A 427 bp long fragment of the red fluorescent protein (dsRed) of the mushroom anemona Discosoma spec. was cloned into pCRII vector by the Bucher Lab. Additionally, 300 bp of Rpt3 were cloned into pJet 1.2 (Thermo Fisher Scientific, Waltham, MA, USA; Ulrich et al., 2015). For RNAi experiments targeting Tc-CRLF3, primer pairs for two different fragments targeting au3.g2971.t1 (Tc000209) were designed (Table 1). The fragment 1 covers 638 bp whereas the fragment 2 is smaller and consists of 527 bp (sequences shown in Supplementary Figure S1). The fragments were cloned into the pCRII vector using the TA Cloning^® Kit (Invitrogen, Life Technologies, Darmstadt, Germany). Subsequently, the plasmids were transformed into DH5α cells and finally purified with the NucleoSpin^® Plasmid Kit (Macherey-Nagel, Düren, Germany).

Template DNA was prepared by PCR using M13F and M13R primers with a T7 RNA polymerase promoter sequence attached to the latter (Table 1). In vitro transcription of dsRNA was performed with the MEGAscript^® T7 Transcription Kit following the user manual (Life Technologies, Darmstadt, Germany). The resulting RNA pellet was resuspended in injection buffer (1.4 mM NaCl, 0.07 mM Na₂HPO₄, 0.03 mM KH₂PO₄, 4 mM KCl). In order to anneal both strands of the RNA, the mix was incubated at 94°C for 5 min and cooled down slowly in a heatblock. Before use in cell culture experiments it was sterile filtered (Millex^®-HV, 0.45 μm, hydrophilic durapore, Millipore, Darmstadt, Germany).

Information about Tribolium gene sequence and structure was gained from http://bioinf.uni-greifswald.de/gb2/gbrowse/tribolium4/ working with the augustus3 prediction. Sequences were identified with Basic Local Alignment Search Tool (BLAST; Altschul et al., 1990) using the matrix BLOSUM62 with standard settings and the mode tblastn. Sequence information of other species was retrieved with BLAST at NCBI data bases (Altschul et al., 1990).

Geneious^® (Biomatter Ltd., Auckland, New Zealand) was utilized for alignments using the algorithm ClustalW with standard settings. To identify orthologs of Tribolium castaneum CRLF3, a phylogenetic tree was built with the bootstrap resembling method and standard settings as follows: The initial protein sequence alignment was created with the first three hits of Tribolium castaneum and Drosophila melanogaster and the first four hits of Mus musculus. Domains with high quality of alignments were extracted and concatenated for tree building.

Phylogenetic analysis of the CRLF3 protein was also performed with Geneious^®. Orthologous amino acid sequences were aligned with ClustalW algorithm with standard settings. The tree was calculated using the Geneious^® tree builder with Neighbor-Joining method and bootstrap resampling with default settings. The analysis includes CRLF3 orthologs of the human CRLF3 protein identified by NCBI BLAST. They belong to Homo sapiens (GenBank accession number NP_057070.3), Mus musculus (NP_061246.1), Danio rerio (NP_001017817.2), Xenopus laevis (NP_001080600.1), Branchiostoma floridae (XP_002607148.1), Ciona intestinalis (NP_001107600.1) and Tribolium castaneum (XP_008190449.1).

Studies were performed with red flour beetles Tribolium castaneum (wild-type strain San Bernadino). The beetles were bred in plastic boxes filled with a mixture of full grain flour (Wheat flour type 405; Rosenmühle GmbH, Landshut, Germany) and yeast at 27°C and 40% air humidity. The study was conducted exclusively with insects. Studies on insects do not require special permit. All experiments comply with the German laws for animal welfare (“Deutsches Tierschutzgesetz”).

