Synthesis and accumulation of amylase-trypsin inhibitors and FODMAPs in bread wheat (Triticum aestivum L.) during grain development

Recent gastroenterology studies assume that amylase-trypsin inhibitors (ATIs) and FODMAPs (fermentable oligo-, di-, monosaccharides and polyols) play an important role in promoting wheat sensitivity. Hitherto, no study has investigated the formation of ATIs during the development of the wheat caryopsis. We collected caryopses of common wheat cv. ‘Arnold’ at eight different grain developmental stages to study compositional changes in ATI and FODMAP content.


Conclusions
The compositional changes of ATIs and FODMAPs during the development of wheat grains were characterized by qualitative and quantitative approaches. The results provide new insights into the complex metabolisms during grain lling and maturation, with particular emphasis on the ATI content as well as the inhibitory potential towards trypsin. The time lag between ATI accumulation and development of their biological activity is supposed to be attributed to the assembling of ATIs to dimers and tetramers, which seems to be crucial for their inhibitory potential.

Background
Amylase-trypsin inhibitors (ATIs) are water-soluble cereal proteins, which were identi ed as causative proteins for non-celiac wheat sensitivity (NCWS) and other wheat-related disorders. The diseases are characterized by both intestinal and extra-intestinal symptoms that are tightly linked to the consumption of wheat and other gluten-containing foods [1,2]. As there are no clear diagnostic biomarkers for NCWS, a precise number for prevalence is di cult to obtain, but based on the few data available, prevalence is estimated to range from 0.6-10% [3].
ATIs were found to initiate innate immune responses by directly activating speci c pro-in ammatory receptors (i.e. TLR4) in human body cells, leading to severe in ammations and immune reactions [4,5].
Additionally, ATIs have the potential to inhibit the activity of two important digestive enzymes in the gastrointestinal system, amylase and trypsin, causing a signi cant impairment of digestion [6]. Due to this inhibitory feature, ATIs and other proteins from the prolamine superfamily are considered crucial for the natural defence mechanism of the plant itself [7]. They are located in the endosperm of cereal seeds, where they defend starch and protein reserves by blocking amylase and trypsin activities of invading parasites and insects [8,9]. ATIs can be classi ed, according to their degree of aggregation, into monomeric, dimeric and tetrameric forms. These aggregates are stabilized by non-covalent interactions, including hydrophobic interactions. Due to the low stability of these bonds, ATI aggregates break down easily into their monomeric forms. The enzymatic activity of ATIs is strongly dependent on the 3Dstructure and the state of aggregation [9].
Besides ATIs, FODMAPs (fermentable oligo-, di-, monosaccharides and polyols) have been implicated in NCWS and are suspected to trigger irritable bowel syndrome (IBS) that causes similar, however exclusively intestinal, symptoms [10]. Among FODMAPs, fructans are most abundant in wheat and wheat-based products. Other FODMAPs, such as free fructose and galacto-oligosaccharides, including ra nose, stachyose and verbascose, were also found to be present in wheat, however in negligible amounts [11].
Although recent studies provided valuable information on the proteomics of the wheat grain [7,12], the metabolism of ATIs during grain lling and maturation has not been investigated thoroughly so far. In the present study, grains of bread wheat (Triticum aestivum L.) cv. Arnold were harvested at eight grain developmental stages from anthesis to maturity in order to display the changes in ATIs and FODMAPs during grain lling. Understanding ATI and FODMAP synthesis and regulation during grain lling might be of fundamental importance for breeding purposes.

Kernel growth
Wheat kernel development was studied from anthesis until 46 days after anthesis (DAA) when maturity was reached. Images of developing spikes and kernels as well as their corresponding kernel sizes and kernel dry weights are shown in Fig. 1. While samples taken 7 to 25 DAA showed rather immature ears, spikes and kernels of samples taken 33 to 46 DAA were close to maturity. Caryopses harvested 7 days after full/late owering (anthesis, BBCH scale 65-67) were in transition from the watery ripe to early milk stage. Caryopses harvested 18 DAA were in transition from late milk to early dough stage. While samples taken 7 to 18 DAA were still decidedly green, samples harvested 25 DAA were right in the soft dough stage, starting to turn yellow. Ears sampled 33 to 46 DAA had lost all their green color. In general, dry kernel weight was increasing from 4.8 mg/kernel to 45.2 mg/kernel with a maximum of 49.0 mg/kernel at 33 DAA. Simultaneous to dry kernel weight, grain size increased steadily during grain lling and maturation from 13.4 mm 2 to 20.8 mm 2 with a maximum of 24.8 mm 2 at 25 DAA. The slight decrease in kernel size during the later growth stages was corresponding to the nal dehydration phase before maturation. Based on kernel growth analyses and compositional monitoring across the eight sampling dates, grain development could be divided into three phases: cell division and expansion (anthesis to 14 DAA), grain lling (14-25 DAA), grain maturation and desiccation (25-46 DAA) (Fig. 2).

