Kluyveromyces marxianus: a promising alternative to baker's yeast Saccharomyces cerevisiae for enhanced oven rise performance during bread baking

Abstract

For millennia, bread making has relied on Saccharomyces cerevisiae, valued for its efficient sugar fermentation, robustness under processing stresses and production of a desirable flavour profile. Nevertheless, there is growing interest in exploring alternative yeast species that show advantageous traits beyond the physiological boundaries of S. cerevisiae. For example, a yeast's ability to stimulate dough oven rise is of particular interest because it determines loaf volume. It was recently shown for S. cerevisiae that dough oven rise strongly correlates with yeast viability and thermotolerance during baking. In this study, we evaluate the impact of Kluyveromyces marxianus, a yeast species known for its exceptional thermotolerance, on dough oven rise.

K. marxianus strains exhibited a broader range in oven rise (13.3-41.5%) than previously reported for S. cerevisiae. The best-performing strain also showed greater oven rise than the industrial baker's yeast and the top-performing S. cerevisiae strain reported in a previous study using the same electrical resistance oven (ERO) setup. Similar to what has been reported for S. cerevisiae strains, we observe a positive correlation (R² = 0.77) between viability and oven rise, suggesting that cell survival is a key contributor in early baking performance. Additionally, differences in CO₂ release behaviour were observed between strains. While no individual CO₂-release parameter directly correlated with oven rise, the strains exhibiting the highest oven rise generally showed slower CO₂ release than the poorest-performing strain. Together, our results demonstrate the potential of non-Saccharomyces strains for improving oven rise performance, which could lead to applications increasing loaf volume.

Graphical abstract

Highlights

• •Different K. marxianus strains yield a broad range of oven rise.
• •One K. marxianus strain outperformed previously tested reference S. cerevisiae strains.
• •Oven rise and yeast viability in dough are correlated.
• •K. marxianus viability-oven rise links align with past S. cerevisiae findings.

Abbreviations

Km	Kluyveromyces marxianus strain
ERO	Electrical resistance oven

1. Introduction

Yeast is a key ingredient for bread production. The release of CO₂ gas during bread fermentation results in leavening, while the production of various other metabolites contributes to the taste, smell, and texture of bread (Aslankoohi et al., 2016; Cauvain, 2015; Delcour and Hoseney, 2010; Rezaei et al., 2015). Consumers often prefer a voluminous bread loaf with an airy crumb. The oven rise, the fast volume expansion during baking, is crucial for a well-risen bread loaf and is at least partly governed by the yeast's activity during the early baking phase (Janssen et al., 2021). In industrial baking, various additives, called bread improvers, are often added to ensure consistent superior bread quality (Gioia et al., 2017). A better understanding of the extent to which yeast can function in breadmaking could reduce the need for such costly additives that are often perceived as unnatural by consumers (Haen, 2014). Recent research has shown that the choice of Saccharomyces cerevisiae strain significantly affects the extent of oven rise, opening the door for the selection of alternative baker's yeasts showing improved oven rise performance (Liza-Diaz et al., 2026; Vandenbosch et al., 2025). More specifically, a strong positive correlation between oven rise and viability of the yeast strain during the baking phase was observed (Vandenbosch et al., 2025), suggesting that bioprospection to find strains with superior temperature resistance might identify novel yeasts that yield better oven rise.

A particularly interesting species in this respect is Kluyveromyces marxianus (Lane and Morrissey, 2010). Originally isolated from dairy sources, such as yoghurt or cheese, the species has received both the QPS (Qualified Presumption of Safety) label in the EU and the GRAS (Generally Recognised As Safe) label in the USA (EFSA Panel on Biological Hazards (BIOHAZ) et al., 2026; US Food and Drug Administration, 2018). Furthermore, it increases in biomass fast and efficiently during its industrial production, even on some waste streams such as hemicellulose (Amrane and Prigent, 1998; Monteiro de Oliveira et al., 2021). Most importantly, the species has several characteristics specific to the dough matrix. It has already been shown that some strains exhibit good fermentative capacity in this system (Laurent et al., 2021; Struyf et al., 2017; Struyf et al., 2018a; Struyf et al., 2018b). Additionally, it has been reported that bread loaves show little to no differences in aroma or flavour compounds when using K. marxianus as baker's yeast (Caballero et al., 1995; Struyf et al., 2018). Importantly, the strain is inherently thermotolerant, with ethanol production recorded at temperatures exceeding 40°C and growth up to 52°C (Fonseca et al., 2008). K. marxianus is maltose-negative but has inulinase activity, meaning the cells cannot ferment maltose naturally present in flour, but it can hydrolyse up to 90% of fructan in dough (Struyf et al., 2017). The hydrolysis of fructan, a so-called FODMAP carbohydrate known to trigger intestinal discomfort in certain consumers suffering from irritable bowel syndrome and certain other disorders, could give them access to bread again, a staple food in Europe (Struyf et al., 2018). Maltose is the most abundantly present sugar in dough fermentation due to the hydrolysis of damaged starch by the flour's α- and β-amylases. However, the lack of maltose consumption is not expected to pose limitations in an industrial bread-making context because amyloglucosidases and α-glucosidases are often added as flour improvers to further hydrolyse dextrins and maltose into glucose (Laurent et al., 2021; Struyf et al., 2017a, Struyf et al., 2017b).

