적두, 녹두, 흑두(팥) -- 항산화, 항당뇨 효과 2025 네이처!!

[문형철]

1. nature  
2. scientific reports  
3. articles  
4. article

Multi-omics for unveiling potential antidiabetic markers from red, green and black mung beans using NIR-UPLC-MS/MS multiplex approach

Download PDF

• Article
• Open access
• Published: 01 June 2025

Multi-omics for unveiling potential antidiabetic markers from red, green and black mung beans using NIR-UPLC-MS/MS multiplex approach

• Dina S. Ghallab, 
• Reham S. Ibrahim, 
• Doaa A. Ghareeb & 
• Nihal M. El Newehy 

Scientific Reports volume 15, Article number: 19213 (2025) Cite this article

• 1542 Accesses

• 3 Altmetric

• Metricsdetails

Abstract

Inspired by the nutritional and biological attributes of mung beans, the current work aims to monitor metabolome patterns of different mung species and their entanglements on antidiabetic potential using NIR-UPLC-MS/MS multiplex approach combined with chemometrics. In this regard, a total of 71 chromatographic peaks spanning sugars, amino acids, flavonoids, fatty acids and their lipid derivatives, and phytosterols were chemically profiled. Coincidently, OPLS-DA underscored an obvious discrimination among the green, red and black mung species suggesting their chemical discrepancies where eriodictyol-O-glucoside, caffeic acid, formononetin-O-glucoside, viniferal and genistin serve as focal discriminators of green mung beans while lysoPC 18:2, lanosterol, gallocatechin, tyramine, petunidin 3-O-glucoside, biochanin A, vigvexin A, vignatic acid B, lysoPC 16:0 and phaseollin were the determining metabolites of red ones. Successively, the differential markers enriched in black mung samples included 10-formyltetrahydrofolate, stearidonic acid, hydroxylinoleic acid, vignatic acid A, campestrol, arachidonic acid and PG (18:2/18:1). Experimentally speaking, all mung samples exerted noteworthy dose-dependent inhibitory potential towards α-amylase and α-glucosidase enzymes. OPLS coefficient plots highlighted gamma-aminobutyric acid (GABA), gallic acid and beta-sitosterol as possible efficacy metabolites harmoniously mediated antidiabetic potential. Equally important, NIR spectroscopic analysis coupled with PLS-R model quantitively predicted the bio-efficient markers from various mung bean samples with a significant level of experimental reliability. These findings pursue concept of nutritional therapy and provide a fresh perspective to probe into mung beans bioactive molecules which might serve as referenced templates for mitigating diabetes. However, future work should be explored to uncover muti-target mechanisms of mung beans-derived compounds and strengthen their relevance.

초록

녹두의 영양학적 및 생물학적 특성에 착안하여,

본 연구는 화학계량학과 결합된 NIR-UPLC-MS/MS 다중 분석법을 활용하여

다양한 녹두 종의 대사체 패턴과

당뇨병 치료 잠재력에 미치는 영향을 모니터링하는 것을 목표로 한다.

이를 위해

당류, 아미노산, 플라보노이드, 지방산 및 그 지질 유도체, 식물 스테롤 등

총 71개의 크로마토그래피 피크를 화학적으로 프로파일링하였다.

동시에 OPLS-DA 분석은

녹두, 적두, 흑두 종간 화학적 차이를 시사하는 뚜렷한 차별성을 확인하였으며,

에리오디크티올-O-글루코사이드, 카페인산, 포모노네틴-O-글루코사이드, 비니페랄, 제니스틴이

녹색 녹두의 주요 차별화 인자로 작용하는 반면,

리소포스파티딜콜린 18:2, 라노스테롤, 갈로카테킨, 티라민, 페투니딘 3-O-글루코사이드,

바이오카닌 A, 비그베신 A, 비그나틱산 B, 리소포스파티딜콜린 16:0, 파세올린이

적색 녹두의 결정적 대사산물로 확인되었다.

이어서 흑두에서 풍부하게 검출된 차별적 마커로는

10-포르밀테트라하이드로폴레이트, 스테아리돈산, 하이드록실리놀레산, 비그나틱산 A,

캠페스트롤, 아라키돈산 및 PG(18:2/18:1)가 포함되었다.

실험적으로,

모든 녹두 샘플은 α-아밀라아제 및 α-글루코시다아제 효소에 대해

주목할 만한 용량 의존적 억제 잠재력을 나타냈다.

OPLS 계수 플롯은

감마-아미노부티르산(GABA), 갈산 및 베타-시토스테롤이

조화롭게 항당뇨 잠재력을 매개하는 가능한 효능 대사산물임을 강조했다.

동등하게 중요한 점은,

PLS-R 모델과 결합된 NIR 분광 분석이 다양한 녹두 샘플로부터

생물학적 효능 지표를 상당한 수준의 실험적 신뢰도로 정량적으로 예측했다는 것이다.

이러한 결과는

영양 치료 개념을 추구하며,

당뇨병 완화를 위한 참조 템플릿으로 활용될 수 있는 녹두의 생리활성 분자를 탐구하는 새로운 관점을 제공한다.

그러나

향후 연구에서는

녹두 유래 화합물의 다중 표적 기전을 규명하고 그 관련성을 강화하기 위한 노력이 필요할 것이다.

Similar content being viewed by others

Metabolomics-based profiling of 4 avocado varieties using HPLC–MS/MS and GC/MS and eval‎uation of their antidiabetic activity

Article Open access23 March 2022

Technological processing of Phaseolus vulgaris and Morus alba leaves to create a new nutritional food product for individuals with diabetes

Article Open access19 November 2024

LC–MS/MS, GC–MS and molecular docking analysis for phytochemical fingerprint and bioactivity of Beta vulgaris L.

Article Open access29 March 2024

Introduction

Mung beans (Vigna radiata L., F. Fabaceae), also called as moong beans or green grammes, were traditionally consumed as a legume food globally for over 3500 years in African and Asian places1. Originally classified as Phaseolus aureus Roxb., a lot of species within the Phaseolus genus were later reclassified to the Vigna genus. The genus Vigna comprises nine crop species that are primarily cultivated for human consumption or as animal feed in regions ranging from tropical to temperate climates2. The most common species are green mung or green gram (Vigna radiata L.), black mung or black gram (Vigna mungo L.) and red mung or Azuki bean (Vigna angularis L.). Previous studies have indicated that bioactive compounds derived from legumes comprising proteins, carbohydrates, polyphenolics, dietary fibres, minerals, fatty acids, sterols and vitamins offer potential functional food in promoting good health, with their consumption increasing annually by 10%3. The World Health Organization (WHO) has also suggested that these bioactive compounds could improve healthcare and help combat numerous chronic degenerative diseases4. These compounds primarily function as antioxidants and have gained popularity recently due to their nutritional and therapeutic significance5. Mung beans have been shown to improve ailments such elevated blood pressure, cholesterol, and blood sugar, and to prevent cancer and skin pigmentation6.

서론

녹두(Vigna radiata L., F. Fabaceae)는

몽콩 또는 그린그램이라고도 불리며,

아프리카와 아시아 지역에서 3500년 이상

전 세계적으로 콩과 식량으로 전통적으로 소비되어 왔습니다1.

원래 Phaseolus aureus Roxb.로 분류되었으나,

Phaseolus 속 내 많은 종들이 이후 Vigna 속으로 재분류되었습니다.

Vigna 속은 열대부터 온대 기후에 이르는 지역에서

주로 인간 소비나 동물 사료용으로 재배되는 아홉 가지 작물 종을 포함한다2.

가장 흔한 종은

녹두(Vigna radiata L.), 흑두(Vigna mungo L.), 팥(Vigna angularis L.)이다.