Accumulation of proteins during grain development
Protein contents of the developing wheat kernels derived by Dumas and Bradford method as well as RP-HPLC are shown in Table 1 and Fig. 2. RP-HPLC chromatograms displaying the separation of non-gluten proteins upon the eight developmental stages can be found in Additional le 1: Fig. S1. In general, crude protein concentration showed considerable variations (12-18%) in the time frame of this study (i.e. 7 to 46 DAA). In the early stages of grain development crude protein content declined from 17.7% to 12.3%, whereas kernel size and dry kernel weight were increasing enormously (Fig. 2). After reaching a minimum of 12.3%, further protein was accumulated resulting in 14.5% at nal maturity. While protein content per 100 g of our was found to be slightly increasing after an initial decrease (Fig. 2a), protein content on a single kernel basis was steadily increasing in the rst phase of grain development until a nal protein content of 6.6 mg/kernel was reached (Fig. 2b). Concentrations on a single kernel basis for other traits are listed in Additional le 2: Table S1 and Additional le 3: Table S2. On the contrary, the amount of saltsoluble proteins (albumins and globulins) remained relatively constant among the eight maturation stages with values from 1.2% to 1.9% (Table 1). ATI proteins extracted from wheat kernels were quanti ed by RP-HPLC. No ATIs could be detected in the rst and only marginal amounts in the second developmental stage. Hence, initial ATI accumulation occurred after one week after anthesis. Subsequently, the concentration increased gradually until grain maturity and resulted in a maximum ATI concentration of 0.7 g/100 g at 46 DAA ( Table 1).

Characterization of ATIs by MALDI-TOF MS
Sample extracts were analysed by MALDI-TOF MS in order to verify the presence of ATIs. Since ATI aggregates are stabilized by rather low non-covalent forces and thus, easily disrupted by low or high pH, denaturating or organic reagents, ATIs were mainly identi ed and quanti ed in their monomeric form.
Spectra of the samples are shown in Fig. 3. The α-amylase inhibitor (AAI) standard from wheat (Fig. 3a), purchased from Sigma-Aldrich, showed a strong peak accumulation at around 12-14 kDa and a small peak around 15.5 kDa, which can be assigned as ATIs and CM3 protein. The broad peak at 26-27 kDa represents 2M+H ions of the detected ATIs. Furthermore, distinct signals with lower molecular weights (around 6 kDa and 10 kDa, respectively) were spotted in the AAI standard spectrum, that were resulting from the high amount of impurities present in the AAI standard [13,14]. The PWG gliadin isolate (Fig. 2b) illustrated a typical gliadin pattern with a majority of α-/ß-gliadins from 30-40 kDa and minor abundance of ω-gliadins from 40-55 kDa. Besides gluten type proteins, the PWG spectrum revealed the presence of ATIs and avenin-like proteins, which was recently con rmed by Lexhaller et al. [15]. As ATIs are rather small proteins that range from 13-18 kDa [13,16,17], sample peaks in this area were assigned predominately to ATIs and to other proteins from the prolamine superfamily (e.g. avenin-like proteins) in a minor extent [7]. The rst sample from 7 DAA revealed a complete absence of ATIs and other storage proteins. Only an unde ned and strong increase of the baseline below 10 kDa was observable, which probably resulted from intermediate peptides and free amino acids [18]. The second sample collected four days later showed weak peak intensities around 13, 16 and 32 kDa, which indicated the biosynthesis of the mentioned proteins. Samples taken 14 and 18 DAA presented a clear onset of ATI accumulation simultaneous to the synthesis of gliadins, which were represented by peaks >30 kDa. Comparison of signal intensities arising from ATIs and gliadins revealed no trend for later developmental stages, which might be in uenced by manual sample preparation for MALDI-TOF analysis.

