An Arabidopsis thaliana (Ler) inbred line AFDL exhibiting abnormal flower development mainly caused by reduced AP1 expression

The genetic network controlling flowering and flower development consists of a set of floral integrator genes that play a role in light sensing, hormone signaling and developmental pathways. These integrators activate the expression of meristem identity genes LEAFY (LFY) and APETALA1 (AP1) to initiate the flowering transition. However, how the expression of key genes, such as AP1, responds to diverse signals during flower development remains largely unknown. Here, we report that an Arabidopsis abnormal flower development inbred line (AFDL) exhibits a phenotype similar to the ap1 mutant, with delayed flowering time and a high frequency of transition of flower meristems into inflorescence meristems after flowering. The flower organs with an abnormal first whorl lack the second whorl and the increased number of inflorescences at the first- and second-whorl positions most closely resembled the phenotypes of ap1/cal double mutants. Interestingly, both normal and abnormal flowers coexisted in a single individual. Microarray and quantitative real-time PCR analysis revealed that the expression of AP1 was significantly reduced, while the expression of its interacting genes TERMINAL FLOWER 1 (TFL1), SHORT VEGETATIVE PHASE (SVP), AGAMOUS-like 24 (AGL24), SEPALLATA (SEP) and CAULIFLOWER (CAL) and upstream genes FLOWERING LOCUS C (FLC) and FLM were increased in AFDL, which could serve to explain its phenotype. The expression of genes responsive to different stimuli dramatically changed in AFDL relative to the wild type, as revealed by the differential display of transcripts, indicating that this expression variation is subject to a threshold, leading to an on/off expression pattern of the master regulatory gene (such as AP1) of flower development.

Plants have developed many mechanisms to sense and respond to environmental cues to achieve a successful transition from vegetative to reproductive growth (for a review see [1]). This transition has been intensively studied in Arabidopsis thaliana, in which photoperiod, day length, vernalization, ambient temperature and nutrient status influence this process synergistically [2]. The flowering time genetic network in A. thaliana represents one of the *Corresponding author (email: lumz@caf.ac.cn) best-studied functional systems in plants, and integrates environmental and physiological information to modulate expression of key downstream genes controlling the appropriate time for flowering [3,4]. These genes include FLOWERING LOCUS C (FLC), whose expression is regulated mainly by vernalization and an autonomous pathway, and CONSTANS (CO), regulated by an autonomous pathway [5]. These two genes further regulate integrators SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1) and FT, which are also regulated by the GA pathway to trigger the floral meristem identity genes LEAFY (LFY)/AP1 [6,7]. LFY and AP1 upregulate each other and play a pivotal role in specifying floral meristems during the conversion of meristem identity from inflorescence to flower under the control of endogenous and environmental cues [8,9].
The floral homeotic gene AP1 specifies floral meristem, sepal and petal identities in Arabidopsis. Flowers of ap1 plants show a transformation of the first-whorl organs into bract-like structures and the development of flower buds in the axil of each bract-like organ. As a result, these secondary flowers may also produce tertiary and quaternary flowers. Secondary and tertiary flowers can be incomplete or irregular in terms of mosaic organs or possess fewer organs than normal [10,11]. In addition to aberrant development of first-whorl organs as bract-like structures, ap1-1 mutants also lack petals. On secondary and tertiary flower buds, organs occasionally develop mosaics of leaf-like and stamen-like tissue. These organs appear to arise from a region in the meristem where the first-and third-whorl primordia are closely apposed [11,12]. More significantly, in an ap1/cal double mutant, flower meristems lose their floral fate, leading to a transition to inflorescence architecture and resulting in a complex branched structure [11].
The floral meristem identity gene AP1 is essential for emerging floral meristems and flower organ formation by repressing the expression of SVP, AGL24 and SOC1, which promotes inflorescence fate rather than flower formation of the meristem [13]. SOC1, FRUITFUL (FUL) and AGL24 play roles in activation of LFY/AP1 during the emergence of flower meristems at the side of SAM, thus forming a feedback regulation loop to regulate spatial patterning precisely in the SAM upon floral transition and the ultimate organization of inflorescence architecture [1]. AP1 and its homologous gene CAL are activated in the stage 1 floral meristem following the domain expressing LFY [14,15]. Although LFY binds in vivo to AP1 and CAL regulatory sequences, it alone is insufficient to activate AP1 [16,17]. FT was shown to induce AP1 expression [18], although the underlying mechanism still remains to be elucidated. AP1 transcripts are also negatively regulated by the AP3/PI heterodimer in combination with other unknown factors to restrict its spatial distribution during early stages of floral development [19]. AP1 also interacts with SEP1/2/3/4, which are MADSbox genes playing roles in the ABCE model of flower development, to specify floral meristem identity and organ identity [20][21][22]. Recently, it was found that PUCHI and BOP are required for specification of floral meristem identity by providing a positional cue for LFY and AP1 to be expressed when induced by flower-promoting signals such as photoperiod [23]. The authors also suggested that negative factor(s), such as TFL1, which is known to limit LFY expression to the floral meristem, may repress PUCHI function in the secondary inflorescence meristem. Therefore, the regulation of AP1 by endogenous and environmental cues warrants further investigation.
Existing evidence points to AP1 having functions for both onset of flowering and first-and second-flower organ development . It can interact with many genes, such as CAL,  FUL/AGL8, LFY, AG, FT, TFL1, SVP, AGL24, SOC1 and AP3/PI, which were discovered by forward genetic screening of Arabidopsis mutants. Further investigation is therefore needed to understand the regulation of AP1, particularly in response to endogenous and environmental cues. However, a mutagenesis approach limits the ability to examine crosstalk among so many genes that participate in a variety of light-sensing, hormone-signaling and developmental pathways [24]. In this paper, we describe an abnormal flower development line (AFDL) that exhibits a similar, but not identical, flower phenotype to the ap1 mutant. This phenotype was primarily caused by significant downregulation of AP1 expression in conjunction with other genes that regulate flowering and flower development. Phenotypic and genetic analysis of AFDL revealed that the regulation of AP1 and its interaction with other flowering regulators are complicated and sensitive to both internal and external stimuli.

