Genotype 3 year interaction and broad-sense heritability of architecturalcharacteristics in rose bushL A U R E N T C R E S P E L

1,3,*, C A M I L L E L E B R A S1, D A N I E L R E L I O N

1 and P H I L I P P E M O R E L2,*

1Agrocampus Ouest, Institut de Recherche en Horticulture et Semences (INRA Agrocampus Ouest Universit�e d’Angers), SFR 149QUASAV, 49045, Angers, France; 2INRA, Institut de Recherche en Horticulture et Semences (INRA Agrocampus Ouest Universit�ed’Angers), SFR 149 QUASAV, 49071, Beaucouz�e, France; 3Corresponding author, E-mail: laurent.crespel@agrocampus-ouest.fr

With 2 figures and 6 tables

Received June 19, 2013 / Accepted December 9, 2013Communicated by T. Debener

AbstractThe effect of genotype factors, year and their interaction was assessed onsix architectural variables of eight cultivars of rose bush. Plants weregrown in pots in a greenhouse in the spring of 2011 and 2012, twohighly contrasted years in terms of the quantity of cumulative radiation,with a relative deviation (for 2012 compared to 2011) ranging from�24.6% (April) to +13.7% (March). Their architecture was digitized attwo observation scales, the plant and the axis. Highly significant geno-type (G) and year (Y) effects were revealed for all of the variables mea-sured, as well as a G 9 Y interaction. Concerning the year effect, it wassignificantly higher in 2012 and for all of the variables measured. The G9 Y interaction was due to (i) different genotype groupings according toyear, (ii) difference response amplitudes between years according togenotype. Broad-sense heritability was calculated for each of these vari-ables. It was moderate to high, ranging from 48% for the length of longaxes to 98% for the number of metamers on long axes.

Key words: plant architecture — Rosa — shape — phenotypicvariation — ornamental woody plant

Plant architecture is the result of growth and branching processesand depends on genetic and environment factors and their inter-action, as has been shown for architectural characteristics linkedto yield, that is, in wheat for plant height and spike length (Wuet al. 2012), in barley for the number of spikes per m2 and thenumber of seeds per spike (Rodriguez et al. 2008) and in cow-pea for the number of branchings and pods per plant (Shimelisand Shiringani 2010). For ornamental plants, the overall shapeof the plant is a major component of visual quality and a keycriterion when purchasing a plant (Boumaza et al. 2009). How-ever, plant shape is the result of the architectural construction ofthe plant, that is, the positioning of the different aerial organs inspace according to the organization rules specific to each spe-cies. Plant shape can therefore be controlled through the geneticpathway via plant breeding and/or control of the environment,the application of cultivation techniques such as water restriction(Morel 2001, Demotes-Mainard et al. 2013), modification of thelight spectrum (Rajapakse and Kelly 1993, Abidi et al. 2012)and mechanical stimulation (Morel et al. 2012). These factorsmake it possible to modify some architectural characteristicssuch as metamer length, the number of branchings and theirposition along the axis. However, these methods are appliedmore or less empirically by plant breeders and horticulturists.

More extensive knowledge about the heredity of architecturalcharacteristics as well as the genotype 9 environment interactionwould lead to a more effective control of plant architecture and,as a result, its shape. This is why we have undertaken work onthe genetic analysis of architecture in rose and the effect of envi-ronmental factors (light, water restriction, mechanical stimula-tion). The architecture can be objectively characterized bybreaking it down into entities arising from the growth andbranching processes (axis, metamer).Two types of axes (short and long) were identified in rose

(Morel et al. 2009), and Crespel et al. (2013) showed that sixvariables – at the scale of the plant and the axis – were neces-sary and sufficient to describe rose architecture and its varia-tions. The aim of this article was to present the initial results onthe effects of genotype and year factors, and their interaction, onthe six architectural variables selected. Their broad-sense herita-bility was also calculated to assess the feasibility of their geneticanalysis. To do this, the six variables were measured in thespring of 2011 and 2012, two climatically contrasting years, par-ticularly in terms of the amount of radiation, on plants of thesame eight cultivars, aged from 5 to 6 months and cultivated ina greenhouse.

Materials and MethodsPlant material: The plant material was comprised of rose bush cultivarsused for garden decoration, repeat-blooming with contrasting shapes(ranging from upright to spreading). The cultivars were ‘Baipome’ PinkGnome�, ‘Blush Noisette’, ‘Old Blush’, ‘Perle d’Or’ and ‘The Fairy’.Three hybrids with intermediate shapes were also included: Hw20,Hw154 and Hw336, the result of a cross between a dihaploid of Rosahybrida L. H190 and a hybrid of Rosa wichurana Cr�ep. (Crespel et al.2002). Some of these cultivars were eventually used as parents for thecreation of connected segregation progeny. Diploid cultivars were chosenfor the genetic analysis of the architecture in order to simplify thisanalysis. The combination of diploid and repeat-blooming character israre in the cultivated rose, thus excluding the most commonly usedcultivars, which are tetraploids.