Characterisation of ATI functionality
The biological function of ATIs was described in terms of trypsin inhibitory activity (TIA) by an enzymatic assay [13]. Despite the presence of ATIs at an earlier stage, the onset of enzymatic activity occurred later as the samples taken 7 to 18 DAA showed no detectable TIA (Table 1). An initial inhibitory activity was obtained at 25 DAA, however not quanti able as the minimum inhibition for a proper quanti cation (i.e. 40%) was not reached. Samples of later developmental stages showed considerable activities from 79.1 mg/kg to 89.7 mg/kg.

Carbohydrate quanti cation by HPAEC-PAD
Extraction of soluble mono-, di-and oligosaccharides from the maturing wheat kernels was performed in water without heating. In order to gain further insights into the composition of carbohydrates, the insoluble residues were subjected to a strong acid hydrolysis with 2.5 M TFA. As a result, insoluble carbohydrates were depolymerized to monomers and readily quanti able by HPAEC-PAD. Changes in carbohydrate compositions of the samples are shown in Fig. 4 and Table 2. In general, the total amount of soluble carbohydrates was decreasing throughout grain development from 18.7% to 1.7%. Concentrations of all mono-, di-and oligosaccharides present were decreasing, except of sucrose, which was found only in traces at the beginning of grain development, but showed a nal amount of 0.7%. The range of ra nose concentrations remained relatively stable over 46 days, whereas other galacto-oligosaccharides such as stachyose and verbascose could not be detected in later developmental stages or only in very small amounts. The course of changes in glucose and fructose concentrations were almost identical and were the highest during the phase of cell division and expansion (7 DAA) with maximum values of 4.8% for glucose and 8.1% for fructose. During grain development both concentrations decreased signi cantly resulting in poor amounts of 0.05% for glucose and fructose in the mature kernels. Sucrose concentrations were similar to glucose and fructose decreasing slightly during grain lling, but signi cantly increasing during grain maturation and desiccation. Due to the opposing behavior of glucose, fructose and sucrose, a shift in the monosaccharide (sum of glucose and fructose) to sucrose ratio could be observed. Besides free glucose and fructose concentrations, changes in the fructan content were found to be the most striking alteration during grain development. The amount of low molecular weight fructans (GF2-GF4) was rapidly decreasing during the milky stage (7 to 11 DAA) from an initial concentration of 1.3 to 0.6% (Table 2).
After the milky stage only a minimal reduction to a nal amount of 0.3% occurred.

Insoluble carbohydrate composition
As described by Pritchard et al. [19], the sum of arabinose and xylose monosaccharides represents the amount of arabinoxylans. In this case, water-unextractable arabinoxylan (WU-AX) content was determined, which is usually twice that of water-extractable arabinoxylans (WE-AX) in wheat our [20]. In general, WU-AX concentrations were increasing during grain lling and maturation from 6.0% to 7.4% with a maximum of 12.5% at 18 DAA. Arabinoxylans can be classi ed according to their molecular structure, which is typically expressed by the arabinose to xylose ratio (ARA/XYL). ARA/XYL of the investigated samples were found between 0.5 and 1.0. Besides WU-AX, low amounts (0.3% to 0.6%) of waterunextractable galactose were detected in the grain samples.
Insoluble wheat starch in the grain samples was represented by the amount of glucose after acid hydrolysis as the rst extraction with water should have removed free glucose and β-glucan present in the grain sample. Thus, glucose in the water-insoluble residue derived predominantly from starch. In contrast to the classical method for starch quanti cation, the approach used in this study might also cover other insoluble glucose-containing polysaccharides, such as cellulose, to a minor extent. Throughout grain lling and maturation starch concentrations were increasing from 19.6% to 87.3% (Table 2). While the amount was enormously increasing during the rst two phases of grain development (from 19.6% to 79.3%), samples of the grain maturation and desiccation phase showed rather constant values between 69.1% and 87.3% (Fig. 2a). A similar course of accumulation was monitored on a single kernel basis (Fig.  2b). In general, the course of starch accumulation in the grain was highly correlated to the corresponding dry kernel weight. The crude protein content was also found to be negatively correlated with starch concentrations (Table 3).