선행 연구에 따르면

단백질, 탄수화물, 폴리페놀, 식이섬유, 미네랄, 지방산, 스테롤, 비타민 등으로 구성된

콩류 유래 생리활성 화합물은 건강 증진에 잠재적 기능성 식품으로 작용하며,

그 소비량은 매년 10%씩 증가하고 있다3.

세계보건기구(WHO) 또한 이러한 생리활성 화합물이

의료 서비스를 개선하고 수많은 만성 퇴행성 질환 퇴치에 도움이 될 수 있다고 제안했습니다4.

이 화합물들은 주로 항산화제로 기능하며,

최근 영양학적 및 치료적 중요성으로 인해 주목받고 있습니다5.

녹두는 고혈압, 콜레스테롤, 혈당 수치 개선과

암 및 피부 색소 침착 예방에 효과적인 것으로 나타났습니다6.

They also offer liver protection and boost the immune system. These health advantages are largely attributed to the presence and characteristics of the powerful mixture of active compounds found in mung beans7. The nutritional value of mung beans is highly regarded; by dry weight, they contain 20–25% protein. The main storage proteins in mung beans are globulin (60%) and albumin (25%). Consequently, alongside other cereals, mung beans have being consumed in greater quantities. Essential amino acids including leucine, isoleucine, phenylalanine, valine, tryptophan, arginine, methionine, and lysine are abundant in mung bean proteins. Therefore, mung beans are considered a significant source of dietary protein (240 g/kg)8. Mung beans are primarily composed of carbohydrates (55–65%, equivalent to 630 g/kg of dry weight), with starch being the main type. The remaining carbohydrates include raffinose, stachyose, and verbascose, which can cause flatulence9. Nevertheless, with sufficient soaking, fermentation, and germination, these oligosaccharides can be decreased and are readily soluble in water. Compared to other legumes, mung beans have less flatulence because their carbs are very easily digested. Additionally, mung beans and their sprouts are less calorically dense than other cereals, which makes them advantageous for people with diabetes or obesity. Additionally, tannins, hemagglutinin, trypsin inhibitors, phytic acid, and other antinutrients found in mung beans have a variety of biological functions that help the body get rid of toxins. Flavonoids (flavones, isoflavonoids, isoflavones, and anthocyanins), phenolic acids (gallic acid, ferulic acid, vanillic acid, caffeic acid, cinnamic acid, protocatechuic acid, p-hydroxybenzoic acid and shikimic acid), and other organic acids are among the numerous secondary metabolites that have been found to have health-promoting properties10. The most abundant flavonoids detected in the mung beans are Flavonols (kaempferol, quercetin, and myricetin) and flavones (luteolin, vitexin, isovitexin, and isovitexin-6″-O-α-l-glucoside). The two main flavonoids found in mung bean seeds are vitexin and isovitexin. Although they are mostly concentrated in the seed coats, these phenolic compounds are present in both the cotyledons and the seed coats. Furthermore, a number of variables, including cultivar, seed coat colour, meteorological and agronomic circumstances throughout growth, and extraction and analysis techniques, affect the amount and makeup of bioactive chemicals in mung beans7.

또한

간 보호 기능과 면역 체계 강화 효과도 제공합니다.

이러한 건강상의 이점은

주로 녹두에 함유된 강력한 활성 화합물 혼합물의 존재와 특성에서 비롯됩니다7.

녹두의 영양가는 매우 높게 평가되며,

건조 중량 기준으로 20~25%의 단백질을 함유합니다.

녹두의 주요 저장 단백질은

글로불린(60%)과 알부민(25%)입니다.

결과적으로 녹두는

다른 곡물과 함께 더 많은 양으로 소비되고 있습니다.

녹두 단백질에는

류신, 이소류신, 페닐알라닌, 발린, 트립토판, 아르기닌, 메티오닌, 라이신 등

필수 아미노산이 풍부하게 함유되어 있습니다.

따라서

녹두는 식이 단백질의 중요한 공급원(240g/kg)으로 간주됩니다8.

녹두는

주로 탄수화물(55~65%, 건조 중량 기준 630g/kg에 해당)로 구성되며,

전분이 주요 유형이다.

나머지 탄수화물에는

라피노스, 스타키오스, 버바스코스 등이 포함되어 있어 복부 팽만감을 유발할 수 있다9.

그러나

충분한 불림, 발효, 발아 과정을 거치면

이러한 올리고당 함량이 감소하며 물에 쉽게 용해된다.

다른 콩류에 비해 녹두는

탄수화물이 매우 쉽게 소화되기 때문에 가스 발생이 적습니다.

또한 녹두와 그 새싹은

다른 곡물보다 열량 밀도가 낮아 당뇨병이나 비만 환자에게 유리합니다.

더불어 녹두에 함유된

탄닌, 헤마글루티닌, 트립신 억제제, 피틴산 및 기타 항영양소들은

신체가 독소를 제거하는 데 도움을 주는 다양한 생물학적 기능을 수행합니다.

플라보노이드(플라본, 이소플라보노이드, 이소플라본, 안토시아닌),

페놀산(갈산, 페룰산, 바닐산, 카페인산, 신남산, 프로토카테츄산, p-하이드록시벤조산, 시키믹산) 및

기타 유기산은 건강 증진 효과가 있는 것으로 밝혀진

수많은 2차 대사 산물 중 일부입니다10.

녹두에서 검출된 가장 풍부한 플라보노이드는

플라보놀(캄페롤, 케르세틴, 미리세틴)과

플라본(루테올린, 비텍신, 이소비텍신, 이소비텍신-6″-O-α-l-글루코사이드)이다.

녹두 종자에서 발견되는 두 가지 주요 플라보노이드는

비텍신과 이소비텍신이다.

이 페놀성 화합물들은 주로 종피에 집중되어 있지만,

자엽과 종피 모두에 존재합니다.

또한 품종, 종피 색상, 생장 과정 전반의 기상 및 농업적 조건, 추출 및 분석 기술 등

여러 변수가 녹두의 생리활성 화합물 함량과 구성에 영향을 미칩니다7.

Diabetes Mellitus (DM) significantly impacts the wellbeing and quality of life of millions globally. While most DM patients have Type 2 Diabetes11. A key therapeutic strategy for managing Type 2 DM consists of preventing the conversion of dietary carbohydrates into simple sugars in order to lower postprandial blood glucose levels, thereby preventing their absorption11. The enzymes α-glucosidase and α-amylase, which hydrolyze carbohydrates, are the primary targets of this discipline. Mung beans has emerged as a promising agent in this context8,12,13,14,15, especially when used by the suitable processing methods as boiling and sprouting, while vitexin and isovitexin are the two phenolic compounds responsible for the high antidiabetic activity16,17. In this regard, mung bean starch is notable for its low glycemic index. Therefore, its application in the creation of such novel products can aid in reducing the risk of diabetes and managing obesity18. But, its standardization depending on metabolomics analysis is essential to ensure perfect chemical consistency and reliable efficacy19. It is worth noting that there is no previous comprehensive qualitative and quantitative metabolomics study on green, black and red mung bean species along with their antidiabetic efficacies. Therefore, this study aims to standardize mung bean samples collected from the three common species. To achieve this, an integrated strategy for enhanced quality control has been implemented. Fingerprint analysis of mung bean samples using Fourier Transform-Near-Infrared spectroscopy (FT-NIR) in addition to comprehensive chemical profiling using UPLC-QqQ-MS technique were established. By combining these two independent techniques with multivariate analysis, a thorough and precise discrimination, along with high-quality assessment, of mung samples collected from various species was achieved. Furthermore, a fingerprint-efficacy relationship analysis was established utilizing an Orthogonal Projection to Latent Structure (OPLS) multivariate model to identify bio-efficient antidiabetic biomarkers. Finally, for the rapid and accurate prediction of efficacy-related biomarkers a Partial Least Squares Regression (PLS-R) model was developed using FT-NIR data as independent variables.

