Image analysis and in vivo imaging as tools for investigation of productivity dynamics in anthocyanin-producing cell cultures of Daucus carota

Authors


Author for correspondence: Flavia Guzzo Tel: +39 45 8027923 Fax: +39 45 8027929 Email: flavia.guzzo@univr.it

Summary

  • • An anthocyanin-producing suspension culture of Daucus carota (L.) cv. Flakkese was used as model system to study secondary metabolite production in cell culture at the individual cell level.
  • • An approach was set up in which growth and production of anthocyanins were investigated using a combination of biochemical analysis, image (colour) analysis and in vivo imaging.
  • • This novel approach was used to segment the culture in different subpopulations and dissect the productive process in the cell culture grown under two different conditions, known to differ mainly for oxygen supply and mixing intensity (volume of 50 ml or 20 ml in 250 ml flasks). The 20 ml batch cultures gave a higher content and yield of anthocyanins, which depended on a complex balance between events that positively or negatively affected anthocyanin production.
  • • A model is proposed in which the different ability of cells to respond to environmental stimuli and stress depends on the different amount of anthocyanins accumulated within cells.

Introduction

Plant cell cultures are artificial cell communities whose existence depends on the retained ability of cells to divide and change their differentiation state over long periods. In vitro cultures have been used as experimental systems for investigation of a large variety of plant biology aspects, including cell physiology, biochemistry and molecular biology, and have also gained popularity over years for their possible use in production of various molecules. Some highly valuable secondary metabolites, whose chemical synthesis is either too costly or occurs at very low concentrations in intact plants, can be produced in plant cell cultures (for a review see Ramachandra Rao & Ravishankar, 2002). Moreover, the safety issues of molecular farming with genetically modified plants make this environmentally safe technology an attractive alternative.

A rational approach to the biotechnological use of plant cell cultures requires a thorough knowledge of these complex systems. However, this goal is far from being reached: plant cell cultures are still largely unknown systems, and their optimization is often based on empirical rather than on predictive approaches. For example, for improvement of secondary metabolite yields through nontransgene-based methods, treatment is applied and the results are recorded as average yield (average content per average biomass production). More detailed investigations, focused on the effect of specific treatments on metabolic pathway enzymes or specific RNA transcript levels, also rely on average determinations on a large number of cells.

However, accumulating evidence has shown that plant cell cultures are a mixture of subpopulations of cells, differing in morphology, gene expression, epitopes and morphogenetic capacity, and thus, in the ability to produce different chemicals (Toonen et al., 1994; McCabe et al., 1997; Schmidt et al., 1997; van Hengel et al., 1998; Guzzo et al., 2002). Each of these cell subpopulations are dynamic entities (Naill & Roberts, 2004).

This means that the design of more predictive approaches for biotechnological utilization of plant cell cultures should consider individual cell behaviour.

The analysis of individual cell behaviour in secondary metabolite-producing cultures with respect to their productive traits is still challenging, if not completely unapproachable. One noteworthy exception is anthocyanin-producing cell cultures: the intense colour of anthocyanins makes production and/or accumulation or depletion of these molecules relatively easy to be followed at an individual cell level, especially with novel computer-based techniques. Anthocyanin-producing cell cultures may thus represent a model system for investigation of the complex behaviour of a secondary metabolite producing culture.

Anthocyanin-accumulating cell cultures have been obtained from several plant species, and much effort has been expended on determining the optimal conditions for anthocyanin synthesis and/or accumulation. This is especially true for carrot and strawberry, which are probably the best-known in vitro anthocyanin-producing systems (for a review, see Zhang & Furusaki, 1999).

In most cases, anthocyanin levels have been determined after pigment extraction, which is a destructive procedure and allows only average determinations. New techniques based on digital image analysis can be used to study anthocyanin accumulation in individual cells. Images obtained by charge-coupled device (CCD) cameras connected to an optical microscope are digitized by means of appropriate image analysis systems, which allows the measurement of a number of parameters on individual cells, including size, shape and colour (Ibaraki & Kenji, 2001).

Only a few reports have used such approaches to investigate anthocyanin production at the level of individual cells. Miyanaga et al. (2000a) attempted to mathematically extrapolate the anthocyanin content of individual strawberry cells through a colour analysis system based on RGB (red, green, blue) values. Smith et al. (1995) used the more intuitive HSI (Hue, Saturation, Intensity) model to segment Ajuga cells into low, medium and highly pigmented objects on the basis of saturation values.

In this study we improved the segmentation of a cell culture into different subpopulations by means of a biparametric hue and saturation analysis of an anthocyanin-producing cell culture of Daucus carota. Time-course colour analysis and live imaging allowed a fine dissection of the different events occurring within each subpopulation, including cell division, expansion, and death, in addition to the accumulation or loss of anthocyanins.

This approach allowed for the determination of individual cell contributions to the overall differences in anthocyanin content and yield of two batch cultures, which varied mainly in oxygen supply and mixing intensity. These observations suggest a model in which the different ability of cells to respond to environmental stimuli and stressors depends on the amount of anthocyanin accumulated within cells. This approach provides a useful tool for a more predictive approach to the improvement of anthocyanin production in cultured cells.