Discussion
In this study, wheat samples were collected during grain development in order to monitor the accumulation of ATIs as well as compositional changes in the carbohydrate pro le. The results revealed important insights into their expression pro les during speci c developmental stages.
Kernel growth and crude protein accumulation The determination of dry kernel weight and kernel size by Verspreet et al. [21] and Gegas et al. [22] revealed similar values for mature wheat grains as reported in this study, with a dry weight of 51.4 ± 1.8 mg/kernel [21] and a mean kernel size of 22.8 ± 3.6 mm 2 [22], respectively.
In general, relative crude protein content, which was calculated based on the total nitrogen content in the sample, was decreasing from 28.4% at early grain lling to 13.0% in the mature kernel (Fig. 4). A similar decline of crude protein has been reported in earlier studies on immature wheat [18,21,23]. Furthermore, Jennings and Morton [18] demonstrated that at the beginning of grain development about 50% of the total nitrogen was present in protein, while the remaining half was mostly present in free amino acids. However, as the amount of free amino acids and peptides was decreasing during grain development, the majority of the total nitrogen compounds were found to consist of proteins, except for kernels in the very rst phase of development [18]. Protein synthesis and assembly were identi ed as principal functions of the wheat endosperm during grain development, with particular predominance in the early stages of grain development (10 DAA) [24]. The strong MALDI TOF MS signals between 5-10 kDa at early grain development which decreased thereinafter con rmed the presence of bigger peptides at the beginning of grain lling and conversion to proteins with higher masses during grain maturation. The Dumas method measured the nitrogen content by combustion and a conversion factor (i.e. 5.7) is used to calculate the protein content. Because of the mentioned procedure, the crude protein content includes all amino acids, peptides and proteins independently of their size.
The decrease of the total crude protein content observed in the rst third of grain development might be due to the extensive NWSC accumulation, mainly starch (Fig. 4). This decrease was followed by a vast accumulation of crude protein during the grain lling and maturation phases, which was mentioned before in other studies [18,21,23]. The slight increase of protein in the later stages of growth might result from the nal desiccation prior to maturation. Correlation analysis of total crude protein content and kernel size indicated a signi cant negative correlation (Table 3), as samples with kernel sizes >21 mm 2 showed the lowest protein concentrations (Fig. 2). The presented values of salt-soluble proteins were slightly higher when compared to albumin concentrations (0.7-1.2%) quanti ed in a wide range of different Austrian bread wheat samples [16]. However, in the mentioned study only water was used for the extraction of albumins, which precludes the simultaneous quanti cation globulins. Although albumin and globulin concentrations remained relatively stable throughout grain development, their amounts are correlated to crude protein contents (Table 3). In general, contents of total protein and salt-soluble proteins of grains from the nal maturity were similar to those of cv. Arnold from the year 2017 and 2018 as presented by Call et al. [25]. The results on kernel growth and protein accumulation were in agreement with general knowledge on grain development.

