Insights into Regulation of C₂ and C₄ Photosynthesis in Amaranthaceae/Chenopodiaceae Using RNA-Seq

Abstract

Amaranthaceae (incl. Chenopodiaceae) shows an immense diversity of C₄ syndromes. More than 15 independent origins of C₄ photosynthesis, and the largest number of C₄ species in eudicots signify the importance of this angiosperm lineage in C₄ evolution. Here, we conduct RNA-Seq followed by comparative transcriptome analysis of three species from Camphorosmeae representing related clades with different photosynthetic types: Threlkeldia diffusa (C₃), Sedobassia sedoides (C₂), and Bassia prostrata (C₄). Results show that B. prostrata belongs to the NADP-ME type and core genes encoding for C₄ cycle are significantly upregulated when compared with Sed. sedoides and T. diffusa. Sedobassia sedoides and B. prostrata share a number of upregulated C₄-related genes; however, two C₄ transporters (DIT and TPT) are found significantly upregulated only in Sed. sedoides. Combined analysis of transcription factors (TFs) of the closely related lineages (Camphorosmeae and Salsoleae) revealed that no C₃-specific TFs are higher in C₂ species compared with C₄ species; instead, the C₂ species show their own set of upregulated TFs. Taken together, our study indicates that the hypothesis of the C₂ photosynthesis as a proxy towards C₄ photosynthesis is questionable in Sed. sedoides and more in favour of an independent evolutionary stable state.

1. Introduction

C₄ photosynthesis is a carbon-concentration mechanism, enhancing CO₂ at the site of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO). This mechanism leads to a decrease in the oxygenation reaction of RuBisCO, which in turn decreases photorespiration, because fewer toxic compounds resulting from the RuBisCO oxygenation reaction need to be recycled [1]. C₄ photosynthesis requires a series of biochemical, anatomical, and gene regulation changes compared with the C₃ photosynthesis ancestor [2,3]. C₄ photosynthesis has been a major subject in life science. Since the discovery of C₄ photosynthesis more than 50 years ago, its evolution is still under debate [4]. The current model of C₄ evolution relies heavily on the C₃–C₄ intermediate (including C₂) photosynthetic types as evolutionary stepping stones towards C₄ photosynthesis [5,6,7]. Most C₃–C₄ intermediate species utilise C₂ photosynthesis, where a photorespiratory glycine shuttle and its decarboxylation by glycine decarboxylase (GDC) concentrate CO₂ in a bundle sheath-like compartment [5]. The establishment of this glycine-based CO₂ pump and the restriction of GDC activity in the bundle sheath cells (BSCs) is considered an important intermediate step in the evolution towards C₄ photosynthesis [8]. However, the absence of C₄ relatives in lineages with C₃–C₄ intermediate phenotypes indicates that C₂ photosynthesis can be an evolutionarily stable state in their own right [9,10]. On the other hand, the hybrid origin of C₂ photosynthesis has been suggested because the anatomy of hybrids obtained from artificial crosses of Atriplex prostrata (C₃) and A. rosea (C₄) resemble the C₃–C₄ intermediate using a glycine shuttle to concentrate CO₂ [11].

Despite its complexity, the C₄ pathway independently evolved in at least 61 lineages of both monocot and eudicot lineages [12]. In eudicots, the Amaranthaceae/Chenopodiaceae alliance has the largest diversity of C₄ syndromes with 15 independent origins of C₄ identified, ten of which belong to Chenopodiaceae sensu stricto [12,13,14,15,16,17,18]. The closely related lineages Camphorosmeae and Salsoleae, belonging to the goosefoot family (Chenopodiaceae), are rich in C₄ phenotypes [18] and contain a number of C₃–C₄ intermediate species, including C₂, proto-Kranz type and C₄-like type species. Both lineages are found in steppes, semi-deserts, salt marshes, and ruderal sites of Eurasia, South Africa, North America, and Australia [18,19]. Camphorosmeae comprise subshrubs and annuals, predominantly with moderately to strongly succulent leaves with a central aqueous tissue [16]. Evolutionary radiation was later in Camphorosmeae (early Miocene) than in Salsoleae (early to middle Oligocene) [18]. In Camphorosmeae, C₄ photosynthesis likely evolved two times in the Miocene and different photosynthetic types are recognised on the basis of leaf anatomy with several C₄ phenotypes [18,20]. C₄ photosynthesis in Salsoleae likely evolved multiple times and most species are C₄ plants with terete leaves and Salsoloid Kranz anatomy in which a continuous dual layer of chlorenchyma cells encloses the vascular and water-storage tissue [18,21,22]. Therefore, these two sister groups constitute a central component allowing the investigation between and within each plant group for understanding the origin of C₂ photosynthesis, the evolution of C₄ photosynthesis and adjustments in gene regulation leading to different photosynthetic types. Indeed, new insight into C₄ evolution were gained from studying Salsoleae lineage using high-throughput sequencing methods. A photosynthetic transition from C₃ pathway in cotyledons to C₄ pathway in leaves of the Salsoleae lineage (Chenopodiaceae, Salsola soda L.) has been identified [23]. This C₃-to-C₄ transition is thought to be a rather exceptional phenomenon since species conducting C₄ pathway in all photosynthetically active tissues/organs are supposed to be the most abundant within this group. In addition, comparative transcriptomics revealed two proposed transporters associated with C₂ and C₄ photosynthesis [24]. However, for the Camphorosmeae lineage, transcriptome analysis and gene expression profiles of different photosynthesis types are still lacking. Moreover, for both lineages, differential gene expression of regulatory genes (e.g., transcription factors, TFs) involved in different photosynthetic pathways remains poorly understood.

The development of complex traits is controlled by the coordination of expression of many TFs and signaling pathways [25]. Thus, TFs play an important role in regulation of gene expression and are certainly responsible for the fine-tuning of the cell-specific expression patterns in C₄ photosynthesis [26]. The characteristic expression pattern of PHOSPHOENOLPYRUVATE CARBOXYLASE (PEPC) in C₄ plants (i.e., high abundance in mesophyll M and low abundance in BS cells), for example, could be controlled by a number of TFs from the ZINC FINGER HOMEODOMAIN (zf-HD) TF family [27]. Therefore, TFs are hot candidates for a stepwise evolutionary change of complex traits such as C₄ photosynthesis. Reviewing nine studies on potential regulators of C₄ photosynthesis in maize, Huang and Brutnell [28] found no TF consistently identified across these studies and suggested that consistent differential expression obtained between C₃ and C₄ sister lineages could be a more effective way to prioritise candidate TFs.

