To curate putative RING-type ubiquitin E3 ligases in Arabidopsis thaliana genome version ARA11, classification made by Kosarev et al. (2002) and Stone et al. (2005) were used. To this end, the whole Arabidopsis proteome was downloaded from ARAPORT (https://www.araport.org/downloads/), and screened with InterProScan for protein families and domain architecture. To confirm that the newly identified RING domain containing protein sequences indeed represented ubiquitin E3 ligase type RING domains, InterProScan 5 (v5.16-55.0) Gene3D, SUPERFAMILY, ProSiteProfiles, SMART, Pfam, and ProSitePatterns signatures were used. Most of InterProScan tools use Hidden Markov Models (HMMs) to detect conserved domains along protein sequences. HMMs have been developed for conserved protein domains and they define for the software, which and where critical residues should be located along the analyzed protein sequence. From the protein domain collection, the ubiquitin E3 ligase type RING domains were filtered according to the criteria provided by Kosarev et al. (2002) and Stone et al. (2005) for canonical RING domains. Once the RING domains were identified, they were aligned with Jalview using ProbCons algorithm with two rounds of pre-training. The metal ligand binding residues were manually inspected and corrected, and small misalignments were edited. Sequences that failed to meet the criteria of InterProScan search engines were not considered in this study.

To associate the curated collection of 509 RING type ubiquitin E3 ligases with flowering the Genevestigator gene expression database software was used (Hruz et al., 2008). The experiments AT-00087, AT-00088, AT-00089, and AT-00090 containing developmental expression data of AtGenExpress initiative microarrays were selected for the analysis (Schmid et al., 2005). In the selected experiments, hybridization probes were available for 393 RING E3 genes out of the 509. From these experiments, the linear expression data was extracted for the developmental stages of developed rosette, bolting, young flower, developed flower, and flower and silique. For flower organs, the gene expression profiles were extracted for categories of shoot apical meristem (SAM), sepal, petal, stamen, and carpels. In these categories, genes were ranked for their at least 2-fold differential expression against the developed rosette. Their relative expression levels were obtained by log₂(FC) = log₂(FL) − log₂(R), where FC is fold of change, FL is flower organ or development stage and R is rosette. The results for each category were sorted by their log₂(FC) and all genes with log₂(FC) > 1 were considered as up-regulated.

For the candidate approach, we used interaction networks from BioGRID (http://thebiogrid.org/) and cross-checked them with flowering pathway genes listed in the Flowering Interactive Database FLOR-ID (Bouché et al., 2016) to identify RING E3 ligases interacting with known flowering time regulators CONSTANS (CO), CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1), and TARGET OF EARLY ACTIVATION TAGGED (EAT) 2 (TOE2).

For functional characterization of the identified top most differentially expressed genes and for the selected candidates, Arabidopsis mutant lines were obtained from the NASC stock center representing CATMA, SAIL, SALK, and GABI-Kat collections (Alonso et al., 2003; Rosso et al., 2003; Schmid et al., 2005; Kleinboelting et al., 2012). Altogether 49 lines were genotyped by combination of segregation analysis and T-DNA PCR with primers listed in Supplemental Table 1. From these, 43 lines represented 30 unique gene accessions (Supplemental Table 1). As a wild type control, Columbia (Col-0) ecotype of Arabidopsis thaliana was used.

For genotyping, plants were grown in vitro on MS media supplemented with the corresponding selection. For phenotyping, seeds were sown directly on soil with 50% peat and 50% vermiculite. Trays were covered with plastic wrap and cold stratified at +4 °C for three nights, after which they were transferred to the growth chamber (FytoScope, PSI, Czech Rep.). Seven days after stratification (DAS) the seedlings were transferred to their own pots, placed on the analysis trays and sand was added on top of the peat to prevent growth of any green algae. From the full water saturation of the soil, the water content was let to decrease until 70% and was kept at this level through daily weighing and watering. Growth conditions in the Arabidopsis growth chambers were 16 h light/8 h darkness and 22 °C. Relative air humidity of the growth chambers was targeted at 60%. The light intensity was set and controlled at 130 μE (MS6610, Mastech, China).

