Thermal Stability and Transformation of Chlorogenic Acids in Stevia (Stevia rebaudiana) Products: Degradation Pathways, Acyl Migration, Metal-Ion Catalysis, and Processing Implications

1. Introduction

Stevia (Stevia rebaudiana (Bertoni) Hemsl.) is a perennial herb of the family Asteraceae native to the alpine grasslands bordering Paraguay and Brazil in South America. Stevia was introduced to China in 1977 and is currently managed as a vegetable crop; the planting area has reached 20,000–25,000 ha, with an annual output exceeding 100,000 tons. Stevia is best known for its diterpene glycosides (steviol glycosides), whose sweetness is 200–300 times that of sucrose; owing to their non-cariogenic properties and suitability for diabetic consumers, steviol glycosides are widely used in the food, pharmaceutical, and personal-care industries, with a global market volume of ~10,000 tons [1].

In addition to steviol glycosides and flavonoids, stevia leaves contain approximately 7% (HPLC) chlorogenic acids, of which the dicaffeoyl-substituted isochlorogenic acids (diCQAs) account for up to ~5% [2,3]. The diCQA content in stevia is markedly higher than that in traditional medicinal/industrial plants such as Lonicera japonica and Eucommia ulmoides [3]. However, the content and composition of chlorogenic acids vary considerably among cultivars [2]. Isochlorogenic acids (dicaffeoylquinic acids) comprise three positional isomers—isochlorogenic acid A (3,5-diCQA), B (3,4-diCQA), and C (4,5-diCQA) [4]—and exhibit multiple activities, including regulation of glucose metabolism [5,6], hepatoprotection [7,8], anti-inflammatory [9], and antibacterial effects [10,11]. Because of the higher degree of caffeoylation and the o-dihydroxyphenol (catechol) moieties, diCQAs generally display stronger antioxidant, anti-inflammatory, and antiviral activities than mono-caffeoyl chlorogenic acids [2,4,9]. Stevia-derived isochlorogenic acids have been approved in China as both a new food ingredient and a new feed additive and are now produced at an industrial scale, where stevia-derived chlorogenic acid is a byproduct/co-product of steviol glycoside production [3].

Caffeoylquinic acids are intrinsically labile. The o-dihydroxyl structure of chlorogenic acids is prone to decomposition under high temperatures, light, and pH changes, which lowers the extraction yield and product integrity [12,13,14,15,16,17]. Quantitative studies on mono-caffeoyl chlorogenic acids have established that 5-CQA degrades in aqueous solution under heating [12], that its thermal stability is strongly pH-dependent [13], that alcoholic extraction can generate esters and amine derivatives [14,15], and that roasting drives CQA degradation and lactone formation in coffee [16,17,18]. For diCQAs, Deshpande et al. demonstrated that acyl migration among the mono- and dicaffeoylquinic acid isomers occurs under aqueous basic, aqueous acidic, and dry-roasting conditions via hydrated intermediates [19], and Clifford et al. established LC–MSⁿ diagnostic rules that discriminate the six diCQA isomers [21]. Xue et al. reported the stability and degradation of CQAs under different storage conditions [22]. Within our own group, earlier work screened the influence of light, temperature, dissolved oxygen, pH, and metal ions on stevia chlorogenic acid content [23], and a subsequent UPLC-QTOF MS/MS study identified dehydration, degradation, hydrolysis, and isomerization as the main degradation routes of stevia CQAs under solvent, light, heat, and pH stresses [24].

Nevertheless, three aspects remain unresolved for industrial practice: (i) a quantitative temperature–time envelope for the solid diCQA product under realistic processing conditions (drying, coating, extrusion) has not been established; (ii) the complete solid-state transformation network (degradation, methylation, acetylation, lactonization) and its temperature dependence, including the transient accumulation and subsequent degradation of these derivatives, are not described quantitatively; and (iii) the interplay between metal-ion complexation (a common unavoidable contaminant in processing water and vessels) and light/oxidative loss, and its mechanistic basis, has not been reported for stevia products.

