirma samples are probably because of the variation of the properties of raw material, the production conditions used in different facilities as well as the storage period in the market. In addition, specifically in our study, microbial counts might have been affected by the cold chain transportation process of samples from the production facility in Afyonkarahisar to our laboratory in Istanbul. Although the loads differ in different studies, the rising trend during the production stages seem to be common.
Other dry-cured products also reported an increase in LAB counts after salting. For example, in dry-cured lacón production, LAB counts increased during the postsalting stage and in the first weeks of the ripening period (Lorenzo et al., 2010;  Vilar et al., 2000). In another dry-cured product el-gueddid, one log increase from 10 6 to 10 7 was observed in LAB load during the ripening stage compared to the after salting stage (Benlacheheb et al., 2019). M17 counts also had an increasing trend during the production process. The counts were similar with no statistical significance during salting and cold-pressing stages, after which they increased to 7.25 (0.16) before çemen coating (p<0.05) and remained similar in the final product (p>0.05).

LAB diversity during the production process
All 82 MRS isolates (named AB#) and only 7 of 80 M17 agar isolates (named AC#) were Gram positive and catalase negative putative LAB. These isolates were then typed by (GTG)5 fingerprinting analysis (Figure 2). Selected isolates from each fingerprinting group were subjected to 16S rRNA gene sequencing for molecular identification. A phylogenetic analysis was conducted using the 16S rRNA gene sequences of the isolates together with the type strains of the species recognized in BLAST search ( figure 3). As a result, five genera and nine species were identified. While the mostly encountered species was Latilactobacillus sakei, renamed after the new taxonomy (Zheng et al., 2020), this species was followed by Weissella viridescens and Weissella halotolerans ( figure 3). The other species encountered were Latilactobacillus graminis, Latilactobacillus curvatus, Carnobacterium divergens, Leuconostoc citreum, Weissella helenica/sagaensis (could not be discriminated using 16S rRNA sequence), and Weissella thailandensis.
There are only a limited number of studies on the identification of LAB species of pastirma. In one of these studies, the most common LAB species in pastirma samples from different manufacturers was L. sakei which was followed by Different stages of the pastirma production process recruited L. sakei isolates of different (GTG)5 fingerprinting clades (figure 2). For example, while isolates of clade II-a and VI were isolated mostly from the first two stages, clades I, II-b and II-c were generally obtained from the last two stages. This finding indicates a selection of strains with certain genetic background during specific stages of the process. This selection process might be related to the pH or aw variations during production. It could also be related to the competitive ability of the strain because total bacterial counts increase from 5 log cfu/g during the first two stages to 7 log cfu/g during the last two stages (Tab 1). The strain with a better competing ability would be expected to be selected during later stages of production. Similar to our finding, in the production of the Italian fermented sausage Ventricina, three different biotypes of L. sakei were observed and during maturation, a specific biotype was selected and outcompeted others (Tremonte et al., 2017). How specific strains are selected during pastirma production process is an interesting question and requires further experiments, such as determination of pH, aw, and salt tolerance, as well as fitness of each strain, which could be performed in future studies.
In all pastirma production stages analyzed, L. sakei/L. graminis was the most dominant group with its relative abundance (RA) changing between 52-73% during different production stages (Figure 4). Weissella species follow this group with W. viridescens (21-28% RA) and W. halotolerans (4-21% RA). During the stages 3 and 4, other Weissella species, W. helenica/W. sagaensis and W. thailandensis were also observed ( Figure 4). Weissella species are heterofermentative and produce ethanol, CO2, and acetic acid from glucose (Kroeckel, 2013 (Hu et al., 2021). To the best of our knowledge, how the diversity of lactic acid bacteria changes during pastirma production has not been investigated before. We present here that in pastirma, L. sakei/L. graminis remained the predominant species group with an RA of 52-73% throughout the production process. However, the strain diversity of this species changed during different production stages. Analysis of microbial dynamics of samples from different producers in future studies and comparison of the results with the current study will be more comprehensive in understanding pastirma microbial ecology. The effects of seasonal variations on the microbiota of pastirma can also be explored. For example, in spontaneously fermented Italian sausages, L. sakei pangenome changed according to the season indicating a strain-level diversity among production batches from different seasons (Franciosa et al., 2021).

Figure 3
The phylogenetic analysis of LAB isolates and closely related type strains using 16S rRNA gene sequence. The analysis involved 830 nucleotides of 16S rRNA gene.

Figure 4
Relative abundance of LAB species in each production stage of pastirma

CONCLUSION
In conclusion, in this study, we have determined certain physicochemical parameters, bacterial dynamics, and LAB diversity during pastirma production process using samples from a commercial producer. During the production, while the pH increased from 5.6 at the curing stage to 5.8 in the final product, aw decreased from 0.92 to 0.86 before çemen coating, after which it increased to 0.89 in the final product. Both TMAB and LAB counts increased towards the final stages, where 7.15 and 6.64 log cfu/g were attained in the final product, respectively. The most common LAB in pastirma production was L. sakei/L. graminis group in all production stages analyzed maintaining 52-73% RA. L. sakei group was followed by W. viridescens and W. halotolerans. Other Weissella species, W. helenica/W.sagaensis, and a species related to W. thailandensis were also observed during production. Leu. citreum was encountered in the final stage and speculated to have originated from çemen coated on meat. Phylogenetic analysis using 16S rRNA gene demonstrated grouping of pastirma L. sakei isolates with L. sakei subsp. carnosus rather than L. sakei subsp. sakei. On the other hand, (GTG)5 fingerprinting analysis indicated a highly differentiated L. sakei/L. graminis population. The different (GTG)5 clades observed for the L. sakei group were specific to different production stages, which indicated selection of strains with certain genetic background towards the final stages of pastirma. Understanding the strain-level dynamics is important for comprehensive analysis of the production process. How specific strains are selected and the attributes they provided to the final product are exciting topics to be analyzed in future studies.