To fill this knowledge gap with new pieces of the puzzle, we (1) report transcriptome de novo assemblies and differential expression analysis between C₂, C₃, and C₄ species of Camphorosmeae (Amaranthaceae) using RNA-Seq, and (2) assess transcriptional regulator elements involved in C₃, C₂, and C₄ photosynthesis in the Amaranthaceae/Chenopodiaceae complex. In this latter objective, we merged transcriptome data generated in this study with the publicly available transcriptome data of C₂, C₃, and C₄ species of the sister lineage Salsoleae.

2. Results

2.1. Descriptive Statistics of RNA Data and De Novo Assembly

Between 27.4 and 37.9 million reads remained after quality filtering (98.68–99.06%) and were de novo assembled for each of the three species (Supplementary Dataset S1). Reduction in contigs by clustering resulted in 26,842 (Sed. sedoides, C₂), 33,1653 (T. diffusa, C₃), and 34,278 (B. prostrata, C₄) contigs with an open reading frame. The number of BUSCOs genes recovered was 88.1, 88.6 and 90.9% for T. diffusa, B. prostrata, and Sed. sedoides, respectively (Supplementary Table S1).

2.2. Differential Expression Genes (DEGs) within Camphorosmeae

In total, 10,513 transcripts were expressed in at least six of the eight species and downstream analyses focused on this dataset (Supplementary Dataset S2). Principal component analysis of log₂ transformed read counts (normalised to TPM) showed that replicates of each species in Camphorosmeae were very similar, whereas different species were clearly distinct from each other (Figure 1). The first principal component explained 51.47% of the total variation and Sed. sedoides (C₂) was positioned somewhere in between B. prostrata (C₄) and T. diffusa (C₃). This result was similar to what was found in Salsoleae [24] as the C₃–C₄ intermediate species was positioned in between the C₃ and the C₄ species. The second component, explaining 42.51% of total variation, separated Sed. sedoides from the other two species.

2.3. Functional Classification and Enrichment of DEGs within the Camphorosmeae

In all three species, transcriptional investment, defined as percentage of all read counts of transcripts (normalised to TPM) belonging to a particular MapMan category, was highest in the MapMan category ‘Not assigned’ (23.10–27.54%) (Figure 2, Supplementary Dataset S3). MapMan category ‘Photosynthesis’ was second highest in all three species; however, the amount differed between species from 11.24% in B. prostrata (C₄), to 15.03% in T. diffusa (C₃), up to 21.58% in Sed. sedoides (C₂) (Figure 2, Supplementary Dataset S3). In general, high transcriptional investment in the category ‘Photosynthesis’ in Sed. sedoides (C₂) was caused by higher transcription of many genes of the sub-category ‘Calvin cycle’. However, transcription of a gene encoding RUBISCO ACTIVASE in Arabidopsis (AT2G39730) was the main driver of the difference among species (Sed. sedoides compared with T. diffusa: log₂ fold change of 2.09; Sed. sedoides compared with B. prostrata: log₂ fold change of 3.72; Supplementary Dataset S2). Transcripts belonging to the categories ‘Protein biosynthesis’ (5.93–8.27%), ‘Protein degradation’ (4.45–5.98%), and ‘Protein modification’ (3.03–4.09%) were also highly abundant in all three species. In B. prostrata, the category ‘C_4′ was higher (transcriptional investment of 4.54%) compared with Sed. sedoides (1.09%) and T. diffusa (0.58%; Figure 2, Supplementary Dataset S3). ‘Photorespiration’ was about twice as high in Sed. sedoides and T. diffusa (2.59% and 2.21%, respectively) compared with B. prostrata (1.14%). This difference was caused by higher transcription of most genes of the category ‘photorespiration’ rather than a few genes (Supplementary Dataset S3). The categories with the lowest transcriptional investment in all three species were ‘DNA damage response’ (0.10–0.14%), ‘Polyamine metabolism’ (0.16–0.29%), and ‘Multi-process regulation’ (0.18–0.34%).

2.4. Differential Expression of C₄-Related Genes in C₃, C₂ and C₄ Camphorosmeae Species

As observed in other C₄ species, most genes encoding for proteins involved in C₄ photosynthesis were significantly upregulated in the C₄ species B. prostrata (C₄) when compared with Sed. sedoides (C₂) and T. diffusa (C₃) (

Table 1. Differential expression of C₄-related enzymes in leaves of Camphorosmeae species. T. diffusa (C₃), Sed. sedoides (C₂), B. prostrata (C₄).