In this study, a commercial stevia isochlorogenic acid product (total acids 61.39%, diCQAs 53.79%) was subjected to solid heat treatment at 80–180 °C and solution treatment at 60–80 °C, metal ions, light, oxygen, pH, and solvent variations following our standardized stability protocols. Transformation products were identified using UPLC-QTOF MS/MS. The results provide a quantitative stability map—thermal thresholds, transformation pathways, and metal-ion catalytic behavior—that guides processing, storage, transport, and end-use of stevia chlorogenic acid ingredients.

2. Materials and Methods

2.1. Materials and Reagents

Stevia chlorogenic acid product (batch no. 2-0969-200613; total acid content 61.39%; diCQA content 53.79%) was supplied by Chenguang Biotech Group Co., Ltd. (Handan, China). Reference standards of 3,5-diCQA, 4,5-diCQA, 3,4-diCQA, 3-CQA, 4-CQA, 5-CQA, and caffeic acid (purity >95%) were obtained from Chengdu Mansite Biotechnology Co., Ltd. (Chengdu, China). Acetonitrile, formic acid, and trifluoroacetic acid were of chromatographic grade. Methanol and ethanol were of analytical grade. FeCl₃ and CuSO₄ (analytical grade) were dissolved in ultrapure water to prepare 1 mol/L stock solutions of each. EDTA-2Na (analytical grade) was used at a concentration of 1 mol/L for decomplexation. Citric acid (0.1 mol/L) and disodium hydrogen phosphate (0.2 mol/L) were used to prepare pH-adjusting buffer solutions.

2.2. Instrumentation

Electric blast drying oven (101-1AB, Tianjin Taiste Instrument Co., Ltd.); artificial climate chamber (Shanghai Yiheng Scientific Instrument Co., Ltd.); peristaltic pump; thermostatic water bath; analytical balance (AUW220, Shimadzu); pH meter (Shanghai Sanxin Instrument Factory); HPLC system (e2695, Waters); and UPLC-QTOF MS/MS system (Xevo G2-XS, Waters) equipped with a Waters BEH C18 column (2.1 mm × 100 mm, 1.7 µm).

2.3. Solid-State Heat Treatment and Content-Loss Calculation

Chlorogenic acid samples were spread evenly on stainless-steel trays (thickness ≤4 mm) and heated in the oven at 80, 100, and 120 °C for 24 h, and at 140, 160, and 180 °C for 4 h. After each treatment, the content was determined by HPLC (Section 2.10). The content loss rate was calculated as follows:

Loss (%) = (A₁ − A₂) / A₁ × 100 (1)

(1)

where A₁ is the total acid content of the untreated original sample, and A₂ is that of the treated sample.

2.4. Solution Heat Treatment

Product (0.5164 g) was dissolved in 100 mL of 80% methanol and heated at 60, 70, and 80 °C for 4 and 8 h. The 80% ethanol solution was heated at 60 and 80 °C for 4 and 8 h for comparison. After heating, the solutions were cooled and weight-replenished with the corresponding solvent before HPLC analysis.

2.5. Metal-Ion Treatment and EDTA Decomplexation

0.5164 g of product was dissolved in 80% methanol to 100 mL. CuSO₄ or FeCl₃ stock solution was added slowly to reach metal-ion-to-diCQA molar ratios of 1:20, 1:10, and 1:5, and samples were left standing for 24 h. To distinguish complexation-induced detectability loss from genuine degradation, an equimolar amount of 1 mol/L EDTA-2Na was added to the 1:5 samples to decomplex; after mixing, the solutions were allowed to stand for 24 h and re-analyzed. In addition, solutions before and after metal-ion addition were filtered through a 0.45 µm membrane and subjected to full-wavelength scanning (300–400 nm) to examine λmax shifts.

2.6. Light-Exposure Tests

Solid: The product was spread on stainless-steel trays (<5 mm); after 1, 2, 4, 8, and 25 days of exposure in the climate chamber (37 °C, 35% RH, 12,000 lx), samples from four corners and the center were mixed for analysis.

