Phenotype-Chemotype Correlation of the Herb Bletilla Rchb. f. Based on a Comprehensive Evaluation of Thirty- Three Geographic Populations


 Background

The Bletilla genus of Orchidaceae includes plants with great economic value, among which B. striata is the main traditional medicinal plant, and its pseudobulb, known as BaiJi, was first recorded in Shennong's Classic of Materia Medica. However, there has been little systemic evaluation of the germplasm quality of Bletilla plants in China. In order to comprehensive evaluate the Bletilla resources in China and screen out the candidate phenotypic traits determining yield and/or quality of Bletilla, the variation of phenotypic indicators (pseudobulb, leaf, stem, inflorescence, flower) and active ingredients contents (polysaccharide, total phenolics and militarine) in different populations of B. striata and B. ochracea were investigated through 4 years' common-garden experiment.
Results

There were abundant phenotypic variations and significant differences among different populations in the morphological phenotypes, pseudobulb weight and main active ingredient contents. Five populations, including HNSZ, AHBZ, HBLT, HBSN and JSNJ, showed good prospects for industrial development, presenting higher quality in terms of yield and main active ingredient content. Pseudobulb yield, polysaccharide and total phenol content are positively correlated with phenotypic traits. Militarine content is negatively correlated with almost all indexes. Plant height, leaf width and stem diameter may be important indicators of potential excellent germplasms.
Conclusions

 Bletilla is not strictly geoauthentic medicinal plants. B. ochracea could be accepted as an alternative resource to B. striata. The best harvest period of Bletilla is the third year after cultivation. Plant height, leaf width and stem diameter may be important indicators of potential excellent germplasms. These results provide important information required for the efficient screening and utilization of Bletilla germplasm resources.

among different populations, especially the plant height, leaf blade length, stem diameter and in orescence height with high F values and extremely low P values ( Table 1). The plant height, leaf blade length, in orescence height and length of B. ochracea were greater than those of B. striata. In addition to leaf blade length in B. striata, the plant height, leaf blade length and width of B. striata and B. ochracea increased with increasing planting years, while the fruit number decreased in B. ochracea and B. striata (Fig. 2a). Min and Max, the minimum value and maximum value in the sample; Mean, the average value of the sample; CV (%), coe cient of variation; F, ratio of MSA/MSE; P, P value; n, sample size.
Comparison of morphological measurements (a), pseudobulb weight (b) and the main active ingredient content (c) in different growth stages and different species of B. striata and B. ochracea. Values are the mean ± SD (standard deviation), and different lowercase letters indicate signi cant differences with the parameter of P < 0.05.
There were signi cant differences in pseudobulb weight among different species and different populations, and the coe cient of variation (CV) of pseudobulb weight was higher than the CVs of other phenotypic traits (Fig. 2