Materials and Methods

Cell culture and time-course experiments

The anthocyanin producing cell lines Rosa1 and Rosa3 were produced by independent visual selection of rare pigmented cells from the established K1 cell line, obtained from D. carota L. cv. Flakkese plantlets. The cell line was maintained in solid B5 Gamborg's medium (Gamborg et al., 1968) supplemented with 2% sucrose, 0.5 mg l−1 2,4-dichlorophenoxy acetic acid (2,4-D), 0.7% agar. The cell lines were kept at 25°C, with light 16 h d−1. In the standard conditions, the suspension culture was grown in the same medium, without agar, refreshing the medium every 14 d, at a cell density of 4% (v : v, 50 ml in 250-ml flasks), and agitated on a rotary shaker at 90 r.p.m. To verify the effect of the increase of oxygen exchange on anthocyanin production, 250-ml flasks containing 20 ml of cell culture were also used. For time-course experiments, two series of 250-ml flasks containing 50 ml or 20 ml of cell culture were prepared. In order to avoid modification of the culture volume, which is known to affect oxygen exchange, the entire content of individual flasks, instead of aliquots from the same flask, was used for time-course image analysis and anthocyanin extraction.

Anthocyanin extraction and analysis

For qualitative and quantitative estimation of anthocyanins, cells were collected by centrifugation, resuspended 1 : 2.5 (w : v) in methanol–HCl (99 : 1, v : v) and extracted for 30 min on ice. These extraction conditions were set up evaluating, by high-pressure liquid chromatography (HPLC) analysis, the extracts obtained with different solvent volumes (from 1 to 4 ml g−1 cells); once the optimum extraction volume was determined, we tested also different extraction times (from 30 min to 2 h) and temperatures (from 0 to 37°C); in our system 30 min on ice are sufficient to give the maximum extraction efficiency (not shown).

Crude extracts were analysed by HPLC/diode array. The column used was a 150 × 4.6 mm C18RP (Alltech Associates, Inc., Deerfield, IL, USA). Solvents were: (A) 10% acetic acid, 5% acetonitrile and 5% phosphoric acid, and (B) 100% acetonitrile. Solvent gradient was 0–15% B in 15 min, followed by 5 min in 15% acetonitrile at a flow rate of 1 ml min−1 at 25°C. The injection volume was 20 µl. Anthocyanins were analysed by absorbance at 520 nm; this wavelength corresponds to the average of maximal absorption wavelengths of the different anthocyanins present in the extracts. For qualitative and quantitative evaluation of anthocyanins, the absorbance of each peak was compared with that of solutions containing known concentrations of commercial cyanidin-Cl as a standard (Polyphenol Laboratory, Sandnes, Norway), ranging from 0.005 to 0.1 mg ml−1. The aglycone analysis was performed after acid hydrolysis of the extracts as described by Narayan & Venkataraman, 2000). Total anthocyanins were expressed as equivalent of cyanidin; the content of anthocyanins was expressed as micrograms of cyanidin equivalents per gram of cells (fresh weight), while the yield was expressed as milligrams of cyanidin equivalents per litre of culture.

Image analysis

Cell images were taken with an Olympus IX70 microscope equipped with a JVC KY F58 3-CCD camera, in standard conditions of light and optics, and stored as colour 24RGB images with the image pro-plus software (Media Cybernetics, Silverspring, MD, USA). For colour analysis with the same software, a free form was traced inside the vacuole of each individual cell of the image, including c. 200–500 pixels. The average HSI values of these pixels were recorded and used for the analysis. In the HSI colour model, H (hue) ranges from 0 (red) to 360 degrees (red, passing through, yellow = 60°, green = 120°, cyan = 180°, blue = 240°, magenta = 300°); S (saturation) ranges from 0 (white) to 1 (saturated, pure colour); I (intensity) ranges from 0 (black) to 1 (absence of black, pure colour). The image pro-plus software transformed the three parameters in 255 channel values.

For cell size analysis, the bidimensional contour of individual cells was measured using the same software. Each analysis included c. 400 individual cells.

For the analysis of HSI data, data were transformed into fcs (flow cytometry standard) format and analysed by win-mdi software (available at http://facs.scripps.edu/software.html). All the biparametric combinations of hue, saturation and intensity were taken into consideration.

Time-course experiments were repeated three times.

In vivo imaging

In vivo imaging was performed using cells from 14-d-old Rosa3 cultures; 1 ml of liquid culture was spread into a thin layer over solid medium (containing 0.7% agar) on 90-mm diameter Petri dishes. In a typical experiment two Petri dishes were prepared and analysed successively, so that the first represented the earlier stage of culture (days 0–5) and the second represented the latter phase (days 5–10).

With the same devices described in the previous section (Image analysis), a series of time-course images of the same cells (30–60 per experiment) were collected. Images were collected every 15 min for 4–5 d.

At the end of the experiment the cells from the first image of the series (t = 0) were labelled with progressive numbers, and the fate of each individual cell was tracked during the entire experiment considering the following events: cell division, cell expansion, cell death, anthocyanin accumulation and loss of anthocyanins. The individual cells were estimated as ‘white’, lightly pigmented (low-pigmented) or highly pigmented (high-pigmented) by visual inspection.

For the induction of oxidative stress, 20-d-old K1 and Rosa3 cells were treated with 0.5 mm glucose and 0.1 U ml−1 glucose oxidase (glc/glc oxidase), as described by Zottini et al. (2002). Experiments were performed with cells in either liquid medium in flasks or in a thin layer of liquid medium over solid medium as described above.

For cells grown in liquid medium in flasks, after 24 h the viability was monitored by in vivo imaging, collecting a series of 100 time-course images every 10 s, and determining the presence of endocellular movements.

For cells grown in Petri dishes, the response to oxidative stress was followed in vivo for 24 h, collecting series of images every 15 min. Viability was assessed as described above.