Plant materials
Arabidopsis thaliana Landsberg erecta (Ler) seeds were obtained from the Nottingham Arabidopsis Stock Centre (NASC). We found a transposon (Ds) insertional mutant (At1g52910) exhibiting abnormal flower development (thus named AFDL) when screening Arabidopsis mutants associated with the development of vascular tissues. Homo-and heterozygosity identification of the mutated gene was performed on AFDL and the results showed that an insertional mutation of the At1g52910 gene was inconsistent with the abnormal phenotype, indicating that the phenotypes of AFDL were not caused by mutation of the At1g52910 gene. RNAi and complementation constructs for this gene were then introduced into wild-type (WT) and AFDL lines, respectively, but the AFDL phenotypes did not emerge in RNAi transgenic plants and the WT phenotype was not observed in transgenic AFDL plants. This provided further evidence that the phenotype of AFDL does not result from mutation of the At1g52910 gene. Furthermore, the result of Tail-PCR analysis did not reveal any other Ds insertion sites, suggesting that AFDL is not caused by a Ds insertional mutation. In addition, we analyzed segregation in one selfing AFDL plant and found that it was not caused by a single nuclear mutation because the ratio of plants with abnormal and normal phenotypes deviated from Mendel's law. Finally, the phenotype of AFDL was mostly stable after multigeneration selfing, pointing to a genetic basis.
AFDL and WT plants were vernalized at 4°C for 2 d and grown on a mixture of turf soil and vermiculite (3:1) at (22 ± 2)°C in short days (SD:8 h light, 16 h dark) and long days (LD:16 h light, 8 h dark), to check flowering time and for phenotypic observation. Plants were also grown on MS media in petri dishes (120 mm) at (22 ± 1)°C under LD conditions for microarray analysis.

Phenotypic observation
Intact plants were visually checked and photographed. The flowering time was measured by counting the number of leaves at bolting with 0.5-1.0 cm stem above the rosette. The data were statistically analyzed using one-way ANOVA analysis of variance and least significant difference tests with SPSS 15.0 software (SPSS, Inc., Chicago, IL, USA). Flower buds were observed with a Zeiss SteREO Discovery V8 stereomicroscope and KYKY-2800B scanning electron microscope (KYKY Technology Development, Beijing, China). Specimens were prepared using the t-butyl alcohol freeze-drying method for scanning electron microscopy (SEM) [25]. Images were documented with software installed on the instruments.