Experimental conditions: This study was carried out in Angers(France), in the experimental facilities of the IRHS (French ResearchInstitute on Horticulture and Seeds).

The experimental protocol described below was applied in the springof the 2 years, 2011 and 2012. These two periods were considered astwo different environments, denoted 2011 and 2012, respectively.

The plants were obtained from cuttings taken from 6-month-oldmother plants grown in pots in a greenhouse. Cutting was preferable to*Both authors contributed equally to this work.

Plant Breeding, 133, 412–418 (2014) doi:10.1111/pbr.12157© 2014 Blackwell Verlag GmbH

grafting because it made it possible to obtain large quantities of homo-geneous batches of plants, while avoiding the rootstock effect.

Cuttings consisted of a single metamer taken from the median zone ofthe stems. Cuttings were taken in January and planted in plugs (diameter:35 mm; height: 40 mm) consisting of a non-woven fabric containing amixture of fine peat and perlite. Rooting took place in a glass greenhouseunder a plastic tunnel. The average temperature was 18°C at night and22°C during the day, and the relative humidity was maintained at satura-tion by a fine mist humidifier.

Young plantlets were planted in 0.5-l pots 5 weeks later, in February,and then in 2-l pots a month later, in a substrate consisting of peat (60%),perlite (20%) and coir (20%). They were placed in a glass greenhouse on4.5-m² (3 9 1.5 m) benches equipped with a nutrient solution tank, withone cultivar per bench. Subsurface irrigation was triggered by a tensiome-ter, at a water potential value of �10 kPa. Mineral nutrition was providedby fertigation with a liquid fertilizer (N/P2O5/K2O) with a 3 : 2 : 6 ratio,a pH of 6.5 and an average electrical conductivity of 1.2 mS/cm. Mini-mum air temperature was maintained at 18°C, with aeration at 20°C. Therelative hygrometry was maintained at 70% using a fine mist humidifier.For each cultivar, the experimental set-up consisted of a bench with 40plants. Fifteen plants were chosen at random from the central part of thebench in 2011 and 12 in 2012, for measurements that took place from 27April to 10 June 2011 and from 22 May to 5 July 2012, respectively. Dig-itizing plants was time-consuming. The plants were placed in a coldchamber at 2°C at the elementary architectural structure stage in order toarrest their growth and thus avoid an experimental bias (Crespel et al.2013). Finally, plants were defoliated beforehand in order to facilitate theidentification of the axes during digitalization.

The measurement of daily average air temperatures in the greenhousemade it possible to calculate the sum of temperatures (in °C.d) accordingto the method used by Huch�e-Th�elier et al. (2011) to define plant devel-opment time in relation to a base temperature of 1.93°C. Similarly, glo-bal solar radiation outside of the greenhouse (M�et�eo France, Beaucouz�e,France) was measured to make it possible to calculate the sum radiationfor the growing period (in J/cm²).

Architectural description: Plant architecture in rose is characterized bytwo entities, the axis and the metamer, the latter consisting of an internode,a node, an axillary bud and a leaf (White 1979). These entities aretopologically related among themselves, either by succession or branching.The architectural analysis was carried out at two observation scales, theplant and the axis. Two types of axes have been distinguished at the plantscale, as defined by Morel et al. (2009): short axes that comprise one tofour metamers and long axes that comprise five metamers or more. Sixvariables identified by Crespel et al. (2013) as being relevant to describeplant architecture were measured: (i) at the plant scale, the number of axesdetermined (i.e. axes that terminate with a flower or flower bud) (NbDetA)and the number of long axes (NbLA) and (ii) at the axis scale, the numberof metamers on the long axes (NbMet_LA), the length of the long axes(L_LA), the branching angle of the cord of the long axes in relation to thevertical axis (AngLA/Cord) and the number of branching orders (NbBrO).

Architectural measurements were carried out using a Fastrack� digi-tizer (Polhemus, Colchester, VT, USA), and data were recorded withPiafDigit software, designed specifically for that purpose (Don�es et al.2006), making it possible to build a database that included all of themeasurements made at both of the observation scales. This software alsomade it possible to provide a 3D representation of the plant architectureduring digitalization in real time (Fig. 1). The file generated by thissoftware has an MTG-type structure, for Multi Tree Scale (Godin andCaraglio 1998). Variables for the architectural analysis were extractedfrom these data using AMAPmod software (Godin et al. 1999).

Data analysis: Analyses of variance, followed by multiple comparisonsof means (Newman–Keuls test), were carried out on the six variablesmeasured, with a probability of P < 0.05, to reveal whether or not therewas a significant effect of genotype and year factors and the interactionbetween them. These statistical analyses were carried out using StatBox6.6 software (GrimmerSoft, Neuilly-sur-Seine, France).

