Plant Science 135 (1998) 87–101

Pinpointing towards improved transformation and regenerationof cassava (Manihot esculenta Crantz)

Tichafa R.I. Munyikwa, Krit C.J.M. Raemakers, Marianne Schreuder, Rosan Kok,Marja Schippers, Evert Jacobsen, Richard G.F. Visser *

Graduate School Experimental Plant Sciences, Department of Plant Breeding, Wageningen Agricultural Uni6ersity, PO Box 386,6700 AJ, Wageningen, The Netherlands

Received 30 December 1997; received in revised form 19 March 1998; accepted 9 April 1998

Abstract

Friable embryogenic callus (FEC) of the cassava genotype 60444 was transformed by particle bombardment withDNA from the plasmid constructs pHB1 and pJIT100. Both plasmids contained the luciferase (luc) marker geneunder the control of the CaMV 35S promoter. In addition pJIT100 had the CaMV35S driven phosphinothricin acetyltransferase (pat) gene, while pHB1 contained the cassava cDNA coding for the small subunit of ADP glucosepyrophosphorylase (AGPase B) in antisense orientation under the control of a double CaMV35S promoter. A totalof 2 weeks after bombardment, luciferase (LUC) positive FEC units (spots) were isolated and subcultured separatelyfor further proliferation. A total of 4 weeks later, those cultures having at least four positive LUC spots weresubjected to three different selection regimes namely: stringent LUC selection, non stringent LUC selection andcombined LUC/phosphinothricin (PPT) selection. A total of 16 weeks after bombardment, stringent LUC selectiongave rise to cultures in which 92% of the FEC units were LUC positive. Within the same time period non stringentlyLUC selected cultures and LUC/PPT selection had only 1 and 41% of the units that were LUC positive, respectively.The number of LUC positive mature somatic embryos formed was similar to the percentage of LUC positive FECunits, within a culture, found with each selection method. Stringent LUC selection enabled transgenic plants to beproduced in 28–36 weeks compared to 32–41 weeks for LUC/PPT selection and 53–78 weeks for non stringent LUCselection. This indicates that stringent selection is a more efficient method for obtaining transgenic cassava plants.Southern blot analysis of transgenic cassava plants revealed that they had between one to seven copies of the pHB1and pJIT100 construct. The production of the first cassava plants carrying an agronomically important trait affecting

Abbre6iations: 2,4-D, 2,4-dichlorophenoxyacetic acid; AGPase B, ADP glucose pyrophosphorylase small subunit; BAP, 6-benzyl-aminopurine; CaMV, cauliflower mosaic virus; FEC, friable embryogenic callus; LUC, luciferase; NAA, a-naphtaleneacetic acid;PAT, phosphinothricin acetyl transferase; PPT, phosphinothricin.

* Corresponding author.

0168-9452/98/$19.00 © 1998 Elsevier Science Ireland Ltd. All rights reserved.

PII S0168-9452(98)00074-0

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–10188

starch biosynthesis is reported. Expression of the antisense AGPase B gene resulted in cassava plants within which theAGPase mRNA steady state levels was greatly decreased or even absent. The plants with no AGPase mRNAexpression also had extremely low levels of starch, compared to control plants, as shown by iodine staining of in vitrothickened stems. In plants exhibiting the highest AGPase B antisense effect, starch formation was limited only to theepidermal layer of in vitro thickened stems. © 1998 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Cassava; Manihot esculenta ; ADP glucose pyrophosphorylase; Luciferase; Friable embryogenic callus;Starch

1. Introduction

The cultivation of cassava, a major crop in thetropics, is beset by many problems including dis-eases, pests, a high cyanide content of the roots anda low nutritional quality and commercial value ofthe starch. Improvement of the cassava germplasmby traditional breeding methods has been ham-pered by the non-availability of necessary genes inthe germplasm, the alloploidy nature of the plant,and the low fertility of cassava, amongst otherfactors [1]. The newer techniques of genetic trans-formation, whilst offering greater hope for im-provement of cassava, have in the past beenhindered by the lack of a reproducible transforma-tion and regeneration system for this crop [2]. Thebreakthrough came with the development of sys-tems of regeneration based on friable embryogeniccallus [3] or on adventitious shoot formation [4].This enabled certain cassava genotypes to be trans-formed using particle gun delivery of DNA intofriable embryogenic callus (FEC) cultures [5,6] aswell as electroporation of FEC derived protoplasts(Raemakers, unpublished results). In addition Li etal. [4] have described a method that allows transfor-mation of cassava with Agrobacterium tumefaciensinfection of cotyledon explants cultured for adven-titious shoot formation.

The methods used for selecting transgenic tissue,in FEC transformation, can basically be dividedinto procedures involving: (I) chemical selectionwith aminoglycosides such as paromomycin [5]; (II)non invasive selection with the firefly luciferasegene [6]; and (III) a combination of chemicalselection with phosphinothricin and use of theluciferase marker gene [7]. While these methodshave established the route by which cassava will be

transformed in future they are not, however, opti-mally efficient for universal application. The maindisadvantages of the methods being that they aretime and labour consuming and inefficient withrespect to selection and/or regeneration.

The FEC cultures used in the above mentionedstudies were derived from the African cultivarcalled 60444. This cultivar was developed in abreeding programme led by Beck in the 1950s usingmaterial originating from Ghana and East Africaduring the same period (Dr R. Dixon, IITA Nige-ria, personal communication). Although FEC cul-tures have now been produced in about 13 differentgenotypes [3,8], (Raemakers et al., unpublishedresults) much work still remains to be done toproduce FEC from the important cassava cultivarsworld-wide.