ATI accumulation and trypsin inhibition
As shown in Table 1, ATI content increased steadily from 7 DAA until nal maturity. These ndings were in accordance with Finnie et al. [26], who described the gradual accumulation of ATIs throughout barley seed development. This pattern of appearance might be attributed to their functional role in the defensive system of the cereal plant. As their role is to defend the starch and protein reserves of the seed, their accumulation is expected to be simultaneous with grain lling. Also 2D gel electrophoresis-mass spectrometry of wheat proteins by Vensel et al. [24] proved the dominance of proteins involved in stress and defense, including amylase and trypsin inhibitors, at later stages of grain development [24]. The detected amounts of ATIs were in accordance, however slightly lower when compared to values obtained for the same wheat variety harvested in 2017 and 2018, which were 1.3 and 1.1 g/100g, respectively [24]. The slight decrease in ATI concentrations might be due to the replacement of soy trypsin inhibitor (TI) as calibration standard to BSA. Generally, the simple RP-HPLC method used for ATI quanti cation is known to overestimate ATI amounts, since avenin-like proteins were coeluating and thus, included in these results.
MALDI spectra of the Arnold samples con rmed the results obtained by RP-HPLC since ATIs were not detected at 7 DAA and only marginal signals were detectable after 11 DAA. The sample taken 14 DAA showed a clear onset of ATI accumulation simultaneous to the synthesis of gliadins. These ndings are in contrast to Carbonero et al. [9], who claimed that ATI synthesis precedes that of the storage proteins. However, quantity of storage proteins has not been determined in the present study.
Although ATIs could be detected at earlier stages by MALDI-TOF MS, quanti cation by RP-HPLC demonstrated that the samples shortly after anthesis contained rather low concentrations. Nevertheless, a certain time lag from initial ATI occurrence until biological activity of ATIs was evident throughout grain development. While the samples taken 25 and 33 DAA exhibited similar ATI concentrations (0.6 g/100 g), they signi cantly differ in their trypsin inhibitory potential (<LOQ and 79.6 mg/kg, respectively). Based on this discrepancy, no correlation could be observed for ATI concentration and their corresponding trypsin inhibition (Table 3) as it was also demonstrated in earlier studies [25]. The time lag that occurred might be due to the preceding assembling of ATIs to dimers and tetramers, which seems to be crucial for biological functions [9]. Based on the acquired results, the formation of biological activity of ATIs could be assigned to a maturation period between 18 and 33 DAA. In order to further narrow down this period, sample collection could be performed more frequently in follow-up studies. In general, trypsin inhibitory potential at later developmental stages were closely to results of Call et al. [13] with 73.21 mg/kg and Call et al. [25] with 66.16-84.63 mg/kg for cv. Arnold . In comparison with TIA values from other varieties, cv. Arnold is characterized by a medium trypsin inhibitory activity [13,25].
Based on the results on ATI synthesis and accumulation, mature seeds could be identi ed to have the highest potential to trigger NCWS, but on the other side late-stage seeds might be a promising source in order to elucidate the molecular nature of ATIs, which is essential for proper diagnosis and therapy.

Compositional changes in the carbohydrate pro le
The levels of soluble sugars found in the mature grains of the present study were comparable to results from Call et al. [27], whose study found 0.03% glucose, 0.07% fructose, 0.8% sucrose and 0.3% ra nose among 19 winter bread wheat cultivars grown and harvested 2016 in Tulln. In general, carbohydrate metabolism is known to be an abundant process throughout grain development [24]. The decrease of soluble mono-and disaccharides during kernel ripening re ected their progressive conversion into storage polysaccharides, as in fact, starch concentrations were increasing from 19.6% to 87.3%.
During grain development the amount of short-chain fructans was reduced by approximately 80%. The shown levels in mature grains were in accordance with results from Peukert et al. [28], who demonstrated degradations during barley grain development from 0.26% to 0.01% for kestose (GF2) and from 0.18% to 0.16% for nystose (GF3). Probably, fructans with higher molecular weights were generated later, which was responsible for this reduction. Fraberger et al. [29] reported degrees of polymerization (DP) of up to 15 in whole meal our of the same wheat cv. Arnold . Fructans are known to be abundant in immature wheat kernels and similar reduction rates of 50-90% have been reported for mature wheat when compared to immature kernels [30]. However, it has to be pointed out that in the present study only low molecular weight fructans (GF2-GF4) have been determined. Fructans have received much attention in the scienti c literature thanks to their health bene ts and physiochemical properties. Due to the high levels of fructans in immature wheat kernels, the employment of wheat grains harvested at the milk stage for the production of functional foods has been discussed [31,32]. However, these products might then again not be suitable for people suffering from irritable bowel syndrome (IBS) as fructans are suspected to trigger symptoms of IBS [10].
Similar to the study conducted by Verspreet et al. [21], AX concentrations increased rapidly during the cell division and expansion phase, remained relatively stable during grain lling and resulted in slightly decreased, however stable AX amounts in the range of 7.4-7.9% in the maturation and desiccation phase.
Comparable results for WU-AX in mature grains were obtained by Cetiner et al. [33] with a mean value of 7.2% for modern wheat varieties. ARA/XYL in the present study was slightly increased when compared to typical average values of 0.5-0.6 as reported by Cleemput et al. [34]. However, ARA/XYL seemed to be increased in WU-AX due to more intensive branching [35]. Furthermore, arabinose from arabinogalactan was not considered in the quanti cation which might lead to adulterated ARA/XYL. AX are known to be the major component of wheat cell walls, which is re ected by the high amounts found in this study. The poor concentrations of water-unextractable galactose (WU-GAL) found in the present study could be traced back to galactooligosaccharides that are embedded in minor amounts in the cell wall structure as also shown by Mares and Stone [36].
In the literature, starch content was found to rarely exceed 72% of wheat dry matter [37]. The increased amounts found in this study might be caused by an incomplete β-glucan or cellulose removal due to the single cold water extraction step applied. However, the total sum of protein and carbohydrate components does not exceed 110% of the dry kernel weight, which is acceptable considering the measurement uncertainties of the numerous analytical traits that were determined for each caryopsis sample.