Electrical resistance ovens (EROs) have proved instrumental for studying yeast performance during the baking phase (Derde et al., 2014; Liza-Diaz et al., 2026; Vandenbosch et al., 2025, 2026). In EROs, the dough is not heated by convection, as in a conventional oven, but is instead placed between two metal plates through which an electric current flows. Heat is generated as current passes through the dough. The temperature is controlled and modelled so that dough heating follows the same time profile as in a conventional oven (Derde et al., 2014). ERO baking does not form a crust that can indirectly affect oven rise, and there is easy access to the dough for sampling. The use of the ERO technique helped uncover a strong correlation between yeast viability and oven rise performance (Vandenbosch et al., 2025), show a link between fermentation and baking performance (Liza-Diaz et al., 2026) and examine the effect of yeast on gas retention (Vandenbosch et al., 2026).

In this study, we evaluated whether the above-discussed observations extend across genera by first assessing the oven rise performance of eight different Kluyveromyces marxianus strains using the ERO setup and compared their behaviour to that of S. cerevisiae. We demonstrate that K. marxianus exhibits substantially greater strain-to-strain variability in oven rise and identified a K. marxianus strain that surpasses all tested S. cerevisiae strains. These findings highlight the potential within non-conventional yeast species and support further investigation of K. marxianus as a promising alternative for improving breadmaking performance. To further unravel the behaviour of K. marxianus, 5 strains were retained for further characterisation. To our knowledge, this is the first systematic study evaluating strain-dependent oven rise performance among genetically diverse K. marxianus strains while simultaneously examining the relationships between yeast viability and CO₂ release behaviour.

2. Materials and methods

2.1. Materials

Wheat flour (Surbi) was purchased from Dossche Mills in Deinze, Belgium. Sugar and salt were sourced from the local supermarket. The K. marxianus strains used in this study were from the Laboratory of Genetics and Genomics, KU Leuven (Table 1). Originally, these wild-type strains were ordered from the Fungal & Yeast Collection of Westerdijk Fungal Biodiversity Institute (CBS), the Industrial Yeast Collection DBVPG, the Natural Collection of Yeast Cultures (NCYC), the NITE Biological Resource Centre (NBRC) or the Russian Collection of Microorganisms (BKM). The strains were selected based on their performance in thermotolerance (defined as growth performance in growth medium at 42°C) at 6% glucose at pH 6.5, and at pH 3.5, as well as having adequate glycerol production (defined as a conversion rate of g glycerol/g glucose > 0.025 after 24 h of shaking incubation), according to Smets et al. (2025) and their genetic diversity as reported by Moreno-Hernández et al. (2026).

Table 1.

Kluyveromyces marxianus (Km) yeast strains used in this study and their substrate origin. Growth characteristics, such as thermotolerance, of these strains are reported by Smets et al. (2025).

Code	Database origin	Substrate origin
Km1	CBS 5668	Bantu beer
Km2	CBS 6873	Soil South Africa
Km3	NCYC 2886	Soil U.A.E.
Km4	DBVPG 3925	Natural rennet (Italy)
Km5	DBVPG 6550	Argentina
Km6	DBVPG 7126	unknown
Km7	NBRC 1777	Soil Japan
Km8	BKM Y-459	Spoilt curd (Russia)

Chemicals for the preparation of yeast culture medium included Agar No.1 Bacteriological (NEOGEN, Lansing, USA), Yeast extract granulated (Merck, Hoeilaart, Belgium), Oxoid™ Peptone Bacteriological and α-D(+)-glucose with 99% purity (both from ThermoFisher Scientific, Brussels, Belgium). Sodium chloride and potassium chloride were purchased from VWR Chemicals (Leuven, Belgium). Chloramphenicol, sodium phosphate dibasic, potassium phosphate monobasic and ethanol 96% (V/V) were purchased from Merck (Hoeilaart, Belgium).

2.2. Flour characterisation

Moisture was assessed using AACC Method 44-15.02 (AACC, A. A. of C. C., 2000), while protein content was analysed via the Dumas combustion method using a nitrogen-to-protein conversion factor of 5.7 (AOAC, A. of O. A. C., 1995). Analysis revealed a moisture content of 13.9 ± 0.0% and a protein concentration of 13.4 ± 0.1%.

2.3. Growth media and isotonic buffer

For yeast cultivation, YPD medium (1.0% yeast extract, 2.0% peptone, and 2.0% glucose in dH₂O) was used, with 1.5% agar added to prepare a solid medium. When analysing dough samples, agar plates were supplemented with a 100-fold dilution of a chloramphenicol stock solution (30 mg/mL in 70% ethanol) to inhibit bacterial growth. A PBS solution (0.137 M NaCl, 0.0027 M KCl, 0.01 M Na₂HPO₄, 0.0018 M KH₂PO₄), of which the pH was adjusted to 7.4 with HCl if required, was used in viability measurements. Sterilisation of all materials was performed using a Systec VX-55 autoclave (Analis, Namur, Belgium).

