TD–GC–MS profiling of volatile organic compounds and potential odor relevance during insect protein production

Abstract

Black soldier fly larvae (BSFL) are increasingly used for sustainable protein production, but volatile organic compounds (VOCs) generated during rearing remain poorly characterized. This study characterized temporal VOC profiles during BSFL rearing using a single rearing unit for each of three commercial feed conditions. VOCs were collected at 5-day intervals over 15 days using passive sorbent tubes and analyzed by thermal desorption–gas chromatography–mass spectrometry, with feed-only controls analyzed in parallel. VOC signals were higher in BSFL-containing containers than in feed-only controls. Summed GC–MS peak areas were highest during early rearing in the dog−/pig-feed treatments, whereas high responses persisted until day 10 in the fish-feed treatment. Carbohydrate-associated VOCs predominated early, while protein-associated VOCs were more prominent in the protein-rich fish-feed treatment. Putative odorants, including sulfur-containing VOCs and volatile fatty acids, were also detected. These findings should be regarded as preliminary observations requiring confirmation with independently replicated rearing units.

Highlights

• •Volatile signals were higher in BSFL-containing containers than feed-only controls.
• •Protein-rich fish feed showed more prominent protein-associated VOCs.
• •Carbohydrate-derived VOCs dominated the early rearing stage.
• •Putative odorant profiles varied among feed treatments.

1. Introduction

Animal (livestock)-derived meat constitutes a central pillar of the global food system and serves as a major source of protein for humans (Ederer et al., 2023). As more than 70 billion livestock animals (including chickens, pigs, and sheep) are currently being raised worldwide, the resulting environmental burden cannot be ignored (Tullo et al., 2019). Indeed, livestock farming is a major driver of climate change and results in the release of large quantities of greenhouse gases into the atmosphere (Xing et al., 2022). In this context, insect farming has increasingly been attracting attention because its environmental footprint is comparatively low for its high protein yield (Siddiqui et al., 2023). Insects are officially classified as livestock, and insect farming is projected to contribute substantially as an alternative to conventional livestock farming for sustainable protein production (Grasso & Bordiga, 2023). In particular, black soldier fly larvae (BSFL; Hermetia illucens) offer a sustainable protein source because they can be reared efficiently on the products of organic waste streams (Kim et al., 2025). In other words, BSFL can valorize wastes into resources such as proteins and lipids.

Although BSFL can be reared outdoors, indoor/semi-indoor systems are generally preferred because they allow exact environmental control independent of external factors such as weather conditions (Givens et al., 2013). The growth and bioconversion efficiency of BSFL are affected by rearing conditions, including temperature and humidity (Deruytter et al., 2025). During the bioconversion of organic waste by BSFL, the elemental components of the waste are redistributed between larval biomass and gaseous emissions, the majority of which comprises carbon dioxide (CO₂) and nitrogen (N₂) (Parodi et al., 2021). Another component of these gaseous emissions is volatile organic compounds (VOCs), many of which contribute to odor formation (Zhang et al., 2025). The composition of these VOCs may reflect the degradation of feed components as well as metabolic activities of the larvae and associated microorganisms.

Characterizing these volatile compounds is important for understanding the chemical environment of BSFL production systems. Some VOCs are odor-active and may affect indoor air quality and occupational exposure (Zhang et al., 2024), while their profiles may also provide information on substrate degradation and larval-microbial metabolism. However, VOC generation during BSFL rearing remains poorly characterized, particularly with respect to temporal changes and the influence of feed composition (Zhang et al., 2025). Because carbohydrates, proteins, and lipids undergo different metabolic and microbial degradation pathways (Cui et al., 2022), feeds with different biochemical compositions may produce distinct volatile profiles and odor-active compounds. Thus, understanding these relationships is important for characterizing the volatile chemistry of BSFL production and for improving the monitoring/management of insect-based protein production systems.

The major objective of this study was to characterize feed-dependent and temporal VOC generation during BSFL rearing. Three commercial feeds with distinct biochemical compositions were used to examine how feed characteristics were associated with larval growth, biochemical composition, and VOC profiles. VOCs from rearing systems with and without larvae were compared to assess the contribution of larval activity and BSFL-associated microorganisms relative to the feed substrate alone. Two types of sorbent tubes were employed to detect VOCs. In addition, temporal changes in carbohydrate-, protein-, lipid-, and lignocellulose-associated VOCs and the occurrence of odor-active compounds were evaluated. Through this approach, the study aims to provide a chemical basis for understanding VOC formation during insect biomass production and to support future strategies for feed selection, odor management, and sustainable insect protein production.