Genomic sequencing of AP1 and its upstream region
Based on the genomic sequence of the AP1 region (TAIR: http://www.arabidopsis.org), we designed primers (Table  S1) to span the whole region of the AP1 genomic sequence to the end of the upstream adjacent gene. Amplified DNA fragments from both AFDL and WT lines were used for direct sequencing. The sequence data were assembled and compared between AFDL and WT.

RNA isolation and labeling
To identify differences in gene expression between AFDL and WT, we used a bulked sampling strategy to minimize individual variation, by collecting and mixing seedlings with six clear rosette leaves in petri dishes and emerging flower buds at bolting from more than 50 plants of both AFDL and WT. The analysis was combined with the differential display method using Affymetrix Arabidopsis ATH1 GeneChips (Affymetrix, Santa Clara, CA, USA). These bulked samples and cauline leaves (for quantitative real-time PCR (qRT-PCR)) were frozen in liquid nitrogen and stored at −80°C. Total RNA isolation was performed with the Plant RNeasy Mini Kit (Qiagen China, Shanghai) following the manufacturer's instructions. Five micrograms of total RNA was used to synthesize double-stranded cDNA using the Affymetrix One-cycle cDNA Synthesis Kit. The cDNAs acquired were used as a template for synthesis of biotinylated cRNA using the GeneChip IVT Labeling Kit (Affymetrix). Biotinylated cRNA was purified with the Affymetrix GeneChip Sample Cleanup Module and fragmented according to the manufacturer's protocol (Affymetrix).

Array hybridization and analysis of expression data
Hybridization of GeneChip arrays was done in an Affymetrix Hybridization Oven 640 following the manufacturer's protocol (Affymetrix). The Affymetrix Fluidics Station 450 was used to facilitate washing and staining. Chips were scanned by a GeneChip ® Scanner 3000 7G. Expression levels in seedlings and flower buds were calculated from Affymetrix hybridization intensity data and differences between AFDL and WT were compared using preinstalled GeneChip Operating Software.

Quantitative real-time PCR
Reverse transcription was performed using the Thermo-Script RT-PCR System (Invitrogen) according to the manufacturer's instructions, starting with 3 µg of the total RNA isolated from flower buds and cauline leaves. Primers specific to the flower-related genes AP1, AP2, AP3, SEP1, SEP2, SEP3, SEP4, FUL, CAL, LFY, AGL24, SOC1, SVP, SAP18, FLC, FD, FT, TFL1, CO, FLM, SVP and TUBULIN were used for qRT-PCR (Tabke S2). Amplification was monitored in real time with an ABI 7500 Real-Time PCR System (Applied Biosystems, Shanghai, China) using SYBR Green PCR Master Mix (Applied Biosystems) as the fluorescence source. The relative expression levels were calculated and statistically analyzed using SDS software v1.4 (Applied Biosystems).

Phenotypic variation of AFDL
AFDL plants grew vigorously but exhibited variable degrees of phenotypic alteration compared to WT ( Figure  1(a)) as follows: (i) more inflorescences stacked at a small angle to the stem (Figure 1 Figure 1(f)). All of these plants produced more inflorescence branches with cauline leaves because of the conversion of the first-whorl organ into a 2nd inflorescence, and 3rd inflorescences emerged from the first-whorl organ on the secondary inflorescence, until seven or more inflorescences were produced. Multiple flowers towards the axillary side showed a cauliflower-like shape, with no petals being observed in AFDL types (Figure 1(h)), which is very similar to ap1/cal double mutants. Compared to the WT line, no obvious phenotypic alteration or growth rate changes of AFDL plants were observed in the vegetative phase starting from young seedlings to adult plants that had not yet bolted (Figure 2   However, delayed bolting and flowering were observed in AFDL plants (Figure 2(g) and (h)), and significant (P < 0.05) differences in leaf number (LN) at bolting (0.5-1.0 cm above the rosette) were observed between AFDL and WT plants of approximately 10 in the SD condition and 3 in the LD condition ( Figure 3). Therefore, AFDL is a late-flowering type under both SD and LD conditions.
The flowers of AFDL plants exhibited similar phenotypes (Figure 4(b)) to the ap1-1 mutant [11]. However, greater variation was observed in AFDL flowers. (i) The first-whorl organ was either normal, absent or converted to a bract-like, carpelloid, and stamenoid sepal (Figure 4(b),   (m)). An extreme phenotype was observed in which more complex structures such as stamens (Figure 4(o) and (p)) appeared in the carpel. This phenotypic variation was most similar to that of agl24/svp double and agl24/svp/ap1 triple mutants [26]. In addition to variation in the 4 whorls, chimeric structures often appeared on inflorescences (Figure 4(q)-(t)). Compared to WT plants (Figure 4(u)), normal (Figure 4(v)) and twisted ( Figure  4(w)) siliques and siliques with anther or filament structures (Figure 4(x)) were also observed in AFDL. We also used SEM to document flower bud development. Similar to a strong ap1 mutant, first-whorl organs in AFDL plants appeared in an order similar to that of WT sepals (Figure 4(y)) but did not grow to enclose the developing flower bud (Figure 4(aa)). The inflorescence primordia were more rounded, growing away from the central floral meristems (Figure  4(z)), and were present in the axils of first-whorl organs, leading to crowded flower buds (Figure 4(aa)).