Determination of active ingredients content
The polysaccharide, total phenol and militarine contents of pseudobulbs were determined among different species and populations. The results showed that signi cant differences were observed among different populations, with a minimum F value of 284.97 and P values much lower than 0.05 (Table 1). Overall, there were no differences in the content of the main active ingredients among different species and different planting years (Fig. 2c).
Most populations had an excellent phenotype in terms of polysaccharide content, which varied from 8.52 to 59.77 (%). 8, 12, 15 and 13 populations accounted for more than 50% of the one-, two-, three-and four-year-old pseudobulbs, and 17, 12, 13 and 9 populations ranged from 40% to 50%, respectively ( Table 3) Obvious variation was observed in the content of total phenol, which varied from 0.90 to 5.00 (%). There were 14, 11, 16 and 13 populations that were higher than 4% of the one-, two-, three-and four-year-old pseudobulbs, and 12, 15, 12 and 12 populations ranged from 3% to 4%, respectively ( Table 3). The average total phenol content of one-, two-, three-and four-year-old pseudobulbs of B. ochracea were 3.96, 3.57, 3.89 and 3.35 (%), which were lower than the 3.98, 4.14, 4.33 and 3.43 of B. striata, respectively. Therefore, the order of the contents of total phenol in the pseudobulb of B. striata was three-year-old > two-year-old > one-year-old > four-year-old, while that in B. ochracea was one-year-old > three-year-old > two-year-old > four-year-old (Fig. 2c). In general, the top ve populations with the highest average total phenol contents in pseudobulbs of four years were JSNJ, HBYL, HBXG, SCWY and HBLT, while the last ve populations were YCWZ, HNLS (B. ochracea), SXXX, YNBS and HBJM. Table 3 Active ingredient contents of 28 populations of B. striata and 5 populations of B. ochracea (listed at the end of the document) The average milliarine contents of the one-, two-, three-and four-year-old pseudobulbs of B. ochracea were 1.43, 1.62, 1.85 and 1.64 (%), which were generally higher than those of B. striata (1.44, 1.39, 1.45 and 1.60 (%), respectively). That is, the order of the militarine contents of the pseudobulbs of B. striata was four-year-old > three-year-old > one-year-old > two-year-old, while that in B. ochracea was three-year-old > four-year-old > two-year-old > one-year-old ( Fig. 2c). Militarine contents varied from 0.51 to 4.12 (%), 4, 4, 6 and 5 populations accounted for more than 2% of the one-, two-, three-and four-year-old pseudobulbs, and 7, 5, 9 and 12 populations accounted for 1.5% to 2%, respectively (Table 3) (Table 4). The populations of unmarked species are B. striata.
Correlation analysis of phenotypic traits, pseudobulb weight and active ingredient content To investigate which phenotypic traits may determine the yield of pseudobulb and active ingredient content, and whether there is a correlation between the yield of pseudobulb and active ingredient content, Pearson correlation analysis was performed for two-, three-and four-year-old plants among 13 indexes, including 9 morphological indexes of aboveground parts, pseudobulb weight indexes and 3 active ingredient indexes. The results showed that except for the negative correlation between the militarine content and other indexes, the other 12 indexes were positively correlated (Fig. 3).
The correlation coe cient of plants in 2018 ranged from -0.401 to 0.861. The correlation coe cients between plant height and leaf blade length, and between in orescence height and in orescence length were greater than 0.8 (0.806 and 0.861, respectively). There were 4 other correlation coe cients greater than 0.7, 3 correlation coe cients greater than 0.6, 5 correlation coe cients greater than 0.5, and 7 correlation coe cients greater than 0.4, and all of them showed extremely signi cant differences with P values of less than 0.01. Overall, there were signi cant correlations between phenotypic traits except leaf number, ower number and fruit number. There were signi cant correlations between pseudobulb weight and plant height, leaf blade width, stem diameter, and total phenol content. There were signi cant correlations between polysaccharide content and leaf blade width, stem diameter, and total phenol content. There were signi cant correlations between the total phenol content and leaf blade width, stem diameter, pseudobulb weight, and polysaccharide content. There was a signi cant negative correlation between militarine content and leaf blade length (Fig. 3a).
The correlation coe cient of plants in 2019 ranged from -0.282 to 0.920. The correlation coe cient of in orescence height and in orescence length was 0.920. The correlation coe cients of plant height with in orescence height and in orescence length were greater than 0.8. There were 6 other correlation coe cients greater than 0.7, 7 correlation coe cients greater than 0.6, and 7 correlation coe cients greater than 0.5, and all of them had extremely signi cant differences with a P value less than 0.01. Overall, there were signi cant correlations between phenotypic traits except leaf number, and fruit number. There were signi cant correlations between pseudobulb weight and all phenotypic traits except for fruit number; however, the three indexes with the strongest correlations were the same as those in 2018, which were plant height, leaf width and stem diameter. The correlation between polysaccharide content and phenotypic traits was lower than that in 2018, and the same trends were observed for total phenol and militerine. However, there was still a signi cant correlation between the polysaccharide and total phenol content (Fig. 3b).
The correlation coe cient of plants in 2020 ranged from -0.331 to 0.931, and the highest value was obtained for in orescence height and in orescence length. There were 2 other correlation coe cients greater than 0.8, 3 correlation coe cients greater than 0.7, 6 correlation coe cients greater than 0.6, and 6 correlation coe cients greater than 0.5, and all of them had extremely signi cant differences with a P value less than 0.01. Overall, there were signi cant correlations between phenotypic traits except leaf number and fruit number. There were signi cant correlations between pseudobulb weight and plant height, leaf blade width, leaf number, and stem diameter. There were signi cant correlations between polysaccharide content and leaf blade width, stem diameter, total phenol content. There were signi cant correlations between the total phenol content and leaf blade width, and polysaccharide content. There was also a nearly negative correlation between the militarine content and other indexes (Fig. 3c).