Solution: 0.5164 g product/100 mL (80% methanol) was sealed; the dark control was wrapped in aluminum foil; both groups were placed in a climate chamber and sampled at 1, 2, 4, 8, and 20 days, with weight replenishment before each sampling. Parallel solution groups containing 1 mL of 1 mol/L CuSO₄ or FeCl₃ (per 100 mL) were run under the same light conditions.

2.7. Oxidative-Loss Tests

Solutions (as in Section 2.6, ± metal ions) were sealed and protected from light; air was bubbled through at a fixed rate using a peristaltic pump. Samples were taken at 1, 5, 10, and 15 days with weight replenishment; a metal-ion-free solution under identical conditions served as the control.

2.8. pH-Stability Tests

Citric acid (0.1 mol/L) and disodium hydrogen phosphate (0.2 mol/L) stock solutions were added to methanol to adjust the pH to 3.0, 4.0, 5.0, 6.0, 7.0, and 8.0. The product (0.5164 g) was brought to 100 mL with each pH-adjusted solution; each was divided into two portions, one left untreated and the other heated at 80 °C for 2 h, then analyzed.

2.9. Process-Liquid Tests

The stevia isochlorogenic acid process liquid (before and after concentration) was refluxed at 55 °C for 8 h and at 80 °C for 10 h, cooled, and weight-replenished with ethyl acetate before analysis.

2.10. HPLC Quantification and UPLC-QTOF MS/MS Identification

HPLC conditions (330 nm): column temperature, 40 °C; mobile phase A, 0.1% trifluoroacetic acid in water; mobile phase B, acetonitrile; gradient, 0–2 min 95% A, 2–10 min 95–80% A, 10–20 min 80–65% A, 20–20.5 min 65–50% A, 20.5–45 min 50–0% A, 45–48.5 min 0% A, 48.5–49.5 min 0–65% A, 49.5–50 min 65–95% A, 50–55 min 95% A; flow rate, 0.2 mL/min (LC–MS) or 1.5 mL/min (HPLC-UV).

The UPLC-QTOF MS/MS conditions (ESI⁻) were as follows: scan range, 50–2000 Da; collision gas, argon; collision energy, 20–40 V; capillary voltage, 2.5 kV; cone voltage, 40 V; source temperature, 120 °C; desolvation gas, nitrogen at 300 °C; cone gas, 50 L/h; and desolvation gas, 700 L/h. Compounds were assigned based on the exact [M−H]⁻/adduct masses and MS² fragment patterns, following the LC–MSⁿ rules of Clifford et al. [21].

2.11. Statistical Analysis

The stability data were obtained as single determinations per condition under the group’s standardized protocols and are reported descriptively; no inferential statistics were applied, and the results are therefore presented without error bars or measures of dispersion.

3. Results

3.1. Thermal Stability of the Solid Product: A Quantitative Processing Window

The effect of solid heat treatment on the total acid content is shown in Figure 1. Three regimes can be distinguished: (i) heating at ≤100 °C for 24 h caused <2.12% loss; (ii) heating at ≤140 °C for 4 h caused <9.24% loss (8.95% at 120 °C/24 h and 9.24% at 140 °C/4 h); and (iii) heating at ≥160 °C for 4 h caused severe loss (45.45% at 160 °C and 93.21% at 180 °C).

Below 140 °C, the loss rates of diCQAs and CQAs were comparable, whereas at 160–180 °C, the loss rate of diCQAs exceeded that of CQAs by ~30%, consistent with the partial conversion of diCQAs into CQAs (Figure 1). Notably, 5-CQA increased slightly in the 80–160 °C range (4.65% → 5.37% at 160 °C) and declined sharply at 180 °C (1.86%, 60.01% loss), suggesting that diCQA degradation during solid heating generates 5-CQA as a transient product.