Microarray analysis of AFDL
No differences in the genomic sequence of the AP1 region, including AP1 and its 5′ regulatory sequences were found between AFDL and WT strains ( Figure S1). To find out which gene(s) could be responsible for the altered flower development in AFDL, we employed a differential display of transcripts in bulked samples of seedlings prior to bolting and flower buds upon bolting, using Affymetrix Arabidopsis whole genome chips, to analyze gene expression. Comparison of the expression data revealed that 95% of the genes showed no significant difference between AFDL and WT (less than 2-fold). Totals of 1274 and 942 genes were differentially expressed (more than 2-fold) in flower buds (Table  S3) and seedlings (Table S4), respectively. We selected the differentially expressed genes involved in the genetic network regulating flowering and flower development (Table 1), and found that AP1 expression was significantly reduced in the flowers of the AFDL strain, whereas AP1 expression was 128-fold higher in WT. Flowering time and organ development regulators, such as FLC, FT, AGL24, FUL, LFY, SEP4, SVP and SAP (Table 1), and several other less-studied MADS-box homeotic genes, such as AGL8, AGL13, AGL42, AGL44, AGL71 and AGL87, were also differentially expressed (Table S3). We also classified these genes using DAVID software [27,28] according to their biological functions; the gene ontology terms and the number of genes for each term are presented in Table S5. A large group of genes responsive to a variety of stimuli (Table  2) were differentially expressed (Table S6), accounting for 20.3% and 9.2% of the total number of differentially expressed genes in seedlings and flower buds, respectively. These are found in much higher proportions compared to the other terms (Table S5). These genes include JAZ1, ARR7, GI and RGL2, which are involved in flowering regulation in response to hormone signaling, and DREB1A, which has been implicated in cold-inducible transcriptional regulation.

Gene expression using qRT-PCR
We further analyzed expression of selected genes both upstream and downstream of AP1 in the network in cauline leaves and flower buds using qRT-PCR ( Figure S2). In cauline leaves, both flower-promoting (FT, FD and CO) and flower-inhibiting (FLC, FLM and SVP) genes were highly expressed ( Figure 5(a)), but they were expressed at normal or moderate levels in flower buds ( Figure 5(b)). Interestingly, putative downstream genes from AP1, such as SEP1, SEP2, SEP3 and SEP4, were highly expressed in AFDL in cauline leaves (Figure 5(a)). Notably, CAL and TFL1 were highly expressed in flower buds ( Figure 5(b)) and AP1 expression was confirmed to be significantly reduced (400-500 fold) in AFDL flower buds by qRT-PCR.