당뇨병(DM)은 전 세계 수백만 명의 건강과 삶의 질에 중대한 영향을 미칩니다. 대부분의 당뇨병 환자는 제2형 당뇨병을 앓고 있습니다11.

제2형 당뇨병 관리의 핵심 치료 전략은

식이 탄수화물이 단순 당으로 전환되는 것을 방지하여

식후 혈당 수치를 낮추고, 그 흡수를 차단하는 것입니다11.

탄수화물을 가수분해하는 효소인

α-글루코시다아제와 α-아밀라아제는 이 분야의 주요 표적입니다.

녹두는

특히 삶기나 발아와 같은 적절한 가공 방법을 사용할 때8,12,13,14,15에서

유망한 물질로 부상했으며,

특히 비텍신과 이소비텍신은 높은 항당뇨 활성을 담당하는 두 가지 페놀 화합물입니다16,17.

이와 관련하여 녹두 전분은

낮은 혈당 지수로 주목받고 있다.

따라서

이러한 신제품 개발에 적용하면

당뇨병 위험 감소와 비만 관리에 도움이 될 수 있다18.

그러나

완벽한 화학적 일관성과 신뢰할 수 있는 효능을 보장하기 위해서는

대사체학 분석에 기반한 표준화가 필수적이다19.

녹두, 검두, 적두 종과 그 항당뇨 효능에 대한

포괄적인 정성·정량 대사체학 연구가 이전에 없었음을 주목할 필요가 있다.

따라서 본 연구는 세 가지 일반 종에서 채취한 녹두 샘플을 표준화하는 것을 목표로 한다. 이를 위해 강화된 품질 관리를 위한 통합 전략을 시행하였다. 푸리에 변환 근적외선 분광법(FT-NIR)을 이용한 녹두 샘플의 지문 분석과 UPLC-QqQ-MS 기법을 활용한 포괄적 화학적 프로파일링을 확립하였다. 이 두 독립적 기법을 다변량 분석과 결합함으로써 다양한 종에서 채취한 녹두 샘플에 대한 철저하고 정밀한 판별 및 고품질 평가를 달성하였다. 또한 잠재 구조에 대한 직교 투영(OPLS) 다변량 모델을 활용하여 지문-효능 관계 분석을 수립함으로써 생물학적 효능을 지닌 항당뇨 생체표지자를 식별하였다. 마지막으로, 효능 관련 바이오마커를 신속하고 정확하게 예측하기 위해 FT-NIR 데이터를 독립 변수로 사용하여 부분 최소 제곱 회귀(PLS-R) 모델을 개발했습니다.

Results and discussion

Metabolites assignment in differently coloured mung beans extracts

In the current investigation, UPLC-QqQ-MS/MS metabolomics approach was incorporated to objectively demonstrate compositional heterogeneity among differently coloured mung beans (red, green and black) samples in a holistic viewpoint. In this regard, a total of 71 chromatographic peaks spanning sugars, amino acids, dipeptides, flavonoids, anthocyanins, phenolic acids, fatty acids, phospholipids and phytosterols were chemically annotated relying on the respective retention time and unique fragments profiles in conjunction with reference standards and pertinent literature. Base peak chromatograms (BPCs) of the investigated samples in positive and negative polarity modes were depicted in Fig. 1A. In a similar vein, Table 1 briefly outlines the complete list of annotated compounds derived from the various mung samples, their corresponding mass data retention times, molecular formula, and chemical classes.

색상이 다른 녹두 추출물의 대사 산물 분석

본 연구에서는 UPLC-QqQ-MS/MS 대사체학 접근법을 통합하여 색상이 다른 녹두(빨강, 초록, 검정) 샘플 간의 구성 이질성을 전체적인 관점에서 객관적으로 입증하였습니다. 이를 위해 당류, 아미노산, 디펩타이드, 플라보노이드, 안토시아닌, 페놀산, 지방산, 인지질 및 피토스테롤에 걸쳐 총 71개의 크로마토그래피 피크를 각각의 유지 시간과 고유한 단편 프로필을 기준으로, 참조 표준 및 관련 문헌과 함께 화학적으로 주석 처리하였습니다. 조사된 샘플의 양극성 및 음극성 모드에서의 기준 피크 크로마토그램(BPC)은 그림 1A에 표시되었습니다. 마찬가지로, 표 1은 다양한 녹두 샘플에서 유래된 주석 처리된 화합물의 전체 목록, 해당 질량 데이터 유지 시간, 분자식 및 화학적 분류를 간략히 요약합니다.

Fig. 1

Respective base peak chromatograms (BPCs) retrieved from different mung species in both positive and negative ionization modes (A). Heat map of all identified compounds in various mung samples with the relative amount of metabolites from high to low represented by a color-coded scale grading from red to blue (B).

Full size image

Table 1 List of annotated compounds extracted from differently colored mung beans species.

Full size table

Complementarily, the relative average content of main chemical classes monitored in different mung species was depicted in stacked bar chart (Fig. S1).

In a broader sense, the following subsections provide the scheme utilized to annotate the key chemical classes in the different mung samples. Crucially, precision and reproducibility of the analytical system were assessed through QC samples.

Sugars

Five peaks were detected under sugars category including 3 disaccharides (1, 2 & 4), one monosaccharide (3) and one sugar acid (7).

Trehalose was the suggested for peak (2) recorded at m/z [M – H]− 341.21. Upon MS/MS analysis, two prominent daughter ions at m/z 179 and 161 consistent with the glycosidic cleavage at C1 and C3 forming two hexose units were assigned which sequentially fragmented through water loss.

Peak (4) was annotated as laminaribiose at m/z [M – H]− 341 undergoes characteristic glycosidic bond cleavages, cross-ring fragmentations (Characteristic of Disaccharides), and neutral losses, especially of water and sugar-related fragments.

Peak (7) revealed an intense deprotonated precursor ion [M − H]− at m/z 195.3 combined with fragment signals at m/z 151 and 133 attributable to the consecutive elimination of CO2 and H2O moieties.

Organic acids

Three peaks (6, 15 & 16) assigned as organic acids were recorded in the current chromatographic runs.

Compound (6) exhibited a protonated signal at m/z 104.2 and daughter ions at m/z 87 and 60 matching with continuous losses of NH3 and CO2 moieties. Correspondingly, peak (6) was annotated as γ-amino butyric acid further confirmed by reference standard data.

Compound (15) forms a deprotonated ion at m/z 191.2 ([M–H]⁻) and fragments via decarboxylation (-CO₂, m/z 147) and dehydration (-H₂O, m/z 173). Further CO₂ loss produces m/z 103, confirm‎ing a characteristic citric acid fragmentation pattern.

Regarding compound 16 of m/z 125.1, it was characterized as imidazole acetic acid. The characterization was farther confirmed by the MS2 data, which unveiled a distinctive product ion at m/z 79 upon decarboxylation along with a minor ion at m/z 41 assigned to C2H3N upon on imidazole ring fission. This data aligned with an earlier record20.

Amino acids, dipeptides and cyclic peptides

Amino acids (AAs) are a basic class of primary metabolite mostly found in mung beans and contribute to their unique nutritional21.

Under the current investigation, nine amino acids (5, 11, 13, 14, 17, 18, 20, 21 & 23) as well as two dipeptides (9 & 12) were detected during mass spectra examination.

From a practical standpoint, a prominent peak (11) found at m/z 166.3 [M + H]+ was tentatively annotated as methionine sulfoxide further confirmed by relevant data22. During its MS2 analysis, two informative ions at m/z 138 and 120 coherent with consecutive loss of CO and NH3 were detected. Further, a distinctive daughter ion of m/z 56 progressively ascribable to an additional loss of CH3SO was recorded.