Homozygous one locus mutant lines were confirmed by segregation analysis and T-DNA specific PCRs. The PCR primers, T-DNA position and line information were summarized in Supplemental Table 1. The transcript levels of the T-DNA targeted genes were verified by quantitative real-time PCR (qPCR) analysis. The sample material for qPCR was harvested from the tissue indicated by Arabidopsis eFP Browser for each gene expression pattern: if the gene expression pattern indicated at least moderate expression in flower parts during floral development, tips of inflorescences with developing and open flowers were pooled from three to five individual plants. If the expression was in the seeds, young to mature siliques were pooled. If no expression data via eFP was found, new leaves were pooled with developing and open flowers. Three to four biological replications were harvested for each RNA preparation. RNA was extracted using InviTrap® Spin Plant RNA Kit (STRATEC Molecular), complementary DNA was prepared with SuperScript® IV Reverse Transcriptase (Thermo Fisher Scientific), and the qPCRs were performed using Roche Lightcycler® 480 Instrument II (Roche Diagnostics) using LightCycler® 480 SYBR Green I Master (Roche Diagnostics) with primers listed in Supplemental Table 1. Primers were primarily designed to locus downstream of the T-DNA. The fold up values (mutant line against Col-0) were calculated using the 2^−ΔΔCT method according to Livak and Schmittgen (2001). Reference genes used in this study were the most stable Arabidopsis genes according to Czechowski et al. (2005): TIP41 LIKE (AT4G34270, forward: GTGAAAACTGTTGGAGAGAAGCAA, reverse: TCAACTGGATACCCTTT∧CGCA), AP2M (AT5G46630, forward: TTGAAAATTGGAGTACCGTACCAA, reverse: TCCCTCGTATACATCTGGCCA) and PTB1 (AT3G01150, forward: TTGAAGGAGTGGAATCTCACG, reverse: ATGTGCGGAAAGCAGATACC). Significance level of the qPCR were set at 0.1–0.5 FC for knock-down; <0.1 FC for knock-out; and >2 FC for up-regulated (Supplemental Table 1).

The plant phenotyping facility at the University of Helsinki Viikki campus (http://blogs.helsinki.fi/nappi-blog/) was used for the phenotypic characterization of the selected Arabidopsis mutant collection. The plants were imaged daily by overhead CCD camera for RGB images positioned in a PlantScreen™ analysis chamber with automated plant transportation between the imaging, weighing and watering stations. The RGB images were obtained for 20 plants at the time and stored in central database. The images were pre-processed online as described in Awlia et al. (2016) to allow collecting binary and RGB data for each plant. The obtained binary images were used for calculating growth parameters of area and perimeter. The obtained parameters of area, perimeter and the convex hull were then used for automatic online calculations of morphometric rosette parameters including: roundness1, roundness2, isotropy, eccentricity, compactness, Rotational Mass Symmetry (RMS), and Slenderness of Leaves (SOL) (PlantScreen™ analyzer, PSI, Czech R.). To characterize the general morphology of the mutant lines these nine morphological parameters were grouped into four categories based on their type: raw, circularity, symmetry and center distance, and compared over time (Figure 1). Raw parameters were represented by area and perimeter of the rosette and they were calculated by counting pixels of a rosette picture and the edge pixels respectively and transformed to millimeters (Figure 1A). The parameters of roundness1 and roundness2 and isotropy represented the circular parameters (Figure 1B). The parameter roundness describes rosette area in comparison to perfect circle with same perimeter and is affected by leaf slenderness, petiole length and leaf perimeter. For wild type plant, this parameter usually takes values between 0.1 and 0.5 while a perfect circle has value 1. Roundness value tends to decay overtime due to leaf development that at the same time increases the rosette perimeter. Roundness 2 uses rosette convex hull area and perimeter for its computation and for wild type plants this parameter appears to have values between 0.7 and 1.0 following an oscillating pattern with less steep peaks over time (Figure 1B). Isotropy uses the area of a drawn polygon on top of the rosette (Figure 1B). Thus, isotropy has a behavior similar to roundness 2 over time, but with less steep peaks and decreasing tendency similar to roundness. The eccentricity and RMS were symmetric parameters (Figure 1C). Eccentricity describes how elliptical the plant rosette is, where a value close to 1 corresponds to a rosette with highly sharp elliptical shape, while a value close to 0 describes a circular shape. Wild type rosette shows a high eccentricity peak that decays over time with a second smaller peak by the end of growth, thus the rosette shape fluctuates between a round and an elliptical shape. On the other hand, RMS describes the symmetry of the plant rosette by making a ratio between the non-overlapping rosette convex hull area and a perfect circle of the same area centered in the plant centroid and the overlapping area of both. RMS shows a similar pattern as eccentricity, but with higher absolute values and a sharper peak. Compactness and SOL were based on the center distance (Figure 1D). Compactness is the ratio between the rosette area and the rosette convex hull area. This parameter tells about petiole length and leaf blade width. The parameter SOL explains how sharp the leaf blades are, but it is also affected by the leaf number. SOL was derived from the ratio between squared rosette skeleton and rosette area. Thus, SOL can take values greater than 0 and below 50 in dimensionless units for wild type plants (Figure 1D).