Discussion
We identified an Arabidopsis line, AFDL, showing a phenotype similar to ap1, as evidenced by the conversion of first-whorl organ primordia into floral meristems, homeotic changes of the first and second whorls of floral organs, the absence of petals and the conversion of the first sepal whorl into leaf-like structures [11]. However, AFDL flowers exhibit obvious differences when compared to ap1 mutants. For instance, a much higher frequency of conversion of the floral meristem into an inflorescence meristem is observed, similar to the phenotype of ap1/cal double mutants [11], and alterations of the 4th floral whorl (Figure 4(l)-(p)) most similar to that of agl24/svp double and agl24/svp/ap1 triple mutants [26]. In addition, considerable phenotypic variation was found within AFDL, ranging from weakly to severely altered phenotypes. In extreme cases, both normal and abnormal flowers coexist in the same plant (Figure 1(d)). These results suggest that the genetic control of the AFDL phenotype is complex and deserves further characterization. The ap1-like phenotype of AFDL could be caused by mutation of AP1. Therefore we examined the genomic  sequence of AP1 and its upstream region but failed to find changes in AFDL relative to WT, suggesting that the phenotype cannot be attributed to genomic mutations in this region. The quantitative flower trait of AFDL suggests that the ap1-like phenotype of AFDL may be caused by polygenic interactions, particularly given the fact that both typical AFDL and WT inflorescences coexist in one individual.
Therefore, we adopted a 'bulk sampling strategy' (see Materials and methods) in combination with Affymetrix Arabidopsis ATH1 GeneChips to identify genes expressed differentially in seedlings and flower buds between AFDL and WT. Not surprisingly, AP1 was found to be significantly reduced in flowers of AFDL using both microarray (Table 1) and qRT-PCR ( Figure 5(b)) data, indicating that there is almost no expression of AP1 in AFDL. Such a change may explain the ap1-like flower of AFDL but is insufficient to explain the delay in flowering, more frequent mosaic flowers and inflorescences in AFDL. The unique phenotype of AFDL flowers most closely resembles the combined phenotypes in mutants for AP1, TFL1, CAL, SVP, AGL24, SEP, SAP or their compound mutants, suggesting that the phenotypic variation may result from crosstalk among these genes to control flower development. Indeed, AP1, TFL1, CAL, SVP, AGL24, SEP1, SEP4 and SAP were among the genes found to be differentially expressed by microarray (Table 1) and qRT-PCR ( Figure 5(b)) in this study. The roles for these genes in flower development have been well characterized, thus implicating their synthetic contributions to development of the AFDL phenotype. In flower buds of AFDL where emerging inflorescence meristems were formed, TFL1 was found to be highly expressed, and this gene can negatively regulate AP1 [29] and specify inflorescence meristem identity [30], leading to a delay of floral meristem formation. Downregulation of AP1 can also contribute to the higher levels of SVP in cauline leaves, which negatively regulates expression of the AP1 activator FT through direct binding to CArG motifs in the FT sequence [8], and correspondingly results in lower levels of FT and AP1 in flower buds. On the other hand, the low levels of AP1 in flower buds cannot repress expression of AGL24 and SOC1, which promote inflorescence fates rather than flower formation in the meristem [13] and result in more abundant and longer inflorescences in AFDL. Clearly, further analysis of their expression patterns in AFDL is required to elucidate the interactions among these genes. Furthermore, many cauline leaves were observed in AFDL on higher-order inflorescences that emerged at places where normal flowers would develop in WT. To investigate if they also play a role in regulating flowering time and development, we also analyzed genes that interact with AP1 in the network in cauline leaves using qRT-PCR, and found that expression of FLC, FLM, SVP, FT, FD, PI, AP2 and SEP1-4 were highly increased ( Figure 5(a)). Therefore, the delay of flowering in AFDL can be explained by high expression of FLC (also in seedlings) and FLM in cauline leaves. These two late flowering regulators are well documented to be responsible for the delayed transition from the vegetative to reproductive phase [31]. However, flower-promoting genes FT and FD were also highly expressed in cauline leaves, while the opposite case was observed in flower buds ( Figure 5). The elevated expression of these antagonistic regulators in AFDL cauline leaves warrants further investigation, although we can speculate that FT or its transcripts are produced in cauline leaves but somehow fail to be transported to the emerging flower meristem.
