Soil temperature, soil water content, light intensity, air temperature, and air humidity were measured in a barley field (9-ha) in Lower Austria in the years 2021 and 2022. For the measurements in the year 2021, a TensioMark® sensor (ecoTech Umwelt-Messsystem GmbH, Bonn, Germany) was used to measure the soil moisture (pF-value), the soil temperature, and the photosynthetically active radiation (PAR). Three sensors run by one data logger were placed within the 9-ha field, with 60 m between measuring points. The sensors were installed at a soil depth of −30 cm. The data were recorded using the data logger “envilog Maxi” (ecoTech Umwelt-Messsystem GmbH, Bonn, Germany). The air temperature was measured with three EasyLog data loggers (Lascar Electronics, Wiltshire, United Kingdom), each positioned in a weather house mounted on a wooden pole, in 80 cm height. The environmental parameters measured in the year 2021 are published on our homepage www.celbics.com and are available in Supplementary Table S1. In the year 2022, we used three TekBox-TBSST04-3 (TR) temperature sensors (Umweltanalytische Produkte GmbH, Cottbus, Germany) measuring soil moisture in soil depths of −20, −30, and −50 cm and three PR2/4 SDI-12 Delta-T profile sensors (Umweltanalytische Produkte GmbH, Cottbus, Germany) accumulating the data in −10, −20, −30, and −40 cm of soil depth, respectively (Supplementary Figure S1A). The data were recorded and saved by one solar powered data logger (yDoc ML-317, Firmware version 4.3 build 8) (Umweltanalytische Produkte GmbH, Cottbus, Germany). Three quantum sensors (Apogee model SQ-421) were used to measure the PAR in the field. Measurements were performed every 15 min and saved to an SD card. After the end of the field trial, the data were exported into. csv format via the Software “ydocTerminal” version 3.13. Air temperature was measured with three EasyLog data loggers (Lascar electronics, Wiltshire, United Kingdom), each positioned in a weather house mounted on a wooden pole, in 80 cm height (Supplementary Figure S1A). The data were imported into “RStudio” version 2022.12.0 + 353 (RStudio Team, 2020) with “R core” version 4.2.2 (R Core Team, 2022). Rows with missing values were filtered out and soil moisture sensor 3, which delivered only very few data points (probably due to voltage drop in the cable) and one of the soil temperature sensors at −20 cm depth, which got damaged during the setup and sent incorrect data, were filtered out as well. Air temperature and humidity data were imported from the three sensors and merged. Data were subset for the first 16 days, the wanted sensor (and depth for soil moisture and temperature), and transformed into long format using the “melt” function from the R-package “reshape2” (Wickham, 2007). Plots were created using the R-package “ggplot2” (Wickham, 2016) with the color palette “Set2” from the R package “RColorBrewer” version 1.3-3 (Neuwirth, 2022). PAR and soil temperature data points aligned almost perfectly for the replicates and were therefore only drawn using the “geom_line” function. Data points of the soil moisture and air temperature showed greater variability; therefore, smoothened means of the data points were plotted using the function “geom_smooth” with the parameters: method = “loess” and a span of 0.1 for the soil moisture and 0.01 for air temperature (Supplementary Data 1).

The spring barley (Hordeum vulgare L.) wild-type variety Golden Promise (GP) and the facultative variety BCC93 (kindly provided by Kerstin Neumann, IPK Gatersleben) were grown in the plant growth chamber (PSI) at 14 °C/12 °C, and maize and wheat (Bobwhite) at 16 °C/14 °C, 12 h day/12 h night cycle with light intensities of 130–220 µmol m⁻² s⁻¹ and 70% humidity. To track early root growth, GP and BCC93 were grown as near as possible to nature, according to Dermendjiev et al. (2021), and the measured data of the field experiments, at 14 °C/12 °C, 12 h day/12 h night cycle with light intensities of 130–220 µmol m⁻² s⁻¹ and 70% humidity, in a plant growth chamber (Conviron Adaptis A1000). Tomato was grown in the glass house at 26 °C/19 °C, 12 h day/12 h night cycle with light intensity between 200–250 µmol m⁻² s⁻¹ and between 50% and 60% humidity.