2. Materials and methods

2.1. Insect rearing conditions

Four-day-old BSFL were obtained from the National Institute of Animal Science, Republic of Korea. Commercial dog, fish, and pig feed were purchased from a local market. Distilled water was added to each dry feed to adjust the moisture content to 60 wt%. A 500 g (wet weight) portion of each hydrated feed was placed in a plastic rearing box of dimensions 40 × 22 × 22 cm. One thousand larvae were introduced into each box. For each feed type, one BSFL-containing rearing box and one corresponding feed-only control box were prepared. The boxes were covered with mesh and maintained at 28 °C under dark conditions. Distilled water was supplied as needed during rearing to prevent substrate desiccation and maintain adequate moisture.

2.2. Biomass analysis

At each sampling point, twenty larvae were randomly collected from each treatment to determine the larval dry weight. Individual larvae represented subsamples from the same rearing unit and were not treated as independent biological replicates. The larvae were washed with distilled water, oven-dried at 100 °C for 24 h, and weighed using a microbalance.

The protein content was estimated as the sum of the quantified individual amino acids and was expressed as wt% of the dried sample. Amino acid analyses were performed in triplicate. Amino acids were analyzed using ion-exchange chromatography. Briefly, 0.2 g of each dried sample was placed in a hydrolysis tube and mixed with 6 N HCl. The samples were subsequently hydrolyzed at 110 °C for 24 h. After hydrolysis, the samples were diluted with 0.2 M sodium citrate buffer and filtered through a 0.2-μm membrane filter. Subsequently, the products were analyzed for methionine and cysteine after oxidation performic acid oxidation and separately analyzed for tryptophan after alkaline hydrolysis.

For lipid analysis, fatty acids were converted to fatty acid methyl esters (FAMEs) and purified using commercial methylation and purification kits, according to manufacturer instructions. The resulting products were analyzed using gas chromatography coupled with a flame ionization detector. Separation was performed using a DB-WAX capillary column with an internal diameter of 0.25 mm, length of 30 m, and film thickness of 0.25 μm. A Supelco 37 Component FAME Mix was used as the calibration standard. FAME-derived lipid content was estimated as the sum of all quantified FAMEs and was expressed as wt% of the dried sample. FAME analyses were performed in triplicate.

Ash content was determined by thermogravimetric analysis. Ten milligrams of each sample was placed in a crucible, heated to 550 °C, and maintained at this temperature for 2 h. The residue was considered to be ash. The sulfur (S) contents of the feeds were determined using an elemental analyzer. Carbohydrate content was calculated as the following difference: 100 − (protein content + lipid content + ash content).

2.3. Sampling and analysis of VOCs

VOCs generated during BSFL rearing were collected using passive samplers. Two types of sorbent tubes were used for VOC collection: Air Toxics (APK) sorbent tubes (KNR Co., Republic of Korea) and Tenax TA sorbent tubes (KNR Co., Republic of Korea). Tenax TA tubes were stainless-steel thermal desorption tubes (89 mm × 6.4 mm o.d., 5 mm i.d.) packed with 220 mg of 60/80-mesh Tenax TA in a 40-mm sorbent bed. Air Toxics tubes contained Carbograph 1 and Carboxen 1000. The tubes were conditioned at 320 °C for 3 h and impurity-checked by the manufacturer before use and were supplied sealed with brass caps and PTFE ferrules. During passive sampling, both end caps were removed, and no diffusion cap was used. Sorbent tubes were installed as passive samplers inside each rearing container (one Air Toxics tube and one Tenax TA tube per container) by fixing them with tape to the undersides of the mesh lids and were retrieved and replaced every 5 days over the 15-day rearing period (i.e., after 5, 10, and 15 days). The 5-day sampling interval was selected to obtain integrated VOC profiles across the major stages of the 15-day larval rearing period, including the early period of rapid larval growth and substrate utilization. This interval also allowed temporal changes in VOC profiles to be evaluated at three consecutive stages of larval development. Successive 5-day sampling intervals represented repeated temporal measurements from the same rearing systems rather than independent treatment replicates. Samplers installed in feed-only control containers were collected as blanks in parallel with those installed in the treatment containers. After retrieval, the tubes were immediately resealed with brass caps and PTFE ferrules, individually packaged, stored at 4 °C, and analyzed within 1–2 days.

The VOCs collected by the sorbent tubes were analyzed using an automated thermal desorption unit coupled with gas chromatography–mass spectrometry (GC–MS). Thermal desorption was performed using a MARKES automated thermal desorption system. Sorbent tubes were desorbed at 300 °C for 10 min following 1 min of purging. The cold trap was initially maintained at −30 °C and subsequently heated to 300–315 °C, with the exact temperature depending on the instrumental configuration. The transfer line temperature was maintained at 180–220 °C to prevent the condensation of semi-volatile compounds. Desorbed analytes were separated using a GC system equipped with a DB-1 capillary column (60 m length, 0.32 mm i.d., 1.0 μm film thickness). Helium was used as the carrier gas. The oven temperature program was as follows: 40 °C (initial temperature) maintained for 5 min, increased to 260 °C at 5 °C min⁻¹, and finally, 260 °C maintained for 5 min. The GC was coupled to a 5977 A mass selective detector through a GC–MS interface maintained at 280 °C. The mass spectrometer was operated in electron ionization (EI) mode at 70 eV, with the ion source and quadrupole maintained at 230 °C and 150 °C, respectively, and was autotuned with perfluorotributylamine (PFTBA) before each analytical sequence. Data were acquired in simultaneous SIM/scan mode after a 4 min solvent delay; full-scan spectra were collected over m/z 20–350 and used for compound identification, while selected-ion monitoring windows (two or three characteristic ions per compound, dwell time 100–150 ms) were recorded concurrently.