Interactions among the aforementioned genes in contributing to the AFDL-specific phenotype may occur in response to sensed endogenous signals and environmental cues, which is evidenced by the finding that 20.3% of the total number of differentially expressed genes belong to those responding to 36 categories of endogenous signals and stress-related stimuli (Table S6). In addition, both AFDL and WT flowers coexisting in one individual indicates that this interaction is sensitive to slight changes in endogenous and environmental signals. In fact, previous studies have shown the expression of key flowering-time genes to be regulated by different cues. TFL1 plays an important role in establishment and maintenance of the inflorescence meristem [29] and its expression has been shown to respond to changes in sucrose levels [32]. The downregulation of the starch synthase gene (At1g32900, see Supporting Information, also for the following genes) and glucose metabolism-related genes (At3g21370, At1g06000 and At1g19940) but upregulation of the sugar transporter (At4g36670) in AFDL might change the sugar level in the seedling, thus affecting the flowering time. SVP-mediated control of FT gene expression may represent a mechanism used by plants to adjust the timing of flower development under fluctuating temperature conditions [8]. The circadian clock proteins LHY and CCA1 may reduce the abundance of SVP to accelerate flowering by activating FT expression under long days [33]. The reduced transcription of both LHY (At1g01060) and CCA1 (At2g46830), thus leading to the increased SVP (At2g22550) transcription in AFDL, may suggest the involvement of a circadian clock pathway in the expression of its flowering phenotype. The GA pathway may also contribute to the control of flowering in AFDL, because GI (At1g22770, integrating signals from the GA pathway) and RGL2 (At3g03450, major repressor of GA responses) and the genes responsive to other hormones such as JAZ1 (At1g19180, a jasmonate signaling repressor), CRFs (At4g23750 and At3g61630, mediating the transcriptional response to cytokinin) and ARR7 (At1g19050, involved in the negative feedback loop for cytokinin signaling) [34][35][36] were differentially expressed in AFDL compared to that in WT. In addition, high expression of the cold-inducible transcription regulator DREB1A (At4g25480) also hints at crosstalk between cold-inducible and flowering pathways, which is another fine-tuning mechanism for control of flowering time [37]. This situation strongly resembles the recent finding that phenotype is a consequence of large variations in gene expression that are thresholded during development to determine cell fate in Caenorhabditis elegans [38].
Discrepancies between phenotypes of ap1 and AFDL may also be caused by the different expression of other MADS-box genes in both cauline leaves and flower buds ( Figure 5). For instance, SEP genes that play a central role in flower meristem and organ identity, by interacting with AP1 and among themselves to form multimers during regulation of gene expression [22,39], were highly expressed in AFDL. It has been suggested that redundancy exists for AP1/CAL in specifying floral meristem identity [14]. However, although high expression of CAL in flower buds was observed for AFDL in our study, this could not compensate for the function of AP1 in specifying flower meristem identity and the first-and second-whorl organs, which agrees with the observation that ap1 mutants show an obvious phenotypic change while cal mutants show only a WT phenotype and ap1/cal double mutants result in a complete transformation of floral meristems into inflorescence meristems [11]. CAL may be regulated differently from AP1 because of the distinct expression for these two genes in flower buds of AFDL ( Figure 5), and it has been suggested previously that CAL and AP1 are regulated differently by LFY and FT [39]. Additional spatiotemporal analysis of CAL expression in AFDL may shed more light on this issue.
In conclusion, the AFDL strain identified in this study is phenotypically similar to ap1 mutants with noticeable deviations. The abnormal flower development of AFDL was mainly caused by significantly reduced expression of AP1, but also is attributable to crosstalk among key genes in the regulatory network controlling the transition of vegetative growth to the flowering phase as well as flower development. AFDL exhibits variation in flowering with complex flower structures and merits further molecular characterization to understand better the regulatory molecular network. Therefore, AFDL may serve as a good model to study the complex regulation of flower development, especially interactions of regulatory genes in response to various endogenous signals and environmental cues.

Supporting Information
Table S1 Primer pairs used for sequencing the AP1 region Figure S1 DNA sequence of AP1 and its upstream region in AFDL (Lansberg)

Table S2
Genes differentially expressed (more than two-fold) in flower buds between AFDL and WT plants Table S3 Genes differentially expressed (more than two-fold) in seedlings of AFDL and WT plants Table S4 Biological function terms and numbers of genes belonging to groups annotated by DAVID software   Table S6) in cauline leaves (a,b) and flower buds (c,d) using qRT-PCR The supporting information is available online at csb.scichina.com and www.springerlink.com. The supporting materials are published as submitted, without typesetting or editing. The responsibility for scientific accuracy and content remains entirely with the authors.
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.