The broad-sense heritability (H2) of each variable was estimated asbeing the ratio of the genotypic variance (r2

G) to the phenotypic variance(r2

P), expressed in percentage. The components of the phenotypic vari-ance were calculated on the basis of the analysis of variance of theexpected mean squares, as described by Nanson (1970) and used by Shi-melis and Shiringani (2010) (Table 1). The phenotypic variance wastherefore calculated as follows:

r2P ¼ r2

G þ r2GY=Y þ r2

e=rY

wherer2GY is the variance of the genotype9 year interaction,r2

e is the resid-ual variance, Y is the number of years, and r is the number of repetitions.

ResultsClimatic characterization of both years

For the period from March to June, the relative differences inthermal times (for 2012 compared to 2011) varied from �2.6%(June) to �10.1% (April), with an average of �6.3%. For thesame period, differences related to cumulative radiation (for

O3

O2

O1

Axis

Metamer

Fig. 1: 3D representation of the plant architecture obtained by digitaliza-tion with two entities (axis and metamer) and three branching orders:Order 1 (O1), Order 2 (O2) and Order 3 (O3)

Table 1: Partial analysis of variance and expected mean squares amongeight rose genotypes tested in 2011 and 2012

Sources of variation Degrees of freedom Expected mean square1

Genotype G�1 rYr2G þ rr2

GY þ r2e

Year Y�1 –Genotype by year (G�1)(Y�1) rr2

GY þ r2e

Residual Y(G�1)(r�1) r2e

1r: number of replications; G: number of genotypes; Y: number of years;r2G: genotypic variance; r2

GY : genotype by year variance; r2e : residual

variance.

Genotype x year interaction for architecture 413

2012 compared to 2011) varied from �24.6% (April) to +13.7%(March), with an average of �10.5% (Table 2).

Correlations between the six variables measured

Of the 15 correlations observed, eight are high (r ≥ 0.50), forexample r = 0.74 for the number of axes determined (NbDetA)and the number of long axes (NbLA); seven are weakly corre-lated (0.18 ≤ r ≤ 0.45), for example r = 0.18 for NbLA and thelength of long axes (L_LA) (Table 3).

Effect of genotype, year and their interaction

ANOVA revealed significant effects of genotype and year for all ofthe variables measured (Table 4).For the genotype effect, the ANOVA carried out on (Table 5):

NbDetA made it possible to distinguish five groups, with ‘Perled’Or’ characterized by the lowest number of axes (15.74) and‘The Fairy’ by the highest number (121.74); NbLA made it pos-sible to distinguish four groups, with ‘Perle d’Or’ characterizedby the lowest number of long axes (7.85) and ‘Blush Noisette’by the highest number (23.25); the number of metamers on thelong axes (NbMet_LA) made it possible to distinguish sixgroups, with ‘Perle d’Or’ characterized by the lowest number ofmetamers (8.19) and ‘Baipome’ by the highest (12.70); L_LAmade it possible to distinguish three groups, with Hw154 charac-terized by the lowest length (15.56 cm) and ‘The Fairy’ by thehighest length (21.32 cm); the branching angle of the cord ofthe long axes in relation to the vertical axis (AngLA/Cord) madeit possible to distinguish four groups, with Hw154 characterizedby the smallest angle (38.70°) and ‘The Fairy’ by the biggest

angle (74.90°); the number of branching orders (NbBrO) made itpossible to distinguish two groups, with ‘Perle d’Or’ character-ized by the lowest order number (3.56) and ‘Blush Noisette’ bythe highest order number (4.50).For the year effect, the ANOVA revealed a significantly greater

effect for 2012 compared to 2011 for all of the variables(Tables 4 and 5), with a difference of +173.1% for NbDetA,+64.1% for NbLA, +2.4% for NbMet_LA, +7.5% for L_LA,+18.2% for AngLA/Cord and +21.4% for NbBrO.The ANOVA revealed a strong genotype 9 year interaction for

the six variables measured (Table 4).This interaction is due to the different groupings of geno-

types according to year, with the exception of NbMet_LA, withfour similar groups in 2011 and 2012. We can therefore distin-guish four groups in 2011 and six groups in 2012 for NbDetA;three groups in 2011 and four groups in 2012 for NbLA; threegroups in 2011 and four groups in 2012 for L_LA; threegroups in 2011 and four groups in 2012 for AngLA/Cord; andthree groups in 2011 and two groups in 2012 for NbBrO(Table 5).This interaction was also due to the different response ampli-

tudes between years, depending on the genotype. The extremeamplitudes are (for 2012 compared to 2011; Fig. 2): for NbDe-tA: +66.4% (‘Old Blush’) and +288.1% (‘The Fairy’); forNbLA: +23.6% (Hw20) and +153.2% (‘The Fairy’); for AngLA/Cord: +3.9% (Hw336) and +64.6% (‘The Fairy’); for NbBrO:+11.0% (Hw336) and +47.4% (Hw154).For the following two variables, the genotype 9 year interac-

tion is due to the opposite response of some genotypes to the2 years, with extreme amplitudes:for NbMet_LA: �7.5% (‘Old Blush’) and +11.2% (Hw20);

for L_LA: �10.1% (‘Perle d’Or’) and +41.7% (‘BlushNoisette’).