Compounds were putatively annotated by comparison of their mass spectra with the NIST mass spectral library. Assignments were evaluated using the combined evidence of library match quality, characteristic ions, and retention-time consistency across samples rather than by library match score alone. Matches ≥80 were generally retained; lower-scoring assignments were included only when supported by diagnostic-ion agreement and consistent chromatographic behavior in other samples. The chromatograms, mass spectra, and compound assignments were subsequently re-examined and cross-checked by experienced GC–MS analysts before data interpretation. To account for potential background contamination (e.g., within the storage tubes) during sample handling and transport, a procedural blank (collected from an empty rearing container without feed and larvae) was analyzed in parallel with each sampling batch (Lin et al., 2017; Yu et al., 2023). The corresponding peak areas were subtracted from the sample peak areas prior to data interpretation. Because compound-specific passive sampling rates and calibration-based concentrations were not determined, GC–MS peak areas were interpreted as relative, time-integrated analytical responses and not as absolute VOC concentrations or emission rates.

3. Results and discussion

3.1. Larval growth on different feeds

The biochemical compositions of the animal feeds were analyzed (Table 1) to determine the effect of feed composition on BSFL growth and metabolite production―as larval growth and metabolite production have been shown depend on substrate composition (Franco et al., 2026). The protein content was the highest in the fish feed (44.78 wt%), followed by those in the dog feed (25.07 wt%) and pig feed (17.40 wt%). The amino acid composition also varied among the feeds. Glutamic acid was the most abundant amino acid in all three feeds (Table S1). Leucine was the second most abundant amino acid in the fish feed, whereas aspartic acid was the second most abundant in the dog and pig feeds. The fish feed had the lowest carbohydrate content. The dog feed had the highest lipid content (7.39 wt%), followed by the fish (5.95 wt%) and pig feeds (5.45 wt%). The ash content of all the feeds was below 1 wt%. These results indicate that the fish feed was protein-rich, dog feed was relatively lipid- and carbohydrate-rich, and pig feed was predominantly carbohydrate-rich. All feeds also contained S, which may be due to the presence of S-containing amino acids, such as cysteine and methionine (Dordevic et al., 2023). As S-containing volatile compounds are major contributors to odor (Fukutani et al., 2023), a range of odorous compounds may be generated during BSFL rearing.

Table 1.

Biochemical composition of the animal feeds used in this study (wt%).

	Dog feed	Fish feed	Pig feed
Carbohydrates	67.23	49.08	76.86
Proteins	25.07	44.78	17.40
Lipids	7.39	5.95	5.45
Ash	0.32	0.20	0.30
Sulfur⁎	0.01	0.50	0.29

^⁎Sulfur represents elemental sulfur content and is reported independently of the proximate composition; therefore, it was not included in the calculation of carbohydrate content.

BSFL growth and metabolism varied among the different feed treatments (Fig. 1). Growth patterns could not be directly explained by the bulk biochemical composition of the feeds. On day 5, larvae reared on the dog feed had a higher mean dry weight (47.68 mg) than that of those reared on fish (39.74 mg) or pig (17.59 mg) feed (Fig. 1A). This result was not associated with the highest carbohydrate or protein content, as the dog feed had intermediate levels of both components (Table 1). Although the dog feed had the highest lipid content, lipid content alone could not explain the observed growth differences. The fish and pig feeds had similar lipid contents, yet the corresponding larval dry weights differed substantially. These findings suggest that early larval growth was influenced by factors beyond total carbohydrate, protein, and lipid contents, such as nutrient digestibility or other feed-specific properties. For instance, fish feed may contain microalgae as additive (Ma & Hu, 2024), whereas pig feed may contain copper as pharmacological additive (Forouzandeh et al., 2022).

Fig. 1.

(A) Mean dry weight of individual black soldier fly larvae (BSFL) reared on different animal feeds. Twenty larvae, representing subsamples from a single rearing unit, were measured at each sampling point for each feed condition (n = 20 larvae). Individual points represent individual larvae and boxes represent the interquartile range. (B) Metabolite composition of BSFL on day 15.