For each of the methods described above it tookfrom 8 months to more than a year to producetransgenic plants after the initial transformationprocedure. This invariably implies that a lot oflabour goes into maintaining the cultures on freshmedia. The longer the time spent in tissue culture,the greater too is the expense of the procedure.Chemical selection with aminoglycosides such asparomomycin seemed to have a negative effect onthe ability of transgenic FEC to regenerate intoplants [9]. Use of the luciferase gene for selectionof transformed tissue remained labour intensive,since no selective advantage was given to thetransgenic tissue over the non transgenic tissue. Theuse of the selection agent phosphinothricin (PPT)while not being detrimental to the ability of FECto form plants, however, led to escapes and addedextra steps to the whole procedure of producingtransgenic plants [7]. There was a need to obtain aselection system which combines the advantages ofselection using paramomycin (no escapes) and of

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101 89

selection using luciferase (no negative effect on theability to regenerate into plants).

All the transgenic plants derived from the aforementioned experiments carry marker genes such asb-glucuronidase (gus), luciferase (luc), phos-phinothricin acetyl transferase (pat), hygromycinphosphotransferase (hpt) or neomycin phospho-transferase II (nptII) and no agronomically inter-esting genes which confer new traits to the cassavagermplasm. The cassava gene encoding the smallsubunit of ADP glucose pyrophosphorylase hasbeen cloned [10]. Use of this gene in antisenseorientation in potato resulted in plants havingreduced starch and elevated sugar levels in theirtubers [10]. If the effect in cassava was similar theseplants could be interesting for further developmentof cassava usage. This paper describes improve-ments in the procedure of cassava transformationusing only luciferase. Some of the plants producedcarried in addition to the luciferase selectionmarker gene, the cassava AGPase small subunitgene in antisense orientation,

2. Materials and methods

2.1. Plant material and media

A 2 year old FEC culture from the genotype60444 (International Institute of Tropical Agricul-ture) was used in the experiments and was kindlyprovided by Dr N. Taylor, University of Bath, UK.The following media were used: solid FEC prolif-eration medium (GD2) which consisted ofGresshoff and Doy salts and vitamins [11], 8 g l−1

micro agar (Duchefa), 20 g l−1 sucrose and 10 mgl−1 picloram, pH adjusted to 5.7 before autoclav-ing; liquid FEC proliferation medium (SH6) madeup of Schenk and Hildebrandt salts and vitamins[12], 60 g l−1 sucrose and 10 mg l−1 picloram, pHadjusted to 5.7; liquid (SH6) to solid (GD2) transfermedium (GD4) is identical to GD2 with 40 g l−1

sucrose added; maturation media which consistedof Murashige and Skoog salts and vitamins [13], 8g l−1 micro agar and 20 g l−1 sucrose (MS2)supplemented with 1 mg l−1 NAA or 1 mg l−1

picloram; first cycle secondary somatic embryogen-esis medium consisted of solid (8 g l−1 micro agar)

MS2 plus 8 mg l−1 2,4D; second cycle secondaryembryogenesis medium was made up of liquid MS2and 10 mg l−1 NAA; germination medium wasmade up of solid (8 g l−1 micro agar) MS2 plus 1mg l−1 BAP; rooting medium: the shoots wererooted on solid (8 g l−1 micro agar) MS2. Thetemperature in the growth chamber was 30°C, thephotoperiod 12 h and the irradiance 40 mmol m−2

s−1.

2.2. Constructs and particle bombardment

Standard molecular biology techniques wereused in DNA manipulations [14]. The constructsused were pHB1 (8.0-kb) and pJIT100 (6.7-kb). Theplasmid pJIT 100 contained the genes coding forluciferase (luc) and phosphinothricin acetyl trans-ferase (pat) both under the control of the CaMV35S promoter and terminated by the CaMVpolyadenylation region [15]. This construct waskindly provided by J. Guerineau of the John InnesResearch Institute Norwich, UK. Plasmid pHB1(Fig. 1) was made by introducing a 1.3-kb fragmentof the cassava AGPase B cDNA (cut with BamHI/HindIII from plasmid pB45-1) in antisense orienta-tion (HindIII 5% and BamHI 3%) between the CaMV35S promoter and the luc gene of pJIT 100. Bothconstructs contained the ampicillin resistance genefor selection of bacteria carrying these plasmids.For particle bombardment 20 mg of plasmid DNAwas coated on to 10 mg of gold particles (1.6 mm).The method of coating the gold particles andbombardment of the FEC cultures was as describedby Raemakers et al. [6]. Plasmids pHB1 andpJIT100 were used independently for the transfor-mation experiments. The main interest, however,was in the plants transformed with pHB1 becausethese contained an agronomically important gene(cassava AGPase B) unlike pJIT100 derived plantswith only the marker/selection genes.

2.3. Luciferase assays

In order to select transgenic tissue, the fireflyluciferase gene (luc) was used as a reporter gene.The luciferase gene encodes the enzyme luciferase,which converts the substrate luciferin into oxyluci-ferin in the presence of ATP, O2 and Mg2+ emitting

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–10190

Fig. 1. Construct pHB1 carrying the cassava antisense AGPase B gene. A 1.6-kb BamHI and HindIII fragment of the cassavaAGPase B gene from pB45-1 [10] was cloned in antisense orientation into the HindIII and BamHI site of pJIT100 [14] behind aCaMV35S promoter. Selection of transgenic tissue was based on the light emitting activity of the luciferase enzyme, encoded by theluciferase (LUC) gene.

light [16]. Petri-dishes (Greiner, 94/16 with cam)with FEC cultures, mature somatic embryos orplants were sprayed (Douglas perfume dispenser)with a solution of 0.25 mg ml−1 luciferin (Promega,E160) in water. Luciferase (LUC) activity wasdetermined for 1 min by measuring the amount ofphotons emitted by the explants, using VIM inten-sified CD camera and an Argus-50 photon countingimage processor (Hamamatsu Photonic Systems) ata maximum sensitivity. The Petri-dishes were posi-tioned 70 cm below the camera in the case ofnon-stringent LUC selection, LUC/PPT selectionand selection of LUC positive mature somaticembryos and plants. The Petri-dishes were posi-tioned 2–10 cm below the camera in the case ofmicroscopically positioning of the LUC positiveclumps in stringent selection and in the case ofmicroscopic evaluation of the percentage LUCpositive FEC units. Superimposition of the elec-tronic image of LUC activity with the normalimage of the explant enabled the LUC positivetissue to be identified and isolated.

