Olives ( Olea europaea subs. europaea var. euro­paea ), wheat and grapes are some of the oldest of all crops ( Zohary & Hopf, 1994 ). Olives are normally cul­tivated between 30º and 45º N and S, and in other areas where the climate is Mediterranean ( Barranco et al ., 2000 ). Spain has 2,554,829 ha under olive cultivation, and is the world's foremost producer of olive oil (Ministerio de Agricultura, Pesca y Alimentación-MA­PA, 2018); its output accounts for 60% of all the EU's olive oil and 45% of that produced worldwide (International Olive Oil Council, 2015 ¹ ). These data provide an idea of the economic and environmen­tal importance of olives in Spain.

Many olive genotypes are grown around the world, and many of those growing in the most important o­li­­ve oil-producing countries have been described ( Barranco et al ., 2000 ; Belaj et al ., 2002 ; Bartolini et al ., 2005 ; Rallo et al ., 2005 ; Fendri et al ., 2010 and 2014 ; Haouane et al ., 2011 ; Lazovic et al ., 2016 ; Sakar et al ., 2016 ). In Spain, over 250 are reported in use ( Barranco et al ., 2005 ; Vargas-Gómez & Ta­lavera-Lozano, 2012 ), but the current number used in the main commercial plantations is small ( Rallo et al ., 2005 ). The variation in Spanish olive germplasm has been studied in certain areas ( Viñuales, 2007 ; Díez et al ., 2011 ; Gómez et al ., 2012 ; Trujillo et al ., 2014 ; Martí et al ., 2015 ). In marginal areas, however, much less work of this kind has been done, and in some pla­ces no surveys or characterisations have been underta­ken at all. Such is the case of Galicia in the NW Peninsu­lar.

The Atlantic-influenced climate of this region is not currently associated with olive cultivation, yet old references cite olives trees being grown here ( Alonso de Herrera, 1513 ; Contreras, 1798 ; Hidalgo-Tablada, 1870 ). The importance of olive production in the past is evident in archaeological finds such as primitive oil mills dating from the 1st-2nd centuries BCE ( Fernández de la Cigoña & Martínez, 2003 ), and numerous references to olive trees and olive oil in the region's toponymy. A strong oral tradition also exists among the region's inhabitants that testify to families having produced their own olive oil for generations. This residual cultivation of olive trees has persisted in the area until the present day, but in the last 10 years there have been several initiatives that have attempted to recover olive production as part of the regional economy. Indeed, between 2008 and 2017, the area under olive cultivation increased from just 10 ha to 272 ha ( MAPA, 2018 ).

While a number of recent studies have examined the olive oils produced in Galicia ( Espinosa-Sánchez, 2010 ; Reboredo-Rodríguez et al ., 2014a , 2014b and 2015 ), most of these oils were not produced by native trees ( Reboredo-Rodríguez et al ., 2015 ) but by recent­ly planted and commonly cultivated genotypes from Andalusia, such as 'Picual' and 'Arbequina'. Indeed, while the agricultural biodiversity of Galicia's woody-plant crops-grapes ( Gago et al ., 2009 ; Martínez et al ., 2018 ), apples ( Pereira-Lorenzo et al ., 2007 ), pears ( dos San­tos et al ., 2011 ; Pereira-Lorenzo et al ., 2012 ) and chest­nuts ( Pereira-Lorenzo et al ., 1997 ) has been studied, that of the region's olive trees is almost unk­nown. Localising, characterising and conserving the genotypes that may still be found in this geographi­cal area is vital to avoid the genetic erosion of the species and to save their traits for use in olive improvement programmes. A recent article by Reboredo-Rodríguez et al . (2018) , and the doctoral thesis of Reboredo-Rodríguez (2015) , identified a number of olive genoty­pes from this region. However, these contributions co­vered only a very small part of the territory and some of the molecular results were contradictory. Wider and more rigorous and systematic surveying is required to catalogue the area's olive tree biodiversity and to allow their inclusion in the Spa­nish list of olive varieties of commercial interest.