Alongside, the acquired m/z values demonstrated the occurrence of valine and citrulline assigned in signals (14 & 17), respectively and had comparable fragmentation paths. essentially manifested with series of neutral elimination of H2O, CO succeeded by a major product ion consistent with NH3 loss.

Comparably, an acylated amino acid (18) displaying a quasi-molecular ion [M + H]+ at m/z 158.3 was putatively annotated as acetyl proline. During its MS2 analysis, it initially formed a promising ion coherent with distinguishing loss of acetyl unit (-43 Da) expelling the respective amino acid at m/z 115 further fragmented upon a collateral loss of NH3 + CO yielding a sharp product signal at m/z 70. Considering the above mass data and previous literature data 23,24, we can conclude that component 18 might be acetyl proline.

Compound (21) forms a quasi-molecular ion [M + H]⁺ at m/z 205.2, displaying a characteristic L-Tryptophan fragmentation pattern. It undergoes deamination (-NH₃, m/z 188) and decarboxylation (-CO₂, m/z 161). Further indole ring fragmentation produces m/z 146, 130, 117, and 103, confirm‎ing its structural identity.

Regarding dipeptides, a compound (9) recorded at m/z 175.3 [M + H]+ was identified as valyl glycine. Its CID-MS/MS initially gave rise to a fragment ion of 101 Da (valine moiety) through the protonated amide bond joint cleavage in the respective dipeptide revealing a prominent ion at m/z 75 furtherly dissociated through an accompanying elimination of NH3 + CO.

Furthermore, two cyclic peptides with representative peaks; 44 & 57 were clearly detected and respectively annotated as vignatic acid B and vignatic acid A representing chemotaxonomic markers of mung beans.

Peak 44 exhibited an intense molecular ion at m/z 520.2 for [M + H]+ along with a prime product ion of m/z 476 forming from loss of CO2. Also, amide bond cleavage expelled N-terminal and C-terminal fragments at m/z 364 and 112, respectively. Considering the mass data stated above, compound 44 could be characterized as vignatic acid B.

Phenolic acids, flavonoids, anthocyanins and coumestans

Phenolic acids and flavonoids constitute vast groups of metabolites significantly existing in mung beans25.

In current chromatographic runs, only two phenolic acids exemplified in signals (22 & 25) were chemically annotated.

Peak (22) displayed (m/z 171 [M + H]⁺) fragments through neutral losses of H₂O (-18 Da, m/z 153), CO (-28 Da, m/z 143), and CO₂ (-44 Da, m/z 127). Further breakdown of the aromatic system produces smaller ions at m/z 109, 97, and 81, characteristic of phenol-derived structures. This pattern helps in the annotation of gallic acid.

The suggested candidate for signal (25) was caffeic acid owing to the molecular ion at m/z 179.2 and the typical ions of m/z 135 and 143 corresponding with CO2 and two H2O units loss.

In addition, a total of 16 flavonoid peaks pertaining the four subclasses primarily flavonols, flavone, flavanones, and flavan-3-ols along with some glycosylated forms were found constituting one of the main groups of metabolites found in the studied mung species. As well, three isoflavonoids were noted in the acquired MS spectra.

For eriodictyol-O-glucoside (24), the deprotonated ion [M − H]− at m/z 449.3 along with a dominating aglycone (base peak) at m/z 287 [M - H − 162] − resulted upon hexose moiety loss were monitored. During MS2, a daughter ion at m/z 243 was formed by loss of CO2 (− 44 Da) from the C ring. Further, retrocyclisation pathway for C-ring expelled two major signals of m/z 135 [1,2A]− and 151 [1,3A]−.

The metabolite at signal (29) revealed a dominant precursor ion [M − H]− at m/z 445.2. Its MS2 profile dissected the occurrence of a major ion at m/z 283 [M–H–162]− coming from neutral loss of glycosyl residue. Glycitein was affirmed as aglycone by the distinctive ions at m/z 268 and 224 gained through consecutive elimination of methyl radical and CO2. Also, a typical RDA ion [1,3B]− of m/z 133 was found suggesting the occurrence of an additional hydroxyl group on ring-A.

In positive polarity mode, the ion at m/z 285.25 corresponded to biochanin A (Peak 30). During MS2 experiment with higher collision energy, main product ions at m/z 271, 227, 152 and 123 respectively formed as a result of methyl radical loss, CO2 and retro-Diels-Alder (RDA) fission. The fragmentation profile recorded for biochanin A were in good alignment with earlier literature data25.

Furthermore, compound (34) (m/z 433 [M + H]⁺) undergoes characteristic C-glycoside fragmentation. It first undergoes cross-ring cleavage of glucose, producing m/z 343 (-90 Da) and m/z 313 (-120 Da). The loss of the glucose moiety (-162 Da) yields m/z 271, corresponding to the apigenin aglycone. Further fragmentation of the flavone core generates m/z 153, 147, and 119, confirm‎ing the structural identity of vitexin (Apigenin -C-glucoside). This pattern is essential for distinguishing C-glycosides from O-glycosides in natural product analysis.The fragment ion spectrum of the deprotonated compound 43 of m/z 563.15 [M–H]− revealed a base peak with m/z 477 formed upon cleavage of malonyl moiety (– 86 Da) along with an ion at m/z 315 after hexose moiety (–162 Da). Tamarixetin was assured as aglycone part by the distinguishing fragment ions of m/z 300, 255, 163 and 151 formed at high collisional energy accounting for neutral elimination of CH3 radical and CO2 along with RDA fragments. Building on the analyses above and the pertinent literature data 24,25, compound 43 might be identified as tamarixetin-3-O-malonylglucoside.

One anthocyanidin with a respective peak; 28 was clearly noted under the applied ESI conditions.

Peak (28) could be assigned as petunidin 3-O-glucosideon based on its precursor ion at m/z 477.35 for [M − 2 H]- and a principal aglycone signal at m/z 315 formed after the loss of neutral hexose unit. This aglycone was confirmed as petunidin by its MS2 analysis manifested by characteristic daughter ions at m/z 301, 273, 203 and 137 further confirmed from previous records26.

Further, one coumestan peak (42) was recorded and annotated as vigvexin A (Phaseol) based on its protonated molecular ion [M + H]+ at m/z 337.3 and dominating mass fragments at m/z 293 and 249 sequentially matching with losses of CO2 and isopropyl (43 Da) moieties further confirmed from previous literature data27.

Fatty acids

Under the current analysis, fatty acids especially unsaturated forms sharply predominate the secondary metabolites in the investigated mung species (Table 1).

Peak 46 revealed mass characteristics consistent with stearidonic acid displaying a parent ion [M + H]+ at m/z 277.3 and major product ions at m/z 255 and 233 arising from water and CO2 moieties loss. As well, fissure of double bond C6-C7 of parent ion generated an extra product signal at m/z 162. MS2 profile outlined above together with the previous data28 affirmed peak 46 identity.

Peak 59 representing fatty acid glyceride was clearly detected under the current analysis and tentatively assigned as linolenoyl-glycerol. This signal gave a parent ion [M-H]− of m/z 353.3 as well as daughter ions at m/z 277, 233 and 59 matching with linolenic acid formation and its distinctive fragments after favourable elimination of glycerol head group and further affirmed by earlier reports 2428.

The compound at signal 68 was matched to oleic acid with a molecular ion [M − H]− at m/z 281.3 and a dominating fragment at m/z 237 after loss of CO2. Tandem MS fragments at m/z 124 (C9H16) [M − H – C9H16O2]− and m/z 59 (C2H3O2) [M − H−C16H31]− were also recorded and consistent with previous reports 29,30.

Phospholipids

Phospholipids are a naturally occurring class of distinct anionic substances that are abundant in mung beans and account for 32.26% of the grains’ lipid components31.