2.4. Selection of transgenic tissue

Prior to bombardment new suspension cultures

were initiated by transferring 0.1 g FEC from solidproliferation (GD2) into Erlenmeyer flasks (300 ml)filled with 75 ml of liquid proliferation medium(SH6). Twice a week half of the medium wasreplaced by fresh SH6 medium. The flasks werecultured on an orbital shaker (LAB-line Instru-ments Inc. Model 3519) at 120 rpm. After 2 weeksthe suspensions were collected and 0.05 g wasspread out in the centre of a Petri-dish (Greiner,94/16 mm with cam) on solid (8 g l−1 micro agar)SH6 medium. In total 50 Petri-dishes were bom-barded with pHB1 and 40 Petri-dishes with theconstruct pJIT100. The bombarded FEC cultureswere then placed in 250 ml plastic pots (Greiner;diameter 6.8 cm) containing 50 ml of liquid SH6medium. After 2 weeks the cultures were collectedon solid GD4 medium and assayed for LUCactivity. Putatively transgenic tissue was identifiedas LUC positive spots. Each spot plus the sur-rounding tissue (0.5–1 cm diameter around theLUC positive spot) was subcultured separately andgrown for a further 4 weeks in 250 ml potscontaining 50 ml SH6 medium. The plastic potswere cultured on an orbital shaker (LAB-lineInstruments Inc. Model 3519) at 120 rpm and twicea week half of the medium was refreshed. Those

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101 91

cultures in which the number of LUCpositive spots had increased to more than fourwere used for the following selection regimes:

2.4.1. Non stringent luciferase selectionNon stringent selection was applied on 16 cul-

tures bombarded with DNA from pHB1 and tencultures bombarded with pJIT100. For this selec-tion procedure LUC positive tissue plus the sur-rounding tissue (0.5–1 cm diameter around theLUC spot) was spread on GD2 medium. Thisselection process was repeated every 2 weeks.Although it was possible to locate precisely LUCactivity, the tissue in a radius of 0.5–1 cm aroundthe LUC spot was also transferred to avoid lossof transgenic tissue. After 10 weeks (16 weeksafter bombardment) the total LUC activity of thecultures was measured. Furthermore, FEC wasspread out as fine as possible to determine thefraction of individual FEC units, which wereLUC positive.

2.4.2. Stringent luciferase selection (pinpointing)Stringent LUC selection was applied on eight

cultures bombarded with pHB1. For this selectionprocedure LUC positive tissue plus the surround-ing tissue (0.5–1 cm diameter around the LUCspot) was spread on GD2 medium (discarding thebulk of LUC negative FEC). A total of 2 weekslater LUC positive tissue plus the surrounding(0.5–1 cm diameter) was again selected and di-vided as finely as possible in individual FEC units.These selected embryogenic units were grown fora further 2 weeks on GD2 medium and thenassayed for LUC activity. Because the FEC hadbeen spread out finely on the Petri-dishes, smallclumps, derived from individual LUC positiveFEC units in each culture, could be pinpointedmicroscopically, isolated and collected together onfresh GD2 medium (positive selection). The cul-tures were grown for a further 6 weeks on GD2medium and assayed every 2 weeks. A total of 16weeks after bombardment total LUC activity andthe proportion of LUC positive FEC units wasdetermined.

2.4.3. Combined luciferase and PPT selectionCombined luciferase/PPT selection was applied

on ten FEC cultures bombarded with pJIT100.For this selection procedure LUC positive tissueplus the surrounding tissue (0.5–1 cm diameteraround the LUC spot) were spread on solid GD2medium containing 20 mg l−1 PPT. Every 2weeks this selection process was repeated untilclumps of PPT resistant tissue became visible(after 6–8 weeks of combined LUC/PPT selec-tion) which than were transferred to freshmedium. After 10 weeks of combined LUC/PPTselection the proportion of LUC positive tissue ineach culture was determined.

2.5. Maturation of FEC

A total of two sets of experiments were con-ducted to determine the best conditions of matu-ration of the transgenic and LUC negative FEC.In the first set of experiments 0.1 g of FEC,derived 16 weeks after bombardment from allthree selection regimes, was subcultured in petri-dishes (Greiner, 94/16 mm with cam) with 20 mlof solid (8 g l−1 micro agar) MS2 medium supple-mented with 1 mg l−1 NAA [9] or 1 mg l−1

picloram [17] and 0, 0.01, 0.1, 1 and 10 mg l−1

ABA. The cultures were transferred after 4 weeksto medium without ABA.

In the second set of experiments, FEC derivedfrom non-stringent selection was further culturedon solid GD2 medium and assayed for LUCactivity for 10 weeks with the same non stringentselection method. Subsequently, highly LUC ac-tive FEC was transferred to liquid SH6 mediumfor 2 weeks, without refreshing after which, it wascultured on solid MS2 medium plus 1 mg l−1

picloram for maturation.All FEC cultures were transferred to fresh mat-

uration medium every 2 weeks until mature so-matic embryos appeared (up to 12 weeks). Maturesomatic embryos are defined as structures with adistinct hypocotyl and green cotyledons. The ma-ture somatic embryos were harvested and assayedfor LUC activity.