Primary cultures were established from brains of late pupae. Brains were extracted from head capsules and collected in 2 ml sterile growth medium (Leibovitz’s L-15 (Gibco, Karlsruhe, Germany) supplemented with 0.5% gentamicin (BioReagent, Munich, Germany)). All brains used in one experiment (20 brains per treatment group) were pooled at this stage. After three exchanges of the medium, the brains were exposed to a mixture of collagenase and dispase (each 2 mg/ml; Sigma-Aldrich, Munich, Germany) for 45 min at 27°C to digest extracellular matrix proteins. Enzymatic activity was terminated by several washes in Hanks’ Balanced Salt Solution (Gibco, Karlsruhe, Germany). Brain tissue was passed several times through the tip of a 100 μl pipette to dissociate the brain cells. The suspension was centrifuged with 3000 g for 1 min and the pellet containing the cells was resuspended in L-15 medium with 0.5% gentamicin (100 μl per treatment group). The dissociated cells were equally allocated onto concanavalin A-coated (Sigma-Aldrich, Munich, Germany) round cover slips (Type DKR0; Hartenstein; Würzburg, Germany; Ø 10 mm; 100 μl per cover slip) that were placed in sterile plastic culture dishes (Corning Inc., New York, NY, USA; Ø 35 mm). After 2 h the culture dishes were filled to a volume of 2 ml with L-15/0.5% gentamicin ± supplements as specified below. Cultures were maintained in an incubator (27°C; humidified atmosphere) and medium was replaced every 2 days.

Primary cell cultures were fixed with 4% paraformaldehyde dissolved in 0.1 M phosphate buffer for 15 min at room temperature. Cells were washed with phosphate buffered saline (PBS) and PBS with 0.1% Triton-X-100 (Sigma-Aldrich, Munich, Germany; PBST), each three times for 5 min. Cells were incubated for 1 h with blocking solution containing 2% normal goat serum (GE Healthcare, Munich, Germany) and 0.25% bovine serum albumin (MP Biochemicals, Heidelberg, Germany) dissolved in PBST. Primary antisera were incubated over night at 4°C. To label neuronal membranes anti-horseradish peroxidase (anti-HRP from rabbit, Sigma Aldrich) was applied in 1:500 dilution in blocking buffer. To label the activated form of caspase-3, anti-cleaved caspase-3 serum from rabbit (Calbiochem, Merck, UK) was applied in 1:300 dilution in blocking buffer. After several washes in PBS, cells were incubated with 1:200 diluted Alexa 488-coupled anti-rabbit secondary antibody (developed in goat, Molecular Probes, Thermo Fisher Scientific, Waltham, MA, USA) for 2 h at room temperature. In some experiments, DAPI (Sigma-Aldrich, Munich, Germany; 1:1000) was applied with the secondary antibody. After washing away excess antibody, cover slips with labeled cells were mounted with 1,4-diazobicyclo[2.2.2]octane (DABCO, Roth, Karlsruhe, Germany) on microscopic slides. Edges of cover slips were sealed with transparent nail polish and stainings were analyzed by fluorescence microscopy.

Four-day-old primary cultures were maintained in L-15 medium supplemented with 0.5% gentamicin. Particular cultures were incubated with different concentrations (0.32 ng/ml, 0.8 ng/ml, 3.2 ng/ml, 8 ng/ml, 32 ng/ml) of recombinant human Epo (rhEpo, NeoRecormon, Roche, Welwyn Garden City, UK) or (0.42 ng/ml, 0.84 ng/ml) of the human non-erythropoietic Epo splice variant EV-3 (IBA GmbH, Göttingen, Germany) for another 3 days. Cell cultures were fixed in paraformaldehyde (4%, 30 min), labeled with DAPI (Sigma-Aldrich, Munich, Germany; 1:1000) and mounted with DABCO (Roth, Karlsruhe, Germany) on microscopic slides to quantify cellular survival (see below).

Primary cultures of T. castaneum brains were maintained in L-15 medium supplemented with 0.5% gentamicin for 4 days. Each experiment, using the cultured cells from the same population of 80 brains, contained four different treatment groups, one control, incubated without any supplement throughout the entire period and three cultures that were supplemented with 10 ng/μl dsRNA to suppress the expression of dsRed (a protein not naturally expressed by the cells), Tc-CRLF3 (T. castaneum ortholog of CRLF3) and (Regulatory particle triple-A ATPase 3 (Rpt3), a protein of the proteasome that is essential for cellular survival), respectively (for dsRNA sequences see Supplementary Figure S1). After 4 days in vitro, cell cultures were fixed and labeled with DAPI as described above.