Mass spectra examination allowed the characterisation of several phospholipid species mainly phosphatidylglycerols (PGs), lyso-phosphoethanolamines (lysoPEs), and lyso-phosphocholines (lysoPCs) with corresponding peaks 45, 48, 49, 50, 53, 60, 67, 69, 70 and 71 (Table 1).

Peak 45 presenting a [M + H]+ ion at m/z 518.2 was characterized as lysoPC 18:3. During its mass analysis, two intense product ions at m/z 239 and 257 acquired from the losses of linolenoyl acyl substituent (18:3) as free fatty acid and as ketene, respectively. The characteristic fragments at m/z 279, 235, and 59 further verified linolenic acid as the bound fatty acid.

Similarly, lysoPE at peak 48 revealed a dominating [M + H]+ ion at m/z 454.6. Its MS/MS profile revealed two signals at m/z 257 and 213 affirming the occurrence of palmitic acid as the bound fatty acid. Asides, two distinctive tandem ions at m/z 197 and 215 consistent with loss of the free acid (16:0) and ketene, respectively were recorded and further confirmed by literature data28. Accordingly, compound 48 could be annotated as lysoPE 16:0.

Moreover, PG at peak 67 formed a dominating [M-H]− ion at m/z 757.15 and was assigned as C16:0 /C18:2 phosphatidylglycerol further verified by the relevant published data32. Taking the same fragmentation route, two major MS2 fragments at m/z 501 and 519 matching with elimination of the free acid (16:0) and ketene, respectively. As well, the product ions m/z 478 and 496 deriving from the comparable losses of the second fatty acyl substituent (18:2) were recorded. Further, ions respective to the fragmentation patterns of palmitic acid (16:0) and linoleic acid (18:2) were also perceived after the polar glycerol head group cleavage.

Similarly, PG at peak 71 generated an intense [M-H]− ion at m/z 778.4 and was assigned as C18:0 /C18:0 phosphatidylglycerol further supported by the previous data32. During MS2 analysis, two main fragments at m/z 495 and 513 consistent with elimination of the free acid (18:0) and ketene, respectively. As well, the typical losses of the fatty acyl substituent (18:0) were recorded yielding signal at m/z 472 and 490. also, the typical fragment ions of stearic acid (18:0) and linoleic acid (18:2) were perceived after the polar glycerol head group cleavage.

Phytosterols

Four peaks (55, 58, 62 & 66) in the phytosterol category were registered under the current analytical conditions.

Peak (55) exhibited a minor parent signal [M + H]+ at m/z 415.2 along with an intense ion at m/z 397 coherent with demethylation from the alkyl chain. Also, an evident fragment at m/z 273 was registered upon scission between C17/C20 of the aliphatic side chain. Two signals at m/z 161 and 107 corresponding to cleavage of the C-ring (C9/C11 and C8/14) were also observed. In alignment with the above observable data and reference standard, peak (55) could be annotated as of β-sitosterol.

Lanosterol (58) offered a negligible molecular ion [M + H]+ at m/z 397.2 alongside a considerable ion at m/z 409 formed upon H2O loss. In tandem mass analysis, a primary product ion at m/z 298 was assigned after cleavage across C17/C20 which in turn fragmented yielding fragment ions at m/z 95, 109 and 123. In addition, the primary cleavage of the C22–C23 bond in the side chain of sterols gave rise the ion at 69 m/z.

Analogously, peak 62 afforded a parent ion [M + H]+ at m/z 401.5 and a base peak of m/z 383 referable to water loss. The product signals at m/z 147 and 161 arising upon scission of ring C at C9/C11 and C8/14 were noted. As well, a minor fragment of m/z 268 was detected upon methyl radical loss from the side chain of dehydrated ion. In light of the aforementioned and consistent with earlier investigations28 peak 62 might be tentatively characterized as campesterol.

Multivariate statistical analysisUnsupervised pattern recognition of mung bean species using dendritic-heat map model

In this part, we conducted a comparison of the chemical profiles of various three mung species, namely red, green and black. This analysis utilized UPLC-MS/MS data alongside multivariate statistical methods. As previously shown, Fig. 1A illustrates notable differences in the chemical compositions of the various Mung species. The dendritic clustering analysis presented in the heat map (Fig. 1B) classified the samples into three distinct clusters: the first comprised samples from red, the second included green samples, and the third from black Mung species. It also showcased the varied distribution of the identified compounds across the three clusters as shown by color code variation from dark blue to dark brown gradient indicating low to high quantity, respectively. It was evident that red Mung samples had the highest amount of linolenic acid, daidzein, linolenoyl glycerol, phaseollin, biochanin A, gallocatechin, valine, vigvexin A, phenylalanine, 1-palmitoyl-2-linoleoyl-GPC (16:0/18:2), retinoic acid, vignatic acid B, quercetin retinoside, ricinoleic acid, lysoPC 16:1, tyramine, quercetin, petunidin 3- O-glucoside, glucitin, lanosterol, citric acid, gluconic acid, valylglycine, among others. In contrast, green mung samples exhibited elevated levels of vitexin, eicosapentanoic acid, hexosides of apigenin, kaempferol and formononetin, eriodictyol-O-glucoside, N-acetylproline, S-methylcysteine, viniferal, genistin and caffeic acid. Additionally, the black Mung had the highest amounts of trehalose, sucrose, LPE 16:0, PG (18:2/18:1), hydroxylinoleic acid, vignatic acid A, tamarixetin-3-O-(6’’-malonyl)glucoside, methionine sulfoxide, glutamyl methionine, campestrol, glycerophosphorylcholine, lysoPC 18:1, 9-Hydroxy-10,12,15-octadecatrienoic acid, arachidonic acid, stearidonic acid, 1-palmitoyl-2-arachidonoyl-GPC (16:0/20:4), oleic acid and laminarribiose, as shown in Fig. 1B.

Supervised pattern recognition of mung bean species using OPLS-DA model

For the sake of inter- and intra-class differentiation of the investigated Mung bean samples, an OPLS-DA model was developed using UPLC-MS data (Fig. 2). The first and second latent variables (LVs) of OPLS-DA model accounted for 60.7% and 38.5% of the sample variability, respectively. The model showed high reliability, with a goodness of fitness (R²) of 0.995 in addition to strong predictive power with a goodness of prediction (Q²) of 0.993. The classification accuracy of the OPLS-DA model was further validated using the ROC curve shown in Fig. S2, which revealed an AUC value of 1 for all classes, confirm‎ing its excellent classification capability.

Fig. 2

OPLS-DA score scatter plot (A). HCA dendrogram (B). Coefficient plots of the constructed OPLS-DA model for green (C), red (D) and black (E) mung samples, respectively.

Full size image

The score scatter plot (Fig. 2A) illustrated inter-class discrimination among mung bean samples along the first latent variable (LV1) where green and black mung samples clustered on the positive side indicating relative proximity in their chemical composition, while red mung samples were located on the negative side. Additionally, the second latent variable (LV2) highlighted intra-class discrimination. For instance, green mung samples were positioned on the positive side, whereas black mung samples were located on the negative LV2 side. This pattern aligned with the OPLS-DA dendrogram (Fig. 2B), which displayed two major clusters: one for green and black Mung samples, each in separate subclusters, and the other for red mung samples.

