Metabolic engineering of beta-carotene and lycopene content in tomato fruit


  • Carlo Rosati,

    1. Ente per le Nuove tecnologie, l'Energia e l'Ambiente (ENEA), Casaccia Res. Ctr., Biotechnology and Agriculture Division, Via Anguillarese 301, 00060 S. Maria di Galeria (Roma), Italy, and
    Search for more papers by this author
  • Riccardo Aquilani,

    1. Ente per le Nuove tecnologie, l'Energia e l'Ambiente (ENEA), Casaccia Res. Ctr., Biotechnology and Agriculture Division, Via Anguillarese 301, 00060 S. Maria di Galeria (Roma), Italy, and
    Search for more papers by this author
  • Sridhar Dharmapuri,

    1. Ente per le Nuove tecnologie, l'Energia e l'Ambiente (ENEA), Casaccia Res. Ctr., Biotechnology and Agriculture Division, Via Anguillarese 301, 00060 S. Maria di Galeria (Roma), Italy, and
    Search for more papers by this author
  • Patrizia Pallara,

    1. Ente per le Nuove tecnologie, l'Energia e l'Ambiente (ENEA), Casaccia Res. Ctr., Biotechnology and Agriculture Division, Via Anguillarese 301, 00060 S. Maria di Galeria (Roma), Italy, and
    Search for more papers by this author
  • Carla Marusic,

    1. Ente per le Nuove tecnologie, l'Energia e l'Ambiente (ENEA), Casaccia Res. Ctr., Biotechnology and Agriculture Division, Via Anguillarese 301, 00060 S. Maria di Galeria (Roma), Italy, and
    Search for more papers by this author
  • Raffaela Tavazza,

    1. Ente per le Nuove tecnologie, l'Energia e l'Ambiente (ENEA), Casaccia Res. Ctr., Biotechnology and Agriculture Division, Via Anguillarese 301, 00060 S. Maria di Galeria (Roma), Italy, and
    Search for more papers by this author
  • Florence Bouvier,

    1. CNRS, Plant Molecular Biology Institute, 12 rue du Général Zimmer, 67084 Strasbourg Cedex, France
    Search for more papers by this author
  • Bilal Camara,

    1. CNRS, Plant Molecular Biology Institute, 12 rue du Général Zimmer, 67084 Strasbourg Cedex, France
    Search for more papers by this author
  • Giovanni Giuliano

    Corresponding author
    1. Ente per le Nuove tecnologie, l'Energia e l'Ambiente (ENEA), Casaccia Res. Ctr., Biotechnology and Agriculture Division, Via Anguillarese 301, 00060 S. Maria di Galeria (Roma), Italy, and
      *For correspondence (fax +39 06 3048 3215; e-mail
    Search for more papers by this author

*For correspondence (fax +39 06 3048 3215; e-mail


Ripe tomato fruits accumulate large amounts of the red linear carotene, lycopene (a dietary antioxidant) and small amounts of its orange cyclisation product, beta-carotene (pro-vitamin A). Lycopene is transformed into beta-carotene by the action of lycopene beta-cyclase (β-Lcy). We introduced, via Agrobacterium-mediated transformation, DNA constructs aimed at up-regulating (OE construct) or down-regulating (AS construct) the expression of the β-Lcy gene in a fruit-specific fashion. Three transformants containing the OE construct show a significant increase in fruit beta-carotene content. The fruits from these plants display different colour phenotypes, from orange to orange-red, depending on the lycopene/beta-carotene ratio. Fruits from AS transformants show up to 50% inhibition of β-Lcy expression, accompanied by a slight increase in lycopene content. Leaf carotenoid composition is unaltered in all transformants. In most transformants, an increase in total carotenoid content is observed with respect to the parental line. This increase occurs in the absence of major variations in the expression of endogenous carotenoid genes.