Results

Selection and characterization of an anthocyanin producing cell line

The pink-pigmented cell lines Rosa1 and Rosa3 were produced by visual selection of rare clusters of pigmented cells that occasionally emerge on the surface of nonpigmented calli of the K1 carrot cell line. Pigmented cells were visually selected over a period of 24 months. These two cell lines have also proven to be rather stable in liquid medium, where they have been grown for more than 12 months without loss of pigmentation. Eight main anthocyanins were detected by HPLC analysis in both lines; the acid hydrolysis followed by HPLC analysis indicated cyanidin as the aglycone (not shown), as described for carrot cultures also by other groups (Glabgen et al., 1992; Narayan & Venkataraman, 2000).

Cells from the same batch of 14-d-old Rosa1 and Rosa3 suspension culture, grown under standard conditions, were inoculated either in the same conditions (50-ml batch in 250-ml flasks) or in a condition designed to increase gaseous exchange (20-ml batch in 250-ml flasks).

The cell weight as well as the anthocyanin content and yield were determined as a function of time, and data from one representative of three independent 16-d experiments with Rosa3 cell culture are reported in Fig. 1. The individual experiments showed completely comparable trends, although absolute values were different.

Figure 1.

Productivity parameters of the 50-ml (open squares) and 20-ml (closed squares) batch cultures of Daucus carota: (a) growth curves; (b) anthocyanin content (± SD of three high-pressure liquid chromatography (HPLC) determinations); (c) Anthocyanin yield (± SD of three HPLC determinations).

The culture cycle can be subdivided in three main phases for descriptive purposes: an initial phase 1 (days 0–2), a middle phase 2 (days 2–9) and a final phase 3 (days 9–16). In a normal culture cycle, cells are subcultured every 14 d.

Cell growth (Fig. 1a) in both conditions shows a similar pattern during phase 1 and 2. As the culture became older, the 20 ml batch culture reached the plateau quicker, although the final cell density was not different in the two cultures.

Both the anthocyanin content (Fig. 1b) and yield (Fig. 1c) showed significant differences between the two culture conditions, with the 20-ml batch culture reaching much higher values after day 4 until day 14. In both the 50- and 20-ml batches, the final days of culture (14–16) were characterized by a rapid decrease in anthocyanin content and yield.

The Rosa1 cell culture showed similar trends but different absolute values, since this culture is less productive than Rosa3 (not shown); this suggests that the observed responses to the improved gas exchange are not limited to a specific cell line.

Further analyses were performed only on Rosa 3 cell lines.

Analysis of biparametric hue-saturation plots

To further explore factors that contribute to the above-observed variation of the anthocyanin content, the production of anthocyanins was also followed at an individual cell level using time-course image analysis. In principle, the two main factors that contribute to culture pigmentation are the relative amounts of pigmented and nonpigmented cells and the degree of pigmentation of individual cells.

The method of analysis was set up by comparison of hue-saturation (HS) biparametric plots of the Rosa3 anthocyanin-producing cell culture with the nonpigmented K1 cell line. Three different cell subpopulations were distinguishable (Fig. 2). Region 1 (R1) was drawn around cells of the nonpigmented K1 culture (Fig. 2a). A second region, R2, was traced around low-pigmented Rosa3 cells, chosen by eye; this region included cells having the same range of H-values of R1, but higher saturation values (Fig. 2a,b). The third region, R3, drawn at the right border of R1, included cells of the pigmented culture having H-value higher than R1 and R2 (Fig. 2b).

Figure 2.

Hue, saturation and intensity (HSI) analysis of pigmented and nonpigmented Daucus carota cell cultures. (a) Biparametric HS plot of K1 nonpigmented cell line. (b) Biparametric HS plot of Rosa3 pigmented cell line: R1, nonpigmented cells; R2, low-pigmented cells; R3, high-pigmented cells. (c) Distribution of HSI values of the pigmented Rosa3 cell culture; this analysis included c. 400 cells.

The analysis of Rosa3 cells through a biparametric H-S plot showed that no sharp borders existed between the three cell subpopulations contained in the R1, R2 and R3 regions (Fig. 2b). However, since R1 was traced exactly around the HS values of nonpigmented K1 cells and R2 around low-pigmented cells of Rosa3, and cells of the anthocyanin-producing culture Rosa 3 fell into all the three regions, R1 should include the nonpigmented cells of the culture, R2 the low-pigmented cells and R3 the high-pigmented ones. Cells that visually appeared low or heavily pigmented fell into the R2 or R3 regions, respectively (not shown).

Subsequently, the percentage of cells, the distribution of H, S and I-values and average H, S and I values within each individual region were determined during time-course experiments. The distribution of H, S and I values of cells falling in the three regions is reported in Fig. 2(c) for day 0. The individual parameters H, S or I were unable to sharply discriminate between nonpigmented, low-pigmented and high-pigmented cells. However, S values clearly increased from R1 (nonpigmented cells) to R2 (low-pigmented cells) and R3 (high-pigmented cells). Therefore, the S value can be considered to provide an indication of the degree of cell pigmentation.

Image analysis of cell subpopulation dynamics

The results of the image analysis performed during the same time-course experiment described above are shown in Fig. 3 and in Table 1, which show the average S-values and the percentages of high-pigmented, nonpigmented and low-pigmented cells in the same cultures, respectively.

Figure 3.

Average saturation (± SD) of low-pigmented (a) and high-pigmented (b) Daucus carota cells, determined in a time-course experiment, in 50-ml (open squares) and 20-ml (closed squares) batch cultures. Image analysis included c. 400 cells.