		C2 vs. C4*		C3 vs. C4*		C3 vs. C2*
Transcripts	Genes	Log2FC	Padj	Log2FC	P	Log2FC	Padj
Bv_006710_gkqg.t1	MDH	1.29	1.01 × 10−10	2.84	9.92 × 10−42	1.55	1.74 × 10−14
Bv1_004490_tyfq.t1	PHT4	1.57	1.56 × 10−11	2.89	1.72 × 10−31	1.32	7.56 × 10−08
Bv1_013550_fjqs.t1	PPdK	3.37	2.32 × 10−54	4.63	2.78 × 10−90	1.26	8.92 × 10−10
Bv2_031080_twkf.t1	AspAt	1.32	2.84 × 10−06	2.15	1.44 × 10−13	0.82	0.003667
Bv3_049110_qgnh.t1	PHT1	3.16	8.05 × 10−08	-	-	-	-
Bv4_072630_xjai.t1	DIT	-	-	1.30	4.88 × 10−06	2.61	3.46 × 10−19
Bv5_117240_yhsk.t1	PPT	1.94	7.13 × 10−13	3.61	5.61 × 10−35	1.67	2.18 × 10−09
Bv6_135140_uyxu.t1	Asn Synthetase	2.56	5.60 × 10−07	2.42	1.93 × 10−06	-	-
Bv6_148840_uffy.t1	CA	3.80	6.66 × 10−38	3.81	4.08 × 10−38	-	-
Bv7_169130_kwer.t1	AlaAT	2.61	4.89 × 10−17	5.17	4.74 × 10−48	2.56	2.34 × 10−15
Bv8_182550_kstq.t1	TPT	-	-	1.48	1.20 × 10−10	2.29	2.71 × 10−22
Bv8_194450_rkme.t1	CA	-	-	-	-	1.48	0.000263
Bv8_195530_sxjq.t1	BASS2	3.51	5.00 × 10−68	3.97	6.26 × 10−83	0.46	0.018625
Bv8_200290_ujgk.t1	TPT			3.25	1.25 × 10−05	1.97	0.005513
Bv9_209750_xeaz.t1	PEPC	2.95	3.29 × 10−17	4.12	6.98 × 10−29	1.18	0.000429
Bv9_215520_prze.t1	DIT	1.48	1.23 × 10−08	2.63	5.15 × 10−22	1.16	2.27 × 10−05
Bv9_224000_xpgi.t1	NHD	2.48	1.17 × 10−21	2.68	7.04 × 10−25	-	-
Bv9_224840_zmjw.t1	NADP-ME	1.29	8.44 × 10−08	3.86	9.54 × 10−47	2.56	1.34 × 10−23

- not significantly expressed Padj > 0.05.

, Supplementary Dataset S2–S4). Out of 18 C₄-related transcripts, 16 and 13 transcripts encoding known C₄ cycle proteins were significantly upregulated (p < 0.001) in B. prostrata (C₄) compared with T. diffusa (C₃) and Sed. sedoides (C₂), respectively. ALANINE AMINOTRANSFERASE (AlaAT, Log₂FC = 5.17) was the most abundant followed by PYRUVATE ORTHOPHOSPHATE DIKINASE (PPdK, Log₂FC = 4.63), PHOSPHOENOLPYRUVATE CARBOXYLASE (PEPC, Log₂FC =4.12), BILE ACID:SODIUM SYMPORTER FAMILY PROTEIN 2 (BASS2, Log₂FC = 3.97), NADP-malic enzyme (NADP-ME Log₂FC = 3.86), PEP/phosphate translocator (PPT, Log₂FC = 3.61) in B. prostrata (C₄) as compared with T. diffusa (C₃). Conversely, BASS2 Log₂FC = 3.51, PPdK Log₂FC = 3.37, PHOSPHATE TRANSPORTER 1 (PHT1, Log₂FC = 3.16), PEPC (Log₂FC = 2.95), and AlaAT (Log₂FC = 2.61) were in the top five of highly upregulated genes in B. prostrata (C₄) when compared with Sed. sedoides (C₂). However, not all C₄-related geneswere significantly upregulated in B. prostrata (C₄) as compared with Sed. sedoides (C₂). TRIOSE PHOSPHATE TRANSLOCATORransporters [TPT (Bv8_194450_rkme.t1), a chloroplast dicarboxylate transporter isoform [DIT (Bv4_072630_xjai.t1)] and a CARBONIC ANHYDRASE isoform [CA (Bv8_194450_rkme.t1)] were significantly upregulated in Sed. sedoides (C₂) when compared with B. prostrata (C₄) (Supplementary Dataset S2). Interesting, PHT1 (Bv3_049110_qgnh.t1) was significantly upregulated in T. diffusa (C₃) as compared with Sed. sedoides (C₂) (Supplementary Dataset S2).

Twelve of the C₄-related enzymes except ASPARAGINE SYNTHETASE (Asn Synthetase), SODIUM:HYDROGEN ANTIPORT (NHD) and a CARBONIC ANHYDRASE (CA) isoform (Bv6_148840_uffy.t1) were significantly upregulated in Sed. sedoides (C₂) compared with T. diffusa (C₃). These enzymes include typical C₄ enzymes such as PEPC, NADP-ME, PPdK, PHT4, Ala-AT, and ASPARTATE AMINOTRANSFERASE (Asp-AT), as well as C₄-associated transporters such as BASS2 and DIT.

2.5. Differential Expression of Key Photorespiration Genes in C₃, C₂ and C₄ Camphorosmeae Species

Out of 14 transcripts associated with photorespiratory enzymes, 12 were annotated and assigned (

Table 2. Differential expression of photorespiratory transcripts in leaves between Camphorosmeae species. T. diffusa (C₃), Sed. sedoides (C₂), B. prostrata (C₄).

		C2* vs. C4		C3 vs. C2*		C3* vs. C4
Transcripts	Genes	Log2FC	Padj	Log2FC	Padj	Log2FC	Padj
Bv5_106360_ipey.t1	GDC-H	1.862968	3.37 × 10−15	0.990549	1.80 × 10−05	0.872432	0.000167
Bv_012000_yknj.t1	GDC-P	1.253616	9.20 × 10−08	-	-	2.068131	5.39 × 10−18
Bv3_059720_tshd.t1	GDC-L	1.281024	1.13 × 10−11	-	-	0.953812	3.95 × 10−07
Bv3_065510_eeis.t1	SHMT	-	-	1.133566	0.000228	-	-
Bv4_073470_iswc.t1	AGT/SGT	1.254967	2.18 × 10−07	-	-	1.130439	2.83 × 10−06
Bv4_074740_miaa.t1	PGP	1.831751	1.99 × 10−16	0.505181	0.019216	1.326602	1.77 × 10−09
Bv4_094290_jgpp.t1	GOX	1.708751	1.51 × 10−18	0.993515	1.94 × 10−07	0.715241	0.000175
Bv5_107350_ydma.t1		2.435018	1.25 × 10−13	-	-	2.32744	1.09 × 10−12
Bv6_127540_qdph.t1	GDC-T	2.464812	2.21 × 10−23	1.030248	1.32 × 10−05	1.434602	2.66 × 10−09
Bv6_148110_nuir.t1	GGT	2.380809	3.21 × 10−21	1.223501	4.63 × 10−07	1.157324	2.02 × 10−06
Bv6_152820_wtfn.t1	SHMT	1.721992	2.64 × 10−18	0.734934	0.000137	0.987067	3.75 × 10−07
Bv8_184280_guso.t1		0.677745	0.006074	-	-	1.06935	3.85 × 10−06
Bv9_213980_zwen.t1	HPR	0.96884	0.000282	-	-	0.531187	0.045313
Bv9_220360_xogt.t2	GLYK	0.689722	0.001207	0.708265	0.000885	-	-

- not significantly expressed Padj > 0.05.