Primary cultures of T. castaneum brains were maintained in L-15 medium supplemented with 0.5% gentamicin for 5 days. Each experiment, using the cultured cells from the same population of 100 brains, contained five different treatment groups: (1) untreated control maintained in normoxic condition; (2) untreated control exposed to hypoxia; (3) rhEpo-treated culture exposed to hypoxia; (4) rhEpo-treated culture exposed to hypoxia after RNAi-mediated suppression of Tc-CRLF3 expression; and (5) hypoxia-exposed culture after RNAi mediated suppression of Tc-CRLF3 expression. dsRNA (final concentration 10 ng/μl) was present in the respective cultures throughout the entire experimental period. To minimize the risk for off-target effects, two dsRNA constructs (see Supplementary Figure S1) that target non-overlapping sequences were used to interfere with Tc-CRLF3 expression. rhEpo (0.8 ng/ml) was added to the respective cultures 12 h before the hypoxic period. Hypoxia was generated by floating an airproof chamber (22.5 × 4.5 × 4 cm, workshop of our institute) with nitrogen (N₂). Hypoxia (O₂ level ≤ 2%) was monitored with an oxygen analyzer (Greisinger GOX 100, Conrad Electronics, Hirschau, Germany) and maintained for 36 h. Afterwards, cultures were reoxygenated, maintained for 12 h in normoxia, and subsequently fixed and labeled as described above.

Assessment of cellular survival was based on chromatin structure revealed by fluorescent labeling with DAPI (Gocht et al., 2009). Using a fluorescence microscope (Zeiss Axioskop, Oberkochen, Germany) with 63× oil objective, equipped with a CCD camera (Spot RT3, Visitron, Puchheim, Germany), two continuous rows of non-overlapping photographs were taken (typically ~120 per culture), that extended over the entire cover slip to the right and the left of the center. Numbers of nuclei from alive and dead cells (at the time of fixation) were manually counted with the freeware software Fiji and its cell counter plugin (ImageJ by NIH). Relative numbers of alive cells were calculated for each culture. Within each experiment (that included cultured neurons from the brains of the same individuals) the values were normalized to survival in the untreated control culture (≙100%). Differences in data population distributions were probed with the non-parametric Kruskal-Wallis test, and the Mann-Whitney U test for unmatched samples was used to determine the differences between two groups (statistics performed with OriginPro 8.5, OriginLab Corporation). To adjust the false discovery rate, P values were corrected with the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995; Groppe et al., 2011).

The T. castaneum gene prediction au3.g2971.t1 (Tc000209, Tc-crlf3) is an ortholog of the human crlf3. It is located on chromosome linkage group 2 (ChLG2) at position 14179075–14181074. The size of the mRNA is 1803 bp and the coding sequence, distributed over five exons, comprises 1272 bp. Conserved domains of the 424 amino acid containing Tc-CRLF3 protein include a fibronectin type III domain, a cytokine receptor motif and interdomain contacts (Supplementary Figure S2).

A phylogenetic tree revealed clear orthology of human, mouse and Tribolium CRLF3 proteins (Figure 1A). Importantly, no further paralogs were detected in Tribolium which could act redundantly and possibly rescue RNAi-induced phenotypes (Figure 1A).

CRLF3 orthologs are present in mammals, amphibia, fish, cephalochordates, tunicates and insects. Among insects, crlf3 genes were identified in the beetle T. castaneum and the cricket Gryllus bimaculatus (¹Gene: GB-isotig00932) but not in fruit flies, mosquitoes and honey bees (Wyder et al., 2007). On the protein level, Tc-CRLF3 shares 28% sequence similarity with its human and mouse orthologs, 29% with X. laevis and 26% with zebrafish (Table 2). Respective proteins of the lancelet B. floridae and the tunicate C. intestinales share 27% and 24% similarity with Tc-CRLF3. Analysis of phylogenetic relationships among CRLF3 proteins of these species reflects the evolutionary lineages (Figure 1B). Orthologs of mammalian (mouse and human) as well as of vertebrate (mammals together with frog and fish) CRLF3 form branches that are well-separated from the other species. The non-vertebrate species differ significantly from each other, with CRLF3 of Tribolium showing stronger similarity with Ciona than with the other chordate Branchiostoma.