Discussion
In the cultivation of B. striata, regardless of the pseudobulb reproduction and direct seed sowing, most of the provenances are wild resources, which are usually applied in the original cycle of harvesting in situ, reserving seeds in situ and replanting in situ, without systematic population selection and breeding [3]. Through comprehensive investigation and collection of Bletilla genus resources in China to establish a germplasm resource nursery, which can be used for resource quality evaluation and correlation analysis based on the important agronomic traits and pharmacodynamic components, and which will To evaluate the quality differences, fuzzy matter-element analysis and gray pattern recognition analysis were used to analyze the contents of extract, polysaccharide, militarine and total phenol of 30 populations of B. striata from main producing areas in Guizhou Province [7]. In this study, 33 Bletilla populations (including 28 B. striata populations and 5 B. ochracea populations), which covered their major distribution areas in China, were used for resource quality evaluation. This is the largest and most comprehensive evaluation of B. striata and B. ochracea in China at present.
Stable yield and quality are the most important value indicators of medicinal materials, and good germplasm resources are the basis for ensuring the quality of medicinal materials. In our research, we continuously evaluated the main value species, including B. striata and B. ochracea, for 4 years. Based on our yearly results, we screened excellent populations with higher yields in terms of four-year-old pseudobulb weight (Table 2) and higher contents of polysaccharide, total phenol and militarine compounds, respectively ( Table 3), indicating that these populations were good germplasm resources for the single-component utilization of Bletilla. Furthermore, HNSZ, AHBZ, HBLT, HBSN and JSNJ populations showed good characteristics in terms of pseudobulb weight, the pseudobulb growth ratio and the main active ingredients contents according to GRA, indicating that these populations showed higher quality, with good values of all the indicators (Table 4). Bletilla is widely distributed in China, and it is not strictly geoauthentic medicinal. Therefore, different geographical populations from different provinces may have good quality and are excellent germplasm resources. Chen et al. also found that Bletilla from different provinces of China have similar HPLC ngerprints, and were classi ed into one group by cluster analysis [28]. In addition, even in the same region, there may be a large difference between different geographical populations. Western Hubei Province is rich in forest resources, which provide good ecological conditions for the breeding and reproduction of medicinal plants. We collected the HBXS, HBSN, XSSC and HBYL populations near the Shennongjia Forestry District and the HBLC, HBXE, HBHF, HBJS and HBWF (B. ochracea) populations near the Enshi Tujia and Miao Autonomous Prefecture. We found that there were differences in the weight of pseudobulbs and the contents of three active ingredients among different populations, whether near Shennongjia Forest District or Enshi Tujia and Miao Autonomous Prefecture, and there were abundant variations among populations. For example, the weights of the pseudobulbs of the HBXS and XSSC populations were signi cantly lower than those of HBSN, and the polysaccharide and total phenol contents in HBXS were higher than those in HBSN. The content of militerine was highest in HBLC, while that in HBXE and HBHF was very low, and the HBHF population had high contents of polysaccharide and total phenol. The determination of total polyphenol in B. striata from 16 areas of northwestern of Hubei Province showed that the content of total polyphenol varied in different areas. The analysis of the polysaccharide content of 18 samples of B. formosana from 8 populations in Baoshan City of Yunnan Province showed that there were signi cant differences among different samples. The polysaccharide contents of B. striata from different areas of Guizhou and Anhui Province were also signi cantly different [18]. All these studies showed that it is necessary to understand the resource distribution, habitat and situation of Bletilla and then to evaluate germplasm resources, which can provide a theoretical basis for the protection and utilization of Bletilla.