3.2. Identification of Solid-State Thermal Transformation Products and Pathways

Untreated, mid-loss (160 °C/4 h), and near-complete-loss (180 °C/4 h) samples were profiled using UPLC-QTOF MS/MS (Figure 2, Table 1). After heating, quinic acid and acetylquinic acid increased strongly, whereas CQA, diCQA, triCQA, caffeic acid, and steviol glycosides decreased. Ferulic acid, feruloyl-caffeoylquinic acid, acetyl-dicaffeoylquinic acid, and the dicaffeoylquinic acid lactone first increased (160 °C) and then decreased or disappeared (180 °C). The observed transformation network (Figure 3) comprises (i) hydrolytic degradation of CQA and diCQA to caffeic acid and quinic acid; (ii) methylation of caffeic acid to ferulic acid; (iii) acetylation of quinic acid to acetylquinic acid; and (iv) methylation, acetylation, and lactonization of diCQA to feruloyl-caffeoylquinic acid, acetyl-diCQA, and diCQA lactone, respectively. Heat exerts a dual effect: below 160 °C, it promotes the formation of the methylated/acetylated/lactonized products, whereas above 180 °C, these derivatives themselves undergo thermal degradation.

3.3. Solution Thermal Stability: Total-Acid Conservation and Acyl-Migration-Driven Isomer Interconversion

In 80% aqueous methanol, heating at 60–80 °C for up to 8 h only slightly changed the total acid content (0.32–0.35 g/100 mL; Figure 4), and the sums of diCQAs and CQAs were essentially conserved. Monomer-level changes were highly consistent across temperatures: 4,5-diCQA decreased (−12.6% to −16.5%), while 3,4- and 3,5-diCQA increased (up to +35% and +17%, respectively); 5-CQA decreased, while 3-CQA and 4-CQA increased (up to +40% and +47%, respectively). The same pattern was obtained in 80% ethanol (Figure 5), showing that the interconversion is solvent-type-independent. These results indicate the conversion of 4,5-diCQA to 3,4- and 3,5-diCQA and of 5-CQA to 3- and 4-CQA during thermal treatment, that is, isomer interconversion toward thermodynamically more stable structures (Figure 6).

3.4. Metal Ions: Complexation Without Direct Loss, but with Catalytic Acceleration of Light/Oxidative Loss

Addition of Fe³⁺ or Cu²⁺ lowered the measured contents of diCQAs and CQAs with increasing metal concentration. However, after EDTA-2Na decomplexation, the detected contents recovered to 98.36–99.81%, demonstrating that complexation itself did not destroy the active ingredients (Figure 7). After the addition of metal ions, the solutions darkened, trace insoluble particles adhered to the vial wall, and filtration markedly reduced absorbance; the maximum absorption wavelength remained near 325 nm throughout, and metal addition did not shift λmax (Table 2). Thus, the immediate “loss” upon metal-ion addition is attributable to the poor solubility of diCQA/CQA–metal complexes and the pre-injection filtration step of the HPLC method, which removes part of the complexed active ingredients.

The complexes are not inert intermediates. Under light exposure for 20 days, the total acid loss of the metal-free control was 40.43%; addition of Fe³⁺ (48.25%) and especially Cu²⁺ (68.05%) accelerated the loss, with diCQA losses exceeding 70% in the Cu²⁺ group (3,4-: −71.83%; 4,5-: −77.67%; Figure 8). Under oxidative conditions (15 d bubbling), the metal-free control lost 9.25% of total acids (4,5-diCQA −15.93%; caffeic acid −16.67%); Fe³⁺ partly suppressed the loss (5.41%), whereas Cu²⁺ markedly increased it (28.17%; Figure 9). Fe³⁺ thus shows a mild protective effect on oxidative loss, while Cu²⁺ is a strong pro-oxidant/catalyst; Cu²⁺ catalyzes both photo- and oxidative degradation more strongly than Fe³⁺.

3.5. pH Dependence of Hydrolysis and Interconversion

Relative to the original solution (80% aqueous methanol), the total acid content was lowest at pH 3.0 and 4.0 (losses of 5.64% and 7.11%), varied only 0.19–1.97% at pH 5.0–7.0, and increased by 3.5% at pH 8.0. After heating at 80 °C for 2 h, the total acid variation in every pH environment was <0.6% (Figure 10). At pH 3.0–4.0, diCQA loss was larger than at pH 5.0–8.0, and heating had essentially no additional effect, indicating acid-catalyzed hydrolysis mainly of 4,5-diCQA. In the pH 5.0–8.0 range, pH promoted conversion of 3,5- and 4,5-diCQA toward 3,4-diCQA, and diCQA toward CQAs (preferentially 3-CQA and 4-CQA): the strong interconversion buffer kept the total acid content almost constant.