Ten days old (10 DAS) Arabidopsis plants were subjected to growth and morphological characterization by top view imaging for the following 10 days. One phenotyping round was designed to accommodate a maximum of 960 Arabidopsis plants representing 36 genotypes at a time in six consecutive experimental rounds called F006–F011. The total number of lines analyzed throughout the six rounds was 43. The maximum of 36 genotypes were divided in three batches that were rotating between the growth area and the PlantScreen™ analysis chamber. Each batch consisted of three experimental units of four mutant genotypes randomized with Col-0, each represented by 20 individual plants. One experiment consisted thus of five trays of altogether 100 plants. Each experimental unit had their own Col-0 wild type in randomized block design to normalize for any local differences in the microenvironments of the PlantScreen™ or the growth area. Each line showing any phenotypic responses was analyzed in at least three independent experimental rounds. Lines that did not show differences as compared to the Col-0 wild type were excluded from the subsequent rounds thus resulting in reduced numbers of genotypes included.

After the image based growth and morphological measurements of the 20 mutant and 20 Col-0 plants in the PlantScreen™ system, the flowering time parameters leaf numbers at bolting (LAB) and days to bolting (DTB), were manually counted for each of the plant individuals. The number of rosette leaves were counted at the appearance of the flower bud (developmental stage 5.10, Boyes et al., 2001) and the DTB was recorded at the same time. The flowering time phenotypes were observed in two to three independent experimental rounds. Finally, flowers of the main inflorescences were photographed and further dissected for floral organ analysis under stereomicroscope (SteREO Discovery.V20, Zeiss). Microscopic pictures of the inflorescence tips, single flowers, sepals and petals were taken with the attached camera (AxioCam ICc3, Zeiss). The analyzed inflorescences and flowers originated from several independent experiments. Flower developmental stages were determined as in Smyth et al. (1990). To confirm pollen viability pollen grain staining according to the modified Alexander method was performed (Peterson et al., 2010). Anther images were captured using Leica DFC420 C camera attached to an optical microscope.

The significance of the differences between mutant lines and Col-0 was computed by contrasting two fitted models to the data points using several order polynomials (Mirman, 2014). First, a model was fitted to all data points and then a second model was fitted including the factor genotype (wild type and mutant). These two models were then compared using the Chi square test to determine if the second model explained more variance than the first one beyond the significance (α = 0.05). If the second model was statistically different from the first one, it implied that the compared genotypes were different. These statistical analyses were conducted in R software (https://www.r-project.org/). LAB and DTB analysis were performed by Analysis of Variance using GLM procedure and pairwise comparisons against Col-0 using option Dunnett in the MEANS statement using SAS/STAT© software version 9.4 (SAS Institute Inc., Cary, NC, USA).