2.6. Germination of mature somatic embryos

LUC positive mature somatic embryos werefirst cut in small pieces of 4–25 mm2 and culturedfor first cycle secondary somatic embryogenesis in

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–10192

Petri-dishes (Greiner, 94/16 mm with cam) contain-ing 20 ml of solid (8 g l−1 micro agar) MS2 plus8 mg l−1 2,4 D medium. After 2 weeks tissuecontaining globular somatic embryos were har-vested and cultured for another 2 weeks in Erlen-meyer flasks (300 ml) filled with 75 ml of liquid MS2plus 10 mg l−1 NAA medium for second cyclesecondary somatic embryogenesis. The flasks werecultured on an orbital shaker (LAB-line Instru-ments Inc. Model 3519) at 120 rpm. The resultingmature somatic embryos were desiccated and thencultured for germination into plants on MS2 plus1 mg l−1 BA as described by Raemakers et al. [6].A germinated plant is defined as a structure witha distinct stem and normal looking leaves whichcould be multiplied by single bud cuttings.

2.7. Southern and northern analysis

DNA and RNA from leaves of LUC positivecassava leaves were isolated, blotted and assayedusing standard techniques [14]. The southern blotwas hybridized with a 32P dCTP-labelled luc genewhile the northern blot was hybridized with a 32PdCTP-labelled cassava AGPase B gene. Identicalamounts of RNA were applied in the northern blotas judged by concentration measurement andrRNA band identification verification.

2.8. Iodine staining of in 6itro thickened stems

Thickened stems were induced by growing trans-genic and non-transgenic cassava plants in vitro for4 weeks on solid MS medium supplemented with80 g l−1 sucrose as described by Salehuzzaman etal. [18]. The presence or absence of starch wasvisualised by iodine staining of cross sections of thein vitro thickened stems with Lugol’s solution(I2:KI). The stained stem sections were visualisedmicroscopically.

3. Results

3.1. Effect of selection regime on the proportionof luciferase positi6e tissue

A total of 2 weeks after bombardment LUC

assays revealed a mean number of one LUCpositive spot per pHB1 culture and two LUCpositive spots per pJIT100 bombarded culture.Each spot was cultured separately and grown forthe next 4 weeks in liquid SH6 medium. At the endof that period 26 of the 50 pHB1 and 46 of the 80pJIT100 LUC positive spots, had disappeared and,24 pHB1 and 18 pJIT100 cultures had more thanfour LUC positive spots (indicating continuedgrowth of the transgenic tissue). The pHB1 bom-barded cultures were used in the experiments withthe stringent and also the non stringent LUCselection regimes and 20 pJIT100 bombarded cul-tures were equally divided over the combinedLUC/PPT selection procedure and the non strin-gent LUC selection procedure.

After 10 weeks of selection (16 weeks afterbombardment) dramatic differences were observedin the LUC activities of the cultures. Stringentlyselected cultures had 50–100 times higher LUCactivity than non stringently selected cultures andseven times higher LUC activity than cultures fromthe combined LUC and PPT selection (Table 1).Microscopic examination of the cultures revealedthat over 90% of the FEC units in a stringentlyselected culture were LUC positive (Fig. 2e and f).This is considerably higher than the 45% LUCpositive FEC units obtained in LUC/PPT selection.Non stringent selection gave rise to cultures withless than 1% LUC positive FEC units. The processand results of the stringent LUC selection regimeare shown in Fig. 2a–h.

3.2. Mature somatic embryo formation andregeneration of transgenic cassa6a shoots

Regeneration of plants from FEC starts with theformation of mature somatic embryos. The per-centage of LUC positive mature somatic embryosafter using the different selection regimes wasdirectly proportional to the percentage of LUCpositive FEC units (Table 1). A total of 0.1 g FECcultured on maturation medium supplemented with1 mg l−1 NAA or 1 mg l−1 picloram yielded 66 and124 mature somatic embryos, respectively while theaddition of ABA did not have

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101 93

Tab

le1

The

effe

ctof

the

sele

ctio

nre

gim

e(n

onst

ring

ent

and

stri

ngen

tlu

cife

rase

sele

ctio

n,an

dco

mbi

ned

luci

fera

sean

dph

osph

inot

hric

inba

sed

sele

ctio

n)on

the

effic

ienc

yof

prod

ucin

gtr

ansg

enic

cass

ava

plan

tsby

part

icle

bom

bard

men

tw

ith

the

cons

truc

tspH

B1

and

pJIT

100

Tim

eto

form

tran

sgen

icM

atur

eso

mat

icem

bryo

sT

otal

LU

Cac

tivi

ty/d

ish

(pho

tons

/min

)af

-Se

lect

ion

regi

me

Num

ber

ofcu

l-%

LU

Cpo

siti

veF

EC

unit

s/di

shte

r10

wee

ksof

sele

ctio

ntu

res

appl

ied

plan

tsin

wee

ks

%L

UC

posi

-N

umbe

rte

sted

tive

1612

50B

153

–78

0.04

×10

6N

onst

ring

ent

B1

LU

C(p

HB

1)90

28–

3692

1.91

×10

6St

ring

ent

LU

C8

432

(pH

B1)

4110

638

4532

–41

0.29

×10

6L

UC

/pho

s-ph

inot

hric

in(p

JIT

100)

B1

845

B1

53–

78N

onst

ring

ent

100.

02×

106

LU

C(p

JIT

100)

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–10194

Fig. 2. Stringent luciferase selection for the production oftransgenic cassava friable embryogenic callus (FEC) andplants. Panel a: a friable embryogenic culture (BH6) bom-barded with the construct pHB1, expressing luciferase, 2 weeksafter bombardment (diameter Petri-dish 9.4 cm). (Panel b)close up (magnification ×20) of luciferase positive FEC (8weeks after bombardment) that was selected from the sur-rounding non transgenic FEC and spread out finely on solidGD2 for further culture. (Panel c) view (magnification ×40)of stringently selected and well spread out FEC units undernormal light, 16 weeks after bombardment. The precise pin-pointing of LUC positive FEC units was possible, as shown inpanel d. Using stringent selection it was possible to obtaincultures with \95% LUC positive FEC units (magnification×40), 16 weeks after bombardment as shown in panel e(normal view) and panel f (viewed in the dark, after addingluciferin). Following mature somatic embryo formation andgermination stringently selected cultures gave rise to luciferasepositive plants (diameter Petri-dish 9.4 cm) shown under nor-mal light (panel g) and expressing luciferase (panel h).

a positive effect on this number (results notshown).