3.6. Solvent and Process-Liquid Behavior

Solid-phase light exposure for 25 days caused only 3.76% total acid loss (3,4-diCQA −5.95%; 3,5-diCQA −3.63%; 4,5-diCQA −3.89%; 4-CQA −4.33%; 3-/5-CQA <1.0%), indicating good light stability of the solid product, with strong direct light to be avoided during storage.

For process liquids, refluxing the pre-concentration liquid at 55 °C/8 h and 80 °C/10 h changed the total acid content by less than 2% (−0.39% and −1.16%), whereas the post-concentration liquid lost 2.49% and 11.07% under the same conditions, with loss of diCQAs exceeding that of CQAs (Figure 11). HPLC comparison revealed a peak near 15 min in the process liquid—assigned, on the basis of earlier data, to methylated/ethylated chlorogenic acid derivatives—that was absent both before and after heating of the purified product in methanol or ethanol. This indicates that during solvent extraction, the process matrix induces (ethyl)alkylation of chlorogenic acids at elevated temperatures, and that the concentrated process stream is more susceptible to thermal loss than the purified isochlorogenic acid product.

4. Discussion

4.1. The Solid-State Transformation Network and Its Temperature Dependence

The solid-state results define a practically exploitable processing window: total acid loss <10% at ≤140 °C/4 h, but ~45% loss at 160 °C/4 h. The LC–MS data show that the losses arise not only from hydrolytic degradation to caffeic and quinic acids [12,16,17] but also from thermally driven covalent modifications—methylation, acetylation, and lactonization—that have a pronounced temperature dependence. The dual (promoting-degrading) temperature effect below 160 °C versus above 180 °C, respectively, is, to our knowledge, newly documented here for a complex industrial chlorogenic acid product. Recent reviews of caffeoylquinic acid chemistry and of processing effects position this behavior within a general structural lability—the higher the degree of caffeoylation, the more readily the acyl ester chain degrades thermally or photochemically—and attribute the bulk of CQA loss during thermal processing to lactonization, polymerization, and hydrolysis [25,26,27]. Under milder conditions, the process is real but slow, with CGA half-lives of ca. 4–5 years estimated for green coffee during long-term storage [28]. Methylated and acetylated derivatives (ferulic acid, acetylquinic acid, feruloyl-caffeoylquinic acid, acetyl-diCQA) and the diCQA lactone behave as metastable pools: they accumulate at moderate temperature and are themselves consumed at higher temperature. The partial conversion of diCQAs to CQAs (and of diCQAs to 5-CQA, inferred from the transient rise of 5-CQA) explains why diCQA loss rates exceed CQA loss rates at high temperature, and echoes the mono/diCQA interconversion reported under aqueous and roasting conditions [19].

4.2. Solution Behavior: Acyl-Migration-Driven Structural Stabilization

In solution, total acid and class sums (ΣdiCQA, ΣCQA) are conserved even at 80 °C; however, positional isomers re-equilibrate. The observed conversions—4,5-diCQA → 3,4-/3,5-diCQA and 5-CQA → 3-/4-CQA—are fully consistent with the intramolecular acyl-migration mechanism (via hydrated intermediates) established for caffeoylquinic acids under neutral–weakly basic aqueous conditions [19,20,26,29,30]. The thermodynamic ranking of the isomers, which parallels their spatial arrangements (adjacent vs. meta, same vs. opposite orientation; e.g., PubChem CID 5281780, 6474310, 10324242), explains the net direction of conversion: 4,5-diCQA (two adjacent, same-orientation substituents) is the least stable and converts preferentially; 4-CQA (para to the quinic carboxyl) is the most stable, driving the preferential 5-CQA → 4-CQA conversion. This “unstable-to-stable” interconversion is an internal quality-preserving buffer: it minimizes total acid loss during mild processing and storage, consistent with the conserved total acid in Figure 4 and Figure 5; it likewise agrees with the degradation kinetics quantified for isochlorogenic acid A in aqueous solution [31] and with the preparative conversion of chlorogenic acid into its more stable isomers during processing [32].