To re-confirm the published RING-type ubiquitin E3 ligase proteins encoded in Arabidopsis genome the 27,667 Arabidopsis proteins from the latest genome annotation release (ARA11) were scanned for RING domains. Through filtering the signatures in Gene3D, SUPERFAMILY, ProSiteProfiles, SMART, Pfam, ProSitePatterns altogether 509 putative RING domain containing protein sequences were obtained (Supplemental Table 2). RING gene names and descriptions were obtained from Araport using Thalemine tool (Supplemental Table 2). Araport used curated but also automatic gene annotation, therefore many RING domain containing proteins were annotated as RING/U-box protein although they did not contain U-box domain. Similarly, some were annotated as RING/FYVE/PHD zinc finger superfamily proteins. The 509 RING sequences were compared to the previously described RING-type protein sequences (Kosarev et al., 2002; Stone et al., 2005). From the 509 identified RING domain proteins 457 matched with the 490 previously described, thus resulting in 31 non-matching sequences (Figure 2). These non-matching sequences were thoroughly analyzed and, 6 of them were found to be merged with other gene models, 10 had no RING domain, 3 were not found in the database, 3 corresponded to pseudogenes, 7 were split and a new locus identifier had been assigned for them, and 2 were transposable elements (Supplemental Table 3A). The 50 additionally identified RING domain proteins were shown to represent diverse RING domains such as, 1 of D type, 4 of C2 type, 20 of H2, 16 of HC, 2 of S/T, and 7 of V (CH) type, according to the Stone et al. (2005) classification (Supplemental Table 2).

To associate the RING domain proteins with flowering or flower development two approaches were followed: (1) identifying those with gene expression enhanced or enriched during flower development or in flower organs, and (2) by searching RING proteins interacting with known flower regulators. For the first approach Genevestigator (Hruz et al., 2008) tool was used to rank differentially expressed genes (DEG) of the identified RING genes over Arabidopsis developmental stages and in flower organs relative to their expression in developed rosette (Figures 3A,B). In the selected experiments in Genevestigator database, probes were available for 393 of the 509 RING E3 ligases analyzed. The cut off for DEGs was set at 2-fold to be included in the selection resulting in lists of genes of interest for each of the categories. This process was repeated to identify gene expression enrichment at each of the development stages of bolting, young flower, developed flower, and flower and siliques. For the developmental categories altogether 71 DEGs were identified (Figure 3C). In addition to the developmental stages, enrichment for shoot apical meristem, sepal, petal, stamen and pistil organs were retrieved and resulted in 109 DEGs (Figure 3D). Some of the RING genes were common between these two categories and in total 122 unique RING genes were up-regulated in the flower related processes. The gene identifying AGI codes of these 122 flower related candidates are provided in the Supplemental Table 3B.

For the second approach we identified 6 additional genes of interest through literature study and from interaction networks of CO, COP1, and TOE2 from BioGRID (http://thebiogrid.org/). Based on these interaction screens 5 RING E3 ligases were selected to the study, represented by the following mutant lines; N656705 (AT5G65683), N686069 (AT1G61620), N372291 (AT3G29270), N2037522 and N67002 (AT4G17680), and N742646 (AT2G44410). In addition, a mutant line for COP1, cop1-6, and RED AND FAR-RED INSENSITIVE 2 (RFI2) for which a role in mediating red and far-red light signaling and ubiquitination activity has been shown in vitro, were included (Stone et al., 2005; Chen and Ni, 2006a). This E3 ligase was selected as a candidate since its expression is regulated by circadian clock and rfi2-1 mutant flowers early (Chen and Ni, 2006b). Thus, one mutant allele for RFI2 (N878610) was included in the study. Mutants representing these genes were analyzed together with the flower up-regulated RINGs and were named flower related UPS candidates in the Supplemental Table 1.

For functional characterization of the 122 flower related UPS candidates and those selected based on literature, a mutant collection was obtained from the NASC stock center. The mutants represented lines from CATMA, SAIL, SALK, and GABI-Kat collections (Alonso et al., 2003; Rosso et al., 2003; Schmid et al., 2005; Kleinboelting et al., 2012). Altogether 43 lines were shown to contain T-DNA insertion in one locus, six were doubtful and were omitted from the analysis. To confirm that the T-DNA insertion had interrupted the gene of interest, their altered expression levels were confirmed by qPCR analysis with primers listed in Supplemental Table 1. For 43 accessions representing 30 unique loci from the 122 flower related UPS candidates and the selected candidates a differential gene expression pattern was analyzed. Altogether 19 lines were knock-outs, and 13 knock-down mutants, and for 10 lines up-regulation of the gene of interested was observed (Supplemental Table 1). For one line, no differential expression was confirmed and this was excluded from the phenotyping. For 14 lines alleles were available with similar or opposite gene expression patterns.