Traditionally, although other species of Bletilla are also used as local medicinal plants, B. striata is the only species o cially approved for use [3]. Considering the high medicinal value, high demand and gradually diminishing wild resources of B. striata, to protect traditional resources and develop new medicinal plant resources, it is necessary to conduct in-depth research on B. striata and its related species, to excavate alternative resources of B. striata and reduce the waste of effective ingredients and resources. However, the distribution of B. sinensis is narrow and limited to the south of Yunnan Province in China. The pseudobulb size of B. formosana is signi cantly smaller than the pseudobulb size of B. ochracea and B. striata, and the polysaccharide content of B. formosana was slightly lower than the polysaccharide content of B. ochracea and B. striata [29]. The pseudobulbs of B. ochracea are larger, and its owering period is later than the owering period of B. striata. Currently, Bletilla is widely planted as a bonsai plant, so in addition to ornamental plants, can underground pseudobulbs of B. ochracea also be harvested as medicinal plants? We chose 5 populations of B. ochracea and 28 populations of B. striata to demonstrate the phenotypic quality of B. striata with higher average pseudobulbs weights, while there was no difference between the two species in terms of polysaccharide, total phenol and militarine contents. These results suggested that B. ochracea could be accepted as an alternative resource to B. striata. E cient screening of excellent genetic resources is important for new germplasm creation and breeding, and e ciently discovering the phenotypic indicators signi cantly related to yield and quality is especially critical for perennial herbs [30]. Usually, B. striata exhibits the best comprehensive traits and is harvested in the third year after cultivation [31,32], which requires at least 3 years to estimate the genetic material and preliminarily screen out a quality germplasm. In this paper, Pearson correlation analysis of two-, three-and four-year-old plants showed that there were signi cant correlations between pseudobulb weight and other phenotypic traits, especially plant height, leaf blade width and stem diameter, which implies that plants with excellent growth can produce better pseudobulb yields (Fig. 3). Interestingly, there was a strong signi cant correlation between the polysaccharide and total phenol contents, and polysaccharide and total phenol showed a similar trend to pseudobulb yield, which was signi cantly related to leaf blade width and stem diameter. These results may provide important and more direct indicators for quickly screening and breeding potential good germplasms. The uncommon positive correlation between yield and the main quality factors may provide hopeful prospects for excellent germplasm breeding of Bletilla herbs; of course, this positive relation and the genetic foundation of this relation need further demonstration and in-depth investigation [26]. In addition, the militarine content was almost negatively correlated with other indexes, which indicated that the accumulation of militarine may occur more often under the condition of limited growth, and further research is needed to reveal the synthesis and metabolic mechanisms of militarine in Bletilla plants. Next, integrating phenomics, genomics and metabonomics approaches for high-throughput mining of the key phenotypes and revealing the genetic mechanism is our aim in recent years.
The growth capacity of yield characteristics can also be selected to evaluate different genetic resources. Populations of HBSN, HBWH, and AHBZ showed excellent yield phenotypes with high yield and growth capacity, and the pseudobulb weight of the HBSN population could even reach 7.29 and 10.13 times in the third and fourth years after cultivation, respectively. Other populations such as HNSZ, SXZP (B. ochracea), HBXS, and HBJM, also showed a good growth capacity. Our results indicated that the mean annual growth ratios of pseudobulbs in 3 years were 2.42, 1.69 and 1.31, which gradually decreased with the cultivation years. Even in some populations, such as GXHZ, HNCL and SCWY, the pseudobulb weight in the fourth or even third year began to decrease. Yearly content variation showed that the average contents of both polysaccharides and total phenol in the one-, two-, three-and four-year-old pseudobulbs of Bletilla varied in the following order: three years old > two years old > one year old > four years old, which con rmed the conclusion regarding the best harvest period of Bletilla [31,32]. However, the average militarine content of one-, two-, three-and four-year-old pseudobulbs of Bletilla varied in the following order: four-yearold > three-year-old > one-year-old > two-year-old, which again indicated differences in polysaccharide and total phenol, and the synthesis and metabolic mechanisms of militarine in Bletilla plants need to be revealed.