Broad-sense heritability

For each variable measured, the phenotypic variance and itscomponents, as well as the heritability, are given in Table 6.The residual variance is low for all of the variables (<6%) andthe genotypic variance accounts for the major part of the pheno-typic variance, with the exception of L_LA (47.63%). The geno-typic variance is therefore very high (>80%) for NbMet_LA andhigh (between 60% and 80%) for NbDetA, NbLA, AngLA/Cordand NbBrO. Consequently, the variance of the genotype 9 yearinteraction is very low (<20%) for NbMet_LA and AngLA/Cordand low (between 20% and 40%) for NbDetA, NbLA andNbBrO.The heritabilities calculated for each of the variables are in

decreasing order: 98% for NbMet_LA, 78% for AngLA/Cord,70% for NbLA and NbBrO, 64% for NbDetA and 48% forL_LA.

DiscussionThe six variables used in this study to describe the architectureof the eight cultivars were chosen by Crespel et al. (2013) forthe following reasons:

1 to represent the different categories of variables that describethe architecture at the plant scale and at the axis scale (mor-phological, topological and geometric variables);

2 to explain the architectural variability observed;3 to be less correlated among themselves.

Table 2: Climatic characterization of both years (denoted 2011 and2012)

Month

Cumulativeirradiance (J/cm2)

Difference1

(%)

Thermal time(°C.d)

Difference(%)2011 2012 2011 2012

March 34289 38988 +13.7 558 533 �4.5April 59073 44552 �24.6 587 528 �10.1May 69738 62812 �9.9 696 641 �7.9June 61948 55074 �11.1 658 641 �2.6Total 225048 201426 �10.5 2500 2343 �6.3

1Difference between the 2 years with respect to 2011.

Table 3: Pearson’s correlations between the six architectural variables

Variables NbDetA NbLA NbMet_LA L_LAAngLA/Cord NbBrO

NbDetA 1.0NbLA 0.37 1.0NbMet_LA 0.78 �0.10 1.0L_LA 0.17 �0.47 0.57 1.0AngLA/Cord

0.49 0.22 0.45 0.13 1.0

NbBrO 0.53 0.54 0.40 0.02 0.53 1.0

NbDetA, the number of axes determined; NbLA, the number of longaxes; NbMet_LA, the number of metamers on the long axes; L_LA, thelength of the long axes; AngLA/Cord, the branching angle of the cord ofthe long axes in relation to the vertical axis; NbBrO, the number ofbranching orders.

414 LAURENT CRESPEL e t a l .

In this study, the correlations observed between variables areon the same order of magnitude as those observed by Crespelet al. (2013). Each of these variables made it possible to signifi-cantly distinguish between cultivars, revealing a strong genotypeeffect.From the climatic point of view, the 2 years, 2011 and 2012,

are characterized by relatively similar thermal times, regardlessof the month, varying from �2.6% to �10.1%. These small dif-ferences can be accounted for by the efficient control of thegreenhouse climate in terms of heating and aeration, limitingvariations in temperature.In contrast, the same is not true for radiation that varies con-

siderably from 1 month to the next, with a sunnier month ofMarch in 2012 (+13.7% compared to 2011) and a much lesssunny month of April in 2012 (�24.6% compared to 2011).However, in our experiment, this period from March to April iscrucial for the growth and branching of the plant and, conse-quently, for its architectural construction.A significant year effect was observed for all of the variables

measured, with higher values for 2012 in comparison with 2011.This effect is particularly evident for the number of axes deter-mined (NbDetA: +173.1%) and for the number of long axes(NbLA: +64.1%). However, the total number of axes determineddepends on the degree of branching of the long axes. In ourstudy, this total number of axes (for all of the plants in the2-year experiment) is positively correlated with the number oflong, order 2 axes (r = 0.42; Pearson’s correlation coefficient,P < 0.05). The difference observed between 2011 and 2012 forthe number of axes is probably due to an increase in the numberof budbreaks on the first axes in 2012, caused by the greateramount of radiation received by the plants in March. This rela-tionship between the number of axes produced and the amountof radiation was largely described by Zieslin and Mor (1990).The year effect was also significant for the length (L_LA), as