4.3. Metal Ions: Complexation, Catalysis, and the Fe³⁺/Cu²⁺ Asymmetry

The immediate content drop upon metal-ion addition is a detectability artifact of insoluble complex formation (recovered by EDTA), not an actual loss—consistent with the well-established catecholate/gallate chelation chemistry of hydroxycinnamic acids, whereby the caffeic-acid catechol coordinates Fe³⁺ and Cu²⁺ into poorly soluble adducts that are dissociated by strong chelators such as EDTA [25,33]. The long-term danger, however, is catalytic: complex intermediates accelerate photodegradation and oxidative loss. Caffeic acid’s catechol and its phenolate anion coordinate Cu²⁺, and the resulting complex can undergo electron transfer to generate o-quinone and superoxide radical anions, which inflict oxidative damage and lower the energy barrier for active-component reactions [34]. This mechanism explains why Cu²⁺ (stronger complexation, per the Irving–Williams-type order Cu²⁺ > Fe²⁺ > Fe³⁺ coordination tendency; e.g., [33,34]) exceeds Fe³⁺ in accelerating photo/oxidative loss, and why Fe³⁺ can paradoxically protect against oxidative loss by trapping/sequestering oxidizable centers or by forming less redox-active complexes [33,34]. In practical terms, avoid introducing metal ions (processing vessels, water, packaging) and avoid combining acidic conditions (pH < 4) with metals, since hydrolytic and metal-catalyzed routes then superimpose (Section 3.4 and Section 3.5).

4.4. Integrated Stability Map and Industrial Recommendations

Combining all factors, the influence ranking for solid product loss is dominated by temperature (≈160 °C onset), followed by prolonged strong-light exposure; for liquid product, the loss drivers in increasing order are pH, oxidation, light, storage time, and metal ions. Recommended and testable operating rules: (1) solid handling at ≤140 °C with minimal time (extrusion ≤160 °C as an upper bound for short processes); (2) avoid metal-ion introduction at every stage; (3) sealed, opaque, cool, and dry storage (as supported by packaging-dependent CQA retention in stored roasted coffee and by the slow, multi-year CGA degradation estimated for stored green coffee [28,35]); (4) use formulated aqueous/alcoholic solutions promptly, avoid open storage, and strong direct light; (5) for concentrated process streams, minimize thermal hold time and control conditions that promote alkylation.

4.5. Limitations

Results are reported as single determinations per recorded point without measures of dispersion, reflecting the screening-oriented design of the stability trials; the study is therefore descriptive in scope. LC–MS components were quantified as relative peak areas (semi-quantitative) by UPLC-QTOF MS/MS, whereas total and monomer acid contents (%) were obtained by the group’s standard HPLC-UV protocol. The regiochemistry of the methylated, acetylated, and lactonized products was assigned from the exact masses and MS² product-ion patterns, according to published rules [21]; however, NMR confirmation of the diCQA lactone was not performed. Metal-complex stoichiometry and stability constants were not determined.

5. Conclusions

Stevia chlorogenic acid products showed good thermal stability in the solid state, with total acid loss below 10% at ≤140 °C for ≤4 h. Above 160 °C, diCQAs and CQAs underwent degradation, methylation, acetylation, and lactonization; these modification products accumulated below 160 °C and degraded above 180 °C. In an aqueous-alcoholic solution, the product is stable at 60–80 °C, and acyl-migration-driven isomer interconversion (toward 3,4-diCQA and 4-CQA) conserves the total acid content. Metal ions (notably Cu²⁺) do not directly destroy the actives but, as complex intermediates, accelerate light- and oxidation-induced loss, while Fe³⁺ mildly protects against oxidation. Hydrolytic loss occurs at pH 3.0–4.0, whereas pH 5.0–8.0 favors buffering interconversion. These findings provide quantitative guidance for the processing, storage, and application of stevia-derived chlorogenic acid products (≤140 °C solid handling; ≤80 °C liquids; metal-ion-free, sealed, opaque, cool, and dry conditions).