The mean dry weight of the BSFL reared on the pig feed increased markedly between days 5 and 10, with a gain of 19.87 mg per larva. Over the same period, larvae reared on the dog and fish feeds gained 4.68 and 5.75 mg, respectively. Despite this increase, larvae reared on the pig feed remained lighter than those on the other feed treatments by day 10. Between days 10 and 15, the mean larval dry weight increased only after the fish-feed treatment, whereas it decreased after the dog- and pig-feed treatments. By day 15, the mean dry weight was the highest in larvae reared on fish feed (52.47 mg), followed by those reared on dog (40.93 mg) and pig (35.28 mg) feeds. This order corresponds to the initial protein contents of the feeds (Table 1). Analysis of the residual substrates revealed a similar pattern in protein consumption: during the 15-day period, the BSFL under the fish-, dog-, and pig-feed treatment conditions consumed 64.91, 40.69, and 25.26 g of protein, respectively. These results suggest that protein availability and consumption contributed to the differences in larval growth among the treatments, which is consistent with the results from previous studies that identified substrate protein content as an important determinant of BSFL growth (Barragan-Fonseca et al., 2018).

Because substrate composition affected BSFL growth, it was also expected to influence larval metabolism and biochemical composition. Consistent with this expectation, the BSFL lipid and protein content differed among the feed treatments (Fig. 1B). Larvae reared on the dog feed had the highest lipid content, followed by those reared on fish and pig feeds. Lipid accumulation in BSFL increases along with the lipid content of the substrate (Kim, Kim, et al., 2024). Indeed, the high lipid content of larvae reared on the dog feed corresponded with the high lipid content of this feed (Table 1).

In contrast, the larval protein content did not directly correspond to the initial protein contents of the feeds. Although the protein content of the fish feed was more than twice that of the pig feed, larvae reared on the pig feed had a higher protein content than that of those reared on the fish feed. This discrepancy indicates that larval protein accumulation was not solely determined by the initial protein concentration of the substrate. Despite the differences in total protein content, the BSFL amino acid profiles did not differ substantially among the feed treatments (Table 2). Glutamic acid was the most abundant among the amino acids in the larvae from all treatments, followed by aspartic acid and leucine. This order is consistent with that reported in our previous study (Kim et al., 2025). These results indicate that the feed characteristics influenced the overall accumulation of major biomass components, particularly proteins and lipids, but had a limited effect on the relative amino acid composition of larval proteins.

Table 2.

Amino acid profiles of black soldier fly larvae reared on different animal feeds (wt% of total amino acids).

	Dog feed	Fish feed	Pig feed
Aspartic acid	10.44	10.85	10.73
Threonine	4.64	4.63	4.67
Serine	5.18	4.86	4.88
Glutamic acid	12.48	12.62	12.61
Proline	6.31	6.16	6.01
Glycine	6.76	6.17	6.43
Alanine	7.43	7.57	7.40
Valine	6.01	6.14	6.32
Isoleucine	4.54	4.40	4.60
Leucine	7.68	7.75	7.75
Tyrosine	6.63	6.69	6.46
Phenylalanine	4.51	4.94	4.84
Histidine	2.95	2.91	2.72
Lysine	5.88	5.88	5.99
Arginine	5.05	4.79	5.03
Cysteine	0.64	0.62	0.61
Methionine	1.82	1.95	1.83
Tryptophan	1.05	1.07	1.10

Differences in larval growth and metabolism may result in different VOC emissions, because VOCs can originate from metabolic intermediates and end products generated through nutrient degradation and biosynthetic pathways (Guo et al., 2021). To examine whether the feed-dependent differences in growth and biochemical composition were accompanied by changes in VOC profiles, VOCs profiles from the BSFL rearing systems were analyzed using passive samplers.

3.2. Volatile compound generation during insect rearing

Diverse VOCs can be produced and emitted during biological processes (PeÑUelas et al., 2014). During insect rearing, these compounds may originate from the insects themselves and the microorganisms in the larval gut microbiome or from the microorganisms present in the substrate. To distinguish VOC emissions associated with BSFL from those produced by the substrate alone, VOCs were collected from rearing boxes with and without larvae and analyzed using GC–MS (Fig. 2). Although microbial metabolism can generate substantial amounts of VOCs (Weisskopf et al., 2021), in this study, only limited VOC signals were detected in the absence of BSFL. The summed GC–MS peak area measured in the rearing boxes containing BSFL was 52-fold higher than that measured in boxes without larvae (Fig. 2B). These peak areas do not represent absolute VOC concentrations or emission rates but provide a relative measure of the overall analytical response under identical analytical conditions. The higher VOC signals in systems containing larvae suggest that the detected VOC profile was associated with BSFL activity and associated microorganisms rather than with microorganisms in the substrate alone.

Fig. 2.

(A) Gas chromatography–mass spectrometry (GC–MS) chromatograms of volatile organic compounds (VOCs) detected in black soldier fly larvae (BSFL) rearing systems with and without larvae. The sampler (Tenax TA) was exposed for 120 h. (B) GC–MS peak areas of the detected VOCs. One BSFL-containing box and one dog-feed only control box were used for this comparison. Following exposure, all tubes were retrieved and subjected to the same sealing, transport, thermal desorption, and GC–MS analytical procedures.