The present work reports the localisation of anci­ent olive trees in Galicia, their characterisation using bo­tanical and molecular markers, and examines whe­ther or not these trees represent unknown native genoty­pes. Back in the 19th century, Hidalgo-Tablada (1870) suggested that olive genotypes might be characteri­sed via certain leaf, fruit and endocarp variables, the shape of the tree, and other features. Nowadays the International Olive Council (IOC) uses the genotype classification system of Barranco et al . (2005) , which employs botanical and agronomic markers. The pre­sent work, however, introduces a further morphome­tric inspection of the leaf. Martínez & Grenan (1999) developed a graphic method for visualizing the dif­ferences that appeared in biometric studies of the gra­pevine leaf. This method provides a highly realistic representation of the foliar morphology and has been used to compare genotypes ( Martínez & Pérez, 2000 ; Santiago et al ., 2005 ; Martínez, 2007 ; Gago et al ., 2009 ; Martínez et al ., 2018 ) and clones ( Martínez et al ., 2005 ). Martínez & Grenan's (1999) method has been adapted in the present work, in order to be used in the study of olive average leaves. Finally, simple se­quence repeats (SSRs) markers were also used in geno­type identifications. Many genetic characterisation stu­dies have used different sets of SSRs, and the results have greatly increased our knowledge of olive genetic heritage in different areas ( Cipriani et al ., 2002 ; Belaj et al ., 2004 and 2011 ; Gil et al ., 2006 ; Sarri et al ., 2006 ; Baldoni et al ., 2009 ; Muzzalupo et al ., 2010 ; Fendri et al ., 2010 ; Diez et al ., 2011 ; Martí et al ., 2015 ; Lazo­vic et al ., 2016; Sakar et al ., 2016 ). Together, all these techniques provide a glimpse of the possibly notable olive diversity of the Spanish Northwest.

This study on the almost forgotten olive trees of northwestern Spain aims to provide their first botanical and molecular characterisation, and to compare this local germplasm with that conserved in databases. The results provide a glimpse of the olive diversity that the region may still hold.

The molecular profiles of the 18 examined trees grouped them into nine genotypes, of which three could be identified: 'Brava Gallega', 'Mansa Gallega' and 'Cobrancoça'. For now, the identity of the other six genotypes remains unknown. These results agree with those of other studies that have genetically or morphologically characterised centuries-old olive trees in peripheral growing areas; where only a small pro­portion of those examined represented genotypes with a commercial use ( Díez et al ., 2011 ; Salimonti et al ., 2013 ; Martí et al ., 2015 ; Lazovic et al ., 2016 ; Sakar et al ., 2016 ). Similar results have been repor­­ted also for centuries-old grapevines ( Martínez & Pérez, 2000 ; Santiago et al ., 2005 ; Gago et al ., 2009 ).

SSRs are widely used as markers in the identifi­­ca­tion of olive genotypes ( Cipriani et al ., 2002 ; Bal­­doni et al ., 2009; Díez et al ., 2012 ; Jakše et al ., 2013 ; Reboredo et al ., 2018 ). In the present work, the loci GAPU059 and UDO99-019 showed low-le­vel polymorphism, and were therefore little informative in identifying the genotypes of the examined trees. Reboredo et al . (2018) reported the same for these two loci. Loci UDO043 and ssrOeUA-DCA9 showed the greatest discriminatory power, in agreement with the results of other authors who examined olive material from different areas ( Baldoni et al ., 2009 ; Salimonti et al ., 2013 ; Trujillo et al ., 2014 ).

The morphological characteristics of the endocarp, which are considered very stable, are also widely used in olive genotype identification ( Barranco et al ., 2000 ; Fendri et al ., 2010 ). It is also usual to make use of the characteristics of the leaves or drupes. Certainly, the size of the leaves and drupes may differ depending upon the edaphoclimatic conditions, but it should be remembered that in grapevine the effect of 'genotype' dominates that of 'edaphoclimatic conditions' ( Martínez & Grenan, 1999 ). In other words, although the size of the leaves and drupes may be different, their shape is constant. Further, the use of relationships between measure­ments of different variables eliminates the effect of growing conditions, and the mean leaves constructed from them provide an excellent identification tool.

In the present work, the trees with the same molecular profiles fell into the same PCA-determined groups based on their endocarp characteristics. This did not always happen, however, with respect to the leaves; indeed, large qualitative and quantitative differences were seen between trees with identical molecular profiles. Such was the case for the 'Cobrancoça' trees; these grouped together in terms of their leaf qualitative variables (leaf blade shape, leaf blade width and longitudinal curvature of the leaf), but not in terms of their quantitative variables (leaf lengths and angles). In contrast, the Unknown Genotype 1 trees (13, 15, 16 and 17) grouped together in terms of their qualitative but not their quantitative variables. The same was true for the 'Brava Gallega' trees (trees 6 and 7). With respect to drupe qualitative characteristics, Unknown Genotype 1 was also heterogeneous, especially in terms of fruit colour (Table 6). This might be explained in that although all fruits were collected on the same day (by different teams), the trees grew in different areas and their fruit may not have been of equal ripeness. Sali­mon­ti et al . (2013) suggests that many of the differences seen within genotypes could be the result of the existence of different clones, as reported for grapevine ( Boso et al ., 2004 ; Martínez et al ., 2005 ).