Within this project, we used five different soils. (I) For the growth of tomato, maize, and wheat, we used sieved (3 mm mesh) Cocopeat supplied with H₂O. (II) To enable a high root highlighting for the imaging, we used sieved (3 mm mesh) Cocopeat that was mixed with activated charcoal 2:1, supplied with H₂O (henceforth “Cocopeat_black”). Peat substrate (Gramoflor) was supplied with H₂O. (III) We used naturally grown bio-organic field soil and (IV) naturally grown conventional field soil (V) (that were obtained from fields in Lower Austria). The bio-organic field soil and conventional field soil was analyzed by the Austrian Agency for Health and Food Safety GmbH, in short AGES (Vienna, Austria). Bio-organic field soil shows a higher pH value (pH = 6.3) and less mineral nitrogen (0.3 mg/100 g) compared to conventional field soil (pH = 5.5; mineral nitrogen, 0.5 mg/100 g) (Supplementary Data 4). For salt stress analysis, we mixed activated charcoal and Cocopeat with water containing 20 g NaCl/l H₂O (electric conductivity (EC) = 30 EC). Soils were adjusted to a pF-value 2–3 to enable a water content mirroring soil environmental parameters as near as possible to nature and were added to the rhizoboxes, respectively. The pF-value was constant for 16 days; thus, for experiments shown in this manuscript, plants were not additionally watered.

For proteomic analyses, root and shoot material of the barley trait GP was used. Barley grains were germinated in rhizoboxes filled with soil (Cocopeat with activated charcoal). Seven grains were sowed per rhizobox. For control conditions, rhizoboxes were inserted into a BIBLOX setup in a climate chamber; therefore, roots would develop in the dark. For root illumination conditions, stand-alone rhizoboxes were put directly in a climate chamber without covering the roots in the rhizobox (Supplementary Data 3). All rhizoboxes were installed at an angle of 60° allowing root growth along the glass front of the boxes. Conditions in climate chambers (Conviron Adaptis A1000, Controlled Environment limited) were set to 12 h day/12 h night cycles of 14 °C/12 °C, with a light intensity of 130–220 µmol m⁻² s⁻¹ and 70% humidity. No difference in the soil temperature was measured between the stand-alone rhizoboxes and the rhizoboxes covered by the BIBLOX.

Seedlings were harvested 8 and 16 days after sowing (DAS), respectively. Root (R) and shoot (S) material was separately harvested, cleaned from soil, and immediately frozen in liquid nitrogen (LN₂). Shoots and roots of plants with light grown roots (LGRs) were harvested under light conditions (S_LGR + LGR). For plants of dark grown roots (DGRs), the shoot was harvested in light, and the root was harvested in a completely dark room with dimmed red light to prevent root illumination (S_DGR + DGR). Combined root or shoot material of 14 plants for 8 DAS and seven plants for 16 DAS would count as one biological replicate, respectively. Three to four biological replicates were taken per category (8 DAS_S_DGR, 8 DAS_S_LGR, 8 DAS_DGR, 8 DAS_LGR, 16 DAS_S_DGR, 16 DAS_S_LGR, 16 DAS_DGR, and 16 DAS_LGR).

Material was homogenized to powder, using LN₂ and mortar and pestle. Proteins were extracted using a sucrose SDS buffer [100 mM Tris–HCl pH 8.0, 30% (w/v) sucrose, 0.5% (v/v) 2-mercaptoethanol, 10 mM EDTA, 2% (w/v) SDS, Protease Inhibitor (Roche, Cat. No. 05 892 791 001)] by adding 1 ml of buffer to 350 mg of sample. Samples were resuspended completely. A total of 750 µl ROTI®Phenol [Roth, Cat. No. 0038.3] was added to the samples for protein extraction. Samples were vortexed for 1 min and incubated for 5 min followed by centrifugation at 20,000×g for 5 min at room temperature (RT). After phase separation, the phenol phase was carefully transferred to a new tube. The phenol fractions were counter-extracted with 750 µl of sucrose SDS buffer, vortexed for 1 min and incubated for 5 min, and then centrifuged at 20,000×g for 5 min at RT. The phenol phase was carefully transferred to a new reaction tube. Proteins were precipitated by adding 2.5 volumes of ammonium acetate in methanol [0.1 M]. After 16 h incubation at −20 °C, proteins were pelleted by centrifugation at 4 °C for 5 min at 5,000×g. Supernatants were discarded, and the protein pellets were washed with ice-cold ammonium acetate in methanol [0.1 M] and 70% methanol, respectively, followed each by centrifugation at 4 °C for 2 min at 18,000×g. The supernatant was removed, and protein pellets were air dried for 60 min and subsequently resuspended in 50 µl urea buffer [8 M urea, 100 mM ammonium bicarbonate, 5 mM dithiothreitol (DTT), Protease Inhibitor] while incubated at 37 °C for 20 min for better solubility. Next, samples were centrifuged at RT at 20,000×g for 10 min.