Another factor affecting VOC analysis is the sampler, thus requiring appropriate selection. As sorbent materials differ in their affinity and retention capacity for individual compounds, the obtained VOC profile can vary depending on the sampler type (Khan et al., 2018). Air Toxics and Tenax TA sorbent tubes were compared (Table 3) to identify a suitable sampler for VOC analysis in BSFL rearing systems. Both types of sorbent tubes are commonly used to collect a broad range of VOCs (Lamaa et al., 2013). Twelve compounds were detected in the Air Toxics tube, whereas 35 were detected using the Tenax TA tube. All compounds detected using the Air Toxics tube were also detected using the Tenax TA tube. For these shared compounds, Tenax TA produced larger GC–MS peak areas except for acetic acid. For the dog-feed sample examined under the present 120-h sampling conditions, Tenax TA detected a greater number of compounds and produced higher GC–MS responses for the compounds detected by both sorbents than Air Toxics.

Table 3.

Comparison of volatile organic compounds detected using two different samplers. One Air Toxics tube and one Tenax TA tube were positioned adjacent to each other in each container and exposed simultaneously for 120 h under identical rearing conditions. One BSFL-containing box and one dog-feed only control box were used for this comparison. Following exposure, all tubes were retrieved and subjected to the same sealing, transport, thermal desorption, and GC–MS analytical procedures.

Chemical name	Molecular formula	Air Toxics peak area (×106)⁎	Tenax TA peak area (×106)⁎
2,3-Butanedione	C4H6O2	–	23.99
2-Butanone	C4H8O	–	96.91
Acetic acid	C2H4O2	65.84	7.48
2-Butanol	C4H10O	–	9.87
Trichloromethane	CHCl3	10.36	154.25
2-Methyl-1-propanol	C4H10O	–	23.37
2-Pentanone	C5H10O	–	64.54
Acetoin	C4H8O2	–	35.19
3-Methyl-3-buten-1-ol	C5H10O	–	5.81
3-Methyl-1-butanol	C5H12O	12.55	65.72
Dimethyl disulfide	C2H6S2	92.29	611.82
2,4,5-Trimethyl-3-oxazoline	C6H11NO	–	3.29
2-Methylpyrazine	C5H6N2	–	6.84
Xylenes	C8H10	–	3.73
2-Heptanone	C7H14O	13.02	26.50
2,3-Dimethylpyrazine	C6H8N2	11.36	13.59
δ-Valerolactone	C5H8O2	–	5.76
Benzaldehyde	C7H6O	–	4.54
α-Pinene	C10H16	–	2.60
Phenol	C6H6O	17.14	45.98
Dimethyl trisulfide	C2H6S3	18.23	41.18
5-Methyl-3-heptanone	C8H16O	–	5.38
Trimethylpyrazine	C7H10N2	–	19.04
Benzyl alcohol	C7H8O	–	3.09
2-Ethyl-1-hexanol	C8H18O	–	7.02
Limonene	C10H16	–	1.77
Acetophenone	C8H8O	1.31	4.78
2,3,5,6-Tetramethylpyrazine	C8H12N2	3.25	17.56
2-Nonanone	C9H18O	6.50	13.94
Phenethyl alcohol	C8H10O	79.15	175.41
4-Ethylphenol	C8H10O	–	5.14
4-Methylguaiacol	C8H10O2	–	11.12
Dimethyl tetrasulfide	C2H6S4	–	3.78
2-Methyl-6-propylpyridine	C9H13N	–	3.66
2-Undecanone	C11H22O	–	9.06

^⁎Peak areas represent the analytical responses obtained from the simultaneously exposed sorbent tubes before subtraction of the corresponding feed-only control.

The detected VOCs primarily consisted of medium- to relatively high-boiling-point, non-polar and moderately polar compounds, including several oxygenated, S-containing, and aromatic compounds, which generally fall within the effective retention and thermal-desorption range of Tenax TA (Marcillo et al., 2017). The hydrophobic nature and low water adsorption capacity of Tenax TA may also limit interference from water vapor under the humid conditions of BSFL rearing systems (Chen et al., 2022). Based on its broader compound coverage under the conditions tested, Tenax TA was selected for the subsequent characterization of VOC profiles from the BSFL rearing systems. However, retention, saturation, back diffusion, humidity effects, and storage stability were not experimentally evaluated for the specific passive-sampling configuration. Tenax TA uptake and retention are known to depend on compound properties and exposure duration, and relatively volatile or weakly retained compounds may be more susceptible to losses during prolonged passive sampling (Jia & Fu, 2017; Walgraeve et al., 2011). Although samples were refrigerated and analyzed within 1–2 days, compound-specific storage stability was not independently verified. Accordingly, comparisons were restricted to relative GC–MS responses obtained using the same sampler type and exposure duration, and differences between sorbent types were interpreted only for the specific samples and conditions tested.