). All 12 photorespiratory transcripts were significantly upregulated in Sed. sedoides (C₂) as compared with B. prostrata (C₄), including the core photorespiratory enzymes GLYCINE DECARBOXYLASE (GDC, T-, H-, P-, L-), GLUTAMATE:GLYOXYLATE AMINOTRANSFERASE (GGT) and SERINE HYDROXYMETHYLTRANSFERASE (SHMT). In Sed. sedoides (C₂), GGT, two SHMTs, GDC-T, GLYCOLATE OXIDASE (GOX), GDC-H, GLYCERATE 3-KINASE (GLYK), PHOSPHOGLYCOLATE PHOSPHATASE (PGP) were significantly upregulated when compared with T. diffusa (C₃). Only one SHMT isoform (Bv6_143730_mggd.t1) was significantly upregulated in B. prostrata (C₄) when compared with Sed. sedoides (C₂) (Supplementary Dataset S2). In T. diffusa (C₃), only one gene GDC-P was significantly upregulated as compared with Sed. sedoides (C₂) (Supplementary Dataset S2). All photorespiratory genes except GLYK significantly expressed in Sed. sedoides (C₂), compared with B. prostrata (C₄), were significantly upregulated in T. diffusa (C₃) when compared with B. prostrata (C₄).

2.6. Regulatory Elements in C₃, C₂, and C₄ Species of the Amaranthaceae/Chenopodiaceae Alliance

To identify regulatory genes putatively involved in the formation/regulation of C₂ and C₄ photosynthesis, expression patterns of TFs from The Plant Transcription Factor Database v.5.0 (PlantTFDB; [30,31]) using 1163 annotated TFs from Beta vulgaris (version ‘BeetSet-2′, [32]) were investigated. From the whole set of TFs included in PlantTFDB, 824 orthologous TFs were found, of which 494 TFs had orthologs in all eight species (three species from Camphorosmeae: B. prostrata (C₄), Sed. sedoides (C₂), T. diffusa (C₃); five species from Salsoleae: H. soparia (C₄), Sal. divaricata Pop-184 (C₂), Sal. divaricata Pop-198 (C₂), Sal. oppositifolia (C₄), Sal. soda (C₃/C₄), Sal. webbii (C₃)) and were further analysed using Clust. Based on the expression pattern, from the initial 494 TFs, 431 TFs were grouped into 11 clusters with between 22 and 71 genes per cluster (Cluster-C0 to Cluster-10; Figure 3, Supplementary Figure S1). Nine of the 11 clusters were of particular interest, because these clusters included TFs that were highly expressed in each photosynthetic type when compared with others (Figure 3). Cluster-C4, Cluster-C5, and Cluster-C6 contained TFs that were highly expressed in C₄ species as compared with C₃ and C₂ species, whereas Cluster-C0, Cluster-C9, and Cluster-C10 comprised TFs that were highly expressed in C₃ species when compared with C₂ and C₄ species. Finally, Cluster-C1, Cluster-C2, Cluster-C3 encompassed TFs that were highly expressed in C₂ species as compared with C₃ and C₄ species.

Cluster-C4, Cluster-C5, and Cluster-C6 consisted of 71, 33 and 35 TFs, respectively, from 41 different TF families (Figure 3, Supplementary Dataset S5). Four TFs, all part of Cluster-C4, were present in all eight species, with the transcripts significantly (adjusted p-value ≤ 0.01) more abundant in all C₄ species compared with the two C₃ species (

Table 3. Differentially expressed C₄-related TFs in Amaranthaceae/Chenopodiaceae.

		Cluster-4
		BBX15 (CO-Like)		SHR (GRAS)		SCZ (HSF)		LBD41 (LBD)
Lineage	Species (C3 vs. C4*)	LF2C	Padj	LF2C	Padj	LF2C	Padj	LF2C	Padj
Salsoleae	Salweb vs. Hsco	8.8	3.84 × 10−24	2.23	1.55 × 10−08	5.59	1.73 × 10−05	3.71	6.05 × 10−13
	Salweb vs. Salopp	11.86	6.56 × 10−61	1.95	2.56 × 10−06	5.09	0.000374	4.26	1.64 × 10−17
	Salweb vs. Salsod	11.57	6.71 × 10−62	2.63	1.85 × 10−13	7.66	4.94 × 10−20	2.84	2.06 × 10−07
Camphorosmeae	Tdif vs. Bpro	10.99	3.77 × 10−51	4.21	9.49 × 10−12	3.4	2.71 × 10−06	6.71	5.11 × 10−10
	Salweb vs. Bpro	10.82	2.50 × 10−47	1.45	0.0011	6.12	1.58 × 10−07	2.22	0.00019
Salso. × Camph.	Tdif vs. Hsco	8.96	2.75 × 10−27	4.99	2.89 × 10−19	2.84	0.000272	8.23	1.45 × 10−21
	Tdif vs. Salsod	11.72	1.85 × 10−63	5.34	5.60 × 10−25	5.25	3.76 × 10−21	6.96	1.91 × 10−12
	Tdif vs. Salopp	12.03	6.61 × 10−65	4.7	1.56 × 10−15	2.44	0.00602	8.68	3.84 × 10−26

). These TFs comprised BBX15 (CO-like family, AT1G25440.1), SHR (TF family GRAS), SCZ (TF family HSF), and LBD41 (TF family LBD). Cluster-C0, Cluster-C9, and Cluster-C10, respectively, comprised 57, 35, and 34 TFs (Figure 3, Supplementary Dataset S5). Here, two TFs of Cluster-10 (HSF, and NAC) and one of Cluster-9 (HD-ZIP) were significantly abundant in the studied C₃ species when compared with C₄ species (

Table 4. Differentially expressed C₃-related TFs in Amaranthaceae/Chenopodiaceae.