Protein concentration was measured via Bradford assay using a Quick Start™ Bradford 1× Dye Reagent (Biorad, Cat. No. 5000205) prior to protein content normalization. Bovine serum albumin (BSA) dilution series (0–10 mg/ml) in the according buffer was used as standard to calculate sample protein concentration. A total of 2 µl of sample or standard was pipetted into 1.5-ml tubes (in triplicates). 1 ml of Bio-Rad Quick Start™ Bradford 1× Dye Reagent was added. Tubes were vortexed and incubated in the dark for 10 min. A total of 200 µl of the solution was transferred into a 96-well plate. The absorbance of standards and samples was measured at 595 nm wavelength using a Thermo Scientific Multiskan Spectrum. BSA standard curve and calculation of protein concentration were done using Microsoft Excel. Cystein residues were reduced by incubating 200 μg protein per sample for 45 min at 30 °C while shaking at 700 rpm. Cysteine residues were alkylated with 55 mM Iodoacetamide (IAA) while shaking at 700 rpm, in the dark, at RT, for 60 min. Increased DTT [10 mM] concentration and sample incubation at RT and shaking at 700 rpm for 15 min stopped the alkylation process.

Furthermore, the urea concentration was diluted to 2 M with 100 mM ammonium bicarbonate/10% acetonitrile (ACN). CaCl₂ was added to a final concentration of 2 mM. Trypsin digestion was performed at 37 °C rotating for 14–16 h using Poroszyme™ Immobilized Trypsin Cartridge (ThermoScientific Cat. No. 8-0087-40-0994) at a ratio of 1:20 v:w.

Peptides were desalted using C18 solid-phase extraction columns (Bond Elut SPEC C18, 96 round-well plate, 15 mg, 1 ml, Agilent Technologies, Santa Clara, USA) and a water-jet (vacuum) pump. Plates were activated with 2×400 µl methanol passing the columns by gravity for 2 min and then aspirated via the pump. Columns were equilibrated with 4×400 µl of ultrapure H₂O, passing the column by gravity for 2 min and then aspirated via the pump. Subsequently, samples were pipetted into column and peptides and salt bound to it while gravity flow for 5 min, followed by aspiration via the pump. Samples were desalted with 5×400 µl of ultrapure H₂O passing the column by gravity for 2 min and then aspirated via the pump, last aspiration to total dryness. Purified peptides were recovered with 2×200 µl of methanol, passing the column by gravity for 5 min and then total aspirated via the pump. Peptides were transferred into new tubes and dried in a SCANVAC CoolSafe Vacuum Concentrator for 5 h at RT. The peptides were resuspended in 0.1% formic acid (FA) in acetonitrile. The final peptide concentration was measured spectrophotometrically via a NanoDrop device (Thermo Scientific).

Liquid chromatography and tandem mass spectrometry analysis was performed on a nano-LC-system (Ultimate 3000 RSLC; Thermo Fisher Scientific) coupled to an Impact II high-resolution quadrupole time-of-flight (Bruker) using a Captive Spray nanoelectrospray ionization source (Bruker Daltonics). The nano-LC system was equipped with an Acclaim Pepmap nanotrap column (C18, 100 Å, 100 µm, 2 cm; Thermo Fisher Scientific) and an Acclaim Pepmap RSLC analytical column (C18, 100 Å, 75 µm × 50 cm; Thermo Fisher Scientific). The peptide mixture was fractionated by applying a linear gradient of 5%–30% solvent B [0.1% FA in acetonitrile] at a flowrate of 250 nl min⁻¹ over a period of 60 min, followed by a linear increase of 30%–45% solvent B within 15 min. The column temperature was set to 50 °C. MS1 spectra were acquired at 3 Hz with a mass range from m/z 200 to 2,000, with the Top-18 most intense peaks selected for MS/MS analysis using an intensity-dependent spectra acquisition time between 4 and 16 Hz. Dynamic exclusion duration was 0.5 min.

Proteomics MS raw files were processed using the MaxQuant software (version 2.0.30; Cox and Mann, 2008). Peak lists were compared against the barley reference proteome (Hordeum vulgare subsp. vulgare (domesticated barley), cv. Morex, UniProt, UP000011116, version March 2022) using the built‐in Andromeda search engine (Cox et al., 2011). Enzyme specificity was set to trypsin, allowing up to two missed cleavages. Cysteine carbamidomethylation was set as static modification, and N-terminal acetylation and methionine oxidation as variable modifications. During the search, sequences of 248 common contaminant proteins and decoy sequences were automatically added. A false discovery rate (FDR) of 1% was set at peptide and protein level. Proteins were quantified across samples using the label-free quantification (LFQ) algorithm (Cox et al., 2014), and the match-between-runs option was enabled.

Uncharacterized proteins were manually identified by using the UniProt BLAST application.

Data were analyzed and visualized using Microsoft Excel (version 2211 Build 16.0.15831.20098 for Microsoft 365 MSO) and RStudio (version 2022.02.2 for Windows).