Conclusions
In this study, the content of proteins, ATIs and major carbohydrates was studied in developing wheat kernels (Triticum aestivum L. cv. Arnold ) from anthesis until maturity. By combining qualitative and quantitative approaches, a detailed overview on compositional changes in carbohydrates and accumulation pattern of ATIs during grain development was obtained. The results provide insights into the complex compositional changes during grain lling and maturation, with particular emphasis on the protein and ATI content, as well as in the inhibition activities of ATIs towards trypsin. In general, ATI concentration and activity were found to increase throughout grain development, while fructans were abundant in immature wheat kernels. Although the accumulation of ATI monomers started in the rst phase of grain development, biological activity in terms of trypsin inhibition could only be detected after a time lag of around three weeks. This result supports the assumption that biological function of ATIs is strongly dependent on their aggregation state and conformation. Changes in environmental conditions during grain development could additionally alter the rate of ATI synthesis and ultimately affect the immunogenic potential of the grain. Although several studies helped to understand the complex gene network that regulates protein expression during grain development, the biochemical mechanisms for ATI synthesis still require further research. The complexity of trypsin inhibition was impressively demonstrated in a recent research study: although total ATI concentrations were reduced by gene silencing of three major ATI genes, TIA was increased [38].

Reagents
All chemicals and reagents used for sample preparation and analysis were of analytical grade, and purchased from Merck (Darmstadt, Germany), VWR (Radnor, PA) and Carl Roth (Karlsruhe, Germany).
Ultra-pure (UHQ) water was used for all preparations and was provided by a Dionex TM IC Pure system (Sunnyvale, CA, USA).  Table S3. Unfertilized ovaries were excluded from analysis.

Determination of dry kernel weight and kernel size
At least 20 freeze-dried kernels were weighted in order to determine the average kernel weight for each developmental stage. Determination of kernel size of the maturing kernel samples was conducted by a graphical evaluation via ImageJ2 software [39]. Calibration with a simultaneous evaluation of a de ned length scale and an ellipsoid model were chosen for proper kernel size quanti cation. The reported values were mean values obtained from the measurement of at least 20 kernels ± SEM.

Sample preparation for protein determination
The caryopses for analyses were manually collected from the spikes and ground under liquid nitrogen using a CryoMill (Retsch GmbH, Haan, Germany), lyophilized and stored at -18°C prior to analysis. For ATI determinations, 1 g of wholemeal our was extracted twice with 10 mL n-hexane in order to remove sample fat and chlorophyll, especially from rather immature samples, which might interact with spectrophotometric measurements. Subsequently, ATIs were extracted with 150 mM NaCl solution containing 1.3 mM phosphate buffer (pH 7). After vortexing the suspension for 20 s, the extraction was completed under vigorous shaking in an overhead shaker for 10 min. Non-soluble particles were separated by centrifugation (3000 g, 10 min) before combining the resulting supernatants. Sample preparation was performed twice for each sample.

Determination of protein content
Nitrogen content of the milled wholemeal ours was determined by the Dumas combustion method according to ICC standard 167 (Determination of crude protein in grain and grain products for food and feed by Dumas combustion principle) using a DuMaster D-480 (Büchi Labortechnik AG, Flawil, Switzerland). Calibration was performed with L-aspartic acid (Sigma-Aldrich), and a factor of 5.7 was used to convert the resulting nitrogen amounts into crude protein content. Additionally, the amount of total soluble nitrogen (TSN) was determined using the sodium chloride extracts with a Roti®-Quant assay (Carl Roth, Karlsruhe, Germany) according to Call et al. [13]. Bovine serum albumin (BSA) and a staining solution containing Coomassie blue dye as proposed by the manufacturer (Carl Roth) were used for external calibration and photometric detection at 595 nm with an In nite ® M200 PRO NanoQuant plate reader (Tecan, Männedorf, Switzerland).