		Cluster-9		Cluster-10
		ATHB13 (HD-ZIP)		HSFA6B (HSF)		NAC083 (NAC)
Lineage	Species (C4 vs. C3*)	LF2C	Padj	LF2C	Padj	LF2C	Padj
Salsoleae	Hsco vs. Salweb	9.02	5.45 × 10−24	1.21	0.00017	0.73	0.00398
	Salopp vs. Salweb	1.71	2.77 × 10−05	1.54	1.32 × 10−05	1.42	4.44 × 10−08
	Salsod vs. Salweb	1.14	0.00077	1.43	1.65 × 10−06	0.76	0.0004
Camphorosmeae	Bpro vs. Tdif	8.89	2.80 × 10−23	1.7	1.08 × 10−06	11.74	3.28 × 10−82
	Bpro vs. Salweb	8.99	8.45 × 10−24	1.91	5.76 × 10−08	11.28	1.31 × 10−72
Salso. × Camph.	Hsco vs. Tdif	8.92	1.78 × 10−23	1	0.00169	1.01	5.11 × 10−05
	Salsod vs. Tdif	0.87	0.00956	1.23	4.09 × 10−05	1.04	6.71 × 10−07
	Salopp vs. Tdif	1.45	0.00038	1.34	0.00011	1.71	4.00 × 10−11

).

Cluster-C1, Cluster-C2, and Cluster-C3 included 22, 32, and 38 TFs, respectively, of which one TF of Cluster-2 bHLH 106 (TF family bHLH) was significantly higher in the C₂ species when compared with the C₃ species and C₄ species (

Table 5. Differentially expressed C₂-related TF in Amaranthaceae/Chenopodiaceae.

		Cluster-2
		bHLH106 (bHLH)
Lineage	Species (C3 vs. C2*)	LF2C	Padj
Salsoleae	Sweb vs. Saldi1	1.68	8.90 × 10−13
	Sweb vs. Saldi2	1.63	4.07 × 10−11
Camphorosmeae	Tdif vs. Sedsed	0.45	0.04616
Salso. × Camph.	Tdif vs. Sdi1	1.06	3.33 × 10−07
	Tdif vs. Sdi2	1.01	2.79 × 10−06
	Sweb vs. Sedsed	1.07	4.19 × 10−05
	Species (C4 vs. C2*)	LF2C	Padj
Salsoleae	Hsco vs. Saldi1	1.37	1.23 × 10−10
	Hsco vs. Saldi 2	1.32	3.58 × 10−09
	Salopp vs. Saldi1	2.17	7.37 × 10−19
	Salopp vs. Saldi2	2.12	1.03 × 10−16
	Salsod vs. Saldi1	1.21	3.24 × 10−10
	Salsod vs. Saldi2	1.16	7.30 × 10−09
Camphorosmeae	Bpro vs. Sedsed	0.56	0.01398
Salso. × Camph.	Bpro vs. Saldi1	1.17	2.14 × 10−08
	Bpro vs. Saldi2	1.12	2.21 × 10−07
	Hsco vs. Sedsed	0.76	0.001176
	Salopp vs. Sedsed	1.56	9.00 × 10−09
	Salsod vs. Sedsed	0.6	0.00483

, Supplementary Dataset S2–S5). To assess the integration of C₃ and C₄ pathways into the intermediate C₂ pathway at the regulation level and vice versa, specific TFs of C₄ (Cluster-C4, Cluster-C5 and Cluster-C6) and C₃ (Cluster-C0, Cluster-C9, and Cluster-C10) pathways were assessed in the following comparisons: C₂ species vs. C₃ species and C₂ species vs. C₄ species. Then, specific TFs of the C₂ pathway (Cluster-C1, Cluster-C2 and Cluster-C3) were estimated in the pairwise comparison of C₃ species vs. C₄ species. Among the four TFs common to all C₄ species, only one TF (BBX15, TF family CO-like) was significantly upregulated in C₂ species when compared with C₃ species (Supplementary Dataset S2–S5). Conversely, no TF of C₃ species was highly expressed in C₂ species when compared with C₄ species.

3. Discussion

3.1. Transcriptome Analysis in Camphorosmeae

Gene expression analysis predominantly paved the way to understand the difference between derived photosynthetic types (C₂, C₄) and the ancestral C₃ photosynthesis [24,34,35]. Much progress in understanding C₄ and C₂ photosynthesis was achieved by comparing differentially expressed genes of closely related species in the genus Flaveria (Asteraceae) considered as a model organism to study the evolution of C₄ photosynthesis [6,7,34,35,36,37]. The goosefoot family (Chenopodiaceae) has a large number of C₂ and C₄ species that differ anatomically and ecologically from Flaveria. This family therefore represents a good supplementary alternative to decipher the convergent evolution of C₄ photosynthesis. With PCA based on gene expression, it was possible to clearly distinguish between T. diffusa (C₃), representing the ancestral condition, and Sed. sedoides (C₂) and B. prostrata (C₄), representing derived conditions. The physiologically C₃-C₄ intermediate Sed. sedoides (C₂) was positioned in a triangle with T. diffusa (C₃) and B. prostrata (C₄) in terms of transcript variation. This result was similar to what was found in Salsoleae [24]. The first three components explained about 76% of the total variation, which was slightly higher than the 73% reported in Salsoleae [24]. Similar to Salsoleae, in Camphorosmeae, the three different photosynthesis types predominantly structure the gene expression pattern in assimilating tissue. Indeed, species with C₃, C₂, and C₄ photosynthesis differ in leaf anatomical structure. T. diffusa (C₃) exhibits the Neokochia type characterised by an undifferentiated chlorenchyma of several layers. Sed. sedoides (C₂) has the Sedobassia type consisting of kranz-like cells near peripheral vascular bundles. B. prostrata (C₄) depicts the Bassia prostrata type with the chlorenchyma differentiated in an outer mesophyll and inner kranz-layer [20]. In contrast, the first three PCA components in a comparable study of Flaveria explained only 27% [35]. This difference could be due to the younger evolutionary age or other confounding factors affecting gene expression in the Flaveria study as suggested by Lauterbach et al. [24].