Proteins of which LFQ values were not detected for any of the measured sample groups and proteins where three out of four or two out of three values of biological replicates of one sample group were missing were dismissed. For missing third (for three replicates) or forth (for four replicates) values, an average value was calculated from the other two of three values of the biological replicates (Supplementary Table S3).

The T.TEST function (heteroscedastic, with two-tailed distribution; Microsoft Excel) was used to find significant differences regarding LFQ values between mean values of different sample groups (Supplementary Table S3). Prior to principal component analysis (PCA), data were logarithmically normalized using log₁₀(x+1). PCAs, loading plots, and contribution plots of all data, and subgroups were calculated and visualized using RStudio (Supplementary Table S4, Supplementary Data 5). Proteins that significantly differed in their abundance when comparing plants (shoots and roots) of LGRs and DGRs were classified regarding subcellular localization and molecular function using Microsoft Excel. Subcellular localization categories were “Cytoplasm,” “Cytosol,” “Nucleus,” “Mitochondria,” “integral component of membrane,” “Ribosome,” “Chloroplast,” “Extracellular,” “Plasma membrane,” “Cytoskeleton,” “Endoplasmic reticulum,” “Golgi apparatus,” “Peroxisome,” “Vacuole,” “Cell wall,” “Apoplast,” and “Plasmodesmata.” Molecular function categories included among others “RNA binding,” “ATP binding,” “Metal ion binding,” “Oxidoreductase activity,” “Defense response activity,” “Cytoskeleton,” and “Actin filament binding” (Supplementary Table S3).

To enhance the contrast between the roots and substrate for further semi-automatic image analyses, we manually traced the roots in every 10th image from each experiment (derived from between three and five biological replicates) with an Apple Pencil on an iPad (Supplementary Figures S3A, B), since our group works paperless and has been using those tools as digital lab notebooks. From the traced images, a binary image of the root system was made using color thresholding. Binary images were skeletonized, and a network graph was constructed using the sknw package (Xiaolong, 2020). From the network graph, the longest root and total root system length were calculated using the network-x package (Hagberg et al., 2008). Primary root angle was calculated by fitting a line through the x and y coordinates of the primary root skeleton pixels. The convex hull area and bounding box width of the root system were calculated from the binary images using OpenCV (Bradski, 2000) (Supplementary Figure S3C).

The root growth angle (RGA) and the seminal root growth angle (SRGA) were measured with the angle tool provided in ImageJ (Schneider et al., 2012) by drawing lines from the grain to the maximum distance of the seminal root to the horizontal level of the grain and between the first two seminal roots (Supplementary Figure S3D). Data were analyzed and visualized using GraphPad Prism (version 9.0 for Mac, GraphPad Software, San Diego, CA, USA, http://www.graphpad.com/).

Recently, we have successfully set up conditions to follow the germination in the lab at parameters as near as possible to nature (Dermendjiev et al., 2021). We monitored the soil temperature and moisture, air temperature, and PAR in a field of an organic spring barley farmer in lower Austria between the period of sowing and harvesting barley within 2021 and 2022, respectively. The recorded data of 2021 (Supplementary Table S1), which are publicly available on our group homepage (www.celbics.com), show a pF-value in − 30 cm depth for the first 16 days between 2 and 2.5. Additionally, the soil temperature was between 6 °C and 13 °C in soil depth of −30 cm. In 2022, the soil temperature was between 7.3 °C and 18.2 °C in −20 cm soil depth resulting in a mean temperature of 12 °C (Supplementary Figures S1A, B). The soil water content was between 14% and 20% (v/v). Thus, the parameters measured in year 2022 were consistent with the environmental parameters measured in the year 2021. Subsequently, the temperature for barley germination in the lab condition was set to the temperature of 14 °C/12 °C, considering that barley grains are sown at -3 cm below the soil surface and that air temperature was measured between 0 °C and 25 °C within these first 16 days. The soil moisture was set for the germination to a pF-value between 2-3 during the first 16 days of germination.

Rhizoboxes were constructed from PVC for the purpose of using them as DRD and as stand-alone devices. To reduce the CO₂ footprint and increase flexibility, subsequently, rhizoboxes were 3D printed with an in-house 3D printer using polylactic acid (PLA), a thermoplastic polymer that is manufactured from renewable and biodegradable plant-based materials (Henton et al., 2005; Bhatia et al., 2007) (Figure 1). Additionally, PLA is able to withstand plant growth conditions. All rhizoboxes are reusable. This 3D-printed version enables high flexibility in terms of construction size and production timepoint. Additionally, using 3D printers is a first small step for more sustainable research in the lab.

Applying the previously measured natural environmental parameters enables us to perform experiments in controlled lab conditions with settings as near as possible to nature.