Lycopene is the main carotene accumulated in ripe tomato fruits. This linear carotenoid is synthesised in plants through a pathway starting from geranylgeranyl diphosphate and is the biosynthetic precursor of most cyclic carotenoids, including beta-carotene (Figure 2a). Lycopene is converted into beta-carotene by the action of lycopene beta-cyclase (β-Lcy), an enzyme introducing beta-ionone rings at both ends of the molecule (Cunningham et al., 1994; Hugueney et al., 1995). A competing epsilon cyclase (β-Lcy), which in tomato is encoded by the Delta gene, introduces a single epsilon-ionone ring (Ronen et al., 1999). The Delta gene is usually silent in fruits, but its de-repression in the Delta genotype results in the accumulation of compounds in the epsilon cyclisation branch (Ronen et al., 1999; Tomes, 1967) (Figure 2a). A second gene, the B gene (Tomes et al., 1956; Tomes, 1967) entails the accumulation of beta-carotene in fruits (Figure 2c). B does not encode β-Lcy (Pecker et al., 1996; J. Hirschberg, personal communication) nor does it increase expression of the β-Lcy gene (see below).

Figure 2.

Metabolic engineering of tomato carotene levels.

(a) Reactions leading to beta-carotene and vitamin A synthesis. The transformation of carotenoids into retinol occurs in animal tissues. Alpha-carotene is only half as efficient as beta-carotene in pro-vitamin A activity.

(b) Colour phenotypes of fruits from line Moneymaker and transformant OE 3, 10 days after the breaker stage.

(c) Histogram representation of the caro tenoid contents in normal and metabolically engineered tomato fruits (10 days after breaker stage). The values and standard deviations are given below.

Beta-carotene is the major dietary precursor of vitamin A, together with other carotenoids containing unsubstituted beta-ionone rings (Lakshman and Okoh, 1993) (Figure 2a). Vitamin A deficiency (VAD) is a major public health problem in over 75 countries, most of them localised in the developing world. Beta-carotene supplementation of the diet in areas at risk of VAD decreases morbidity and mortality related to several pathological conditions (Fawzi et al., 1993; Glasziou and Mackerras, 1993; Humphrey et al., 1992; West et al., 1999). The US Recommended Dietary Allowance (RDA) is 1 mg/day retinol equivalents (approximately 6 mg beta-carotene equivalents). Lycopene does not have pro-vitamin A activity, but it is a good dietary antioxidant. High plasma lycopene levels have been associated with a decreased incidence of prostate cancer (Gann et al., 1999).

Metabolic engineering is one of the possible approaches to improve the levels of vitamins/antioxidant compounds in crop plants. The engineering of high vitamin E (alpha-tocopherol) levels has been obtained in Arabidopsis seeds (Shintani and DellaPenna, 1998). A major recent breakthrough has been the engineering of beta-carotene in the endosperm of rice, a major staple food in areas at high risk of VAD (Ye et al., 2000). The dose reachable with this engineered rice (2 mg beta-carotene/kg dry rice endosperm), while being sufficient to prevent severe VAD, is insufficient to reach the RDA. Therefore, further engineering efforts in edible plants are required to provide optimal vitamin A supplementation from diversified sources.

Our understanding of the regulation of genes mediating plant carotenoid biosynthesis has made great progress in the past few years (Bartley et al., 1994; Cunningham and Gantt, 1998). The genes controlling lycopene synthesis in tomato, such as Psy1 and Pds (encoding, respectively, the fruit-specific phytoene synthase and the phytoene desaturase, Figure 2a) are up-regulated during fruit ripening (Corona et al., 1996; Giuliano et al., 1993; Pecker et al., 1992), while those controlling lycopene cyclisation, like β-Lcy or ε-Lcy are down-regulated (Pecker et al., 1996; Ronen et al., 1999). As a result, beta-carotene in ripe tomato fruits does not exceed 15% of the total carotenoids.

Several attempts have been made to engineer higher lycopene levels in tomato fruit, but none of them has met success. In one of these attempts, a bacterial phytoene desaturase (crtI, able to transform phytoene into lycopene, Figure 2a) was fused to a plastidic transit peptide and introduced in tomato plants under the control of the CaMV 35S promoter (35S/tp/crtI). This experiment, aimed at increasing lycopene levels, has unexpectedly resulted in a threefold increase in beta-carotene, but not in lycopene (Roemer et al., 2000). The total carotenoid levels are in fact decreased in the transformants, and the reasons for this result are still poorly understood (Giuliano et al., 2000). A slight enrichment of beta-cyclic carotenoids in leaves is observed in the transformants due to the constitutive nature of the promoter used.

The promoter of the tomato phytoene desaturase (Pds) gene shows high levels of expression in tomato fruits and negligible levels of expression in both tobacco and tomato leaves (Corona et al., 1996). Therefore, it is a good candidate for driving the expression of exogenous genes in a fruit-specific manner. Here, we report the results of the overexpression and antisense repression of the β-Lcy gene under the control of the Pds promoter. As expected, the overexpression increases the levels of beta-carotene up to sevenfold, to a level sufficient to cover the vitamin A RDA with 120 gm of tomato product. Total fruit carotenoid levels are slightly, but consistently, increased in most of the transformants. In some of the antisense transformants, an increase in lycopene content is observed. Leaf carotenoid composition is essentially unaltered.


To modify β-Lcy gene expression in tomato fruits we made two constructs (Figure 1a). In the overexpression (OE) construct, the Arabidopsis β-Lcy cDNA (Scolnik and Bartley, 1995) was fused to the tomato Pds promoter, which is up-regulated in ripening fruits (Corona et al., 1996). We used an Arabidopsis gene since co-suppression effects have been reported previously when using tomato coding sequences (Fray and Grierson, 1993); in a second antisense (AS) construct, the 3′ portion of the tomato β-Lcy cDNA was cloned in antisense orientation under the control of the same promoter.

Figure 1.

Molecular characterisation of the transformants.

(a) Schematic representation of the OE and AS constructs.

(b) RT–PCR analysis of the expression of β-Lcy and Ef-1α mRNA in fruits 10 days after the breaker stage.

The constructs were introduced into tomato (cv. Moneymaker) via Agrobacterium-mediated transformation and primary transformants were brought to maturity in the greenhouse. The presence of the transgene was assayed on leaf DNA via PCR and the chromosome complement via flow cytometric analysis of leaf nuclei (data not shown). Only PCR-positive euploid plants were subjected to further studies. These plants showed no significant alteration of growth habit or leaf colour phenotype, and provided normal fruit set. Among the OE transformants, many showed an altered fruit colour phenotype, varying from the red of the parental Moneymaker (MM) line to bright orange (transformant OE 3, Figure 2b). The AS transformants displayed a red fruit colour although, upon visual examination, some showed a slightly darker hue. For the present study, five OE and six AS transformants were chosen. β-Lcy expression was assayed, by semi-quantitative RT–PCR, on RNA extracted from fruits 10 days after the breaker stage, using the expression of the housekeeping Ef-1α mRNA (Mahe et al., 1992) as an internal control (Figure 1b). For biochemical analysis, samples from leaves and from four different fruits from each transformant, harvested 10 days after breaker stage, were subjected to HPLC analysis of carotenoids (see Experimental procedures).

Three OE transformants (3, 4 and 5) show high levels of expression of the Arabidopsis β-Lcy transgene (Figure 1b). All three transformants show increased beta-carotene content, reaching sevenfold in transformant OE 3 (Figure 2c). These transformants display different fruit colour phenotypes: from ‘orange-red’ of OE 4 to ‘orange’ of OE 3 (Figure 2b). The difference in the fruit colour phenotypes reflects the beta-carotene/lycopene ratio, which ranges from 0.16 in the parental MM line to 1.36 in line OE 3. Total carotenoid content is increased in most transformants, from 66 μg gm−1 fresh weight in the parental Moneymaker line up to 112 μg gm−1 fresh weight in transformant OE 5. A non-transformed tomato line carrying the dominant B gene (Tomes et al., 1956; Tomes, 1967) was subjected to the same analysis. The total carotenoid content is approximately threefold higher in the B genotype (which has a different genetic background than Moneymaker). The beta-carotene/lycopene ratio reaches 3.77 in the B genotype (Figure 2c).

Transformants AS 1, 3 and 5 show minor, but consistent, changes in their molecular and biochemical phenotypes: we observed a slight decrease in β-Lcy expression (Figure 1b and Table 3), accompanied by an increase in lycopene levels (Figure 2c) which is almost doubled in transformant AS 3 (97 µg gm−1 fresh weight compared to 54 µg gm−1 fresh weight in the parental Moneymaker line). The beta-carotene/lycopene ratio is decreased to 0.07 and the total carotenoids again show a consistent increase, up to 110 µg gm−1 fresh weight in line AS 3.

Table 3.  Endogenous gene expression in tomato fruits 10 days after the breaker stage (normalised against the EF transcript)
  • *

    Not determined.

MM1.00 ± 0.161.00 ± 0.151.00 ± 0.061.00 ± 0.091.00 ± 0.10
B1.12 ± 0.170.98 ± 0.101.11 ± 0.190.64 ± 0.100.69 ± 0.08
OE 11.33 ± 0.171.20 ± 0.141.30 ± 0.180.81 ± 0.121.09 ± 0.14
OE 21.19 ± 0.130.99 ± 0.181.22 ± 0.240.74 ± 0.120.73 ± 0.06
OE 31.30 ± 0.201.15 ± 0.181.06 ± 0.150.60 ± 0.090.74 ± 0.07
OE 41.52 ± 0.221.37 ± 0.231.25 ± 0.150.65 ± 0.110.74 ± 0.15
OE 51.13 ± 0.140.98 ± 0.170.94 ± 0.081.10 ± 0.180.40 ± 0.10
AS 1n.d.*n.d.*n.d.*n.d.*0.42 ± 0.09
AS 2n.d.*n.d.*n.d.*n.d.*0.68 ± 0.07
AS 30.71 ± 0.070.90 ± 0.190.76 ± 0.160.80 ± 0.120.45 ± 0.06
AS 40.97 ± 0.050.86 ± 0.120.84 ± 0.170.78 ± 0.080.72 ± 0.10
AS 51.25 ± 0.341.11 ± 0.341.12 ± 0.220.98 ± 0.140.32 ± 0.04
AS 61.14 ± 0.111.27 ± 0.271.33 ± 0.291.08 ± 0.080.75 ± 0.11

Leaf chlorophylls and carotenoids were measured spectrophotometrically using the method of Lichtenthaler (1987). The results show that the levels are essentially unaltered in all of the lines (Table 1). As for fruits, the B genotype was found to also have a higher leaf pigment content than Moneymaker. HPLC analysis of leaf carotenoids confirmed the absence of substantial alterations in the transformants (data not shown), confirming that the low level expression of the Pds promoter in leaves (Corona et al., 1996) is not sufficient to alter carotenoid composition in this tissue.

Table 1.  Leaf pigment composition (µg gm−1 fresh weight)
MM112 ± 1127 ± 2
B156 ± 1334 ± 2
OE 1100 ± 1421 ± 2
OE 2140 ± 4531 ± 5
OE 3102 ± 1527 ± 7
OE 4130 ± 728 ± 2
OE 5127 ± 826 ± 1
AS 1115 ± 525 ± 1
AS 2120 ± 1325 ± 1
AS 3128 ± 827 ± 2
AS 4106 ± 2924 ± 2
AS 5104 ± 325 ± 2
AS 6120 ± 1324 ± 1

Unlike the 35S/tp/crtI tomatoes described by Roemer et al. (2000), not only do most of the OE and AS transformants not show a decrease in fruit carotenoid content, but they also show a more or less consistent increase. This could be due to an up-regulation of the activity of the endogenous genes, occurring at either the transcriptional or the post-transcriptional level. In their 35S/tp/crtI transformants, Roemer et al. (2000) found slight increases in Pds, Zds and β-Lcy mRNA levels. Thus, transformation with exogenous genes seems to be able to perturb the expression of endogenous genes.

We measured the expression of the endogenous genes for the whole pathway from GGPP to beta-carotene (Figure 2a): Psy1 and Psy2 (encoding, respectively, the fruit-specific and leaf-specific phytoene synthases (Bartley and Scolnik, 1993; Fraser et al., 1999; Fray and Grierson, 1993), Pds, Zds and β-Lcy. Table 2 shows the oligonucleotides and the PCR conditions utilised for each gene. The data were normalised for the expression of the housekeeping Ef-1α gene (Mahe et al., 1992). Only minor alterations were found in the expression of these genes in the various transformants (Table 3). Zds was slightly, but consistently, inhibited in most of the OE transformants, with the exception of OE 6. The endogenous β-Lcy message was also somewhat inhibited, not only in the AS transformants, but also in some OE transformants, probably due to partial co-suppression by the introduced Arabidopsis β-Lcy gene. Interestingly, B fruits, which contain high levels of beta-carotene (Figure 2c) show, if anything, lower levels of expression of the β-Lcy gene (Table 3).

Table 2.  Primers and PCR conditions used for RT–PCR analysis
GeneUpstream primerDownstream primerCyclesTa*
  • *

    Annealing temperature.


It is possible that the genes we analysed do not represent the complete complement of endogenous genes mediating beta-carotene biosynthesis in the tomato fruit. A ‘Blast’ search in the GenBank resulted in an EST (accession number AW944994), showing differences at 11 out of 510 positions with respect to the Zds cDNA (accession number AF195507). Furthermore, the laboratory of J. Hirschberg (personal communication) has cloned a second cyclase, which is active in tomato fruits and is encoded by the B gene.


Over-expression of the Arabidopsis lycopene beta-cyclase in tomato fruits under the control of the Pds promoter results in the accumulation of beta-carotene. This result, along with that of Roemer et al. (2000), paves the way for increasing the pro-vitamin A content of tomato fruits without the long breeding efforts and undesirable agronomic traits brought by the introgression of the B gene from wild tomato species (Tomes et al., 1956; Tomes, 1967). It also opens the possibility for the metabolic engineering of compounds further downstream in the carotenoid pathway, for which ‘natural’ accumulating mutants or genotypes are not available. An elegant example of this type of metabolic engineering is provided by the overexpression of the Hematococcus pluvialis ketolase in tobacco under the control of the Pds promoter (Mann et al., 2000). These plants accumulate astaxanthin, a non-plant carotenoid, in the nectary tissue, which is virtually the only tissue in tobacco plants to contain chromoplasts. No alteration of leaf carotenoid content is observed. These data and those presented here confirm that the chromoplast-associated expression of the Pds promoter (Corona et al., 1996) is ‘tight’ enough to leave leaf carotenoids unaltered.

Carotenogenic enzymes have been proposed to be present in multi-enzyme aggregates (Cunningham and Gantt, 1998). Therefore, the alteration of the levels of one enzyme may affect the activity of other enzymes in the complex. This hypothesis may explain the unexpected increase in beta-carotene levels found in crtI overexpressors of rice (Ye et al., 2000) or tomato (Roemer et al., 2000). A more detailed discussion of the possible regulatory phenomena occurring in crtI transformants can be found in Giuliano et al. (2000).

Unlike what was observed by Roemer et al. (2000) in 35S/tp/crtI tomatoes, in our transformants we did not observe a decrease in fruit carotenoid levels. On the contrary we observed a slight, but consistent, increase both in the OE and AS transformants. This increase is independent of the construct and is not due to a transcriptional induction of the endogenous carotenoid mRNA levels (Table 3). The hypothesis which is in best agreement with the data is that the regulation is effected by sequences found in both the constructs used, acting in trans and at the post-transcriptional level. Since the Pds promoter construct we used contains the whole 5′ untranslated leader of the Pds gene, we are currently investigating the hypothesis that this region contains sequences acting in trans at the post-transcriptional level on the function of other carotenoid genes. Of course, an alternative explanation is that an up-regulation of endogenous genes is indeed occurring, but it is confined to early fruit development. More detailed studies are needed to discriminate between these hypotheses.

In one of the OE transformants (not shown) a much higher increase (approximately threefold) in lycopene levels was observed. Besides the increase in carotenoid content, this transformant showed, however, other dominant, pleiotropic phenotypes, including the total lack of seed germination. We are currently checking the hypothesis that the phenotype was due to an extensive rearrangement during TDNA integration.

The results observed with the AS construct are less clear-cut, some inhibition of β-Lcy is observed in some AS transformants, accompanied by an up to twofold increase in the amount of lycopene. These data, if confirmed, would indicate that the low levels of β-Lcy expression found in the fruit (Pecker et al., 1996 and this paper) contribute to beta-cyclisation in this organ, probably acting together with the product of the B gene. Our data indicate that B does not encode a positive regulator of the β-Lcy gene, since the mRNA levels of the latter are, if anything, decreased in B fruits. It will be interesting to determine if the pools of beta-carotene produced in crtI fruits, B fruits and β-Lcy fruits are metabolically comparable, as judged by their ability to be converted into downstream compounds.

The transformants reported here are able to deliver 6 mg of beta-carotene (the equivalent of the vitamin A RDA, which is able to reduce by 50% the pregnancy-related mortality in areas at high risk of VAD (West et al., 1999)), in combination with high doses of lycopene, a dietary antioxidant and cancer-preventing agent, with 120 gm of tomato product, a dose fully compatible with a normal diet.

Experimental procedures

Basic molecular biological experiments were performed as described previously (Sambrook et al., 1989). The Arabidopsis β-Lcy cDNA (nucleotides 103–1680 of the published sequence (GenBank accession number U50739) was amplified via RT–PCR for 30 cycles from Arabidopsis leaf mRNA, using the oligonucleotides TCGGGATCCGGTTAGAAATCGTGTCCGGTG (forward) and TGCGAGCTCTTAGTCTCTATCTTGTACC (reverse), introducing BamHI and SacI restriction sites at the end of the molecule. The BamHI–SacI fragment was cloned in the plasmid pBSK+ , sequenced to check the lack of mutations, and cloned under the control of the tomato Pds promoter in a pBI101-based transformation vector (Corona et al., 1996). The 3′ portion of the tomato β-Lcy cDNA (nucleotides 324–1637 of the published sequence (GenBank accession number X86452) was amplified via RT–PCR for 30 cycles from tomato leaf mRNA using the oligonucleotides TGAGCTTCCTATGTATGAC (forward) and GACAAGATTCCG AATTACTC (reverse). The amplified fragment was blunt end-cloned in the pGEM4Z SmaI site, sequenced to check the orientation, excised using the flanking BamHI and SacI sites and cloned under the control of the tomato Pds promoter as described above. The constructs were introduced into tomato transgenic plants using previously published techniques (van Roekel et al., 1993). Semi-quantitative RT–PCR was performed as previously reported (Giuliano et al., 1993) with the following modifications: for each sample, 1 µg total RNA was retrotranscribed in 30 µl RT reaction volume containing 2.5 µm of oligo-dT17; 2 µl of RT mix were subjected to PCR, using the oligonucleotides and conditions indicated in Table 1; each PCR cycle consisted of three steps-denaturation 45 sec at 94°C, annealing 45 sec at the appropriate temperature (Table 1) and extension 45 sec at 72°C; the number of PCR cycles was adjusted for each transcript to obtain a detectable signal without reaching saturation (Table 1). For detecting the sense β-Lcy mRNA in the AS plants, the downstream β-Lcy primer was substituted for the oligo-dT primer in the RT reaction. The amplification products were loaded on a 1% (w/v) agarose/ethidium bromide gel and quantified using a digital camera and the ‘NIH image’ software run on a Macintosh G3 computer.

Plants were grown in the greenhouse, in the spring and early summer, at temperatures controlled between 15°C and 25°C. Fruits were visually inspected daily for change of colour (breaker stage) and harvested 10 days after breaker stage. Pericarp tissue was immediately harvested, quick-frozen and lyophilised before carotenoid or RNA extraction.

For HPLC analysis, the tomato fruit pericarps were chilled in liquid nitrogen and powdered in the presence of magnesium carbonate (1% of the sample weight). The resulting powder was extracted first with methanol and subsequently with acetone until discoloration. Pigments were transferred to diethyl ether by adding an equal volume of ether to the combined methanol-acetone extract. The resulting ether fraction was concentrated to dryness and dissolved into chloroform before chromatographic analysis. Due to the overwhelming amount of the carotene fraction compared to the xanthophyll one, we routinely made use of a rapid chromatographic separation (less than 8 min) using a Nova-Pak C18 column (3.9 × 15 mm) eluted with acetonitrile: ethyl acetate: chloroform (70/20/10, v/v) as described previously (Camara, 1985). Leaf carotenoids and chlorophyll were extracted and quantified spectrophotometrically according to published procedures (Lichtenthaler, 1987).

For cytometric analysis, nuclei were extracted from leaves and analysed as described previously (Lucretti et al., 1997) using a Beckton-Dickinson Facstar Plus flow cytometer.


This work was supported by EU research contracts FAIR CT96 1633 and BIO4 CT97 2077 and by a research contract from the Italian Ministry of Agriculture. We thank Dr Sergio Lucretti (ENEA, Rome, Italy) for performing the flow cytometric analysis, Mrs Elena Nebuloso for help with sequencing, and Prof. P. Bramley and Dr Paul Fraser (University of London, London, UK) and Prof. J. Hirschberg (University of Jerusalem, Jerusalem, Israel) for the communication of unpublished results.