3.2. C₄ Key Enzymes in C₄ and C₂ Camphorosmeae Species

Analyses of differential gene expression between C₃ and C₄ species of Camphorosmeae showed that core C₄ cycle proteins were highly abundant in B. prostrata (C₄). Similar results were found in Cleome [38], Flaveria [34,35] and Salsoleae [24]. Traditionally, three biochemical subtypes of C₄ photosynthesis are classified according to the predominant type of decarboxylation releasing CO₂ around RUBisCo in the BSCs: NAD-ME, NADP-ME, and PEP-CK. However, PEPCK should be considered as a supplemental subtype to either NAD-ME or NADP-ME [39]. Significant expression of NADP-ME indicates that B. prostrata (C₄) uses a NADP-ME type C₄ cycle. Asparagine synthetase (ASN) and NHD were found significantly expressed and upregulated only in B. prostrata (C₄) as compared with T. diffusa (C₃) and Sed. sedoides (C₂). ASN was reported upregulated in C₄ species Gynandropsis gynandra when compared with closely related C₃ species Tarenaya hassleriana (Cleomaceae), as well as in C₄ leaves of Sal. soda when compared with its C₃ cotyledones [23]. On the other hand, NHD was found upregulated in C₄ species compared with C₃ and C₂ species of Flaveria [35]. Moreover, the top three highly expressed C₄ enzymes in B. prostrata (C₄) as compared with T. diffusa (C₃) were Ala-AT, PPDK, and BASS2. ASN is involved in ammonium metabolism and asparagine in nitrogen transport [24]. Achievement of the C₄ cycle requires the transport of pyruvate to the mesophyll cell (MC) for regeneration of PEP. While Ala-AT plays an important role in pyruvate generation, PPDK intervenes in the regeneration of PEP. Pyruvate transport is mediated by the BASS2/NHD transport system [40]. Taken all together, this indicates not only a possible functional connection between nitrogen metabolism and the switch from C3 to C4 pathway as suggested by Lauterbach et al. [24] and Mallmann et al. [35], but also the capacity to shuttle pyruvate from the BS plastid. In this regard, the pyruvate shuttle ensures the regeneration of the CO₂ acceptor (PEP), and therefore maintains the C4 pathway.

Eleven C₄-related genes were found significantly upregulated in Sed. sedoides (C₂) compared with T. diffusa (C₃), including, for example, PEPC, NADP-ME, PPdK, and PHT4. Upregulation of C₄ typical enzymes such as PEPC, NADP-ME, PPdK, PPT was also reported in the C₂ species when compared with the C₃ species in studies of Flaveria and Salsoleae [24,35]. This result suggests that genes associated with the C₄ cycle are present in Sed. sedoides (C₂) and play an important role in C₂ metabolism. A DIT isoform (Bv4_072630_xjai.t1), NADP-ME and Ala-AT were the three most upregulated C₄ enzymes in Sed. sedoides (C₂) as compared with T. diffusa (C₃). Moreover, we found two transporters (TPT and DIT) upregulated in Sed. sedoides (C₂) when compared with B. prostrata (C₄). These transporters were found highly expressed in some C₂ species when compared with C₄ species in Flaveria [35]. DIT is a putative plastidial dicarboxylate transporter and TPT is the chloroplast envelope triose-phosphate/phosphate translocator (TPT) [41]. Based on simulated data, it was shown that a high TPT capacity is required to obtain high assimilation rates and to decrease the CO₂ leakage from BSCs to MCs [39]. The most likely reason for upregulation of these genes is their involvement in decreasing the CO₂ leakage from the Kranz-like cells back to the MCs due to the presence of RuBisCo, which is not the case for C₄ plants. This explains the low CO₂ compensation observed in C₂ species [42]. Thus, C2 plants upregulate a distinct set of C₄ enzymes to handle constraints related to the C₂ pathway and not an entirely congruent set. This does not support their interpretation as an intermediate state towards C4 photosynthesis, but is more in line with their interpretation as an independent evolutionarily stable state ([43] and refs. therein).

3.3. C₄ Key Enzymes in C₄ and C₂ Camphorosmeae Species

Transcripts associated with photorespiration were about twice as abundant in C₃-C₄ intermediate (Sed. sedoides) and T. diffusa (C₃) compared with B. prostrata (C₄). Likewise, we found key photorespiration enzymes were differentially expressed and upregulated in the C₂ species (Sed. sedoides) when compared with T. diffusa (C₃) and B. prostrata (C₄). This corroborates the expression patterns reported in the C₂ species of Flaveria [34,35] and Salsoleae [24], implying a successful integration of C₂ photosynthesis in Sed. sedoides (C₂). Our transcript data showed that GDC-P and a SHMT isoform (Bv6_143730_mggd.t1) were downregulated in Sed. sedoides (C₂) when compared with T. diffusa (C₃) and B. prostrata (C₄), respectively. Similar results were obtained in the C₂ species of the genus Flaveria [35]. Schulze et al. [36] showed that downregulation of GDC-P was closely linked to the establishment of the C₂ pathway in Flaveria. Since GDC-P and a SHMT isoform are known to be involved in glycine decarboxylation, their downregulation in Sed. sedoides (C₂) might have similar consequences. It is worth noticing that the number of significantly upregulated photorespiratory genes in Sed. sedoides (C₂) was equal to T. diffusa (C₃) when compared with B. prostrata (C₄).

A significant reduction in almost all photorespiratory genes was observed in B. prostrata (C₄). All photorespiratory genes except GLYK were downregulated in B. prostrata (C₄) as compared with T. diffusa (C₃). Mallmann et al. [35] reported significant downregulation of all photorespiratory genes in C₄ Flaveria except the transport proteins DIT1 and DIT2 and one isoform of GLDH. On the other hand, GLYK was expressed in the M of C₄ Sorghum bicolor [44]. GLYK catalyses the regeneration of 3-phosphoglycerate (3-PG). The localisation of GLYK within the leaf cells of B. prostrata (C₄) could clarify its high expression and role.