VOC profiles varied over time and among the feed treatments (Fig. 3). For interpretive purposes, putative biochemical origins, including carbohydrate-, protein-, lipid-, and lignocellulose-derived categories, were assigned based on previously reported metabolic and degradation pathways. These assignments should be regarded as tentative rather than experimentally confirmed sources. In the dog- and pig-feed treatments, the summed peak area was highest during the early rearing period (days 0 to 5) and decreased rapidly thereafter. In contrast, the fish-feed treatment maintained a high summed peak area until day 10, followed by a decline by day 15. On day 5, compounds classified as carbohydrate-derived VOCs represented the largest VOC class in all treatments, accounting for 59.02_area%, 42.07_area%, and 59.34_area% in the dog, fish, and pig feeds, respectively. This pattern was consistent with the high carbohydrate contents of the feeds (Table 1) and may reflect the preferential utilization of readily degradable carbohydrates during the early rearing stages. However, because the commercial feeds differed in multiple compositional characteristics, the contribution of carbohydrate content cannot be isolated from other feed-specific factors. Simple sugars can enter glycolysis directly (Khamkeaw & Phisalaphong, 2016); by contrast, protein utilization requires proteolytic hydrolysis to release peptides and amino acids, followed by reactions such as deamination and decarboxylation during amino acid catabolism (Song et al., 2025). This difference in metabolic processing is consistent with the possibility that readily available carbohydrates were utilized earlier than proteins under the present rearing conditions (Yang et al., 2015).

Fig. 3.

Putative biochemical origin of volatile organic compounds (VOCs) detected in black soldier fly larvae rearing systems supplied with dog, fish, and pig feeds. The peak area obtained from the corresponding feed-only control was subtracted from that of the BSFL-containing treatment. Detailed peak area values are provided in Tables S2–S4, and retention time and mass spectral identification information is provided in Table S5.

The early predominance of carbohydrate-derived VOCs was consistent with rapid carbohydrate utilization during the initial rearing period. Among these VOCs, acetoin showed the highest relative abundance in all feed treatments. Following glycolysis, acetoin can be produced by the conversion of pyruvate to α-acetolactate and then to acetoin (Li et al., 2023). Acetoin can also be reversibly converted to 2,3-butanediol by acetoin reductase (Ji et al., 2011). Consistent with this pathway, 2,3-butanediol was detected in all three feed treatments during the analysis of VOCs. The co-occurrence of acetoin and 2,3-butanediol was consistent with the possible involvement of glycolytic and fermentative pathways during the early stages of BSFL rearing. These pathways may involve both the larvae and associated microorganisms, and the detected compounds could not be assigned exclusively to carbohydrate metabolism or to a single biological source.

The total peak area of protein-derived VOCs on day 5 was the highest in the fish-feed treatment, followed by those in the dog- and pig-feed treatments. This order corresponds to the initial protein contents of the feeds (Table 1). In the fish-feed treatment, the total peak area of protein-derived VOCs increased further by day 10 (Table S3), whereas that of carbohydrate-derived VOCs decreased over the same period. This temporal change is consistent with the fact that protein degradation is slower than carbohydrate utilization (Song et al., 2025). In contrast, protein-derived VOCs decreased from day 5 to 10 in the dog- and pig-feed treatments. This difference was consistent with the lower protein contents of the dog and pig feeds; however, other compositional and physicochemical differences among the commercial feeds may also have contributed to the observed VOC patterns. The protein content of the fish feed was approximately twice that of the dog feed and more than twice that of the pig feed, thus providing a larger protein pool for continued degradation and VOC formation.

Lipid-derived VOCs accounted for only a small fraction of the total VOC profile in all of the feed treatments. This may partially reflect the lower lipid content of the feeds relative to their carbohydrate and protein contents. Moreover, triacylglycerols and most long-chain fatty acids have high molecular weights and low volatility and are therefore unlikely to be detected directly as VOCs without prior degradation or derivatization (Kim, Park, et al., 2024). Lipid-derived VOCs are formed mainly through the oxidation and subsequent decomposition of unsaturated fatty acids. During lipid oxidation, unsaturated fatty acids are initially oxidized to form hydroperoxides, which are further degraded into volatile secondary products, such as aldehydes, alcohols, and ketones (Shahidi & Hossain, 2022). Elevated temperatures accelerate these reactions and increase the formation of lipid-derived VOCs (Xie et al., 2022). However, in this study, BSFL were reared at a relatively low temperature of 28 °C under moist and dark conditions. These conditions may have limited the oxidative degradation of substrate lipids, thereby reducing the formation of lipid-derived VOCs.