Table 1.  Percentage of nonpigmented, low-pigmented and high-pigmented Daucus carota cells during a representative time-course experiment, in 50-ml and 20-ml batch cultures
Time of culture (d)Nonpigmented cells (%)Low-pigmented cells (%)High-pigmented cells (%)
50-ml Batch20-ml Batch50-ml Batch20-ml Batch50-ml Batch20-ml Batch
0272762621111
1522130651814
2254564451110
4174063352025
7212362531724
9232161521727
11274664331021
14292960441127
16463236551713

The S values of low-pigmented cells were very similar for both cultures and varied very little during the course of the experiment (Fig. 3a). However, the S values of high-pigmented cells and percentages of the different subpopulations varied between the two cultures and with time.

Phase 1: two main events occurred during phase 1 both in 50- and 20-ml batch cultures: (a) a transient accumulation of anthocyanins in high-pigmented cells (Fig. 3b), which closely resembled that of anthocyanin content determined from pigment extraction (Fig. 1b) and (b) a shift of low-pigmented to nonpigmented cells (Table 1), probably as a result of pigment loss.

Phase 2: image analysis of 50-ml batch culture cells revealed that the slow decrease in anthocyanin content during this phase (Fig. 1b) is mainly caused by the decrease in average saturation of high-pigmented cells (Fig. 3b). Nonetheless, in the 20-ml batch culture, the increase in anthocyanin content (Fig. 1b) was accompanied by a relative increase in high-pigmented cells (Table 1) and in their average saturation values (Fig. 3b).

It is worth noting that in the 20-ml batch culture the increase in anthocyanin content occurring during this phase (Fig. 1b) represents a balance between events with opposite effects: at early times (days 2–4), the average pigmentation of high-pigmented cells (Fig. 3b) and their relative amount increased sharply, and the relative amount of low-pigmented cells decreased (Table 1); afterwards (days 4–9), the average pigmentation of high-pigmented cells and their relative amount no longer increased, but the relative amount of low-pigmented cells increased sharply, probably owing to de novo accumulation of pigments in nonpigmented cells, which sharply decreased.

Phase 3: the average saturation of high-pigmented cells decreased in the 50-ml batch culture and remained at a high level in the 20-ml batch culture.

Pigmented and nonpigmented cells measured through image analysis showed very similar average sizes (not shown). Therefore, individual growth rates of the different subpopulations of cells could be extrapolated based on the curve in Fig. 1a and on the percentage of nonpigmented, low-pigmented and high-pigmented cells of Table 1. The extrapolated data are reported in Fig. 4.

Figure 4.

Growth curve of nonpigmented (a), low-pigmented (b) and high-pigmented (c) Daucus carota cells, in 50-ml (open squares) and 20-ml (closed squares) batch cultures. The curves were extrapolated from the growth curve of Fig. 1a and the percentage of the three subpopulations of cells reported in Table 1.

The three cell subpopulations showed variable, time-dependent growth rates in the two different experimental conditions, particularly during phase 3. With respect to 50-ml batch cultures, in the 20-ml batch cultures high-pigmented and nonpigmented cells showed a higher growth rate, while low-pigmented cells showed a lower growth rate (Fig. 4a,c). During the last 2 d of culture, there was a sharp decrease in the anthocyanin content and yield in both culture conditions. Image analysis revealed that this was caused, in the 50-ml batch culture, by a decrease in low-pigmented cells accompanied by an increase in nonpigmented ones. In the 20-ml batch culture, the decrease in high-pigmented cells was accompanied by an increase in low-pigmented cells (Fig. 4, Table 1).

This strongly suggests that the decrease in anthocyanin content and yield at the end of the experiment is caused by loss of pigmentation of the cells, which shifted between different subpopulations. In the 50-ml batch culture, anthocyanin loss occurred in low-pigmented cells and generated nonpigmented ones, while in the 20-ml batch culture the loss occurred in high-pigmented cells and generated low-pigmented ones.

In vivo imaging of anthocyanin-producing cultures

Since anthocyanin-producing cell cultures are heterogeneous mixtures containing different kind of cells, in vivo imaging was used to visualize the dynamic behaviour of the single cells. Some representative sequences are shown in Fig. 5. The events tracked during the observation period were cell division (Fig. 5g–l,r–u), cell expansion (Fig. 5f–h,v–z), new anthocyanin accumulation (Fig. 5m–q), and loss of viability and pigmentation (Fig. 5a–e,e–g,r–u,y–z). Cells were considered no longer viable when intracellular movements ceased.

Figure 5.

Sequences of events occurring in an anthocyanin-producing Daucus carota cell culture, as determined by in vivo imaging. (a–l) Arrows indicate two cells in which loss of anthocyanin accompanied by cell death occurs, the arrowhead indicates a dividing cell and the asterisk labels an expanding cell. (a) Time 0 (beginning of the sequence); (b) 19 h; (c) 19 h 15 min; (d) 23 h 45 min; (e) 33 h 30 min; (f) 34 h; (g) 69 h 15 min; (h) 69 h 45 min; (i) 75 h 30 min; (j) 75 h 45 min; (k) 76 h; (l) 80 h. (m–q) Two initially nonpigmented cells accumulate anthocyanin (arrow); (m) time 0 (beginning of the sequence); (n) 55 h 45 min; (o) 60 h 45 min; (p) 80 h 30 min; (q) 114 h 30 min. (r–u) Arrow indicates a pigmented cell in which a gradual anthocyanin degradation occurs, while the arrowhead indicates a dividing cell; (r) time 0 (beginning of the sequence); (s) 2 h 45min; (t) 68 h 30min; (u) 73 h. (v–z) Arrow indicates a single cell that became less pigmented during cell expansion, probably because of anthocyanin dilution, while the asterisk indicates another cell in which the loss of pigmentation is accompanied by cell death; (v) time 0 (beginning of the sequence); (y) 23 h 45 min; (w) 68 h 15min; (z) 92 h. Bar, 20 µm

Cell death was often followed by plasmolysis (which was probably due to a decrease in the turgor pressure) or to cell wall breakage (Fig. 5f,g). Rapid cell contraction was only rarely seen. The observed pigmentation loss could be caused by: loss of viability (Fig. 5a–e,e–g,v–z), in which case it occurred rapidly (less than 15 min, which is the time occurring between two subsequent images); cell expansion and thus gradual pigment dilution (Fig. 5v–z); gradual pigmentation loss without concomitant cell death or expansion (Fig. 5r–u). The last pattern was rare.