From the genotypically and qPCR confirmed T-DNA insertion mutant lines, 43 were subjected to phenotypic characterization by top view RGB imaging using the PlantScreen™ system. Image series of each analyzed line were collected daily allowing analyzing the growth and changes in morphology over time. For scoring those lines showing phenotypes, we fitted general additive models (GAM) to each parameter of each analyzed lines (data not shown). Most of the lines showed no differences to their corresponding Col-0 controls. However, three lines were consistently different across the experiments compared to Col-0 in both growth and rosette morphology: csu1-4 (cop1 suppressor 1-4, N686069), sinal7-2 (seven in absentia like 7-2, N833574) (Peralta et al., 2016) and rha1a-1 (ring-h2 finger a1a-1, N2045046) (Table 1, Supplemental Tables 1, 4). The csu1-4 mutant rosette was clearly smaller than Col-0 and showed a yellowish coloration (Figure 4). The mutant line rha1a-1 seemed to have smaller leaves than Col-0, however, at the end of growth it appeared to have more leaves that resulted in similar final rosette area as compared to Col-0. This line also had shorter petioles and leaf serration. The third line sinal7-2 rosette was clearly larger than Col-0 but did not show major differences in color, shape or number of leaves (Figure 4).

To further analyze these three lines, mixed non-linear models were fitted to their data using several order polynomials for parametric analysis of the models. This analysis confirmed the earlier observations of significant changes in growth and development for these lines over time (Table 1). Line csu1-4 showed slower growth, reduced rosette area and perimeter compared to Col-0 along the complete measured period (Figures 5A,D). For line rha1a-1 the rosette area was very similar to Col-0 being, however, slightly but significantly larger over time probably due to its higher number of leaves (Figure 5B, Table 1). Although the differences between rha1a-1 and Col-0 were small the statistical model was able to capture those. Conversely, sinal7-2 showed both area and perimeter larger than Col-0 indicating more vigorous growth (Figures 5C,F).

Morphological data for parameters of circularity that include roundness, roundness 2 and isotropy were also evaluated for these lines (Figure 1). Line csu1-4 showed increased roundness over the total period analyzed in comparison to Col-0 (Figures 6A–C, Table 1). However, csu1-4 roundness curve had similar pattern to Col-0 but shifted to the right (Figure 6A). Similar situation was observed for sinal7-2, where the roundness curve shape was almost identical to Col-0 but in this case was shifted to the left, showing lower roundness along the total time period (Figure 6C, Table 1). Roundness curve of rha1a-1 was neither shifted nor similar to Col-0 curve. This line showed a lower roundness than Col-0 at the beginning of the analysis, reaching a stabilization point around 16 DAS (Table 1). For Col-0 plants roundness continued decreasing until it become lower than rha1a-1 (Figure 6B).

Line csu1-4 showed a similar roundness 2 pattern as Col-0 that is shifted to the right by approximately 2 days (Figures 6D–F). Line rha1a-1, showed an oscillating pattern too, however, its roundness 2 values were constantly close to 0.9 with less steep peaks than Col-0, presenting the highest differences between days 12 and 16 (Figure 6E, Table 1). Similarly, to line csu1-4, line sinal7-2 presented an oscillating pattern very similar to Col-0, however, this time the curve had shifted to the left by approximately 1 day (Figure 6F).

Isotropy showed similar results as roundness and roundness 2, where line csu1-4 and sinal7-2 had similar oscillating pattern as Col-0, but csu1-4 curve is shifted to the right, while the curve for sinal7-2 is shifted to the left (Figures 6G–I). Line rha1a-1 showed a constant high isotropy value decreasing over time until reaching Col-0 pattern by day 23 (Figure 6H, Table 1).

The morphological parameters describing symmetry were eccentricity and rotational mass symmetry (RMS) (Figure 1). For eccentricity, line csu1-4 showed a similar pattern as Col-0 plants with a large and a small eccentricity peak, but shifted to the right (Figure 7A). Line rha1a-1 presented no shift in its curve, but it showed a rather flat peak around days 11 and 15, remaining lower than Col-0 until the end of the analysis (Figure 7B, Table 1). This result shows that rha1a-1 is less eccentric than Col-0 along the complete analysis. Line sinal7-2 showed also a similar pattern to Col-0 plants with two eccentric peaks, but slightly shifted to the left (Figure 7C). For RMS line csu1-4 showed similar pattern as Col-0 plants, but shifted again to the right about 1 day for the highest peak and remained higher than Col-0 in the last days of the analysis (Figure 7D). On the other hand, rha1a-1 presented no shift in its curve, but it showed a decrease in the peak around days 11 and 15, decaying faster and remaining lower than Col-0 plants (Figure 7E, Table 1). Like in eccentricity, sinal7-2 was almost indistinguishable from the Col-0 plants, except for a slight shift to the left captured by the model (Figure 7F).