High numbers of mature somatic embryoswere obtained with FEC cultures, derived fromnon stringent selected cultures, that were grownfirst in liquid SH6 medium without refreshingfor 2 weeks and then transferred to maturationmedium. The FEC units became primed formaturation i.e. instead of initiating new FECunits the existing units became bigger. Theseprimed cultures gave rise to high numbers ofmature somatic embryos.

About half of the mature somatic embryoshad a morphology similar to that observed formature somatic embryos derived from secondarysomatic embryogenesis [19]. The mature somaticembryos possessed two cotyledons that were inmost cases fused together. Malformations thatwere observed included mature somatic embryoswith pinnulate, oval, and serrated cotyledons. Inmany cases a Petri-dish (with FEC derived fromone transgenic line) contained predominantlymature somatic embryos from one aberrant typewhile another Petri-dish, from the same trans-genic line and on the same medium containednormal looking mature somatic embryos.

3.3. Efficiency of production of transgenic plants

In total 50 and 40 Petri-dishes carrying 0.05 gof FEC were bombarded, respectively withpHB1 and pJIT100. In total 24 independentpHB1 and 20 independent pJIT100 cultures weredivided over the three selection regimes. Thetime required to obtain transgenic plants wasdependent on the amount of transgenic FECnecessary for mature somatic embryo formationand the time taken to obtain transgenic plantsfrom the FEC cultures. Using stringent LUCand LUC/PPT selection it was possible to getlarge amounts (between 0.5 and 4 g) of trans-genic FEC within a period of 16 weeks (Table1), which after culture for maturation yieldedLUC positive mature somatic embryos. Only 4of the 26 non stringent selected lines yieldedLUC positive mature somatic embryos whenFEC was cultured for maturation, 16 weeks

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101 95

after bombardment. In the other 22 lines non-stringent selection was continued for ten moreweeks before it again was cultured for maturation.All 48 transgenic lines yielded mature somaticembryos. A total of 42% of the in total 156 LUCpositive mature somatic embryos, cultured on 2,4-D supplemented medium, formed secondary so-matic embryos. Although this was notinvestigated systematically there seemed to be arelation between the morphology of the maturesomatic embryos and their ability to form sec-ondary embryos.

All transgenic lines yielded secondary embryosand there were no significant differences in thisrespect between mature somatic embryos derivedfrom FEC bombarded with pHB1 and pJIT100cultures and also no significant differences be-tween mature somatic embryos derived from thethree selection regimes. Globular somatic embryo-genic tissue, isolated from the secondary somaticembryogenic cultures, formed at high frequenciesnew mature secondary somatic embryos when cul-tured in NAA supplemented medium. The num-ber of mature somatic embryos per originalmature somatic embryo varied from 20 to over100. All 20 pJIT100 cultures and 21 of the 24pHB1 luciferase positive cultures yielded a trans-genic plant carrying the LUC selection markergene. The duration of the whole process of matu-ration, secondary somatic embryogenesis of LUCpositive mature somatic embryos, desiccation andgermination varied from 12 to 52 weeks.

Stringent selection led to a reduction in the timerequired to obtain transgenic FEC and maturesomatic embryos. The time required to producetransgenic plants with stringent selection was only28–36 weeks compared to 32–41 weeks for LUC/PPT and 53–78 weeks for non-stringent selection(Table 1).

3.4. Morphology of transgenic plants

The morphology of the 21 pHB1 plants wascompared with those from three control groups,that is: 20 pJIT100 plants carrying the luciferasegene (AGPase control), non transgenic FECderived plants (regenerant control), and nontransgenic controls propagated in vitro via cut-

tings (overall control). Most of the transgenicsand the regenerant controls did not grow as vigor-ously as the overall control. Some of the transgen-ics had an aberrant growth type: highly branched(probably due to a carry-over effect of BAP asthis phenotype disappeared with subsequent mi-cropropagation), zig-zag stem (overall control hada straight stem), plants with curled leaves, andfleshy thick stems with small shoots. All thesephenotypes, except the last one, were also ob-served in the regenerant control plants. None ofthe aberrations was observed in the overall con-trol in vitro propagated by cuttings. A total of90% of the overall control and 40% of the trans-genics and the regenerant controls survived thetransfer to the greenhouse

3.5. E6idence of an antisense AGPase effect incassa6a stems

The introduction of the AGPase B gene inantisense orientation is expected to result in adecrease in AGPase B expression. This shouldhave a negative impact on starch formation, asreduced levels of AGPase would be available forADP-glucose formation. To test this expectationit was necessary to check for starch formation intuberous cassava roots. However, in the green-house it takes at least 4–8 months before tuber-ous roots are formed. It was necessary to analysesome other part of the antisense AGPase B plantsthat had large amounts of starch and for thispurpose in vitro thickened stems were used [18].

Sections of induced thickened stems of 16 of the21 antisense AGPase B plants and controls (20pJIT100 transgenics, 20 non-transgenic regenerantand 20 overall control plants) were stained withiodine to determine the presence or absence ofstarch. The antisense AGPase B plants could bedivided into three categories based on iodinestaining of in vitro thickened stems, as shown inFig. 3. Of the 16 transgenic plants containing theAGPase B antisense gene there were four whichbelonged to category III (Table 2). These plantsexhibit staining characteristics similar to the threegroups of control plants i.e. all tissue in the cortexregion and epidermal area stained blue/black withiodine (Fig. 3a).