The fate of each individual cell in several in vivo sequences was determined for more than 600 cells. Percentages of cells undergoing the above mentioned events are reported in Table 2 (days 0–5) and Table 3 (days 5–10).

Table 2. In vivo imaging of a Rosa 3 Daucus carota cell culture
 Nonpigmented cellsLow-pigmented cellsHigh-pigmented cellsTotal
  1. The fate of 329 cells was followed during the first 5 d of culture. As only 30–60 cells could be observed in each film, seven different films were recorded and analysed. In the first row the total number of cells (viable and nonviable), as well as the percentage, in parentheses, are reported. The cells were visually recognized as non-, low- or high-pigmented and their viability at the beginning of the experiment was determined by the presence of intracellular movements. The percentages of cell division, cell expansion and anthocyanin loss/accumulation refer to initially viable cells. The pattern of anthocyanin loss is reported as percentage of the total anthocyanin loss events. Viability at the end of the experiment refers to the percentage of initially viable cells.

Beginning of the experiment
Total cells (%)99 (30.1) 86 (26.1)144 (43.8)329 (100)
Viable cells percentage65.6100.0100.0 89.6
Events occurring during the experiment
Cell division percentage18.5 22.1  7.6 14.2
Cell expansion percentage41.5 41.9 38.2 40.0
Anthocyanin accumulation percentage 7.7 19.8  4.9  9.8
Anthocyanin loss (%)  30.2 28.5 22.7
 Due to cell death (%) 38.5 17.1 24.4
 Due to cell expansion (%) 61.5 68.3 65.7
 Due to pigment degradation (%)  0.0 14.6  9.0
Viability at the end of the experiment (%)90.8 88.4 95.1 92.2
Table 3. In vivo imaging of Rosa 3 Daucus carota cell culture: the cell fate of 329 cells was followed during the second part of culture (days 5–10)
 Nonpigmented cellsLow-pigmented cellsHigh-pigmented cellsTotal
  1. The fate of 329 cells was followed during the second part (days 5–10) of culture. As only 30–60 cells could be observed in each film, seven different films were recorded and analysed. In the first row the total number of cells (viable and nonviable), as well as the percentage, in parentheses, are reported. The cells were visually recognized as non-, low- or high-pigmented and their viability at the beginning of the experiment was determined by the presence of intracellular movements. The percentages cell division, cell expansion and anthocyanin loss/accumulation refer to initially viable cells. The pattern of anthocyanin loss is reported as percentage of the total anthocyanin loss events. Viability at the end of the experiment refers to the percentage of initially viable cells.

Beginning of the experiment
Total cells (%)77 (23.4)66 (20.1)186 (56.5)329 (100)
Viable cells percentage27.395.5100.0 82.1
Events occurring during the experiment
Cell division percentage 0.0 0.0  3.2  2.2
Cell expansion percentage 4.825.4  32.3  2.2
Anthocyanin accumulation percentage 0.0 4.8  5.4  4.8
Anthocyanin loss 14.3 17.7 15.5
 Due to cell death (%)88.9 93.9 92.9
 Due to cell expansion (%)11.1  6.1  7.1
 Due to pigment degradation (%) 0.0  0.0  0.0
Viability at the end of the experiment (%)71.487.3 83.3 83.3

During the first 5 d of culture (Table 2) growth depended both on cell division and cell expansion, while in the subsequent 5 d (Table 3) cell division nearly stopped. Moreover, during the final 5 d cells showed a lower viability and anthocyanin loss was mainly the result of cell death. In individual subpopulations, nonpigmented cells showed a lower viability compared with pigmented cells: they halted division, while some pigmented cells continued to divide. The number of cells undergoing cell expansion dramatically decreased, while pigmented cells were still expanding; pigments did not accumulate in these cells (Table 3).

Effect of the oxidative stress on cell viability

To determine whether different types of cells have different abilities to respond to oxidative stress, cells of the Rosa3 cell line and of the nonpigmented K1 line were treated with the glc/glc oxidase H2O2-generating system. Viability was determined in treated and untreated cells after 24 h of culture in liquid medium by observing, by short (15 min) in vivo imaging analysis, the presence of intracellular movements.

The results are reported in Fig. 6. The percentage of dead cells after 24 h was much higher in the treated K1 cell culture compared with the untreated one; in the treated Rosa3 line, the percentage of dead cells was very similar to that of the untreated control (Fig. 6a).

Figure 6.

Effect of the oxidative stress, induced by the glc/glc oxidase H2O2-generating system, on K1 and Rosa3 Daucus carota cells. (a) Percentage of dead cells, determined after 24 h of culture in treated (closed bars) and untreated (control, open bars) K1 and Rosa3 cells. About 400–900 cells were included in each analysis. (b) Percentage of cells dying during 24 h of culture in treated K1 and Rosa3 pigmented (closed bars) and nonpigmented (open bars) cells. About 250 cells were included in each analysis.