well as for the number of metamers on the long axes (NbMe-t_LA), but to a lesser degree than for the number of axes. Thedifference is +7.5% for L_LA and +2.4% for NbMet_LA for2012 compared to 2011. In our study, the axes that develop inMarch then lengthen in April. The difference observed between2011 and 2012 for these two variables is probably due to thesmall amount of radiation received by the plants in April 2012(�24.6%). This observation, which links length or the numberof metamers on the axes to the amount of radiation, is consistentwith that of Bredmose (1998). In his study, a significant increasein the number of metamers (+2.3%) and of the length (+5.8%)of the axes was observed with a decrease of �15.2% of thedaily radiation.The year effect is also significant for the branching angle of

the long axes (AngLA/Cord), with a greater slope of the axes(+18.2%) in 2012, leading to a more plagiotropic growth habitof the plants. Branching is mainly distal in rose (Morel et al.2009). In our study, the difference observed between 2011 and2012 for AngLA/Cord is probably due to the load borne by thelong axes, caused by the higher degree of branching in 2012, aswas shown by Alm�eras et al. (2004) in apricot.This study showed a significant global year effect on rose

bush architecture, although this effect varies strongly dependingon the variable and the genotype considered. In fact, a signifi-cant genotype 9 year effect was shown for all of the variables.This was due to (i) different groupings of genotypes accordingto year, (ii) different response amplitudes between years for allof the variables, depending on the genotype, and (iii) oppositeresponses for some genotypes, particularly ‘Perle d’Or’ forT

able

4:Su

mmaryof

combinedanalysisof

variance

onsixarchitectural

variablesof

eigh

trose

genotypestested

in20

11and20

12

Source

ofvariation

df

Num

berof

determ

ined

axes

Num

berof

long

axes

Num

berof

metam

ers

ofthelong

axes

Lengthof

thelong

axes

Branching

angleof

the

cord

ofthelong

axes

Num

berof

branching

orders

MS

Fprobability

MS

Fprob

ability

MS

Fprob

ability

MS

Fprob

ability

MS

Fprobability

MS

Fprobability

Genotyp

e7

42198.20

**76

3.33

**76

.43

**10

7.08

**38

27.45

**2.77

**Year

115

9448

.93

**38

58.52

**3.77

**10

8.49

**65

29.58

**34

.19

**Genotyp

eby

year

715

201.57

**22

7.80

**1.89

**56

.08

**84

2.89

**0.84

**Residual

197

118.80

13.66

0.57

5.88

91.42

0.13

Total

212

df,degreesof

freedo

m;MS,

meansquares.

**significantly

differentat

P<0.01.

Genotype x year interaction for architecture 415

NbMet_LA and L_LA. These amplitudes are very high for twovariables, NbDetA and NbLA, particularly for ‘The Fairy’,whose number of axes was at least twice as high in 2012 as in2011. This genotype is among those that are the most responsiveto the environment. In contrast, the response amplitude of somecultivars is much lower, as in the case of Hw336 for AngLA/Cord and NbBrO, and Old Blush for the number of axes andtheir length.In our study, the genotypic variance is moderate to high, rang-

ing from 47.6% for L_LA to 97.5% for NbMet_LA. The numberof metamers per axis is a much more stable trait than axislength. Moreover, similar observations were reported by Costeset al. (1992) in apricot, by Ducousso et al. (1995) in commonwalnut and by Raffaele et al. (1998) in Nothofagus in ArgentinePatagonia. In rose, Kawamura et al. (2011) obtained genotypicvariances that were globally lower for the architectural variablesof the flowering axis, ranging from 32.6% to 62.7%. Theseresults were obtained on a segregation progeny resulting fromthe cross between ‘The Fairy’ and a hybrid of R. wichurana.However, ‘The Fairy’ is itself related to R. wichurana by itsfather, ‘Lady Gay’. This relationship between the parents of theprogeny leads to a lower contrast in terms of their architecture,which could partially explain the lower genotypic variancesobtained in this study, as suggested by Segura et al. (2006) inapple. Contrary to the results obtained in our study, the variancesof the genotype 9 year interaction were globally lower and theresidual variances were much higher, ranging from 21.4% to47.7%. This progeny was cultivated outdoors, directly in theground, with three plants per genotype and only three axes perplant. This difference can therefore certainly be partiallyaccounted for by the quality of our experimental set-up that ischaracterized by (i) a high number of repetitions (at least 12plants per genotype), (ii) an environmental homogeneity betweenplants of the same genotype, cultivated in a greenhouse and

sampled from the centre of the culture bench, and (iii) the preci-sion of the measurements obtained using digitalization of all ofthe axes of each plant.The moderate-to-high genotypic variances observed in our