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–10196

Fig. 3. Iodine staining of in vitro thickened stems of cassava plants containing the pHB1 construct (a) cross section of the in vitrothickened stem of HB-10, a category III plant stained with iodine. Note the strong staining of all tissue layers in the cortex andepidermal areas indicating the presence of starch (magnification ×5). (b) Category II transgenic plant (HB-10) had little or nostaining in the cortex region but intense blue/black staining in the epidermal layer (magnification ×5). (c) category I transgenicplants (example is HB1) had no staining in the cortex region and very few starch granules in the epidermal region (magnification×5) as shown in the close up (magnification ×50) transverse section in panel d).

There were nine plants (56%) designated ascategory II antisense plants (Table 2) with little orno staining in the cortex region but retainingmore stain in the epidermal layer (Fig. 3b). Cate-gory I plants of which there were three transgenicplants (Table 2) had no staining at all in thecortex region (Fig. 3c). They showed some stain-ing in the epidermal layer that was less than 10%of that found in category II sections. When exam-ined microscopically and only a few iodine stain-ing starch granules were found in the cells of theepidermal layer as shown in Fig. 3d for plantHB1.

Staining of the induced in vitro thickenedstems, revealed that in total 74% of the AGPase Bantisense plants had reduced levels of starch com-pared to the three groups of control plants. Atotal of seven AGPase antisense plants and threecontrols were isolated and analysed by northernblotting. In Fig. 5 the results are given which

clearly show that whereas the control plants pro-duce a strong signal, most of the AGPase anti-sense plants do not show any or only a weaksignal indicating the absence or reduction of AG-Pase mRNA in these plants. Only HB13 showed asignal comparable with the controls. The numberof pHB1 inserts in the antisense plants was deter-mined by southern analysis using the LUC geneas a probe. This revealed that the plants con-tained between 1–7 inserts of the AGPase Bantisense construct and between 1–6 inserts of thepJIT100 construct (Fig. 4). There was no clearcorrelation between a high number of pHB1 in-serts and the degree of antisense effect (Table 2).

4. Discussion

Previously transgenic cassava plants using FECwere produced using either selection by paro-

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101 97

Table 2Identity of transformants obtained after introduction of the plasmids pHB1 and pJIT100 into cassava friable embryogenic callus

Plant Stage of regener- Minimum num- Morphology of regenerants (number of plants in paren-Starch inhibitionthesis)levelber of insertsation

4 IHB-1 DPlant1 IPlant CHB-2

PlantHB-3 1 I NPlantHB-4 1 II C

1 IIPlant DHB-6PlantHB-7 2 II A

7 IIPlant NHB-8PlantHB-9 2 II NPlantHB-10 1 II N

1 IIPlant CHB-11PlantHB-12 1 II D

2 IIIPlant NHB-131 IIIHB-14 DPlant1 IIIPlant NHB-151HB-16 IIIPlant D2 ndPlant DHB-17

PlantHB-18 3 nd DPlantHB-19 1 nd A

nd ndPlant NHB-20HB-21 ndPlant nd C

1Mature somatic ndHB-22 ndembryosMature somatic 6HB-23 nd ndembryosMature somatic ndHB-24 nd ndembryos

1–6JT:1–20 IIIPlants N (10), A (1), B (2), C (4), D (3)0 IIIPlants N (10), A (3), B (7)FEC:1–20

60444:1–20 0Plants III N (20)

Starch inhibition level morphology of regenerated plants: I, complete inhibition in cortex; N, normal region some starch inendodermal layer; A, zig-zag; II, incomplete inhibition in cortex region; B, curled leaves no inhibition in endodermal layer; C, thickstems and numerous shoots; III, no inhibition in cortex and endodermal layer; D, stunted growth with tiny leaves.

momycin [5] or luciferase [6]. Paromomycin selec-tion is efficient with respect to obtain pure trans-genic FEC, however, the regeneration efficiency islow [9]. The opposite holds true for the luciferaseselection system [17]. Paromomycin selection re-quires large numbers of transgenic lines and luci-ferase selection large quantities of tissue of atransgenic line to produce the desired amount oftransgenic plants. The amount of FEC necessaryfor transgenic plant formation is dependent on thechance that a FEC unit develops into a maturesomatic embryo, the chance that a mature somaticembryo is transgenic and the chance that a ma-ture somatic embryo develops into a plant. The

chance that a mature somatic embryo was trans-genic was the only factor that differed amongstthe three selection regimes used. In the stringentselection procedure based on two cycles of micro-scopic identification, selection and growth ofLUC positive FEC units, 90% of the maturesomatic embryos were LUC positive. Less than1% of the mature somatic embryos in non strin-gent selection was LUC positive, compared to 90and 45% of the stringent and LUC/PPT selectedcultures, respectively. Therefore the amount ofFEC needed to produce a transgenic plant forstringent selection is a factor of two lower thanthat of LUC/PPT selection and more

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–10198

Fig. 4. Southern blot analysis of transgenic cassava plants containing the plasmids pJIT100 and pHB1. A total of 10 mg of DNAwas applied per lane, blotted on to a hybond filter and hybridized with a 32P-labelled 1.5-kb luciferase gene. M, 1-kb DNA markerlane. Sizes are as shown on the left-hand side of the panel. C, control DNA from non-transgenic in vitro propagated 60444 plants.Lanes 1–9 contain DNA from the transformants J1–J9. Lanes 11–19 contain DNA from the transformants HB10, HB19, HB13,HB8, HB2, HB5, HB3, HB18 and HB23. DNA from all the transformants was derived from leaves of in vitro plantlets except forlane 19 (HB23) were DNA isolated from mature somatic embryos was used.

than a factor of 90 lower in comparison to nonstringent selection. As a result stringent selectionrequires a shorter time period to obtain the neces-sary amount of FEC needed to regenerate a plantwith less labour input. Cultures of FEC on a MS2medium supplemented with 1 mg l−1 NAA didnot improve embryo maturation, in comparisonto MS2 medium supplemented with 1 mg l−1

picloram. High numbers of mature somatic em-bryos were obtained by using FEC, derived fromnon stringent selected cultures, which were‘primed’ for maturation. This was done by leavingthe FEC cultures in liquid proliferation mediumwithout refreshing, over a period of 2 weeks, andthen transferring the cultures on to maturationmedium. These primed cultures gave rise to rela-tively high numbers of mature somatic embryos.Further research is required to determine whether

this priming effect is also observed in stringentselected cultures and whether it is due to nutrientstarvation or to density.