2.4. Growing yeast cultures

Yeast colonies were obtained by incubating strains from −80°C glycerol stocks (50%) on YPD agar for 48 h at 30°C. Following this, a 5.0 mL YPD pre-culture was grown overnight (30°C, 250 rpm) and then scaled up via a 100-fold dilution into 250 mL of fresh YPD. Once optical density readings indicated that the culture had reached the stationary phase, the biomass was harvested by centrifugation for 3 min at 870 g. After a washing step with sterilised dH₂O, the yeast pellet was ready for use in dough preparation. For all baking experiments, yeast addition was standardised on a dry weight basis, with 0.60 g yeast (dry weight equivalent) added per dough. Consequently, all doughs contained the same yeast biomass.

2.5. Dough preparation and electrical resistance baking

Bread doughs were made in triplicate for each analysis using the ERO as described in Vandenbosch et al. (2025). Dough was prepared utilising 100.0 g of flour (standardised to a 14% moisture basis), 2.0 g sucrose, 1.5 g NaCl, 57.0 mL H₂O, and 0.60 g (dry weight equivalent) of freshly harvested yeast. Mixing was performed for 6.0 min using a National Manufacturing pin bowl mixer. After resting at 30°C and 90% relative humidity (RH) for 15 min, the dough was punched and moulded before being placed into the acrylic holding trays (12.5 cm × 6.5 cm x 23.0 cm) of the ERO system and fermented (30°C, 90% RH) until 400 mL of CO₂ was generated. This volume was selected because it reflects standard conditions for bread making. It furthermore ensured the dough reached a height of at least 4 cm, allowing height to serve as a reliable proxy for volume (section 2.6). The time required to reach this CO₂ level was pre-determined using the Risograph (National Manufacturing, Lincoln, NE, USA) (Table S1). Each biological replicate consisted of an independently grown yeast pellet that was used to prepare an independent dough, which was subsequently fermented and baked as a separate experiment.

Following fermentation, the dough was baked for 24 min in the ERO, utilising a setup as defined in Derde et al. (2014). The dough, in the ERO holding tray, was placed between two conductive metal plates. These were linked to a proportional-integral-derivative (PID) controller (Imago 500, Jumo Automation, Eupen, Belgium), which regulated the current; as electricity passed through the dough, resistance converted this energy into heat. A type T thermocouple monitored the internal temperature, allowing the PID system to adjust voltage and follow a precise heating trajectory. This profile was set as follows: from 30°C to 42.2°C between 0 and 6 min, from 42.2°C to 67°C between 6 and 10 min, from 67°C to 87.3°C between 10 and 13 min, from 87.3°C to 98.1°C between 13 and 16 min and from 98.1°C to 100.1°C between 16 and 24 min. The time-temperature profile was calibrated to match the core temperatures of dough baked conventionally at 215°C (National Manufacturing), verified using a Datapaq MultiPaq 21 temperature logger (Datapaq, Cambridge, UK).

2.6. Dough pH at the start of baking

At the onset of the baking phase, before heating was initiated, the dough pH was recorded. This was done with an FC2023 dough probe (Hanna Instruments, Temse, Belgium) inserted directly into the sample matrix, measuring three biological replicates.

2.7. Height measurements, determining oven rise and collapse

Monitoring of dough expansion was done through the transparent ERO holding tray (Vandenbosch et al., 2025). Height measurements with a ruler were performed every minute as a proxy for volumetric change, in triplicate. The percentage increase in height over time was determined by comparing each measurement to the baseline height at 0 min of baking. Oven rise was characterised as the peak percentage increase during the baking phase, while collapse was quantified as the percentage drop from this maximum value to the final loaf height recorded immediately after 24 min of ERO baking.

2.8. Yeast viability during baking

Yeast viability during ERO baking was measured as described by Vandenbosch et al. (2025). For three loaves, every 3 min, about 3 g of dough was extracted from the ERO tray. To avoid border effects and ensure thermal consistency, the outer 0.5 cm layer was removed as per Derde et al. (2014), resulting in about a 2 g sample for analysis. The exact weight of the dough sample was taken into account in further analysis. The dough sample was homogenised in a 1:10 ratio with PBS using a Waring 8011E blender (Waring Products, Torrington, CT, USA) for 20 s, with the release of starch serving as a visual indicator of successful homogenisation. Samples were then serially 10-fold diluted and spread on YPD plates containing chloramphenicol (30 μg/mL). Following 48 h of growth at 30°C, viable counts were recorded for plates within the 30 to 300 colony range. Viability was defined as the ratio of the cell count at a given time point to the count at the onset of baking.

2.9. CO₂ release from dough

The CO₂ released from the dough during baking was measured in triplicate as in Vandenbosch et al. (2026). Briefly, the ERO tray was tightly closed, and a near-infrared CO₂ sensor (CO₂Meter, Ormond Beach, Florida, USA) was connected to the top of the box, which recorded CO₂ concentrations in the headspace at 6-s intervals. A rapid increase in headspace CO₂ concentration indicated the onset of bulk CO₂ release from the dough. To pinpoint this event, we identified the inflexion point between the tangent of the initial linear baseline and the subsequent exponential increase in the headspace CO₂ profile. These data were corrected for background environmental CO₂.