It is possible that a larger number of SSR markers might have led to different genotype identifications, though this is unlikely given that 13 were examined. This has been reported in grapevine, although a redu­ced number (just six) of highly discriminatory SSRs are now recognised that can identify nearly all genotypes ( OIV, 2009 ).

Recently, Reboredo et al . (2018) published an arti­cle in which cultivated olive material from the same region was examined, and three different genotypes were found among a 32-olive-tree sample; also, using a set of 14 SSRs loci, a total of 37 alleles were reported in the cited work. In the present work, nine different SSRs profiles were found in an 18-olive-tree sample, and a total of 55 alleles detected with a set of 13 SSRs loci. This might be explained in that the present survey covered a much wider sampling area where locations with different numbers of olive trees are present. Historical records for these locations confirmed their past association with active olive cultivation. In addition, the present work selected centuries-old olive trees; these were documented as such in some cases, and at least referred to as such by oral tradition in others.

The molecular profile assigned to the genotype 'Brava' by Reboredo-Rodríguez et al . (2018) matches that of 'Brava Gallega' in the present work (in both cases the SSR profiles were compared to those held in the WOGBC and UCO databases). However, the molecu­lar and morphological (which included only endocarp information) profiles assigned by Reboredo-Rodríguez et al . (2018) to the genotype 'Mansa' (reported as 'Unknown' by Reboredo-Rodríguez, 2015 ) did not match those of 'Mansa Gallega' as determined in the present work and in the consulted WOGBC and UCO databases. It is important to note that the molecular profile and botanical characterisation reported here as identifying the genotype 'Mansa Gallega' correspond exactly to those recognized by the Spanish Department of Agriculture ( MAPAMA, 2017 ).

The correct molecular characterization of genotypes is important to prevent confusion with other genoty­pes with similar morphological characteristics and also to use this plant material in breeding programs and in commercial propagation. SSR analysis is a powerful tool for genotype characterization. In olive, intra-genotype genetic diversity has been reported using SSR markers ( Muzzalupo et al ., 2010 ; Caruso et al ., 2014 ; Trujillo et al ., 2014 ), for these authors, SSR profiles that are differentiated by one or several dissimilar alleles are classified into the same genotype. These are classified as 'molecular variants' and are treated as 'clones' within the main variety due to somaclonal mutations. But in other woody species SSR markers are not considered as an effective approach to detect genetic differences among clones ( Imazio et al ., 2002 ; Bouhadida et al ., 2007 ; Pereira-Lorenzo et al ., 2007 ).

The 'Mansa Gallega' identified in the present work was located in the south of the Province of Ponteve­dra − a long way from the sampling area studied by Reboredo-Rodríguez et al . (2018) . However, the trees studied that were identified as belonging to 'Brava Gallega' were located in the same area studied by the latter authors. Finally, the molecular profile assigned to the genotype 'Picuda' by Reboredo-Rodríguez et al . (2018) was not found among those detected in the present work. Indeed, neither 'Picual' nor 'Arbequina', nor indeed any other genotype cultivated in Spain's most important olive-producing regions, was represented by the examined trees. The olive-growing area closest to Galicia is in northern Portugal; the detection of the Portuguese genotype 'Cobrancoça' (trees 1, 2, 4, 5 and 10) is therefore not very surprising. Fig. 1 shows all these 'Cobrancoça' trees to be located within a few kilometres of the Portuguese border. It is rather more surprising that no other specimen of this genotype was found away from this area. It is also of note that no specimens of a genotype extensively grown in Portugal, known as 'Galega' ( Cordeiro et al ., 2008 ) - a name that suggests it originated in Galicia - were found in the present study.