Coefficient plots were created to pinpoint specific metabolites unique to each sample, which contributed to their segregation patterns (Fig. 2C and E). The plot for the green mung samples (Fig. 2C) were characterized by S-methylcysteine, N-acetylproline, eriodictyol-O-glucoside, caffeic acid, kaempferol-3-O-glucoside, formononetin-7-O-glucoside, viniferal, genistin and eicosapentaenoic acid. In contrast, the plot for red mung samples (Fig. 2D) highlighted metabolites such as lysoPC 18:2, lanosterol, valine, gallocatechin, tyramine, phenylalanine, naringenin, dihexosylquercetin, petunidin 3- O-glucoside, glycitin, biochanin A, quercetin, quercetin rutinoside, vigvexin A, vignatic acid B, LPE 18:2, LPC 16:0, phaseollin and retinoic acid. Lastly, for the black mung samples (Fig. 2E), prominent markers included lysoPC 16:1, glutamylmethionine, S-oxide, glycerophosphorylcholine, methionine, 10-formyltetrahydrofolate, tamarixetin-3-O-(6’’-malonyl)glucoside, stearidonic acid, 9-hydroxy-10,12,15-octadecatrienoic acid, hydroxylinoleic acid, vignatic acid A, lysoPC 18:1, campestrol, arachidonic acid, PG (18:2/18:1) and 1-palmitoyl-2-arachidonoyl-GPC (16:0/20:4).

In vitro anti-diabetic activities of mung bean samples

Three distinct species of mung beans were eval‎uated for their ability to inhibit α-amylase and α-glucosidase, thereby reducing carbohydrate breakdown and glucose absorption, using in-vitro enzyme inhibition assays. As shown in (Fig. S3) the three studied mung species extracts exhibited a significant variable inhibitory effect against α-amylase, and α-glucosidase enzymes. The concentration-dependent nature of the biological effects was demonstrated by the way they increased with the extract’s concentration. In comparison to the inhibitor medication acarbose (IC50 = 0.431 ± 0.43 mg/ml), the red mung beans had the strongest α-glucosidase inhibitors among the studied extracts (IC50 = 0.178 ± 0.7 mg/ml), followed by the black mung beans (IC50 = 0.466 ± 0.58 mg/ml) and the green mung beans (IC50 = 0.6 ± 0.56 mg/ml) (Fig. S3). Regarding the α-amylase inhibitory effect, it was observed that, in comparison to the inhibitor drug acarbose (IC50 = 0.38 ± 0.47 mg/ml), the red mung beans had the strongest enzyme inhinition (IC50 = 0.248 ± 0.47 mg/ml), followed by the black mung beans (IC50 = 0.251 ± 0.58 mg/ml) and the green mung beans (IC50 = 0.297 ± 0.44 mg/ml) (Fig. S3). These findings demonstrated that the varying antidiabetic action of mung bean species could be attributed to the qualitative and quantitative differences in their phytochemical composition, particularly their phenolic content, flavonoids, and other bioactive compounds. The red mung beans showed the strongest inhibitory activity against the α-amylase and α-glucosidase enzymes, then the black, and finally the green mung beans showed the least inhibition activity.

Bioactivity-guided discrimination of mung bean species and identification of potential antidiabetic markers via OPLS model

To investigate the segregation patterns of the tested mung samples based on their anti-diabetic properties and identify the specific metabolites responsible for these effects, an OPLS model was developed. This model combined data on anti-diabetic activities (Y variables) with MS analysis data (X variables). The explained variance (R²Y) and predicted variance (Q²) for the OPLS model were found to be 0.980 and 0.985, respectively, indicating a high level of reliability. The OPLS-derived biplot (Fig. 3A) revealed that red mung samples exhibited the strongest antidiabetic activity, as they were closest to the activity data.

Fig. 3

OPLS biplot chart for discrimination of mung samples according to α-amylase and α-glucosidase inhibitory activities (A). Coefficient plots of OPLS model revealing putative biomarkers correlated with α-amylase (B) and α-glucosidase (C) inhibitory activities.

Full size image

To pinpoint the biomarkers linked to the antidiabetic effects of the extracts, OPLS coefficient plots were generated (Fig. 3B and C). These plots highlighted key metabolites, including gamma-aminobutyric acid (GABA), gallic acid, beta-sitosterol, laminaribiose, trehalose and L-tryptophan which significantly inhibit both alpha-glucosidase and alpha-amylase activities (Fig. 3A, B and C).

The identified bioactive antidiabetic markers align with previous studies which evidenced the effective inhibitory activities of gamma-aminobutyric acid (GABA) against α-amylase and α-glucosidase enzymes33. Another leading investigation demonstrated that GABA administration to patients with type 2 diabetes improves insulin resistance through modulating insulin signalling pathway in skeletal muscle in diabetic patients34. Equally important, gallic acid has been credited with noticeable inhibitory efficacy against starch-hydrolysing enzymes, i.e. α-amylase and α-glucosidase, and serves as a promising nutraceutical product for type 2 diabetes control35. Another study has demonstrated that beta-sitosterol exerted noticeable inhibitory actions on α-amylase and α-glucosidase in dose dependent manner and modulated early stages of type II spontaneous diabetes in mutant mice models36. Essentially, previous reports have suggested that trehalose is a potential therapeutic agent to diminish blood glucose level and rectify diabetes-related complications37.

Our experimental findings along with OPLS analysis illuminated that these compounds stated above were possibly set as potentially efficacy compounds that function together as safe dietary inhibitors for preventing digestion of carbohydrates to control diabetes.

Targeted UPLC quantitation of antidiabetic biomarkers revealed from OPLS model

A UPLC method for quantification of unravelled bioactive compounds, namely; gamma-aminobutyric acid, gallic acid and beta-sitosterol; in mung samples was developed and validated in terms of specificity, linearity, limit of detection (LOD), limit of quantification (LOQ), precision, and accuracy by adhering to ICH regulatory criteria38. The validation parameters of the developed method are depicted in Table S1. Meanwhile, the concentration of gamma-aminobutyric acid, gallic acid and beta-sitosterol in different mung samples are represented in Table 2.

Table 2 Quantitation of GABA, Gallic acid and beta-sitosterol in red, black and green mung bean species (mg/g dry weight).

Full size table

NIR spectroscopic analysis of different mung samples and PLSR for prediction of quantified biomarkers

NIR has proven to be an effective tool being advantageous over other analytical techniques, including rapidity, ease of use, and price-effectiveness. It can record spectra from both solid and liquid samples with minimal pretreatment and enables the parallel analysis of multiple components39. However, identifying specific absorption bands linked to particular functional groups or chemical compounds can be challenging40.

As illustrated in Fig. 4A, distinguishing between the raw NIR spectra of different mung samples is difficult because of the significant overlap of bands. However, preprocessing techniques like SNV (Fig. 4B). and first derivative treatment (Fig. 4C) of the raw spectra highlighted seven distinct absorption peaks. PCA of the processed NIR data, shown in Fig. 5A and corresponding HCA dendrogram Fig. 5B showed distinct separation between red, green and black mung samples in a manner similar to the pattern recognized from UPLC-MS data (Fig. 5A and B). The most significant wavenumbers as revealed from the loading line plots for various variables of the first and second principal components (Fig. 5C and D), respectively were identified as 4420–4366 cm− 1 (associated with CH stretching second overtone of –CH2), 5029–4990 cm− 1 (C = O second overtone of flavonoid nucleus)41, 5768–5749 cm− 1 (involving C-H stretching first overtone of –CH2, –CH3, –CH = CH–), 6950 cm− 1 (O-H first overtone of flavonoid)42 and 7709–7529 cm − 1 (C-H second overtone stretch vibrations models in –CH3)43.

Fig. 4

Superimposed NIR spectra of the 24 mung samples; raw spectra (A), after preprocessing using SNV (B) and 1st derivative (C).

Full size image

Fig. 5

PCA score scatterplot using processed NIR data (A), HCA dendrogram (B) along with the line loadings plots of PC1 (C) and PC2 (D).