The last two morphological parameters analyzed were compactness and slenderness of the leaves (SOL), which were based on the center distance (Figure 1). Here line csu1-4 showed a decay of compactness overtime in a similar way to Col-0 plants, but its curve was shifted to the right (Figure 8A). Lines csu1-4 and sinal7-2 presented quite normal compactness curves, while for rha1a-1 the pattern that was less compact than Col-0 plants at the beginning of the analyzed period (Figure 8B, Table 1). The compactness later rises above Col-0, showing higher compactness values. Like for the previously described parameters, sinal7-2 compactness curve showed slightly lower values than Col-0, except for the last 2 days where Col-0 plants reached sinal7-2 compactness (Figure 8C).

Line csu1-4 showed lower SOL values than Col-0, while rha1a-1 and sinal7-2 showed higher SOL values than Col-0 (Figures 8E,F, Table 1). The main differences in SOL could be observed during the exponential growing phase of the rosette and reaching a plateau at the end of the analyzed period where the differences to Col-0 plants become insignificant (Figures 8D–F).

Flowering time mutants identified in the screen represented both with reduced and increased leaf numbers at bolting (Table 2). Line csu1-4 (AT1G61620) was clearly early-flowering in both experimental replications. AT5G63970, a putative forkhead box protein, mutant line was early flowering in one of two experimental replications. SBP (S-ribonuclease binding protein) family protein (AT4G17680) was late flowering in both experimental replications. As already shown by others, cop1-6 mutant was early flowering in both LAB (7) and DTB (22). In most of the mutant lines, LAB did not differ from Col-0 in all experimental replications, but the trend was observed in both or all. LAB or DTB of sinal7-2 did not differ from Col-0 in either of the experimental replications.

Flower morphology of the analyzed mutants was observed under stereomicroscope. The mutant line sinal7-2 was found to produce flower buds of abnormal shape, characterized by presence of cavities in the bud tips (Figures 9A,B). These openings were present at one or both sides of the affected buds and were caused by tips of the lateral sepals bending inwards (Figures 9E,F). Also medial sepals frequently showed altered morphology: their tips covered the buds to a lesser extent than in Columbia, resulting in their “blunt” appearance. Whereas these phenotypes were present in all 18 analyzed inflorescences of sinal7-2 plants, regardless of the plant age—only two out of 13 analyzed wild type inflorescences showed similar sepal features, restricted to the first six flowers on the main stems. Scoring flowers stage late 12–15 (located between positions 1st and 20th on the main inflorescences) revealed that in 54% of the mutant flowers (43/80) at least one lateral sepal tip was bent inwards—as compared to 6/50, i.e., 12% in Col-0 (the analyzed flowers came from 13 to 9 individual plants, respectively).

Dissecting flower buds at the end of stage 12 revealed that the occurrence of ingrown lateral sepal tips was accompanied by petal wrinkling, as the sepal shape interfered with elongation of the petals (Figures 9G,H). Indeed, in some of the mature flowers with bent lateral sepal tips, the petal blades remained wrinkled; in several cases also pistil or stamen shape was affected (Figures 9C,D,I,J).

SINAL7 has been shown to mediate ubiquitination of glyceraldehyde-3-phosphate dehydrogenase 1 (GAPC1) enzyme in vitro and to affect its enzymatic activity and subcellular localization in Arabidopsis (Peralta et al., 2016). In plants lacking GAPC1 male sterility was observed (Rius et al., 2008). To investigate whether deficiency of SINAL7 impairs male fertility in the sinal7-2 mutant, pollen viability was inspected according to the modified Alexander method (Peterson et al., 2010). Anthers of 12 mutant and 11 Col-0 flowers in the developmental stages late 12 and 13 were stained (early and late flowers, originating from at least five individual plants per line). However, no difference between the mutant and Col-0 pollen was observed: anthers of both lines contained almost exclusively viable pollen grains (Figures 9K,L).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.