2.10. Statistical analysis

Data were analysed for significance using a one-way ANOVA in JMP Pro 15 (SAS Institute Inc., Cary, NC, USA). To further evaluate differences, a Tukey HSD post-hoc test was performed for all-pairs comparisons (n = 3 biological replicates per strain). This was done for the oven rise and collapse data, the viability data at each time point, the pH measurements, as well as the CO₂ release parameters. The ANCOVA model used to compare correlations in section 3.3. was performed in JMP Pro 15 using oven rise as the response variable, viability as a continuous covariate, genus as a categorical factor, and the viability × genus interaction term to test for differences in slope between genera. All statistical conclusions were based on a significance level of p < 0.05.

3. Results and discussion

3.1. Selection of genetically diverse K. marxianus strains with adequate dough fermentative power

To obtain a broad range of performance, 8 genetically diverse K. marxianus strains were selected from the study of Moreno-Hernández et al. (2026). In that study, the phylogenetic relationships among 178 K. marxianus strains were examined (Fig. 1A). Strains indicated in red originate from natural environments and agave fermentations, whereas those indicated in blue are used in dairy and beer fermentations.

Fig. 1.

K. marxianus strains (Km) which are used in this study. A. The strains were received from the study of Moreno-Hernández et al. (2026) and were chosen to be a genetically diverse representation from their collection of 178 K. marxianus strains. They were either isolated from natural environments and agave fermentations (indicated in red) or from dairy and beer fermentations (indicated in blue). B. The CO₂ production (ml) of the selected strains in a dough environment as a function of fermentation time.

The selected strains were subsequently screened for their CO₂ fermentation capacity in a dough matrix (Fig. 1B). A standardised fermentation was developed where fermentation was stopped when 400 mL of CO₂ was formed, ensuring that only baking-phase effects were compared. Because strains Km2 and Km4 required more than 200 min to reach 400 mL CO₂, these strains were not included in further tests. Fermentation times exceeding 200 min would not be industrially relevant, nor, more importantly, would they make it more difficult to compare these results to earlier literature. The exact time required for each strain to reach 400 mL CO₂ is provided in Table S1. Nevertheless, because yeast addition was standardised on a dry weight basis rather than on cell number, strain-dependent differences in CFU counts may have contributed to small differences in CO₂ production during baking.

3.2. Some K. marxianus strains outperform S. cerevisiae baking strains in oven rise performance

The effect of various K. marxianus yeast strains on the oven rise of the dough was assessed by monitoring dough height throughout the ERO baking procedure. Because of the large inconsistency in oven rise observed for strain Km3 (coefficient of variation was 54%, whereas all other strains showed a coefficient of variation smaller than 15%), this strain was excluded from further analysis (oven rise and collapse performance using strain Km3 is visualised in Supplementary Fig. S1). Oven rise profiles of the other strains are shown in Fig. 2A, and the corresponding oven rise and collapse percentages are shown in Fig. 2B.

Fig. 2.

Oven rise and collapse characteristics when baking with different K. marxianus (Km) strains. A. Dough height increase (%) during baking relative to the starting height of the dough at the end of fermentation. B. Oven rise and collapse data (%) derived from panel A. Mean values for both oven rise and collapse were analysed separately by one-way ANOVA followed by Tukey HSD post-hoc testing. Values with a different capital letter are significantly different (p < 0.05).

Statistical differences in oven rise and collapse profiles were observed for doughs made with the selected K. marxianus strains. In terms of oven rise, a rise of only 13.3 ± 2.0% was observed for strain Km5, compared to 41.5 ± 1.0% for strain Km8. Additionally, the dough inoculated with strain Km5 continued to rise for approximately 9 min during baking, whereas the dough inoculated with strain Km8 continued to rise for about 11 min. For collapse, smaller differences were observed. Km5 again performed worst, with a collapse of 6.5 ± 1.1%, while baking with strains Km1, Km7 and Km8 resulted in less collapse, with Km7 leading to only 2.0 ± 0.9% collapse. These values were compared with the performance of the 14 industrial S. cerevisiae strains used in the study by Vandenbosch et al. (2025), which includes an industrially used baker's yeast as a reference strain (Fig. 3).

Fig. 3.

Oven rise (%) performance of 5 K. marxianus (Km) strains compared with 15 S. cerevisiae strains. Oven rise performance values of S. cerevisiae, including the black dots representing an industrially used baker's yeast, are taken from Vandenbosch et al. (2025).

Even though almost twice as many S. cerevisiae strains were evaluated compared to K. marxianus, the latter showed a broader range of performance (13.3–41.5%). Moreover, Km8 showed a significantly stronger oven rise compared to the industrial baker's yeast benchmark (which showed an oven rise of 23.0%) and the best-performing S. cerevisiae strain (with an oven rise of 31.5%) included in the study of Vandenbosch et al. (2025). Similarly, on the other end of the spectrum, the worst-performing K. marxianus strain, Km5 (with an oven rise of 13.3%), yielded a lower oven rise than the worst-performing S. cerevisiae strain. Because these data were obtained using the same yeast dosage, dough formulation, fermentation standardisation, and ERO baking protocol as those used by Vandenbosch et al. (2025), the observed differences likely reflect biological variation between the tested strains rather than methodological differences. Nevertheless, the absolute oven rise values should not be directly extrapolated to other baking conditions or industrial processes.