Full size image

PLSR models were developed to predict the concentrations of identified biomarkers i.e. gamma-aminobutyric acid, gallic acid and beta-sitosterol in the samples. The samples were divided into two groups: a calibration set with 15 samples and a test set with 9 samples. First derivative of the NIR spectra from the entire range (9000–3500 cm− 1) served as the predictor variables (X matrix), while UPLC quantification results of the three bioactive markers were used as the dependent variables (Y matrix). NIR spectroscopy was selected as the preferred analytical method over UPLC for several reasons: it offers quicker analysis, requires fewer reagents and involves simpler procedures. The models created from the calibration samples were eval‎uated based on their R² values, observed versus predicted plots, and the root mean square error of calibration (RMSEC). To validate the PLS model, a LOO cross-validation method was used, where each sample is excluded in turn, allowing the model to be constructed from the remaining samples to predict the excluded one44. LOO cross-validation was implemented to calculate RMSECV (Table 3 and Fig. 6A) in addition to the intercept values of permutation plots (Table 3 and Fig. 6B). To assess the model’s fitting quality, we utilized RMSEE. A well-constructed PLS model is indicated by low RMSEE and RMSECV values. Moreover, it is essential to minimize the regression factors used for an effective calibration model (Table 3 and Fig. 6C). To measure the PLS model’s predictive capability, we eval‎uated parameters such as RMSEP45. Given the favourable metrics obtained in Table 3; Fig. 6, this model can reliably predict efficacy-related markers from various mung bean samples using a simple and rapid NIR run.

Table 3 Performance statistics of NIR-PLSR model for prediction of the three antidiabetic markers in different mung bean samples. *Values are average of three measurements ± SD.

Full size table

Fig. 6

Validation of the constructed PLSR model for prediction of concentrations of GABA, gallic acid and beta-sitosterol in mung bean samples represented by; observed versus predicted plots (A), permutation plots (B) and correlation between RMSECV and optimal number of latent variables (C).

Full size image

Conclusion

Mung beans are common legumes consumed worldwide because of their financial and medical advantages working as a potential functional food in promoting good health. Previous studies reported their role in treatment of many metabolic illnesses, particularly type 2 diabetic mellitus (T2DM), because of its ability to reduce blood glucose levels. Therefore, in this study 24 mung bean samples collected from different red, black and green mung species have been imposed to an integrative metabolomic chemical profiling through UPLC-QqQ-MS analysis assisted by Fingerprint analysis of mung bean samples utilising Fourier Transform-Near-Infrared spectroscopy (FT-NIR). By combining these two independent techniques with multivariate analysis, a thorough and precise discrimination, along with high-quality assessment, of mung samples collected from various species was successfully achieved. A total of 71 chromatographic peaks spanning flavonoids, phenolic acids, fatty acids sugars, amino acids, and their lipid derivatives, and phytosterols were chemically profiled. Multivariable statistical models were built to reveal discriminatory compounds between red, black and green mung bean species including eriodictyol-O-glucoside, caffeic acid, formononetin-O-glucoside and genistin serve as focal discriminators of green mung beans while lysoPC 18:2, lanosterol, gallocatechin, tyramine and retinoic acid were the determining metabolites of red ones. Successively, the differential markers enriched in black mung samples included 10-formyltetrahydrofolate, stearidonic acid, vignatic acid A and PG (18:2/18:1). These compounds after being precisely identified using UPLC-QqQ-MS have been utilised to produce X-matrix of OPLS model, whereas α-amylase and α-glucosidase inhibition were allocated by the Y-matrix. The In vitro anti-diabetic assays revealed that the red beans were the most potent inhibitory activity against both α-amylase, and α-glucosidase enzymes followed by black beans and finally the green mung beans. Furthermore, it was found that gamma-aminobutyric acid (GABA), gallic acid and beta-sitosterol were the main health-relevant biomarkers. In order to predict of these biomarkers’ concentrations, a validated PLS regression model was built using FT-NIR data as predictor variables. Finally, this method may be expanded to use as a quick and easy FT-NIR run to accurately predict these active biomarkers from any red, black or green mung bean sample.

결론

녹두는 경제적 및 의학적 이점으로 전 세계적으로 소비되는 일반적인 콩과 식물로, 건강 증진에 기여하는 잠재적 기능성 식품으로 작용합니다. 기존 연구들은 혈당 수치를 낮추는 능력으로 인해 특히 제2형 당뇨병(T2DM)을 비롯한 여러 대사성 질환 치료에서의 녹두의 역할을 보고했습니다.

따라서 본 연구에서는 푸리에 변환 근적외선 분광법(FT-NIR)을 활용한 녹두 샘플 지문 분석을 보조로, UPLC-QqQ-MS 분석을 통해 다양한 적색, 흑색, 녹색 녹두 종에서 채취한 24개 녹두 샘플에 통합적 대사체 화학 프로파일링을 수행하였다. 이 두 독립적인 기법을 다변량 분석과 결합함으로써, 다양한 종에서 채취한 녹두 샘플에 대한 철저하고 정밀한 판별과 고품질 평가를 성공적으로 수행하였다. 플라보노이드, 페놀산, 지방산, 당류, 아미노산 및 그 지질 유도체, 식물 스테롤을 아우르는 총 71개의 크로마토그래피 피크가 화학적으로 프로파일링되었다. 다변량 통계 모델을 구축하여 붉은, 검은, 녹색 녹두 종 사이의 차별화 화합물을 규명했는데, 에리오디크티올-O-글루코사이드, 카페인산, 포르모노네틴-O-글루코사이드, 제니스틴이 녹색 녹두의 핵심 차별화 인자로 작용하는 반면, 리소포스파티딜콜린 18:2, 라노스테롤, 갈로카테킨, 티라민, 레티노산이 붉은 녹두의 결정적 대사산물로 확인되었다. 그 결과, 검은콩 샘플에 풍부하게 함유된 차별적 마커로는 10-포밀테트라하이드로폴레이트, 스테아리돈산, 비그나틱산 A 및 PG(18:2/18:1)가 포함되었습니다. 이 화합물들은 UPLC-QqQ-MS를 통해 정밀하게 동정된 후 OPLS 모델의 X-매트릭스 생성으로 활용되었으며, α-아밀라아제 및 α-글루코시다아제 억제 활성은 Y-매트릭스로 할당되었다. 체외 항당뇨병 시험 결과, 적색 콩이 α-아밀라아제 및 α-글루코시다아제 효소에 대해 가장 강력한 억제 활성을 보였으며, 그 다음으로 검은콩, 마지막으로 녹두 순이었다. 또한 감마-아미노부티르산(GABA), 갈산 및 베타-시토스테롤이 주요 건강 관련 바이오마커로 확인되었다. 이러한 바이오마커 농도를 예측하기 위해 FT-NIR 데이터를 예측 변수로 사용하여 검증된 PLS 회귀 모델을 구축하였다. 결론적으로, 이 방법은 적색, 검정색 또는 녹색 녹두 샘플로부터 이러한 활성 바이오마커를 정확히 예측하기 위한 빠르고 쉬운 FT-NIR 분석법으로 확장될 수 있다.

Materials and methods

Chemicals and reagents

The external standards: γ-Aminobutyric acid (GABA), gallic acid and β-sitosterol were acquired from Sigma-Aldrich (St. Louis, USA). Other chemicals as methanol, formic acid and acetonitrile were of analytical grade and procured from Fisher Scientific, UK and Merck India.

Sample preparation and extraction

Three representative mung species with different bean colours (500 g), including Vigna radiata (green mung beans), Vigna mungo (black mung beans), and Vigna angularis (red mung beans) were selected for this study and supplied by Agricultural Research Centre, Cairo University, Egypt during September 2022. Differently coloured mung beans were vacuum freeze-dried (Edwards Lyophilizer, ModE2MB, Brazil), ground using an electric blender, and passed through a 50-mesh sieve. Subsequently, 100 g of each powder was weighed and kept in a desiccator until NIR analysis which the remaining fractions (400 g) of each mung variety were subjected to extraction process according to previous protocols 30,46. In short, 800 ml of 70% ethanol was used to extract 400 g of each lyophilised mung powder twice, using an ultrasonic bath apparatus (3 L Alpha Plus, Japan) at 35 °C for 60 min. Using a rotary evaporator (BuchiRotavapor Model R-200, Flawil, Switzerland), the different extracts were vacuum-concentrated until dryness.