study led to broad-sense heritabilities that were also moderate tohigh (ranging from 48% for L_LA to 98% for NbMet_LA). Theestimation of this heritability could be consolidated by the appli-cation of a broader range of environments by modulating factorsthat influence plant architecture such as light, water restrictionand mechanical stimulation. The heritabilities calculated byKawamura et al. (2011) were very high, ranging from 82% to93%, which is surprising given the low-to-moderate genotypicvariances calculated. A similar study was conducted by Seguraet al. (2006) on an apple progeny produced from grafted plants.For architectural variables comparable to ours, calculated broad-sense heritabilities were lower than those measured in our study(ranging from 6% to 56%). These lower values can be explainedaccording to Segura et al. (2006) by (i) a contrast in the architec-ture of weaker parents, like in the case of Kawamura et al.(2011), (ii) a reduction in the total variability due to the root-stock effect, which we avoided by using cuttings.These results have led us to think that genetic analysis based

on segregation progeny can be seriously considered for the sixtraits selected for our study. The experimental set-up describedin this article will therefore be applied, but the number of plantsper genotype will be reduced to the benefit of the number ofgenotypes, as proposed by Knapp and Bridges (1990) andrecently applied by Pauly et al. (2012) in ray grass. In fact, aswas shown by Kawamura et al. (2011), non-genetic variance ismainly due to intraplant variance rather than the variancebetween plants, and the increase in the number of axes measuredmust therefore be given priority rather than the number of plantsper genotype. This approach will be applied to two connectedsegregation progeny resulting from a cross of parents with con-

Table 5: Means of six architectural variables of eight rose genotypes measured tested in 2011 and 2012

Genotypes

Number of determined axes Number of long axesNumber of metamers of the long

axes

2011 2012 Mean 2011 2012 Mean 2011 2012 Mean

Baipome 55.42e1 145.27f 98.00e 12.97bc 24.09c 18.41 cd 12.72d 12.69e 12.70fBlush Noisette 24.43 cd 66.10d 47.92c 18.50d 27.00c 23.25e 8.60ab 8.89ab 8.74abHw154 14.35ab 29.35ab 23.67a 10.30b 14.17ab 12.52b 8.72ab 9.36bc 9.00bHw20 22.02bc 59.35 cd 41.26bc 14.77 cd 18.25b 16.81c 9.20b 10.23c 9.65cHw336 32.02d 107.43e 68.19d 11.97bc 24.25c 17.93 cd 10.93c 11.53d 11.19dOld Blush 24.55 cd 40.85bc 34.44b 18.10d 24.09c 21.26de 9.16ab 8.47ab 8.85abPerle d’Or 6.68a 21.10a 15.74a 5.97a 9.09a 7.85a 8.23a 8.15a 8.19aThe Fairy 52.22e 202.68 g 121.74e 13.17bc 33.34d 22.63e 11.85 cd 12.23de 12.01eMean 31.73 86.67 13.58 22.28 9.95 10.19

Genotypes

Length of the long axesBranching angle of the cord of the long

axes Number of branching orders

2011 2012 Mean 2011 2012 Mean 2011 2012 Mean

Baipome 20.59d 19.87 cd 20.30 cd 61.13d 67.34 cd 64.58d 3.95bc 4.70ab 4.33bBlush Noisette 14.99a 21.24d 18.14bc 60.62d 68.14d 65.06d 4.20c 4.70ab 4.50bHw154 16.12abc 14.79a 15.56a 35.46a 41.19a 38.70a 3.02a 4.45ab 3.70aHw20 15.82ab 16.96ab 16.36ab 43.29ab 56.24bc 49.73bc 3.89b 4.54ab 4.22bHw336 18.12abcd 20.09 cd 19.02c 56.07 cd 58.26bcd 57.73 cd 4.09bc 4.54ab 4.33bOld Blush 19.26 cd 20.31d 19.75 cd 48.43bc 61.66 cd 55.00c 3.95bc 4.45ab 4.22bPerle d’Or 19.59d 17.61bc 18.74c 43.52ab 47.58ab 46.01ab 3.02a 4.12a 3.56aThe Fairy 19.01bcd 24.14e 21.32d 57.66 cd 94.91e 74.90e 3.89b 4.95b 4.41bMean 18.04 19.40 51.21 62.60 3.79 4.60

1Multiple comparison test; the means followed by different letters are significantly different (Newman–Keuls test, P < 0.05).

416 LAURENT CRESPEL e t a l .

trasting architecture (Crespel et al. 2013): ‘The Fairy’ with ‘OldBlush’ and ‘The Fairy’ with ‘Perle d’Or’.Plant architecture and, subsequently, plant shape can be con-

trolled through the genetic pathway via plant breeding and culti-vation techniques such as the modification of the light spectrum

using selective films (Rajapakse and Kelly 1993). The six vari-ables used in this study demonstrated their relevance for charac-terizing architecture in rose (Crespel et al. 2013). Most of thesevariables are heritable and can therefore be used to characterizegenitors, like UPOV-type descriptors, but with a greater discrim-

Table 6: Variance components and broad-sense heritability estimates on six architectural variables of eight rose genotypes tested in 2011 and 2012