Since cell death is always accompanied by pigment loss, as observed in the previous in vivo imaging experiments, the above approach did not allow to understand whether different sensitivity to H2O2 exists between pigmented and nonpigmented cells within the Rosa3 cell culture. To this purpose, K1 and Rosa3 cells were directly tracked for 24 h by the in vivo imaging technique. The results of these experiments, reported in Fig. 6b, showed that, under the induced oxidative stress, cell death occurred at a much higher frequency in the K1 cells than in Rosa3 cells (as shown also in the previous experiments), and in the nonpigmented cells of the Rosa3 line than in the pigmented ones.

Discussion

Image analysis

Only a small number of papers have investigated anthocyanin accumulation in individual cultured plant cells through image analysis, although interesting attempts to investigate these complex systems have been made by counting pigmented and nonpigmented protoplasts derived from cell suspension using an optical microscope. With this latter technique, it has been shown that high pH in the growth medium increases the anthocyanin content of strawberry cultures (biochemically determined) without changing the percentage of pigmented cells, suggesting that anthocyanin accumulates inside pigmented cells (Zhang & Furusaki, 1997). However, this approach relies on the subjective perception of the operator. An objective description of cell pigmentation can be performed using digital image analysis with specific colour models.

Two main colour models have been used, namely RGB (Miyanaga et al., 2000a,b) and HSI (Smith et al., 1995). The RGB model describes colours as three selected clusters of light (red, green and blue). Miyanaga et al. (2000a) described the colour characteristics of anthocyanins as having red component values that were much larger than the green or blue ones. It was thus proposed that the difference between the red and green component be used as an index for anthocyanin accumulation. Using this method, the above authors demonstrated a correlation between the pigmented cell ratio and the average anthocyanin content in strawberry cultures (Miyanaga et al., 2000a). Moreover, it was found that the anthocyanin content of individual cells underwent sharp variations during time (Miyanaga et al., 2000b).

The HSI model is more intuitive than RGB, because it describes colours as they are perceived by the human eye. Hue is the pure colour, described by a number from 0 (red) to 360 degrees (red), passing from yellow, green, cyan, blue and magenta. Saturation is the measure of the degree to which pure colour is diluted by white light, higher amounts of white corresponding to lower saturation, and gives an estimation of the purity of the colour. The intensity value estimates the darkness of the colour, with higher amount of black corresponding to lower intensity. The values of S and I are usually expressed in 0–1 range. In our system the three HSI values are all converted into 255 channel values.

In transmitted light microscopic images of individual pigmented cells, S and I values are not completely independent, in accordance with the observations of Smith et al. (1995). This is due to the fact that the colour of the cell depends on the amount of pigment and on the interaction between the pigmented cell and the incident light of the microscope. Highly pigmented cells, for example, absorb and reflect more light, and this results in less light arriving at the eye or the camera, and thus in more saturated (less amount of white) and less intense (greater amount of black) images.

Smith et al. (1995) used the HSI colour model to analyse Ajuga anthocyanin-producing cell cultures. Hue values were used to segment the total image pixels into pigmented (H > 240) and nonpigmented pixels, while saturation values were used to further segment pigmented pixels in lowly (0 < S > 0.1), medium (0.1 < S > 0.2) and highly pigmented (0.2 < S > 1) ones.

We used the HSI model and biparametric HS plots of pigmented and nonpigmented cells in order to evaluate the accumulation of anthocyanins in carrot cell cultures. Our observations suggested that the hue value alone is not sufficient to discriminate between white and pigmented cells, since a certain number of cells of the pigmented cultures showed the same H-value range of cells from a nonpigmented culture, but with higher saturation values. The analysis of selected cells perceived as low-pigmented by visual inspection confirmed that these cells have the same hue values of white cells, but higher saturation values. Segmentation based on HS two-dimensional plots instead of an one-dimensional H-value allowed us to identify two different populations of pigmented cells: one that has lower and more scattered hue values (included in R2 region), and the other with higher hue values and higher and more scattered saturation values (included in R3 region). According to the analysis of visually selected pigmented cells, the latter cell population includes more strongly pigmented cells.

Since saturation distribution within the individual regions roughly correlated with the pigmentation level perceived visually (Fig. 2c), the average saturation value of each region was used to evaluate the accumulation/loss of anthocyanins within the population of low-pigmented and high-pigmented cells (Fig. 3b).

Dissection of the anthocyanin producing process at individual cell level

The analysis described earlier was applied, in combination with in vivo imaging, to unravel the individual cell contribution to the anthocyanin content and yield of the culture during the growth cycle and in different culture conditions. The early peak in anthocyanin content at the beginning of the culture cycle correlated with a peak in the frequency and average saturation of high-pigmented cells. Differential pigmentation of high-pigmented cells, differential growth of the three cell subpopulations, with differences in cell division and cell expansion frequency, and shifts of cells from one population to another were revealed during the culture cycle. In particular, two aspects were apparent. First, the absence of sharp borders between nonpigmented and low-pigmented cell subpopulations and the specularity of the variations of their relative amounts suggest that cells can easily shift between regions 1 and 2 through loss or gain of pigmentation. Second, these shifts were directly observed by in vivo imaging, which also revealed that the loss of pigmentation may be accompanied by cell death. Different factors, such as pH and glycosidic substitution, can affect colour properties and stability of anthocyanins (for a recent review see Wrolstad, 2004); therefore, the shift between low and nonpigmented cells could depend on synthesis, degradation and modification of anthocyanins and/or on changes in the chemical properties of the vacuolar environment. The latter possibility could explain the rapid loss of pigmentation observed during cell death. The loss of tonoplast integrity during cell death could enhance the pH of the vacuolar sap, and this could in turn generate the colourless carbinol pseudo-base from the coloured flavylium ion.