Comparing these data to other previous research is difficult because oven rise values are often not reported, and because different experimental conditions, such as different dough formulations, different dosages of yeast, added sucrose, or the addition of enzymes, can also strongly influence the results (Dobraszczyk, 2004; Hemdane et al., 2018; Masure et al., 2019). For example, while Struyf et al., 2017a, Struyf et al., 2017b did not focus on oven rise, they reported a lower specific volume when using a K. marxianus strain compared to a S. cerevisiae strain. Part of the difference was attributed to the use of equal fermentation times for both strains, despite the K. marxianus strain's lower CO₂ production rate. Furthermore, the ERO method used in the current study prevents crust formation, thereby increasing oven rise potential relative to conventional baking. Although Liza-Diaz et al. (2026) also used ERO baking to compare the performance of several industrially produced S. cerevisiae strains during fermentation and baking, they standardised the fermentation phase differently, either by dough height (80 mm) or by a fixed time (70 min), yielding oven rise values of 36 to 90% and 43 to 113%, respectively. Moreover, several other parameters in their study also differed from the present study, including (i) a three-fold higher yeast concentration (5.8% commercial cream yeast), (ii) an addition of 5% soybean oil and 0.5% improver, likely enhancing gas retention and (iii) the different fermentation standardisation methods independent of CO₂ production. These variables seemingly lead to broader variation in oven rise performance. Standardising CO₂ production to 400 mL at the onset of baking in the current study resulted in variable fermentation times and dough heights across strains. By contrast, the CO₂ levels in dough in the Liza-Diaz et al. (2026) study likely varied at the onset of baking. According to the ideal gas law, making the assumption that in both studies the pressure and temperature rise in both systems would be equal, where the volume of a gas is proportional to the amount of substance present and the temperature, a greater initial CO₂ volume would result in a higher dough volume, an effect magnified at baking temperatures. While this ignores gluten network resistance, dynamic yeast activity, and the limitations of ideal gas assumptions in a complex dough matrix, it underscores how seemingly trivial choices in experimental design can induce significant differences in oven rise results between studies. Therefore, the superior oven rise observed for K. marxianus strain Km8 should be interpreted within the context of CO₂-standardised fermentation conditions, in which this study used the same method as Vandenbosch et al. (2025) described using S. cerevisiae strains, but does not necessarily translate directly to industrial baking processes. Although the ERO provides a reproducible platform for mechanistic investigations of the baking phase, its heating characteristics differ from those of conventional industrial baking. Therefore, validation under commercial baking conditions remains necessary in future work. For S. cerevisiae strains, Vandenbosch et al. (2026) performed an initial correlative comparison, assessing the reproducibility between ERO and conventional baking based on over-rise performance following a standardised fermentation that produced 400 mL CO₂. The results showed that yeast strain rankings remained consistent across both baking methods. Further work is needed to confirm this relationship and to investigate it using a standardised fermentation time, as is common industry practice.

3.3. Yeast viability in dough and oven rise performance are positively correlated

The viability of the K. marxianus strains was evaluated as an indirect proxy for the impact of thermotolerance on the oven rise performance. To assess survival dynamics, yeast viability was measured at 3-min intervals during baking (data provided in Table S2), with the resulting relative viability for each selected strain visualised in Fig. 4A.

Fig. 4.

Viability of K. marxianus (Km) yeast strains in dough during baking. A. Relative viability (%) of each strain throughout the baking phase was determined by dividing the cfu/g dough at a specific time point by the initial cfu/g dough for that dough at the start of baking. A T-coupled temperature probe at the centre of the dough measured the dough temperature. Significant differences in strain viability were observed at 9 and 12 min of baking (*). B. A zoom-in on relative viability at the 9- and 12-min baking mark. Significant differences between mean values at one time point are indicated by different capital letters (one-way ANOVA followed by a post-hoc Tukey HSD test, p < 0.05).

All yeast strains survived the first 6 min of baking (dough temperature of 42°C). At the 9-min sampling point (a dough temperature of 59°C), viability declined for some strains. Strains Km5 and Km7 showed significantly lower viability (71.4 ± 4.2% and 80.2 ± 2.8%, respectively) than strains Km1 and Km6 (102.5 ± 11.0% and 101.7 ± 4.1%, respectively). After 12 min of baking, at a dough temperature of 77°C, these initial differences in viability increased further, with strain Km8 showing a viability of 37.1 ± 3.0%, compared to 7.1 ± 1.6% for strain Km1 (4B).

When examining the viability profile of the S. cerevisiae strains reported by Vandenbosch et al. (2025), after 9 min of baking (a dough temperature of 59°C), all strains, including the industrial baker's yeast, showed steady survival. The decline in viability was only visible at the 12-min sampling point, at a dough temperature of 77°C, where the tested S. cerevisiae strains showed viability between 42.1% and 5.8%. At this 12-min sampling point, the range in viability is similar for both the K. marxianus and the S. cerevisiae strains (Fig. S2), suggesting that the inherent thermotolerance of the K. marxianus strains does not lead to a higher overall plating-based viability during the baking phase under these very harsh conditions. However, this does not exclude differences in metabolic activity, as viability determined by plating also includes cells that remain alive but contribute little to fermentation, while other cells may still be active in dough but fail to multiply after plating. It is therefore possible that K. marxianus maintains higher fermentative activity despite comparable viability. To validate this hypothesis, a method is required to accurately monitor CO₂ production throughout the baking phase directly within the dough matrix, which remains a significant experimental challenge.