Variance componentand heritability

Number ofdetermined axes

Number oflong axes

Number of metamersof the long axes

Length of thelong axes

Branching angle of thecord of the long axes

Number ofbranching orders

Genotypic 1124.86 (63.98) 22.31 (70.16) 3.11 (97.53) 2.13 (47.63) 124.36 (77.98) 0.08 (69.68)Genotypic by year 628.45 (35.74) 8.92 (28.05) 0.06 (1.73) 2.09 (46.88) 31.31 (19.63) 0.03 (25.63)Residual variance 4.95 (0.28) 0.57 (1.79) 0.02 (0.75) 0.25 (5.49) 3.81 (2.39) 0.01 (4.69)Phenotypic variance 1758.26 31.81 3.18 4.46 159.48 0.12Heritability (%) 64 70 98 48 78 70

Values in brackets are percentages over the total phenotypic variance.

(a) 3.0

2.5

2.0

1.5

1.0

0.5

0.0TheFairy

Hw336 TheFairy

Hw336Perled’Or

Perled’Or

BlushNoisette

BlushNoisette

Genotypes Genotypes

Hw20 Hw20Baipome BaipomeHw154 Hw154OldBlush

TheFairy

Hw336 Perled’Or

BlushNoisette

Genotypes Genotypes

Hw20 BaipomeHw154 OldBlush

TheFairy

Hw336Perled’Or

BlushNoisette

Genotypes

Hw20 BaipomeHw154OldBlush

TheFairy

Hw336Perled’Or

BlushNoisette

Genotypes

Hw20BaipomeHw154 OldBlush

TheFairy

Hw336 Perled’Or

BlushNoisette

Hw20 Baipome Hw154OldBlush

OldBlush

3.0

2.5

2.0

1.5

1.0

0.5

0.0

NbD

etA

(20

11–2

012)

'201

1

NbL

A (

2011

–201

2)'2

011

NbM

et_L

A (

2011

–201

2)'2

011

Ang

LA/C

ord

(201

1–20

12)'2

011

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

–0.1

–0.2

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

(b)

(c)

L_LA

(20

11–2

012)

'201

1

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

–0.1

–0.2

(d)

(e)

NbB

rO (

2011

–201

2)'2

011

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

(f)

Fig. 2: Differences obtained for the eight genotypes and the six variables measured (for 2012 compared to 2011): NbDetA (a), NbLA (b), NbMet_LA(c), L_LA (d), AngLA/Cord (e) and NbBrO (f)

Genotype x year interaction for architecture 417

inatory power (Crespel et al. 2013). They can therefore be usedto choose genitors for breeding programmes designed to improveplant shape. Moreover, a significant environment effect wasobserved for these six variables, making it possible to considerthe viability of controlling plant shape by acting on certain envi-ronmental factors, as demonstrated by Demotes-Mainard et al.(2013) in rose, for example imposing temporary restrictionphases in the case of water supply, and low irradiance phases inthe case of the amount of radiation.These results also showed a strong genotype 9 environment

interaction for some variables and some genotypes that shouldbe taken into account by the horticulturist in view of more effec-tively controlling plant shape.

AcknowledgementsThe authors would like to thank R�emi Gardet and Jacky Granger of theIRHS experimental station, Herv�e Autret, technician, and CedricDumont, intern, for their contribution to experiments carried out withinthe framework of this study, as well as Gail Wagman for the Englishtranslation of this article.

ReferencesAbidi, F., T. Girault, O. Douillet, G. Guillemain, G. Sintes, M. Laffaire,H. Ben Ahmed, S. Smiti, L. Huch�e-Th�elier, and N. Leduc, 2012: Bluelight effects on rose photosynthesis and photomorphogenesis. PlantBiol 15, 67—74.

Alm�eras, T., E. Costes, and J. C. Salles, 2004: Identification of biome-chanical factors involved in stem shape variability between apricot treevarieties. Ann. Bot. 93, 455—468.

Boumaza, R., S. Demotes-Mainard, L. Huch�e-Th�elier, and V. Gu�erin,2009: Visual characterization of the esthetic quality of the rosebush.J. Sens Stud. 24, 774—796.

Bredmose, N. B., 1998: Growth flowering and postharvest performanceof single stemmed rose (Rosa hybrida L.) in response to light quantumintegral and plant population density. J. Am. Soc. Hortic. Sci. 123,569—576.

Costes, E., P. de Reffye, J. Lichou, Y. Gu�edon, A. Audubert, and M.Jay, 1992: Stochastic modelling of apricot growth units and branching.Acta Hort. 313, 89—98.

Crespel, L., M. Chirollet, C. E. Durel, D. Zhang, J. Meynet, andS. Gudin, 2002: Mapping of qualitative and quantitative phenotypictraits in Rosa using AFLP markers. Theor. Appl. Genet. 105, 1207—1214.