Conclusion
There were abundant phenotypic variations and signi cant differences among different populations in the morphological phenotypes, pseudobulb weight and main active ingredient contents. Bletilla is not strictly geoauthentic medicinal plants. Excellent germplasm resources may be distributed in distant geographical locations. There may be a large difference between different geographical populations even in the same region. B. ochracea could be accepted as an alternative resource to B. striata. Five populations, including HNSZ, AHBZ, HBLT, HBSN and JSNJ, showed good prospects for industrial development, presenting higher quality in terms of yield and main active ingredient content. Pseudobulb yield, polysaccharide and total phenol content are positively correlated with phenotypic traits. Militarine content is negatively correlated with almost all indexes. Plant height, leaf width and stem diameter may be important indicators of potential excellent germplasms. The best harvest period of Bletilla is the third year after cultivation. These results provide important information required for the e cient screening and utilization of Bletilla germplasm resources.

Experimental material
Thirty-three natural populations of Bletilla (including 28 B. striata populations and 5 B. ochracea populations) were randomly collected from its main natural range across ten provinces in China (Additional le 1; Fig. 1

Phenotypic measurements
The weight of the pseudobulb was measured and marked as one-year-old in winter 2017. The morphology of the aboveground parts and the weight of the pseudobulbs of 10 individuals were measured every year for the next three years (2018-2020).
The recorded morphological parameters of the aboveground parts included plant height (cm), leaf blade length (cm), leaf blade width (cm), total leaf number, stem diameter (mm), in orescence height (cm), in orescence length (cm), total ower number and total fruit number. Leaf phenotypes and stem diameter were measured in August or September. Plant height was the vertical distance between the top of the highest leaf and the ground. Leaf length and leaf width were measured using the largest leaf. The phenotypes of the in orescences were measured in April or May. In orescence height was the vertical distance from the base to the tip of the in orescence stem.
In early November 2018, the withered aboveground parts of 10 individuals were removed, and the pseudobulbs (two years old) were excavated and cleaned to measure the weight. The corresponding fraction of pseudobulbs was removed for the subsequent determination of polysaccharide, total phenol and militarine content. The same method was used to measure the weight of pseudobulbs in 2019 (three years old) and in 2020 (four years old) using the other 10 individuals.

Polysaccharide content analysis
After removing the brous roots, the sampled pseudobulbs were sliced, steamed in a water bath, and dried in the shade according to the Chinese Pharmacopeia. Dried pseudobulb slices were crushed for polysaccharide content, total phenol content and militarine content analyses. All the samples were collected in early November 2017, 2018, 2019 and 2020.
The phenol-sulphuric acid method was used to analyze polysaccharide content as described previously with minor modi cations [23]. Approximately 2 g of dried pseudobulb powder was treated with 50 mL H 2 O for 120 min at 100 o C twice using the conventional heating re ux extraction method. The 100 mL extract was concentrated to 10 mL by rotary evaporation, 25 mL of 95% ethanol solution was added to half of the extract solution (5 mL) for precipitation, and they had to rest overnight at room temperature. After centrifugation (3000 rpm/min) for 15 min, the precipitate was dissolved in 25 mL H 2 O to obtain the crude polysaccharide solution and quanti ed by the phenol-sulfuric acid method. 10 mg of standard glucose was dissolved in 100 mL H 2 O at a concentration of 100 µg/mL and diluted to six different concentrations of 0, 20, 40, 60, 80 and 100 µg/mL. 1 mL glucose solution, 1 mL 5% phenol and 6 mL concentrated sulfuric acid were added into a plug test tube in turn. After shaking and mixing, the samples were left at 25 o C for 40 min, and the absorbance was determined at 484 nm (A 484 ). The regression equation was A 484 = 0.0088 C (µg/mL) + 0.0122, r = 0.9998 (n = 6). Then, 1 mL polysaccharide solution was measured by the same method as glucose solution. The absorbance was compared with the regression equation, multiplied by the dilution multiple (50) and divided by the weight of dried pseudobulb powder. The results were expressed as polysaccharide content (%).