At the 12-min baking point, a positive correlation between viability and oven rise performance of the strains was observed (R² = 0.77, p = 0.050; Fig. 5A). No significant correlation was observed for any other sampling points. Notably, the correlation found in this study is broadly consistent with the one reported by Vandenbosch et al. (2025) (R² = 0.95, p = 0.005), as visualised in Fig. 5B.

Fig. 5.

Correlation between the oven rise and strain viability at 12 min of baking for A. the K. marxianus (Km) strains in this study, B. the S. cerevisiae (Sc) strains from the study of Vandenbosch et al. (2025), C. data of K. marxianus and S. cerevisiae strains combined, and D. the combined data of both studies in separate correlations (K. marxianus strains: dotted line, S. cerevisiae strains: dashed line, combined data from panel C: dash-dot line). For each correlation, regression analyses were performed using strain mean values (n = 5 strains, with three biological repeats for each strain).

The positive relationship between oven rise and yeast viability observed in K. marxianus is consistent with previous observations in S. cerevisiae, suggesting that a similar mechanism of yeast functionality may affect oven rise beyond a single yeast genus. To assess whether this trend was similar across both genera, the correlation was calculated for all measurements combined from S. cerevisiae and K. marxianus (Fig. 5C). When combined, the correlation holds, although the R² value decreases to 0.55 (p = 0.014). In Fig. 5D, each tested genus shows a specific trend. While the correlation between yeast viability and oven rise during baking was observed in both datasets, the differing trendlines may indicate differences in how the two species interact with the dough system during baking. However, as the datasets originate from separate studies, this observation should be interpreted cautiously. This was further observed in an ANCOVA test, which shows that the interaction between genus and viability was significant (p = 0.044) (See Supplementary Table S3 for the full model). Moreover, the slope of the viability-oven rise correlation is steeper for the K. marxianus strains than for the S. cerevisiae strains (Fig. 5A compared to Fig. 5B), indicating that a 1% increase in viability translates into a larger predicted increase in oven rise for K. marxianus. While it is a cross-study comparison, this suggests that strain-to-strain differences in viability contribute more strongly to the oven rise performance for K. marxianus than for S. cerevisiae strains, although other strain-dependent factors could be involved in causing this viability-oven rise link. However, direct comparisons across more genera are needed to confirm this effect.

It is worth noting that, while Liza-Diaz et al. (2026) assessed the viability of three commercial strains, they did not identify a significant difference in viability or a direct correlation with oven rise performance. As discussed before, this discrepancy may stem from their fermentation standardisation based on dough height, which likely led to varying CO₂ concentrations at the onset of baking. Oven rise occurs only partly due to yeast activity and partly due to the lower solubility and greater expansion of the CO₂ already present from the fermentation phase (Janssen et al., 2021). Furthermore, differences in fermentative power could lead to varying levels of secondary metabolites, potentially influencing dough stability (Vandenbosch et al., 2026). These complex mechanisms show the need to validate these findings further, for example, by confirming this correlation by testing other non-conventional yeast genera, such as Wickerhamomyces anomalus, which is found in sourdough (Pradal et al., 2025; Pradal and De Vuyst, 2025) or by evaluating these trends under conventional baking conditions without standardised fermentation, a point also noted by Liza-Diaz et al. (2026).

Since the different trendlines suggested that the mechanisms by which the K. marxianus strains affected the dough matrix differed from those of the S. cerevisiae strains, the dough pH at the end of fermentation and CO₂ release during baking were monitored using the methods in Vandenbosch et al. (2026), in which the effect of the same S. cerevisiae strains as in Vandenbosch et al. (2025) on the gas retention of dough whilst baking was studied.

3.4. Different K. marxianus strains lead to different rates of CO₂ release from dough during baking

The dough pH at the end of fermentation was 5.5 ± 0.0 and did not statistically differ across strains (data shown in Supplementary Table S4). To examine how the yeast strains affected the timing of dough structure setting and overall gas-cell coalescence, CO₂ release from the dough during baking was measured. The release of CO₂ occurs in two distinct phases during baking, as visualised in Fig. 6. All CO₂ release data are presented in Table 2.

Fig. 6.

CO₂ release in the headspace above the dough during electrical resistance oven (ERO) baking using the different K. marxianus (Km) yeast strains. The error bars given are the representation of the standard deviation of three biological repeats.

Table 2.

Parameters describing CO₂ release to the headspace of the electrical resistance oven (ERO) during baking of dough with different K. marxianus (Km) strains. The CO₂ release during the baking phase can be described by a sigmoidal curve with an initial linear segment followed by an exponential segment. The transition between the two segments is the time point at which bulk CO₂ release begins. Statistical differences between mean values (columns) were tested with a post-hoc Tukey test (p < 0.05) and are indicated by capital letters.