Effect of the reduced batch volume on cell growth and anthocyanin content

The time-dependent effects on growth and anthocyanin content and yield were determined in anthocyanin-producing cells originated from the same batch and cultured in two different experimental conditions, namely 50-ml batch culture in 250-ml flasks, which is also the standard growing condition for this cell line, and 20-ml batch culture in 250-ml flasks. Cells in the latter condition were characterized by a much higher anthocyanin content and yield during a normal culture cycle of 14 d.

The reduction of the batch volume is known to increase gaseous exchange and therefore the oxygen supply, owing to the higher volumetric oxygen transfer, which depends on the surface/volume ratio and on the mixing intensity (Gao & Lee, 1992). This latter parameter, in turn, is higher for smaller volumes because of the increase of the empty space in the flasks for liquid movement during agitation. Therefore, cells in the 20-ml batch culture should have a higher oxygen supply and a higher mechanical stress compared with 50 ml batch cultures.

An increase in oxygen concentration has been reported to increase the growth rate in many plant cell cultures, including tobacco (Gao & Lee, 1992), Catharanthus roseus (Tate & Payne, 1991) and carrot (Jay et al., 1992).

The effect of fluid mixing on cell growth is, however, more complex, since poor mixing causes oxygen limitation with consequent limitations on growth (Zhong et al., 2002), while mixing that is too vigorous causes hydrodynamic stress that can result in cell damage and a decrease in biomass production (Ho et al., 1995). In general, mechanical stress in cell cultures can generate oxidative stress, as described, for example, in Taxus cuspidata (Han & Yuan, 2004) and soybean cell cultures (Legendre et al., 1993) even in the absence of apparent mechanical cell damage.

In our system, the reduction of the batch culture volume led to an increased growth rate during phase 3 in 20-ml batch cultures, which reached a growth plateau at earlier times. According to literature data, this could be an effect of the improved oxygen supply.

However, the analysis of individual cells revealed a much more complex situation: growth of the high-pigmented cell subpopulation was higher in the 20 ml batch culture up to day 14, which had positive effects on the anthocyanin yield. By contrast, the growth rate of nonpigmented cells during phase 3 increased. These events are expected to affect the anthocyanin yield in an opposite manner.

The analysis of the relative abundance (%) of the three types of cells during the same phase also suggested that at least part of the new, nonpigmented cells are derived from the low-pigmented subpopulation. In vivo imaging showed that in the late phases of culture (final 5 d) the shift of cells from pigmented to nonpigmented subpopulations is mainly (more than 90%) the result of cell death. The experimental conditions during in vivo imaging were quite different with respect to the time-course experiments in which cells were growing in liquid batch in shaken flasks. For in vivo imaging, cells were plated on a thin layer of liquid medium over solid medium within a large dish. Because of these experimental conditions, colour analysis could not be performed, and cells were visually classified as white, low-pigmented and high-pigmented. Therefore, in vivo imaging allowed only a rough estimation of the possibility and frequency of these events. Nevertheless, the above observations suggest that the reduction of the batch volume could cause cell death of low-pigmented and nonpigmented cells, while specifically improving the growth of high-pigmented cells.

The increase in the oxygen concentration, which in our system was achieved through the volume variation of the batch cultures, has been reported to increase the production of secondary metabolites such as ajmalicine from Catharanthus roseus (Schlatmann et al., 1994), berberine from Thalictrum minus (Kobayashi et al., 1989) and anthocyanins from Perilla frutescens (Zhong et al., 1993).

In our system, the effect of the culture volume on anthocyanin content appeared complex. At the time of subculture (14 d), the anthocyanin content of the 50 ml batch was still increasing. The very early anthocyanin content peak in both 20-ml and 50-ml batch cultures could be seen as a sort of continuation of the anthocyanin content peak of this late phase, which is characterized by an abundant biomass still growing and accumulating anthocyanins. Alternatively, the initial anthocyanin peak could be a specific effect of the new environmental conditions following the reinitiation of the culture cycle, characterized by fewer cells in fresh medium. During phase 1, other parameters, such as the average S-values of the different cell subpopulations and their relative abundance, also showed that same trends were present at day 14 of a 50-ml batch culture, indicating the former hypothesis is more plausible. The fact that conditioned medium and abundant biomass, rather than fresh medium, is generally described as promoting secondary metabolites production (Lee & Shuler, 2000; Zhang et al., 2002) supports the same conclusion.

Therefore, the anthocyanin content during phase 1 seemed to be strongly affected by the productivity and the physiological state of the inoculated cells.

The enhancement in both the content and yield of anthocyanin during phases 2 and 3 in the 20 ml batch culture compared with 50 ml batch culture could be ascribed to two synergistic events. These include the enhanced high-pigmented cell growth and the enhanced accumulation of pigments into high-pigmented cells. A third important event, the enhancement of nonpigmented cell growth during phase 3, negatively affects the anthocyanin content and yield.