Putatively lignocellulose-derived VOCs were absent or present at low levels during early rearing and were detected more prominently at later stages. Lignocellulosic biomass is relatively resistant to biodegradation because the cellulose and hemicellulose therein are embedded within a structurally complex lignin matrix, which restricts microbial and enzymatic access to these polysaccharides (Zoghlami & Paës, 2019). Thus, readily available nutrients may have been preferentially used during early rearing, whereas the gradual disruption and degradation of lignocellulosic structures may have increased the release of lignin- and plant-derived VOCs during the later stages. These results are consistent with evidence that BSFL-associated microbial communities contribute to lignocellulose degradation during rearing (Xiang et al., 2024). Collectively, the highest summed GC–MS responses were observed during the early stages of BSFL rearing, whereas compounds associated with the degradation of the more recalcitrant feed components became more prominent during the later stages. These temporal shifts demonstrate that VOC profiles vary throughout the rearing process, thus necessitating monitoring across the entire rearing period rather than at a single sampling point.

3.3. Odor and food-quality implications of VOC profiles during insect rearing

Some of the VOCs tentatively identified during BSFL rearing have been reported to possess characteristic odors and low odor detection thresholds. S-containing compounds, volatile fatty acids, phenolic compounds, and indolic compounds are associated with odor formation in biological systems (Trabue et al., 2011). The tentatively identified VOCs with reported odor characteristics are summarized in Table 4. However, the presence of these compounds does not demonstrate their actual contribution to perceived odors in the present rearing systems. GC–MS peak areas reflect the analytical response under the applied conditions, whereas odor perception depends on compound concentration, odor detection threshold, interactions among odorants, and other sensory properties (Hwang et al., 2018). Therefore, a compound detected at a relatively low abundance may contribute more to the odor than a compound with a larger peak area but a higher odor-detection threshold. For example, acetoin exhibited a high GC–MS peak area (Tables S2–S4), but its contribution to the overall odor may have been lower than that of the less abundant S-containing or indolic compounds with greater odor potency. Indeed, quantitative concentration measurements combined with olfactometric or sensory analyses would be required to determine the actual odor contribution of individual compounds.

Table 4.

Tentatively identified VOCs with reported odor characteristics during black soldier fly larval rearing.

Chemical name	Molecular formula	Reported odor description	Detection
			Dog feed	Fish feed	Pig feed
Methanethiol	CH4S	Rotten cabbage, sewage, putrid sulfurous odor		O
Dimethyl disulfide	C2H6S2	Garlic-like, rotten cabbage, decaying fish odor	O	O	O
Dimethyl trisulfide	C2H6S3	Rotten vegetables, onion-like, sewage, swampy odor	O	O	O
Dimethyl tetrasulfide	C2H6S4	Strong sulfurous and putrid odor	O
Trimethylamine	C3H9N	Fishy odor	O	O
Dimethylamine	C2H7N	Fishy, ammonia-like odor		O
3-Methylbutanoic acid	C5H10O2	Sweaty, foot-like, cheesy, rancid odor		O
Acetic acid	C2H4O2	Vinegar-like, sour, pungent odor	O	O	O
Phenol	C6H6O	Medicinal, disinfectant-like, tar-like, fecal odor	O	O
4-Ethylphenol	C8H10O	Horse-sweat, barnyard, medicinal, fecal odor	O	O
Indole	C8H7N	Fecal and putrid at high concentrations; floral at low concentrations		O
3-Methylbutanal	C5H10O	Malty, cheesy, sweaty odor	O	O	O
2-Methylpropanal	C4H8O	Malty, pungent, fermented odor	O	O	O
3-(Methylthio)-1-propanol	C4H10OS	Cooked potato, cabbage-like, sulfurous odor	O	O	O
Guaiacol	C7H8O2	Smoky, tar-like, medicinal, burnt-wood odor	O

The diversity of the detected odorants varied among the feed treatments. Thirteen VOCs with reported odor characteristics were tentatively identified in the fish-feed treatment, compared with six in the pig-feed treatment. The greater odorant diversity observed with fish feed as substrate may be related to its higher protein content. Several odorants that were detected only in the fish-feed treatment, including dimethylamine, 3-methylbutanoic acid, and indole, can be formed through the biological degradation of amino acids and other nitrogen-containing compounds (Roager & Licht, 2018; Serrazanetti Diana et al., 2011). These results indicate that feed composition influenced both the number and types of odorants generated during BSFL rearing.

Despite these feed-dependent differences, S-containing odorants (including dimethyl disulfide and dimethyl trisulfide) were detected in all treatments. Although S accounted for less than 1 wt% of each feed and was present at a lower concentration than lipids, S-containing VOCs produced strong chromatographic signals (Tables S2–S4). These results may reflect the efficient microbial conversion of bioavailable S-containing amino acids, particularly methionine, into low-molecular-weight volatiles, such as methanethiol, dimethyl disulfide, and dimethyl trisulfide. Because various S-containing VOCs have low-odor detection thresholds and produce strong, unpleasant odors, their consistent formation may result in a substantial source of odor during insect rearing. Repeated or sustained exposure to these emissions may also impede indoor air quality in insect-rearing facilities and pose occupational health concerns, depending on the compound concentrations, exposure duration, and ventilation conditions. Thus, further studies are required to quantify VOC emissions, characterize occupational exposure, and assess the potential health effects of VOCs released from BSFL rearing systems. Recent advances in sensor-based analytical platforms have demonstrated the potential for rapid chemical profiling and on-site detection in complex food and environmental matrices (Xu et al., 2025; Xu et al., 2026). The development of comparable rapid-monitoring approaches for insect-rearing environments, together with quantitative chromatographic analysis, could facilitate routine surveillance of odor-active and potentially hazardous VOCs.