Yeast strain	Rate of CO2 release during the linear part (mL/L*min)	Time point of onset of bulk CO2 release (min)	Rate of CO2 release during the exponential part (mL/L*min)	CO2 concentration at the end of the baking phase (mL/L)
Km1	1.09 ± 0.33AB	9.6 ± 1.4A	12.5 ± 0.9AB	123.5 ± 16.3AB
Km5	1.81 ± 0.61A	10.1 ± 0.7A	14.7 ± 0.4A	156.5 ± 12.7A
Km6	0.35 ± 0.47B	11.1 ± 0.9A	11.1 ± 0.9B	120.2 ± 18.4AB
Km7	0.17 ± 0.17B	11.4 ± 1.2A	11.0 ± 1.2B	95.1 ± 6.4B
Km8	0.50 ± 0.37B	10.8 ± 0.7A	11.9 ± 0.5B	113.4 ± 6.4B

The rate of CO₂ released from the dough varied with the yeast strain used, suggesting that the strains (de)stabilise the dough matrix differently or to varying degrees. Based on the overall CO₂-release behaviour across the measured parameters, the strains appeared to fall into three qualitative categories, with the best-performing strains generally showing slower CO₂ release than the poorest-performing strains. Km5 performs poorly and leads to more CO₂ release, while Km1 and Km6 show intermediate performance, and Km7 and Km8 are the best-performing strains. Surprisingly, despite large differences in the rate of CO₂ release, no differences in the onset of CO₂ release were observed between the best and worst-performing strains. The CO₂ release during baking was significantly faster when using Km5 than for Km7 or Km8, both in the early baking phase and during structure setting. Likewise, the CO₂ concentration at the end of baking was significantly higher for Km5 than for Km7 or Km8. Notably, Km5 also resulted in a significantly lower oven rise than Km7 and Km8, the strains that produced the highest oven rise among the K. marxianus strains tested (Fig. 2). The strains with intermediate CO₂ release performance also showed intermediate oven rise performance.

The release of CO₂ from the baking dough into the headspace is a two-step process driven by structural changes induced by the temperature increase. First, at lower baking temperatures, gas production and CO₂ expansion increase, leading to a very slow, gradual increase in CO₂ released into the headspace, either due to evaporation of CO₂ at the surface of the dough or as overextended gluten strands break and allow small groups of gas cells to coalesce. Second, once the heat increase triggers water redistribution due to starch gelatinisation and gluten polymerisation, the network becomes rigid, and coalescence occurs throughout the dough, leading to a surge in gas escaping from the matrix. When baking with each of the yeasts, those two processes occur, as can be seen in Fig. 6. The intersection between the two steps is defined as the moment at which the “foam”-like dough structure becomes a fixed “sponge”-like bread structure (Grenier et al., 2021; Janssen et al., 2021; Monterde et al., 2025). Consistent with the literature, the timing of this structural setting (CO₂ release) was unaffected by the different yeast strains used in this study (Alpers et al., 2022). Even more so, the time point of structure setting using the K. marxianus strains did not statistically differ from the time point of structure setting when using the S. cerevisiae strains, as reported by Vandenbosch et al. (2026). Connecting the CO₂ release data directly to the oven rise performance of the strains remains challenging, as no single parameter correlates directly with oven rise for either the K. marxianus or the S. cerevisiae strains. The data suggest that differences in CO₂-release behaviour may be associated with differences in oven-rise performance. One possible explanation is that strain-dependent differences in dough matrix stability contribute to these observations, although this hypothesis requires further experimental validation.

To better understand the factors associated with the observed differences in oven rise performance, a comparison was made between the best- and worst-performing strains, Km8 and Km5, respectively, taking their viability during baking into account. Strain Km5 led to a significantly faster CO₂ release throughout the baking process, as well as to a poor oven rise. Although Km5 viability dropped significantly after 9 min of baking (59°C dough temperature), it did not possess the lowest viability by the 12-min mark (77°C dough temperature). These observations indicate that viability alone cannot fully explain the differences in oven rise performance observed between strains. One possible explanation is that strain-dependent cell death may influence the concentration of yeast-derived compounds that affect dough properties during baking. At the point of cell death, reducing agents could hypothetically be released from the yeast's cytosol, such as glutathione, which could negatively impact dough integrity (Lagrain et al., 2007). This train of thought could explain, in part, the dough's apparent instability when baking with Km5. Strain Km8 leads to the highest oven rise and shows high viability at 9 and 12 min (88.4 ± 4.6% and 37.1 ± 3.0%, respectively), suggesting residual metabolic activity during the critical phase. While Km8 released CO₂ significantly more slowly than Km5, it did not have the lowest overall release rate. In absolute values, Km7 led to the slowest CO₂ release, and even though the strain maintained relatively high viability (80.2 ± 0.7% at 9 min and 23.4 ± 3.2% at 12 min baking, respectively) and the slowest CO₂ release, it did not achieve the highest oven rise. Together, these observations suggest that oven rise is likely determined by multiple interacting factors rather than by viability or CO₂ release alone. There is a need for more precise CO₂ measurements to completely unravel the influence of yeast's CO₂ activity during baking. As discussed before, when temperatures increase, the volume expansion of CO₂ in the gas cells is much larger than during the fermentation phase. Under these conditions, even the smallest differences in CO₂ production by yeast during the first part of baking probably can lead to much larger, visible changes, such as a significant difference in oven rise.