Our observations might be explained by considering that a decreased culture volume leads to an increased oxygen supply, with a subsequent positive effect on cell growth. However, this condition may cause some type of cell stress, for example mechanical stress, which may result in oxidative stress. Considering the above, anthocyanins have the ability to protect against oxidative damage, and high-pigmented cells, rich in anthocyanins, would thus be able to withstand such oxidative damage. Moreover, oxidative stress would induce further anthocyanin accumulation in cells that have the anthocyanin biosynthetic pathway activated. As these cells would be able to overcome the oxidative damage, they could take advantage of the extra oxygen and increase their rate of growth. Conversely, nonpigmented cells would be less able to prevent the oxidative damage and this would result in more frequent cell death, as was observed during the in vivo imaging experiments. The latter situation could in turn affect the growth of the high-pigmented cells, which would have fewer competitors for nutrients in the culture medium.

For this to be true, the culture conditions in the 20-ml batch culture must be stressful and anthocyanins must also have an antioxidant effect in vivo.

To address the first point, it should be considered that the in vitro culture condition itself is considered by some authors to impose oxidative stress (for a review see Hagege, 1995; Cassells & Curry, 2001). High salt, water stress, mineral deficiency, overexposure to auxin and mechanical stress are situations that can cause oxidative burst in more sensitive cell lines (Cassells & Curry, 2001; Han & Yuan, 2004). In D. carota cell lines, oxidative stress was observed to occur during routine subculture of callus, as suggested by the presence of thiobarbituric acid-reactive substances (Robertson et al., 1995); in the same species, various lipid peroxidation products have been identified in dedifferentiated cell cultures, also indicating the occurrence of an oxidative stress (Bremner et al., 1997). Therefore, carrot cell lines can be considered quite sensitive to the stresses imposed by the culture conditions.

Another cause of oxidative stress in plant cell culture might be the high partial pressure of O2 compared with many plant tissues. As plants lack specialized systems for efficient oxygen distribution, many plant cells in vivo are normally exposed to very low O2 concentrations (Geigenberger, 2003).

In our system, cells are routinely grown in 50-ml volume batches, and thus are probably adapted to this situation. When cells are exposed to higher oxygen concentrations in 20 ml volumes, this could impose an additional oxidative stress similar to re-exposure to air in hypoxic plant tissues (Biemelt et al., 1998). The enhanced mechanical stress in the agitated 20-ml batch suspensions thus likely acts as a further agent contributing to cellular stress.

Anthocyanins are well-known antioxidant molecules, with an acknowledged ability to scavenge many reactive oxygen species (ROS) in vitro (Tsuda et al., 1996, 2000; Yamasaki et al., 1996; Lazzéet al., 2003). In vivo they have a role in attracting pollinators and seed dispersers, and evidence is emerging that they might also protect plant tissue against photodamage caused by ultraviolet and visible light (Klaper et al., 1996; Feild et al., 2001; Steyn et al., 2002). However, their possible role as antioxidants in vivo is less clear. Evidence exists that many plant tissues respond to different stresses by accumulation of anthocyanins (Chalker-Scott, 1999). Moreover, in vitro cultured plant cells respond to stress, such as nutrient deprivation and osmotic stress, by anthocyanin accumulation (Sato et al., 1996). Many stressing conditions, such as altered salinity (Vaidyanathan et al., 2003), herbivore or pathogen attack (de Gara et al., 2003; Mithofer et al., 2004), heavy metal exposure (Mithofer et al., 2004), and water deficiency (Farrant et al., 2003) cause oxidative stress and generation of ROS in plants. Thus, the accumulation of anthocyanins following exposure to stress could help to protect against oxidative damage. Recent evidence has suggested that anthocyanins may have a H2O2 scavenging role in planta (Gould et al., 2002). Moreover, Philpott et al. (2004) showed in situ antioxidant activity in anthocyanin-containing cells of sweet potato storage roots.

In our system, the viability of nonpigmented cells was much lower, as observed by in vivo imaging, suggesting a greater ability of the pigmented cells to adapt to the culture conditions. This variation in the viability between pigmented and nonpigmented cells, accompanied by a general decrease in cellular activities such as division, expansion and pigment accumulation, was restricted to the final stages of culture. These late stages of culture are generally characterized by sugar and mineral starvation; the latter situation is reported to be a cause of oxidative stress in plants (Malusa et al., 2002; Tewari et al., 2004). This further supports the idea that pigmented cells are better protected against oxidative damages.

In order to directly verify this hypothesis, cells from the Rosa3 culture and the K1 nonpigmented culture were treated with the glc/glc oxidase H2O2-generating system.

These experiments supported nicely the proposed hypothesis since (a) the anthocyanin-producing Rosa3 cell line is much more resistant to the induced oxidative stress than the nonpigmented K1 cell line, and (b) within the Rosa3 cell line, the nonpigmented cells are more sensitive than the pigmented ones.

The fact that the nonpigmented Rosa3 cells appeared to be much less sensitive than the nonpigmented K1 cells to the induced oxidative stress could be due to the efficient H2O2 scavenging action performed by the pigmented cells.

In conclusion, colour analysis and in vivo imaging, in combination with growth determination and biochemical analysis, allowed us to study anthocyanin production of carrot cell lines at individual cell level and to propose a model in which anthocyanin production in cell culture represents a stress response. This hypothesis, in turn, has also been verified by in vivo imaging.

This methodological approach can therefore provide a deeper knowledge of the dynamics of productivity of plant cell cultures, which is necessary in order to set up rational production strategies, since these complex systems are a source of important and high valuable secondary metabolites. Anthocyanins as well are of pharmacological and nutraceutical importance for their well-known biological properties (for a recent review see Kong et al., 2003).

However, this approach could, in principle, be extended not only to other naturally pigmented molecules but also to fluorescent molecules or molecules that can be labelled to give a coloured or fluorescent product.

Acknowledgements

This paper was partly supported by the Consorzio ZAI and by University of Verona.

Ancillary