The time taken to produce transgenic cassavaplants in stringent LUC selection was only 28–36weeks compared to 32–41 weeks for combinedLUC/PPT selection and 53–78 weeks for nonstringent LUC selection. In the stringent selectionprocedure it took 10 weeks before a (almost)complete transgenic line was obtained. This wascultured for six more weeks on proliferationmedium which resulted in 0.5–4 g FEC, before itwas cultured for maturation (the first step of theplant regeneration process). It is estimated thatalready 0.1 g of completely transgenic FEC issufficient to obtain a transgenic plant. This wouldreduce the duration of the proliferation phase byat least 4–6 weeks. In the stringent selected cul-

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101 99

tures the process of plant regeneration from FECtook 12–20 weeks. The procedure used for plantregeneration consisted of four steps (maturation,multiplication of LUC positive mature somaticembryos by secondary somatic embryogenesis,desiccation and culture of, mature somatic em-bryos for germination). Using improved methodsof regeneration the procedure can be shortened.Schopke et al. [9] obtained plantlets in a one stepprocedure when FEC was cultured on rafts float-ing in liquid medium. This method is currentlytested on transgenic FEC lines derived from luci-ferase selection.

There is only one other crop, Dendrobium [20],known to be recalcitrant for plant transformation,where the luciferase gene has been used to obtaintransgenic plants. Luciferase based selection hasas advantage that it does not interfere with theregeneration process as has been shown withchemical selection. The results presented here fur-ther show that even in friable embryogenic calluscultures in which the embryogenic units are verysmall and therefore difficult to isolate, completetransgenic lines can be obtained via a stringentselection regime. The stringent selection systemhas been applied successfully in FEC of two othergenotypes of cassava showing its universal appli-cability (Raemakers, unpublished results).

A total of 21 independent transgenic cassavaplants were produced by particle bombardmentmediated transformation of cassava 60444 FECwith the plasmid pHB1. This plasmid carried thecassava starch gene AGPase B in antisense orien-tation under the control of a double CaMV 35Spromoter. This is the first reported transformationof cassava with a gene conferring a new trait tothe crop other than the previously describedtransgenic cassava plants, which carried reporteror marker genes [4–7]. Southern analysis revealedthat these plants carried between one and seveninserts of the pHBI gene-construct. LUC assaysconducted on the plants and explants confirmedthat they were indeed transgenic. In actual fact itwas even possible to visualise the light emitted bythese plants in the dark, upon spraying them withluciferin, using the naked eye. Other researchersworking with the Luciferase firefly gene in plantshave not previously reported this phenomenon.

This result is clearly correlated to the high activityof the luciferase enzyme in the antisense AGPasecassava plants.

Most of the transgenic and non-transgenic FECderived plants did not grow as fast as the overallcontrol, in vitro propagated, plants. In about 50%of the cases continued in vitro propagation re-sulted in growth rates similar to those of overallcontrol plants. The other 50% of the plants hadmore serious aberrations such as zig-zag stems,stunted growth, curly leaves, thick stems and nu-merous small shoots. The latter two were onlyobserved in the transgenics. It remains unclearwhether or not this is due to the introduction ofthe gene constructs, or to somaclonal variationinduced by the regeneration process.

The transgenic plants containing the constructpHB1 contain the AGPase B gene in antisenseorientation as shown by the southern hybridiza-tion (Fig. 4). The AGPase B gene encodes thesmall subunit of the heterotetrameric AGPase en-zyme. This enzyme plays a critical role in starchformation where it is involved in forming ADP-glucose from ATP and glucose-1-phosphate [21].Suppression of the expression of this gene wouldlead to reduced levels of the enzyme and ulti-mately to a reduction in starch formation. In factthe northern analysis (Fig. 5) showed that in anumber of plants expression of the gene wascompletely inhibited which should result in agreatly decreased activity of the AGPase enzymeand thus in the amount of starch deposited inthese plants. Analysis of the starch present in invitro thickened stems of the antisense AGPase Bplants revealed that 74% of the transgenic plantshad reduced levels of starch compared to thecontrol non transgenic and pJIT100 transformedplants. This functionally confirms the identity ofthe AGPase B cDNA cloned from a cassava tuberspecific library [10]. It is also a clear indicationthat the introduction of the antisense AGPase Bgene blocks the formation of ADP-glucose, themajor glucosyl donor for starch formation [21].The high frequency (74%) of transgenic plantsexhibiting an antisense effect is in stark contrastto what was observed when the same cDNA wasintroduced into potato where the frequency ofplants exhibiting an antisense effect was only 5%

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101100

Fig. 5. Northern blot analysis of total RNA from leaves of transgenic cassava plants containing the plasmid pHB1. A total of 40mg of total RNA was loaded per lane and the cassava AGPase B gene was used as probe. Lanes 1, 2 and 10 contain RNA fromcontrol non transgenic plants.

(as judged by reduced starch levels), [10]. Thisdifference in response can be attributed to thereduced sequence homology between the cassavaand potato AGPase B genes of only 68% henceprobably the reduced response when the het-erologous cassava gene was introduced intopotato.