Trees 1-10, all known locally under the name of 'Brava', were found in areas where olive growing has more of a tradition. However, only trees 6 and 7 had a molecular profile that matched with the profile recorded for the genotype 'Brava Gallega' in the WOGBC and UCO databases. Trees 1, 2, 4, 5 and 10 were found to belong to the genotype 'Cobrancoça' ( Cordeiro et al ., 2008 ), and others belonged to unknown genotypes (both in terms of their molecular profile and botanical characteristics). The name 'Brava' appears to be used locally to refer to many different genotypes; only one of them, of course, is the 'Brava Gallega' genotype. The term ' brava ' in fruticulture is used to refer to plant grown from a seed and normally used as a seedling rootstock, but in this particular case the olive growers in this area use this term to refer to a number of genotypes with a high agronomic quality and clearly distinct from a wild olive or a rootstock and that they propagate using cuttings.

The second most locally used genotype name was 'Mansa', but only one tree (tree 11) actually had a molecular profile that matched that deposited in the WOGBC and UCO databases.

The problems of homonyms and synonyms affec­ting Galicia's olive trees is not the same as that which affects grapevine genotypes ( Martínez et al ., 2018 ). While grapevine genotypes may have synonyms, they always identify the same genotype. For example, the genotype that goes by the name 'Tempranillo' in the Rioja winemaking region, is called Tinta Fina in the Ribera del Duero region, and has different names in other areas. However, even though viticulturists may use these different names, they all identify the same genotype through association with the same leaf and cluster characteristics. 'Brava' and 'Mansa', in contrast, are not terms that identify respective olive genotypes in Galicia. In conversations with growers in the present work, it was noted that they used the terms with entirely different genotypes. The affirmation by Reboredo-Rodríguez et al . (2018) that Galicia 'Mansa' is a homonym of the genotypes 'Brava' and 'Mansa', and that 'Mansa' is a synonym of the genotype 'Brava', seems not to hold up.

Trees 3, 8 and 9, which were located very close to one another, each represented an unknown genotype (Unknown Genotypes 3, 4 and 5 respectively), with each showing different molecular and botanical differences. The presence of different unknown genotypes in such a small area hints at the diversity yet to be discovered. Also, tree 18, which was located close to tree 11, was of another unknown genotype (Unknown Genotype 2).

Trees 13, 15, 16 and17 all belonged to Unknown Genotype 1. The age of these trees, plus their being found over a wide area, suggests that the vegetative propagation of olive trees has long been performed in the region. Tree 14 (Unknown Genotype 6) was found in the same cultivation area that trees 13, 16 and 17 (Unknown Genotype 1) but it has a molecular profile that differs in one SSR locus from this genotype (trees 13, 16 and 17). Tree 12 was also found in the same area but its molecular profile differs in one allele for two loci from the Unknown Genotype 1; in addition, this tree also differs from trees of Unknown Genotype 1 in some morphological characteristics, as the absence of nipple in the fruit or the high weight in the endocarp.

The results suggest that Galicia may be a reservoir of olive diversity. This agrees with the thinking of other authors (Trujillo et al ., 1990; Zohary & Hopf, 1994 ; Claros et al ., 2000 ; Cordeiro et al ., 2008 ) who suggest the majority of the region's olive genotypes to be native and to have spread little to other areas. Apart from providing new genetic material, such native genotypes could provide information of use in other scientific studies. For example, studies on the domestication and parentage of olive trees ( Trujillo et al ., 2014 ; Diez et al ., 2015 ) have normally examined genotypes native to more Mediterranean areas. Galicia's native genotypes could add new variability and molecular heterogeneity to be considered in such studies.

The present work provides the molecular profiles and complete botanical descriptions of some unreported, local olive genotypes surviving in Galicia. The results identified two potentially native genotypes 'Brava Gallega' and 'Mansa Gallega', and clarified certain misidentifications of the latter by other authors. Six unknown genotypes were also detected, as well as the Portuguese genotype 'Cobrancoça'. The evidence suggests that olive trees have been cultivated in the region for centuries, and that the diversity of native genotypes is high. This diversity should be preserved as part of Europe's agricultural heritage, but also because it may offer scientific and commercial opportunities. A larger survey should be performed to determine the full range of Galicia's olive tree diversity, followed by agricultural studies that might indicate the potential of the region's rediscovered genotypes.

Dr. I. Trujillo provided assistance in SSR analysis during a period at the Laboratorio de Elaiografía y Marcadores Moleculares at the Dept. of Agronomy, University of Córdoba. Dr. Trujillo also compared the profiles obtained with those in the WOBG and UCO databases (performed in 2015). Iván González and Elena Zubiaurre are thanked for technical assistance, as is Adrian Burton for the English translation of the text.