From a food chemistry perspective, the VOC profiles observed during rearing may also be relevant to the sensory quality of insect-derived ingredients. Volatile composition is an important determinant of the flavor characteristics of alternative protein materials, including edible insects (Park et al., 2024). Of particular relevance, trimethylamine, acetic acid, and 3-methylbutanoic acid, which were detected in the present BSFL rearing systems (Table 4), have previously been identified by GC–olfactometry as key off-flavor compounds in BSFL meal, contributing fishy, sour, and cheesy notes (Huseynli et al., 2023). The overlap between VOCs detected in the rearing headspace and compounds previously reported in BSFL meal provides relevant food chemistry context; however, the present measurements do not demonstrate transfer or retention of these compounds in harvested larvae. Because VOCs and sensory properties of the harvested or processed biomass were not analyzed in this study, their contribution to final product flavor cannot be determined.

Post-harvest processing can modify the volatile and sensory characteristics of insect-derived ingredients. Defatting has been shown to reduce the intensity of off-flavors in BSFL meal (Huseynli et al., 2023), while fermentation has also been proposed as an approach to reduce unpleasant odors in insect-derived foods (Perez-Santaescolastica et al., 2022). Processing conditions must be carefully selected because they can also generate new volatile compounds. In addition, blanching, drying, extraction, and other processing technologies can alter the nutritional, physicochemical, and sensory properties of insect ingredients (Liang et al., 2024). Therefore, paired characterization of VOCs in the rearing headspace, harvested larval biomass, and processed ingredients, together with olfactometric or sensory analysis, is needed to determine whether rearing-associated VOCs are retained or transformed during processing and whether they affect final product flavor.

The present study demonstrated that diverse VOCs are generated during BSFL rearing and that their profiles vary with feed type used as substrate. Although consistent temporal and feed-dependent VOC patterns were observed across the three feed conditions and corresponding controls, these patterns should be further confirmed using independently replicated rearing units. Future studies incorporating biological replication will allow statistical evaluation of treatment effects and improve the generalizability of the observed VOC profiles. Also, the 5-day passive sampling interval provided integrated temporal profiles over the larval rearing period; however, future studies should evaluate shorter sampling intervals and sampler-specific uptake characteristics to optimize temporal resolution and quantitative VOC monitoring. In addition, because various VOCs are mainly produced through microbial metabolism, clarifying the contributions of BSFL-associated microorganisms to their production will improve our understanding of VOC formation during rearing. VOCs may also affect larval physiology and behavior through the BSFL chemosensory system (Park et al., 2026). Future studies should examine both the microbial origins of VOCs and their potential effects on BSFL.

4. Conclusions

VOCs generated during BSFL rearing were monitored using three different animal feeds as substrate. Larval growth varied with feed composition, and the highest mean dry weight on the final day (day 15) was observed in the fish-feed treatment, followed by those in the dog- and pig-feed treatments. VOC signals were substantially lower in the control rearing systems without larvae than in those with larvae, indicating that the observed VOC profiles were closely associated with BSFL activity and associated microorganisms rather than with the feed substrate alone. Tenax TA detected a greater number of compounds and produced higher GC–MS responses than the Air Toxics sampler under the conditions tested. Higher summed GC–MS peak areas were observed during the early rearing period, particularly during days 0–5, although high analytical responses persisted until day 10 in the fish-feed treatment. Carbohydrate-related VOCs were more prominent during the early rearing period, while the protein-rich fish feed showed a greater contribution of protein-associated VOCs. Several putative odorants, including S-containing compounds, volatile fatty acids, phenolic compounds, and indolic compounds, were also detected, with their profiles varying among the feed treatments. Overall, these results reveal temporal and feed-dependent patterns in VOC profiles during BSFL rearing. Further studies incorporating quantitative VOC measurements and sensory or olfactometric analyses are required to determine emission rates and the actual contributions of individual compounds to perceived odor.

CRediT authorship contribution statement

Dong-Jun Lee: Writing – review & editing, Writing – original draft, Visualization, Investigation. Ka Young Kim: Writing – review & editing, Investigation. Dongkyu Park: Writing – review & editing, Methodology. Min-Ho Oak: Writing – review & editing, Resources. Pil-Gon Kim: Writing – review & editing, Writing – original draft, Validation, Conceptualization. Jee Young Kim: Writing – review & editing, Writing – original draft, Supervision, Conceptualization.

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

Supplementary material

Data availability

Data will be made available on request.