Interestingly in the three plants exhibiting thehighest antisense effect a cell layer within the stemstill had some starch. This could be due to the factthat the CaMV35S promoter is not active in thistissue layer. The reduction in starch levels inAGPase antisense cassava plants is similar towhat was observed with the introduction of anti-sense cassava AGPase B in potato. This gave riseto plants with reduced starch levels and increasedlevels of fructose and glucose in their sugars [10].Similar reductions in starch levels and increase inglucose levels had been reported in potatoes car-rying the antisense potato AGPase B gene [22].From the antisense work in potato it can beinferred that sweet cassava plants might havebeen produced. This, however, awaits further in-vestigation which will be carried out once tubersare produced.

Acknowledgements

We would like to thank the Department ofInternational Co-operation (DGIS) of TheNetherlands Ministry of Foreign Affairs forproviding financial support through the Cassavaand Biotechnology Project Zimbabwe.

References

[1] A.M. Thro, R.N. Beachy, M. Bonierbale, et al., Interna-tional research on biotechnology of cassava and its rele-vance to Southeast Asian economies: a cassavabiotechnology network review, Asian J. Trop. Biol. 2(1996) 1–31.

[2] C.C.J.M. Raemakers, E. Jacobsen, R.G.F. Visser, Regen-eration and transformation in cassava, Euphytica 96(1998) 153–161.

[3] N.J. Taylor, M. Edwards, R.J. Kiernan, C. Davey, D.Blakesley, G.G. Henshaw, Development of friable em-bryogenic callus and embryogenic suspension cultures incassava (Manihot esculenta Crantz), Nat. Biotechnol. 14(1996) 726–730.

[4] H.Q. Li, C. Sautter, I. Potrykus, J. Pounti-Kaelas, Ge-netic transformation of cassava (Manihot esculentaCrantz), Nat. Biotechnol. 14 (1996) 736–740.

[5] C. Schopke, N. Taylor, R. Carcamo, et al., Regenerationof transgenic cassava plants (Manihot esculenta Crantz)from microbombarded embryogenic suspension cultures,Nat. Biotechnol. 14 (1996) 731–735.

[6] C.C.J.M. Raemakers, E. Sofiari, N. Taylor, G. Henshaw,E. Jacobsen, R.G.F. Visser, Production of transgeniccassava (Manihot esculenta Crantz) plants by particlebombardment using luciferase activity as selectionmarker, Mol. Breed. 2 (1996) 339–349.

[7] S.C.H.J. Snepvangers, C.J.J.M. Raemakers, E. Jacobsen,R.G.F. Visser, Optimization of chemical selection oftrangenic friable embryogenic callus lines using the luci-ferase reporter gene system, Afr. J. Root Tuber Crop(1998) in press.

[8] N. Taylor, R.J. Kiernan, D. Blakesley, G.G. Henshaw,Strategies for the production of friable embryogenic tis-sues across a range of cassava cultivars, Afr. J. RootTuber Crop (1998) in press.

[9] C. Schopke, R. Carcamo, R.N. Beachy, C. Fauquet,Plant regeneration from transgenic and non-transgenicembryogenic suspension cultures of cassava (Manihot es-culenta Crantz), Afr. J. Root Tuber Crop (1998) in press.

T.R.I. Munyikwa et al. / Plant Science 135 (1998) 87–101 101

[10] T.R.I. Munyikwa, PhD Thesis, Wageningen AgriculturalUniversity (1997) 1–127.

[11] P.M. Gresshoff, C.H. Doy, Development and differenti-ation of haploid Lycopersicon esculentum (tomato),Planta 107 (1974) 161–170.

[12] R.U. Schenk, A.C. Hildebrandt, Medium and tech-niques for induction and growth of monocotyledonousand dicotyledonous plant cell cultures, Can. J. Bot. 50(1972) 99.

[13] T. Murashige, F. Skoog, A revised medium for rapidgrowth and bioassay with tobacco cultures, Physi-ologium Plantarum 15 (1962) 473–497.

[14] J. Sambrook, E.F. Fritsch, T. Maniatis, MolecularCloning: A Laboratory Manual, 2nd, Cold Spring Har-bor, New York, 1989.

[15] F. Guerineau, P.M. Mullineaux, Plant transformationand expression vectors, in: R.R.D. Croy (Ed.), PlantMolecular Biology Labfax, BIO.S Scientific, Oxford,1993, pp. 121–148.

[16] M. De Luca, W.D. McElroy, Purification and propertiesof firefly luciferase, Methods Enzymol. 57 (1978) 3–15.

[17] C.C.J.M. Raemakers, M.G.M. Rozeboom, E. Jacobsen,R.G.F. Visser, Regeneration of plant from somatic em-

bryos and friable embryogenic callus of cassava, Afr. J.Root Tuber Crop (1998) in press.

[18] S.N.I.M. Salehuzzaman, E. Jacobsen, R.G.F. Visser,Expression patterns of two starch biosynthesis genes inin vitro cultured cassava plants and their induction bysugars, Plant Sci. 98 (1994) 53–62.

[19] C.C.J.M. Raemakers, C.M. Schavemakers, E. Jacobsen,R.G.F. Visser, Improvement of cyclic somatic embryo-genesis of cassava (Manihot esculenta Crantz), Plant CellReports 12 (1993) 226–229.

[20] T.F. Chia, Y.S. Chan, N.H. Chua, The firefly luciferasegene as a non invasive reporter for Dendrobium trans-formation, Plant J. 6 (1994) 441–446.

[21] J. Preiss, K. Ball, J. Hutney, B. Smith-White, L. Li,T.W. Okita, Regulatory mechanisms involved in thebiosynthesis of starch, Pure Appl. Chem. 63 (1991)535–544.

[22] B. Muller-Rober, U. Sonnewald, L. Willmitzer, Inhibi-tion of the ADP-glucose pyrophosphorylase in trans-genic potatoes leads to sugar-storing tubers andinfluences tuber formation and expression of tuber stor-age protein genes, EMBO J. 11 (1992) 1229–1238.

.