Total phenol content analysis
The sampling procedure was the same as above. Total phenol content was determined by Folin-Ciocalteu's phenol reagent [33]. Gallic acid was used as the standard to draw the regression equation, and 100 mg gallic acid was dissolved in 100 mL 60% ethanol at a concentration of 1000 µg/mL and diluted to ve different concentrations of 10, 20, 30, 40 and 50 µg/mL. Then, 1 mL of gallic acid solution and 5 mL 10% (v/v) Folin-Ciocalteu's phenol reagent were added into a test tube to react for 10 min, and 5 mL 2% (m/v) Na 2 CO 3 was added. After shaking, the sample was allowed to rest at 25 o C for 1 h, and the absorbance of each sample was determined at 760 nm (A 760 ). The regression equation was A 760 = 0.0132 C (µg/mL) -0.0004, r = 0.9999 (n = 6). Approximately 2 g dried pseudobulb powder was treated with 80 mL 60% ethanol for 120 min at 90 o C by using the conventional heating re ux extraction method. Cooling down to RT, the solution was made up to 80 ml with 60% ethanol and then ltered using lter paper. The residue was extracted with 60% ethanol again, and the ltrates were mixed as the test solution, which was measured by the same method as the standard sample, to obtain the A 760 , compared with the regression equation, multiplied by the dilution multiple (160) and divided by the weight of dried pseudobulb powder. The results were expressed as total phenol content (%).

Militarine content analysis
The sampling procedure was the same as above. HPLC was used to measure the militarine content according to the Chinese Pharmacopeia 2020 edition. 11.4 g millitarine was dissolved in 6 ml 52.9% ethanol to obtain a 1.9 mg/mL standard solution, which was diluted to six different concentrations of 1900, 950, 475, 237.5, 118.75 and 59.375 µg/mL. The standard solutions were analyzed by reversed-phase HPLC using a mobile phase of acetonitrile and water with 0.1% phosphoric acid (elution ratio: 22:78, v/v; wavelength: 223 nm; ow rate: 1 mL/min; injection volume: 10 µL) and the column oven temperature was set at 30 o C. The regression equation of the standard curve was peak area = 22287199 C (µg/mL) + 137858, r = 1 (n = 6). Approximately 0.2 g dried pseudobulb powder was extracted with 25 ml 52.9% ethanol at 30 o C for 30 min by an ultrasound system (power = 300 W, frequency = 37 Hz). The solution was brought up to 25 ml with 52.9% ethanol and then ltered through lter paper. The ltrate was analyzed by the same method as the standard sample, compared with the regression equation, multiplied by the dilution multiple (25000) and divided by the weight of the dried pseudobulb powder. Ultimately, the results were expressed as militarine content (%).

Gray relation analysis
The means of the values in Table 2 and Table 3 were used as the raw data for GRA. For pseudobulb weight, in addition to considering the weight of the pseudobulb measured directly in every year, pseudobulb growth ratios of one, two and three years were also used as indexes. A total of 19 indexes constituted the evaluation matrix, {X ik } (i = 1, 2, 3, … m; k = 1, 2, 3, … n, m = 33, n = 19) and calculated the average value (`X K ) of each index. The original data were normalized by the formula Y ik = X ik /`X K . The maximum and minimum values of each index were selected as the optimal {Y sk } and the least {Y tk } reference sequences respectively. The correlation coe cients relative to the optimal and the least reference sequences were calculated with the formulas (∆min + ρ∆max)/(Y sk -Y ik + ρ∆max) and (∆min + ρ∆max)/(Y ik -Y tk + ρ∆max), respectively, where ρ is the discrimination coe cient, and its value is generally approximately 0.5, ∆min = min | Y sk -Y ik |, ∆max = max | Y sk -Y ik |. The uniform average of the correlation coe cient in each index was used to calculate the correlation degree relative to the optimal and the least reference sequences, which is expressed as r i(s) and r i(t) , respectively. The relative correlation degree (r i ) of the evaluation matrix {X ik } relative to the optimal and least reference sequences was de ned as r i = r i(s) / (r i(s) + r i(t) ), which wes used to rank (Table 4).

Correlation analysis
The means of the values in Additional le 2, Table 2 and Table 3 were used to calculate the Pearson correlation coe cient with IBM SPSS Statistics 25 software, which was used to construct a heat map in Microsoft O ce Excel 2019. * or ** represent signi cant differences with the parameters of P < 0.05 and P < 0.01, respectively.

Data analysis
Two-tailed ANOVA was performed to test signi cant differences in measured variables among different populations and different growth years. ANOVA and Pearson correlation analysis were carried out with IBM SPSS Statistics 25. Table 3 Active ingredient contents of 28 populations of B. striata and 5 populations of B. ochracea