3.4. Regulation of C₃, C₂ and C₄ Photosynthesis in Amaranthaceae/Chenopodiaceae

Transcription factors are proteins that bind to the DNA promoter or enhancer regions of specific genes and regulate their expression. They have a crucial role on plant growth, development and adaptation under various stress conditions, and therefore are excellent candidates for modifying complex traits in plants [45]. C₃, C₂ and C₄ species of Salsoleae and Camphorosmeae are widely spread in desert, semi-desert, saline, and arid regions [18,19]. In former Chenopodiaceae, C₄ photosynthesis evolved as an adaptation to hot, dry, or saline areas from the C₃ ancestor which was already preadapted to grow in these harsh environments [15]. We focused on TFs that were differentially expressed between C₃, C₂, and C₄ species/states irrespective of the lineage, to further reduce the amount of differentially expressed TFs to a small subset of actually C₄-, C₂-, and C₃-related changes. Indeed, a small number of TFs were found differentially expressed between C₃, C₂, and C₄ species/states.

Cluster analysis showed that BBX15, SHR, SCZ, and LBD41 were co-regulated and significantly more abundant in all C₄ species irrespective of the lineage when compared with C₃ species. The families to which these TF families belong play an important role in regulatory networks controlling plant growth and development, and plant adaptive responses to various environmental stress conditions [46,47,48,49,50]. Except the LBD TF family, the SHR, HSF, and CO-like families have been shown to be involved in the development of C4 Kranz anatomy in Zea mays L. and potentially involved in the establishment of C₄ M and Kranz cell identities [51,52,53,54,55]. However, members of the LBD TF family are key regulators of plant organ development, leaf development, pollen development, plant regeneration, stress response, and anthocyanin and nitrogen metabolisms [50,56]. Since mRNA of all the four TFs was highly abundant in C₄ species and co-regulated, our data suggest a critical role of these TFs in the development of any C₄ Kranz anatomy in the Amaranthaceae/Chenopodiaceae complex.

We found that C₃ species enhanced different TFs compared with C₄ species. Three TFs (ATHB13, HD-ZIP family), HSFA6B (HSF TF family), and NAC083 (NAC TF family), in which two TFs (HSF6B and NAC083) are co-regulated, were significantly higher in all C₃ species when compared with C₄ species. As C₄ TFs, they are involved in plant growth, development, and stress tolerance. The NAC TF family was shown to contribute to root and shoot apical meristems formation in Arabidopsis [57,58], organogenesis [59], salt and drought tolerance in Arabidopsis [60], leaf senescence in tobacco [61], and secondary cell wall formation in cotton [62]. The HD-Zip TF family was reported to regulate plant growth adaptation to abiotic stress such as salt and drought in apple and Arabidopsis [63,64]. Interestingly, HD-ZIP, HSF, and NAC TF families were suggested to control the C₄ photosynthesis in maize and rice [55,65]. However, in these studies, these TF families were detected using development gradient transcriptome comparison only on C₄ maize and rice plants. This may imply higher activity of these TF families in C₃ species. Nevertheless, different transcripts of these TFs families were involved when compared with the present study. Thus, a significant expression of these TFs in C₃ species could indicate a potential function of these TFs in the C₃ pathway.

In C₂ species, one transcript of the BASIC HELIX-LOOP-HELIX (bHLH106) protein from the bHLH TF family was found to be upregulated compared with C₄ and C₃ species. Two TFs of the bHLH TF family were shown to regulate a C₄ photosynthesis gene in maize [66]. This upregulation of bHLH106 in all C₂ species may suggest its possible role in the development and establishment of the C₂ photosynthesis specificities relative to other photosynthesis types. Interestingly, one C₄-specific TF (BBX15, TF family CO-like) was significantly higher in C₂ species when compared with C₃ species. Thus, this TF could be responsible for similarities of C₄ photosynthesis found in C₂ species such as the Kranz-like anatomy. Surprisingly, no C₃-specific TF was significantly expressed in C₂ species when compared with C₄ species. This indicates that C₂ and C₄ photosynthesis represent more derived types of photosynthesis compared with C₃ photosynthesis. Nonetheless, this seems to be inconsistent with the current model of C₄ evolution which relies heavily on the interpretation of the physiological intermediacy of C₂ photosynthesis as an evolutionary stepping stone to C₄ [8]. One would expect C₃-specific TFs to be higher in C₂ species when compared with C₄ species if the C₂ photosynthetic type represents an intermediate step along the evolution of C₃-to-C₄ photosynthesis as revealed by differential expression analysis of core photorespiratory genes in C₂ and C₃ species of Camphorosmeae (this study) and Salsoleae [24].

Taking the results of this study together, the unique derived TF profile of the C₂ intermediate species suggests an evolutionarily stable state in its own right. Similarities with C₄ relatives might result from a hybrid origin involving C₃ and C₄ parental lineages, parallel recruitment of a number of TFs in C₄ and C₂ lineages or common ancestry, and later divergent evolution. The position of Sedobassia as sister to Bassia (all C₄) allows all of these three scenarios [16]. For C₂ species in Salsoleae, however, phylogenomic evidence points to a hybrid origin of the Sal. divaricata agg. (Tefarikis et al., in prep.). Further phylogenomic analyses are needed to discern if an early hybridisation event of a C₄ (or ancestral preadapted C₄) lineage and a C₃ lineage led to the origin of the Sedobassia lineage which then evolved towards stable C₂ photosynthesis.

4. Materials and Methods

4.1. Plant Material

Plants of three Camphorosmeae species (Bassia prostrata (L.) Beck (C₄), Sedobassia sedoides (Schrad.) Freitag and G. Kadereit (C₂), and Threlkeldia diffusa R.Br. (C₃) (Figure 4, for voucher information see Supplementary Table S2) were grown from seeds in custom mixed potting soil in a glasshouse at the Botanic Garden, Johannes Gutenberg University Mainz, Germany at a minimum temperature of 18 °C in the night. Daytime temperatures varied from 25 to 35 °C in the summer and from 20 to 25 °C in the winter. Plants were watered once a week in the winter and twice a week in the summer and kept at 16 h light/ 10 h dark with natural light and an additional light intensity of ca. 300 µmol m−2 s−1. Leaf samples of the three species were harvested between 16th April and 16th May 2014 between 10:30 a.m. and 13:00 p.m., immediately frozen in liquid nitrogen, and stored at −80 °C for RNA extraction.