Chemical characterization of differently coloured mung beans extracts through UPLC-QqQ-MS/MS analysisPreparation of different mung samples and reference standards solutions for LC-MS analysis

In accordance with previous sample preparation protocols 29,44, 1 mg/mL from each mung extract were individually resuspended in methanol (HPLC-grade), filtered using a Millipore membrane and 10 µL of each sample were loaded onto the column. Similarly, stock standard solutions of γ-Aminobutyric acid (GABA), gallic acid and β-sitosterol were prepared. These stock solutions were then diluted to working concentrations ranging from 0.01 to 2 mg/mL, as shown in Table S1. Each concentration level of the standard solutions was injected three times into the chromatographic column using 5 µL aliquots. Calibration curves of reference standards were constructed wherein the analytical parameters of linearity, limit of detection (LOD) and limit of quantification (LOQ) were tracked and assessed building on ICH criteria for analytical validation38 (Table S1).

UPLC-MS/MS analysis and mass data processing

For detailed information regarding UPLC chromatographic parameters, Electrospray ionisation- triple quadrupole mass spectrometry (ESI-QqQ-MS/MS) conditions, and the method used for metabolite annotation, please refer to the supplementary materials.

The pre-processing of the generated chromatograms, features detection, integration and alignment was performed via MZmine 2.0 (http://mzmine.sourceforge.net/) software. Mass detection utilized the local maxima approach to determine the minimum signal intensity necessary for a signal to be considered a chromatographic component. Additional MS data processing techniques, including background subtraction and adjacent peak deconvolution, were employed to consolidate nearby peaks into single peaks. Spectral alignment was achieved using the join aligner technique with one minute tolerance window for retention time, allowing for the matching of corresponding peaks across different samples while accommodating variations in retention times and m/z values during chromatographic runs47.

Subsequently, metabolites assignment was conducted through spectral matching with reference standards data. As well, our self-built database, typical mass spectra (quasi-molecular ions and characteristic fragmentation profiles), pertinent literature data and some public metabolite databases mainly MassBank and the Dictionary of Natural Products database were exploited in order to attain a high degree of annotation confidence.

Multivariate statistical analysis

Hierarchical cluster analysis (HCA) heat map was generated using Metaboanalyst 6.0 (http://www.metaboanalyst.ca/) to eval‎uate the relative quantities of annotated metabolites in different Mung species. Multivariate statistical analysis was carried out using SIMCA-P software (Version 14.0, Umetrics, Sweden), which was utilized to develop an orthogonal projection to latent structures-discriminant analysis (OPLS-DA) model for segregation of Mung species based on their chemical compositions in addition to an OPLS model for clustering based on antidiabetic activities. Additionally, OPLS correlation coefficient plots were generated to identify phytoconstituents that were significantly correlated with the in vitro biological activities under investigation.

FT-NIR spectral acquisition and data preprocessing

Near-infrared spectra were collected by multi-purpose Fourier Transform Near Infrared (FT-NIR) spectrometer (Bruker Optics GmbH, Ettlingen, Germany) equipped with an integrating sphere module and an InGaAs linear photodiode array detector, covering wavenumber range of 9000–3500 cm− 1. Spectra were recorded using OPUS software (version 6.5, Bruker Optics Inc.) at a resolution of 16 cm− 1, averaging over 64 scans per spectrum. Each sample was analysed in triplicate to minimize instrumental noise, utilizing uniform glass vessel (20 × 50 mm) filled with approximately 1 g of powder. Data pre-processing involved standard normal variate (SNV) transformation then 1st derivative methods, where SNV centres and scales the data while the 1st derivative corrects baseline drifts and highlights subtle differences. These processes were executed using SIMCA 14.0 software.

PLS-R models for predicting bioactive markers from NIR metabolome

Partial Least Squares Regression (PLS-R) serves as a vital tool for analysing multivariate data, primarily for prediction. PLS-R calibration models were developed to correlate predictor variables from the NIR metabolome with the bioactive markers found in mung samples. The X-matrix contained concentrations of bioactive markers; namely, gamma-aminobutyric acid, gallic acid, and beta-sitosterol; previously determined by UPLC in each mung sample, while the X-matrix comprised NIR spectra from 24 mung samples (9000 –3500 cm− 1) in accordance with a previous protocol48. Model accuracy was assessed using metrics such as root mean square error of estimation (RMSEE), regression coefficient (R2, root mean square error of prediction (RMSEP), and root mean square error of cross-validation (RMSECV). The constructed PLS model was validated through Leave-One-Out (LOO) cross-validation, systematically excluding one sample at a time to predict it using the remaining data. High R2 values and low RMSEE, RMSEP and RMSECV indicate a robust model, with a minimal number of regression factors preferred for effective calibration.

In vitro anti-diabetic activity testing of the studied mung speciesPancreatic α-amylase inhibitory activity

With a few minor modifications, the assay was performed in compliance with a previous approved protocol49. The supplementary information contains the assay’s specifics.

α-Glucosidase inhibitory activity

In this assay, slight modifications were made to the previously published method50. The supplementary information contains the assay’s specifics.

Statistical analysis

The data were interpreted using the statistical software GraphPad Prism v8 (GraphPad Software, San Diego, CA, USA). The statistical analysis of the data involved one-way analysis of variance, with P < 0.05 being deemed statistically significant. The data are expressed as the mean ± standard deviation.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

References

1. Ganesan, K. & Xu, B. A critical review on phytochemical profile and health promoting effects of mung bean (Vigna radiata). Food Sci. Hum. Wellness. 7, 11–33 (2018).

   Article Google Scholar 

2. Takahashi, Y. & Tomooka, N. in Taxonomy of Mungbean and its Relatives BT - The Mungbean Genome. 27–41 (eds Nair, R. M., Schafleitner, R. & Lee, S. H.) (Springer International Publishing, 2020). https://doi.org/10.1007/978-3-030-20008-4_3

3. Milenkovic, D. et al. Interindividual variability in biomarkers of cardiometabolic health after consumption of major plant-food bioactive compounds and the determinants involved. Adv. Nutr. 8, 558–570 (2017).

   Article PubMed PubMed Central Google Scholar 

4. Kowalska, K. Natural compounds involved in adipose tissue mass control in in vitro studies. Postepy Hig Med. Dosw (Online). 65, 515–523 (2011).

   Article PubMed Google Scholar 

5. Burton-Freeman, B. M., Sandhu, A. K. & Edirisinghe, I. Mangos and their bioactive components: adding variety to the fruit plate for health. Food Funct. 8, 3010–3032 (2017).

   Article CAS PubMed Google Scholar 

6. Singh, B. & Kaur, G. Mung beans: bioactive compounds and their potential health benefits. in Handbook of Cereals, Pulses, Roots, and Tubers 449–460 (CRC, 2021).

7. Hou, D. et al. Mung bean (Vigna radiata L.): bioactive polyphenols, polysaccharides, peptides, and health benefits. Nutrients 11, 1238 (2019).

   Article CAS PubMed PubMed Central Google Scholar 

8. Kudre, T. G., Benjakul, S. & Kishimura, H. Comparative study on chemical compositions and properties of protein isolates from mung bean, black bean and Bambara groundnut. J. Sci. Food Agric. 93, 2429–2436 (2013).

   Article CAS PubMed Google Scholar 

9. Nair, R. M. et al. Biofortification of Mungbean (Vigna radiata) as a whole food to enhance human health. J. Sci. Food Agric. 93, 1805–1813 (2013).

   Article CAS PubMed Google Scholar 

10. Anwar, F., Latif, S., Przybylski, R., Sultana, B. & Ashraf, M. Chemical composition and antioxidant activity of seeds of different cultiv