Crespel, L., M. Sigogne, N. Don�es, D. Relion, and P. Morel, 2013: Iden-tification of relevant morphological, topological and geometrical vari-ables to characterize the architecture of rose bushes in relation toshape. Euphytica. doi:10.1007/s10681-013-0902-6.

Demotes-Mainard, S., L. Huche-Thelier, P. Morel, R. Boumaza, V. Guerin,and S. Sakr, 2013: Temporary water restriction or light intensity limita-tion promotes branching in rose bush. Sci. Hortic. 150, 432—440.

Don�es, N., B. Adam, and H. Sinoquet, 2006: PiafDigit software. Avail-able at: http://www1.clermont.inra.fr/piaf/fr/telechargement/telecharger.php. (last accessed on April 25, 2011).

Ducousso, I., S. Sabatier, D. Barth�el�emy, and E. Germain, 1995: Com-paraison de quelques caract�eristiques morphologiques des pousses an-

nuelles et des branches de la cime de sept vari�et�es de noyer commun,Juglans regia L. (Juglandaceae). In J. Bouchon (ed), Architecture desarbres fruitiers et forestiers, 91—108, Les colloques n°74. INRA,Paris.

Godin, C., and Y. Caraglio, 1998: A multiscale model of plant topologi-cal structures. J. Theor. Biol. 191, 1—46.

Godin, C., Y. Gu�edon, and E. Costes, 1999: Exploration of a plant archi-tecture database with the AMAPmod software illustrated on an appletree hybrid family. Agronomie 19, 163—184.

Huch�e-Th�elier, L., R. Boumaza, S. Demotes-Mainard, A. Canet, R. Sy-moneaux, O. Douillet, and V. Guerin, 2011: Nitrogen deficiencyincreases basal branching and modifies the visual quality of the rosebushes. Sci. Hortic. 130, 325—334.

Kawamura, K., L. Hibrand-Saint Oyant, L. Crespel, T. Thouroude, D.Lalanne, and F. Foucher, 2011: Quantitative trait loci for floweringtime and inflorescence architecture in rose. Theor. Appl. Genet. 122,661—675.

Knapp, S. J., and W. C. Bridges, 1990: Using molecular markers to esti-mate quantitative trait locus parameters: power and genetic variancesfor unreplicated and replicated progeny. Genetics 126, 769—777.

Morel, P., 2001: Growth control of Hydrangea macrophylla throughwater restriction. Acta Hort. 548, 51—58.

Morel, P., G. Galopin, and N. Don�es, 2009: Using architectural analysisto compare the shape of two hybrid tea rose genotypes. Sci. Hortic.120, 391—398.

Morel, P., L. Crespel, G. Galopin, and B. Moulia, 2012: Effect ofmechanical stimulation on the growth and branching of garden rose.Sci. Hortic. 135, 59—64.

Nanson, A., 1970: L’h�eritabilit�e et le gain d’origine g�en�etique dans quel-ques types d’exp�eriences. Silvae Genet. 19, 113—121.

Pauly, L., S. Flajoulot, J. Garon, B. Julier, V. B�eguier, and P. Barre,2012: Detection of favourable alleles for plant height and crown rusttolerance in three connected populations of perennial ryegrass (Loliumperenne L.). Theor. Appl. Genet. 124, 1139—1153.

Raffaele, E., J. Puntieri, P. Martinez, J. Marino, C. Brion, andD. Barth�el�emy, 1998: Comparative morphology of annual shoots inseedlings of five Nothofagus species from Argentinian Patagonia.C. R. Acad. Sci. 321, 305—311.

Rajapakse, N. C., and J. W. Kelly, 1993: Influence of spectral filters ongrowth and postharvest quality of potted miniature roses. Sci. Hortic.56, 245—255.

Rodriguez, M., D. Rau, R. Papa, and G. Attene, 2008: Genotype byenvironment interactions in barley (Hordeum vulgare L.): differentresponses of landraces, recombinant inbred lines and varieties to Medi-terranean environment. Euphytica 163, 231—247.

Segura, V., C. Cilas, F. Laurens, and E. Costes, 2006: Phenotyping prog-enies for complex architectural traits: a strategy for 1-year-old appletrees (Malus x domestica Borkh.) Tree Genet. Genomes 2, 140—151.

Shimelis, H., and R. Shiringani, 2010: Variance components and herit-abilities of yield and agronomic traits among cowpea genotypes. Eu-phytica 176, 383—389.

White, J., 1979: The plant as a metapopulation. Ann. Rev. Ecol. Syst.10, 109—145.

Wu, X., X. Chang, and R. Jing, 2012: Genetic insight into yield-associ-ated traits of wheat grown in multiple rain-fed environments. PLoSONE 7, e31249.

Zieslin, N., and Y. Mor, 1990: Light on roses. A review. Sci. Hortic. 43,1—14.

418 LAURENT CRESPEL e t a l .