4.2. RNA Isolation and Sequencing

Total RNA extraction, library preparation, and mRNA sequencing were performed as described by Lauterbach et al. [23,24]. Total RNA was extracted from 16–55 mg leaf tissue of B. prostrata, Sed. sedoides, and T. diffusa. Sequencing of 101 bp single-end reads was performed on an Illumina HiSeq2000 platform. For each species, three individuals were sequenced (i.e., biological triplicates). Sequencing reads of these three species are available under study accession PRJEB36559.

4.3. Data Access

RNA-Seq data of the five Salsoleae species were retrieved from Lauterbach et al. (2017 a, b; study accession numbers PRJNA321979 and PRJEB22023) (Figure 4, Supplementary Table S2). These data comprise: cotyledons, and first and second leaf pair of Salsola soda (C₃/C₄), cotyledons and leaves of the Salsola divaricata population 184 (C₂, Pop-184), Salsola divaricata population 198 (C₂, Pop-198), and Salsola oppositifolia (C₄); leaves of Salsola webbii (C₃); and assimilating shoots of Hammada scoparia (C₄). For all of these samples, triplicates per species and organ were available [23,24].

4.4. RNA-Seq Data Processing

Single-end sequencing reads were checked for quality using the FASTQC tool (www.bioinformatics.babraham.ac.uk/projects/fastqc/, accessed on 15 March 2021), and filtered and trimmed using Trimmomatic v.0.38 [67]. For each species, de novo assembly was conducted using quality-filtered reads of all replicates of leaves and, where present, cotyledons of the respective species with default parameters in Trinity v.2.1.1 [68]. Quality of assemblies were assessed with BUSCO v.3.0 (Benchmarking Universal Single-Copy Orthologs, [69]) using the ‘Eudicotyledons odb10′ dataset [70]. Number of contigs of de novo assemblies were reduced by clustering via CD-HIT-EST v.4.7 [71,72] and only contigs with an open reading frame were included in the downstream analysis, which was conducted with TransDecoder v.5.3.0 (github.com/TransDecoder/TransDecoder accessed on 29 February 2020) followed by another round of CD-HIT-EST. Orthology assignment between the nine de novo assemblies was carried out by conditional reciprocal best (crb) BLAST v.0.6.9 [26] run locally using protein-coding sequences of Beta vulgaris (version ‘BeetSet-2′, [32]) as a reference. Only contigs were included in downstream analyses that had ortholog assignments between at least six of the eight species. Besides ‘BeetSet-2′ from Beta vulgaris, contigs were annotated using Arabidopsis (TAIR10). Reads of each of the replicates were separately mapped against these reduced data sets via bowtie2 v.2.3.4.1 [73]. Re-formatting and final extraction of read counts (excluding supplementary alignments) were carried out in Samtools v.1.3 [74].

4.5. Differential Gene Expression Analysis

Read counts were normalised into transcripts per million (TPM) and used for differential gene expression analysis. Here, pairwise comparison between all eight species was statistically evaluated using edgeR [75] in R (R Core Team, 2018). Hierarchical clustering using Pearson’s correlation and principal component analysis of log₂ transformed read counts were carried out with Multiexperiment Viewer (MeV) v.4.9 (http://mev.tm4.org/ accessed on 5 February 2020). Co-expressed gene clusters of (1) all expressed transcripts and (2) transcripts annotated as transcription factors were carried out with Clust v.1.10.7 [33]. Pathways were defined in MapMan4 categories [29] with the additional category ‘C_4′. To identify TFs putatively involved in the formation/regulation of C₂ and C₄ photosynthesis, the 1163 annotated TFs from Beta vulgaris (version ‘BeetSet-2′, [32]) from The Plant Transcription Factor Database v.5.0 (PlantTFDB; [30,31]) were used. Here, two different datasets were combined (1: leaf transcriptome data of the three Camphorosmeae species T. diffusa (C₃), Sed. sedoides (C₂), and B. prostrata (C₄); 2: leaf transcriptome data of the five Salsoleae species Sal. webbii (C₃), Sal. divaricata Pop-184 (C₂), Sal. divaricata Pop-198 (C₂), H. scoparia (C₄), Sal. oppositifolia (C₄), and Sal. soda (C₄); for study accession numbers see above) and transcribed TFs grouped using Clust v.1.10.7 [33] and grouping all samples based on the photosynthetic type into the three conditions C₃, C₂, and C₄. VENNY v. 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/ accessed on 12 July 2021) were deployed to find intersected TFs across all pairwise comparisons.

5. Conclusions

The transcriptome data of the Chenopodiaceae family provided new insight into C₄ evolution. Proteins encoding for C₄ transporters (DIT and TPT) were found significantly upregulated in Sed. sedoides (C₂) when compared with B. prostrata (C₄). Upregulation of those transporters reduces CO₂ leakage from BSCs to MC, which could otherwise be detrimental to C₂ photosynthesis due to the presence of RuBisCo in the MC. This suggests evolution of a stable C₂ photosynthesis independent of C₄ photosynthesis. Combined analysis of TFs of the sister lineages provides further support of this result. Indeed, while one C₄-specific TF (BBX15) was significantly higher in C₂ species when compared with C₃ species, no C₃-specific TFs were higher in C₂ species compared with C₄ species. Finally, apart from well-known TFs involved in the development of C₄ Kranz anatomy such as SHR, BBX15, SCZ, and LBD41 may also be associated with its development and physiology. Furthermore, bHLH106 could be related to specific C₂ anatomy and BBX15 to a characteristic C₄-like expression pattern found in species with C₂ photosynthesis. This study sheds light on the differentiated regulation and evolution of transcription factors in C₂ and C₄ photosynthesis.