There is a need for more in-depth observations of the K. marxianus strains’ metabolite profiles, both metabolites such as organic acids which are continuously excreted, as well as reactive stress metabolites such as glutathione which might be released during cell death to get a grasp on the causal reasons for the observations made in the current study. Only then can the possible effects on the gluten matrix during breadmaking of these metabolites uncover the reason for differences in, or the extent of, (de)stabilisation of the dough matrix. While the identical dough pH across all strains at the end of fermentation suggests no major differences in organic acid production, Vandenbosch et al. (2026) showed that small strain-dependent concentration differences could still affect the gluten matrix, even when working with standardised fermentation volumes (400 mL CO₂). In the present study, the K. marxianus strains required substantially longer fermentation times to reach 400 mL CO₂ (ranging from 110 min for Km7 to 150 min for Km6), whereas the S. cerevisiae strains reported by Vandenbosch et al. (2026) reached the same CO₂ level in approximately 90 min. The ranking of fermentation times did not mirror the ranking of oven-rise performance, suggesting that fermentation duration alone is unlikely to explain the observed differences between strains. These longer fermentation times may have amplified strain-specific metabolite differences in K. marxianus compared with S. cerevisiae. These larger strain-specific differences in metabolite concentrations could thus lead to a greater impact on gluten network (de)stabilisation than previously observed.

The correlations between viability and oven rise performance in this study seem to suggest that viability (or otherwise framed cell death) has a larger influence on the oven rise performance for K. marxianus than for S. cerevisiae (Fig. 5). One possible explanation is that the inherent thermotolerance of K. marxianus may be associated with differences in intracellular stress-response metabolites. However, this possibility remains speculative and will require direct biochemical or metabolomic investigation (Kosaka et al., 2022). However, quantitative studies confirming this hypothesis were not found, especially not in a dough matrix, where quantifying reducing agents such as glutathione remains technically challenging.

4. Conclusions

This study identified a positive association between yeast viability during baking and oven rise among the K. marxianus strains evaluated. This observation is consistent with previous findings in S. cerevisiae, although additional strains and species are needed to determine the broader applicability of this relationship. Moreover, we show that specific K. marxianus strains, showing high thermotolerance in the dough matrix, could serve as a promising alternative to traditional baker's yeast that demonstrates significantly greater oven rise performance, using the same ERO setup. A positive correlation between the K. marxianus strains' oven rise performance and their viability during baking was found (R² = 0.77), similar to what was observed for a set of S. cerevisiae strains (Vandenbosch et al., 2025) which may indicate differences in how these species interact with the dough system during baking. These findings support the hypothesis that maintaining cell viability during the early heating phase contributes to improved oven rise performance and suggest that this relationship may extend beyond S. cerevisiae. Further validation across additional yeast species and strains is required.

While differences in CO₂-release behaviour were observed between strains, no individual CO₂-release parameter directly correlated with oven rise performance. Together, the results suggest that oven rise is determined by multiple interacting factors. The poorer performance of strain Km5 was characterised by faster CO₂ release and higher final headspace CO₂ concentrations, whereas the best-performing strains combined relatively high viability with slower CO₂ release. However, the mechanisms underlying these differences remain unresolved and require direct biochemical and metabolomic investigation.

Ultimately, these findings highlight the potential of the inherently thermotolerant species K. marxianus as a promising baker's yeast to improve oven rise performance. From an industrial perspective, identification of superior K. marxianus strains may represent an interesting avenue for future investigation aimed at improving loaf volume and evaluating opportunities for reducing dependence on dough improvers, thereby further supporting clean-label bread production. However, validation under conventional baking conditions remains necessary.

CRediT

Sara Vandenbosch: Conceptualisation, Methodology, Formal analysis, Investigation, Validation, Data Curation, Writing – original draft, Writing – review and editing, Visualisation. Yamina De Bondt: Conceptualisation, Methodology, Writing – review and editing. Steven J. Simmonds: Conceptualisation, Methodology, Writing – review and editing. Jan Steensels: Conceptualisation, Methodology, Writing – review and editing. Kevin J. Verstrepen: Writing – review and editing, Supervision. Christophe M. Courtin: Conceptualisation, Methodology, Writing – review and editing, Supervision, Resources, Project administration, Funding acquisition.

Data availability

The data is available via 10.5281/zenodo.20511476.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work, the authors used CoPilot (Microsoft, CA, USA) to enhance the language quality. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Funding

The authors of the Laboratory of Food Chemistry and Biochemistry acknowledge financial support from the project HealthFerm, which is co-funded by the European Union under the Horizon Europe grant agreement No. 101060247 and the Swiss State Secretariat for Education, Research and Innovation (SERI) under contract No. 22.00210. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or European Research Executive Agency (REA). Neither the European Union nor REA can be held responsible for them. Christophe Courtin and Sara Vandenbosch (Laboratory of Food Chemistry and Biochemistry) acknowledge Internal Funds KU Leuven BOF for the financial support of the C1 Yeastbake project (grant number: C14/18/069). Yamina De Bondt thanks the FWO for a postdoctoral research fellow position (FWO 12B3723N).

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: