Colour break in daffodils can be either a partial phenotype, with mottled colour, or a complete loss of the vivid colour attributed by carotenoids. This lack of uniformity can be a significant problem for commercial production of this important ornamental crop. Therefore, determining whether virus infection is indeed the cause of colour break will help daffodil growers to manage the issue, as well as furthering our understanding of virus interactions with plant metabolic pathways. In this study, a clear association of daffodil flower colour break with viral presence in the perianth has been demonstrated. The association of colour break with virus was confirmed by a number of independent tests such as ability of daffodil tepal sap from colour broken flowers to cause symptoms on indicator plants, direct visualisation of viral particles with TEM, and the ability to amplify viral sequences from the RNA of colour-broken but not non-colour-broken tepals.
The sensitive RT-PCR-based profiling of viral families was a powerful method for determining whether virus (es) were associated with colour break in the reverse bicolour daffodils. To give the best chance of detecting as many different types of viruses as possible, the procedure was carried out at a low annealing temperature with degenerate primers shown previously to amplify successfully members of the potex-, poty- and carlaviral families [11, 12]. The clear and consistent differences in the amplification profiles obtained from the six colour-broken flowers compared with the non-colour-broken flowers (Figure 4A, B, C) was striking as the 30 cycles used in the PCR was expected to enable detection of very small quantities of viral RNA. We had initially thought that the profiles would be similar and that the amplification would just be greater in the colour-broken flowers reflecting the presence of higher viral titre in these flowers. However, the inability to amplify products of expected size in most of the non-colour-broken flowers is suggestive of these flowers either not containing or containing extremely low quantities of viruses.
All of the independent tests carried out pointed to NMV being the virus most closely associated with colour break in the reverse bicolour daffodils. For instance, the red ringspot lesions seen in G. globosa after inoculation with colour-broken, but not non-colour-broken flowers is characteristic of NMV infection , and the TEM observations indicated that the morphology and the lengths of the viral particles in the colour-broken flowers fitted with those reported for NMV [9, 13]. However, the most compelling evidence came from sequence analysis and virus-specific RT-PCR profiling. The sequencing of the PCR products amplified from colour-broken flowers with the carla-, poty- and potex-virus degenerate primers all revealed that NMV was the only virus present in all flowers displaying colour break (Table 4). It was particularly telling that the potyviral and carlaviral primers did not predominantly amplify their target family members in PCR, but instead mis-primed to closely related sequences in the NMV genome. The mis-priming to NMV likely occurred because of the absence or very low presence of their respective target family members and the reduced stringency of the PCR achieved by performing it at the low annealing temperature of 47°C. The strong association of NMV with colour break was then able to be confirmed by observing a strong correlation between the occurrence of colour break and the ability to amplify at high stringency (60°C) a PCR product with primers designed specifically to a short region in the NMV genome (Figure 5). This could be contrasted with the poor correlation between the amplification of the potyvirus NLSYV and occurrence of colour break (Figure 5).
NMV was first described and in detail by Brunt . The virus was reported to be widespread in British crops of trumpet, large cupped and double daffodils and NMV was able to be isolated from 27 of the 48 commercial daffodil stocks tested (cultivars 'Actaea', 'Aranjuez', 'Brunswick', 'Carlton', 'Cheerfulness', 'Fortune', 'Golden Harvest', 'Inglescombe', 'King Alfred', 'Magnificence', 'Minister Talma', 'Mount Hood', 'Royal Bride', and 'Zero'). Interestingly, the plants where NMV was the only virus detected were either symptomless or showed inconspicuous mosaic symptoms at the base of their leaves. No mention of colour break was reported for these daffodils, despite their flowers being the best source of NMV particles. However, notably, none of the flowers examined by Brunt  were of the reverse bicolour type (J.A. Hunter, Historian of the New Zealand Daffodil Society, personal communication).
In a separate study, Clark and Guy  surveyed Narcissus spp. for viral infection in the Otago Province of New Zealand using ELISA and mechanical transmission tests. A high incidence of viral infection was found in the five sites examined. Indirect ELISA showed that NMV was one of the five viruses detected and that it was present at all five sites in single or mixed infections. They reported that plants infected only with NMV (based on their detection methods) showed mild to conspicuous leaf mosaic symptoms and reported that no colour break symptoms were observed. The absence of colour break among the daffodils surveyed by Clark and Guy  is presumably explained by the absence of reverse bicolours amongst the flowers they examined.
Unpublished data by Brunt  that "some isolates [of NMV] are associated with flower mottling in cultivars such as Chanter and Spellbinder (A.A. Brunt and S. Phillips, unpublished information)" support our findings that NMV infection is closely linked to colour break in reverse bicolours as both of these cultivars are of that type (J.A. Hunter personal communication). However, to demonstrate causality will require inoculation experiments. We also cannot rule out that it could be a viral titre phenomenon and not specific to a particular virus.
The ubiquitous occurrence of NMV in daffodil cultivars is intriguing given no vector for transmission has yet been discovered and direct spread by mechanical transmission is slow. Brunt  reported it took 17 months after inoculation of daffodil leaves for the virus to be able to be recovered from the leaves. This makes demonstrating Koch's postulate for NMV and colour break in daffodils extremely difficult. First, it would require a set of proven virus-free plants, difficult in itself given the ability of NMV to lie latent at very low titre, secondly, it would require maintaining them in a virus-free environment for a very long period (as the negative controls) and thirdly, it would require having the correct temperature and other conditions for successful virus inoculation.
The widespread occurrence of NMV and the inconspicuous symptoms it produces in many daffodil cultivars makes it difficult for the grower to prevent colour break from occurring as it appears that there is a 'silent' reservoir of the virus which can serve as inocula for infection of reverse bicolours. Because of this, growing this type of flower away from other cultivars may be the best way of preventing their colour break.