Quanti cation of ATIs by RP-HPLC
The prepared ATI extracts were ltered through a Chroma l ® CA 0.45 µm membrane and injected on a Hitachi HPLC system from 5000 series equipped with a DAD detector. Chromatographic separation was performed on a HALO C18 column (Advanced Materials Technology, Inc., Wilmington, DE) with 1000 Å pore size (150×2.1 mm with 2.7 µm particle size) including a corresponding guard column. A gradient with water (solvent A) and HPLC-grade acetonitrile (ACN) (solvent B), both modi ed with 0.1% tri uoroacetic acid (TFA), was applied and is shown in Additional le 5: Table S4. Gradient elution for the quanti cation of ATIs was conducted at 45°C with a ow rate of 0.35 mL min -1 . Retention times of ATIs were assigned by using an AAI (alpha-amylase inhibitor) standard from T. aestivum L. purchased from Sigma-Aldrich (Type I, lyophilized powder). Since the purity of this AAI standard was not suitable for accurate quanti cation [14], BSA with a purity of at least 98% (Carl Roth) was chosen to establish an external calibration as it is also used for the quanti cation of total soluble nitrogen in the sample extracts.

Determination of trypsin inhibitory activity
Trypsin inhibitory activity (TIA) of the grain samples was determined by an enzymatic assay according to a recently published method by Call et al. [13]. It is based on the ability of salt-water soluble proteins, mainly ATIs, to inhibit the activity of trypsin towards a chromogenic substrate (L-BAPA). As only the range between 40% and 60% trypsin inhibition was found to be suitable for proper TIA calculation, sample extracts were varied in order to obtain inhibitions in that range. Samples that showed low inhibition but did not reach minimum inhibition of 40% were declared as lower LOQ. TIA values were expressed in mg inhibited trypsin per kg sample.
ATI characterization by MALDI-TOF MS MALDI-TOF MS was performed according to Call et al. [13] with some modi cations. Proteins were extracted from defatted wholemeal our with a chloroform-methanol solution (equal ratio). The obtained extracts were evaporated at 50°C under nitrogen stream to dryness and resuspended in an acetonitrilewater solution (equal ratio) containing 0.1% TFA. Puri cation and enrichment of proteins was performed with C4 ZipTips purchased from Merck (0.6 µL bed volume, 10µL total pipette volume) according to instructions of the manufacturer. Proteins were applied directly on MALDI steel sample plates and covered after drying with SA matrix solution (10 mg sinapic acid in 1 mL acetonitrile-water containing 0.1% TFA). 100% laser energy was used and at least ve spots were measured with 200 laser shots each in the range from 5 to 70 kDa.

Carbohydrate analysis
In order to determine changes in the carbohydrate metabolism throughout grain development, 50 to 100 mg of sample were extracted in 5 mL water for 10 min in an overhead shaker. Soluble proteins were precipitated by adding 10 µL of each Carrez I and Carrez II reagent. Insoluble components such as starch and precipitated proteins were removed by centrifugation (3000 g, 10 min). The extracts were ltered through a 0.2 µm lter prior to chromatographic analysis. To gain further information on the insoluble carbohydrate components, the insoluble residue after centrifugation was hydrolysed by 5 mL 2.5 M TFA in a shaking water bath for 2 h. As previously described, proteins were removed by Carrez precipitation and subsequent centrifugation. Samples extracts were adjusted to a pH >6 by adding NaOH, ltered and injected into the chromatographic system.
To quantify carbohydrate concentrations in the sample extracts, HPAEC-PAD was applied according to Fraberger et al. [29]. A Dionex TM ICS-6000 DC system (ThermoFisher Scienti c, Sunnyvale, CA) equipped with a CarboPac TM PA210 (2×150 mm, 4 µm) and a CarboPac TM PA20 Fast (2×100 mm, 4 µm) with corresponding guard columns was used for the separation of different carbohydrates. A gradient elution with 200 mM NaOH and 200 mM NaOH/500 mM Na-acetate was used at a ow rate of 0.2 mL/min and at a column temperature of 30°C. Details of the chromatographic conditions are shown in Additional le 6: Table S5. The calibration for the method was performed with galactose, glucose, fructose, sucrose, ra nose, maltose and smaller fructans from DP 3-5 (Megazyme, Bray, Ireland) in a range from 0.5 to 50 mg/L.  c-j Arnold grains sampled 7,11,14,18,25,33,39, and 46 days after anthesis, respectively.