Freshly dissociated primary cell cultures contained round shaped cells that were typically devoid of processes (Figure 2A). Neurons and eventually also glia lose their axons, dendrites and other processes during the dissociation procedure. It is assumed that a large portion of cells is irreversibly damaged during this process preventing their attachment to the concanavalin A-coated cover slips and subsequent survival in vitro. After 1 day, cultures in full medium with 5% FBS contained approximately 65% alive cells while survival in serum-free cultures was reduced to approximately 40% of adherent cells. Primary cultured neurons regenerated neurites that became visible after 3 days in vitro and subsequently increased in length and complexity of arborizations (Figure 2B). Whether anatomical contacts of processes developed into functional synapses was not studied. Neurons and glia were distinguished by the presence or absence of anti-HRP immunoreactivity (Figure 2C). Anti-HRP has been demonstrated to label neuron-specific membrane proteins in various arthropods and anti-HRP immunocytochemistry is a simple, reliable and widely used method to identify insect neurons (Jan and Jan, 1982; Sun and Salvaterra, 1995; Gocht et al., 2009). However, since cytoplasmic membranes disintegrate in dying or dead cells, anti-HRP immunocytochemistry cannot distinguish between dead neurons and glia (e.g., the nucleus marked with “o” in the center of Figure 2C). As it was previously described for primary cultures from locust brains (Vanhems and Delbos, 1987; Gocht et al., 2009), fresh cultures from T. castaneum brains contained low numbers of glia and no glial cells were detected after 4 days in vitro. Viability of primary cultured cells was reflected in the pattern of the DNA-binding fluorescent dye DAPI (Figure 2D). Nuclei of intact cells (both neurons and glia, Figure 2C) display a patchy staining pattern that corresponds to intact chromatin structures. DNA fragmentation, that goes along with apoptotic cell death, leads to condensed and uniformly DAPI-labeled nuclei. Apoptotic cell death in cultured insect neurons can be induced by serum-deprivation, hypoxia and other physiological stressors (Jenkins et al., 2013; Miljus et al., 2014). Caspase activation involved in apoptosis can be detected in T. castaneum brain cell cultures by anti-cleaved caspase-3 immunocytochemistry (Figure 2E).

Cellular survival in untreated control cultures and cultures exposed to different concentrations of rhEpo or EV-3 was assessed after 7 days in vitro. Pharmacological treatment started after 4 days in vitro, when cultures were devoid of glia and exclusively contained neurons that could be labeled with anti-HRP immunocytochemistry (Figures 2B,C). rhEpo concentrations of 0.8 and 3.2 ng/ml significantly increased neuronal survival in cultures deriving from identical pools of brain cells (both with p < 0.001 compared to controls) while 32 ng/ml rhEpo reduced viability (p < 0.05 compared to control; Figure 3A). The human non-erythropoietic Epo splice variant EV-3 also significantly (p < 0.05 for 0.42 ng/ml EV-3 and p < 0.01 for 0.84 ng/ml EV-3 compared to control) increased the survival of T. castaneum brain neurons in vitro (Figure 3B).

RNAi by application of dsRNA to the extracellular space is frequently described as soaking RNAi. In order to test, whether the mechanisms underlying dsRNA uptake and RNAi are retained in cultured brain cells, dsRNA fragments targeting the expression of three different proteins (one that does not exist in T. castaneum, one that is essential for cellular survival and two fragments targeting Tc-CRLF3) were applied to the culture medium for 4 days. DsRNA targeting the expression of Rpt3 dramatically reduced cellular survival in comparison to untreated control cultures that derived from the same pools of brains (Figures 4A,B), indicating effective dsRNA uptake and knockdown of expression of this essential proteasomal protein. In contrast, cell survival was not affected by dsRNA targeting dsRed, a protein that is not naturally expressed in insect cells (Figure 4A). Thus, uptake of dsRNA per se did not compromise the viability of cultured brain cells. Similarly, dsRNA targeting the expression of Tc-CRLF3 (fragment 2: Figure 4A and fragment 1: Figure 4B) had no impact on cellular survival, suggesting that the reduction of Tc-CRLF3 expression is not critical for cell survival under normal culture conditions.

In order to evaluate the role of Tc-CRLF3 in Epo-mediated neuroprotection, we performed two independent series of experiments (experimental procedure shown in Figure 5A) in which Tc-CRLF3 expression was suppressed by interference with two different dsRNAs, targeting different, non-overlapping fragments of the transcript (fragment 1 and fragment 2). Exposure to hypoxia for 36 h significantly reduced cellular survival in comparison to cultures that were maintained under normoxic conditions (Figure 5B, p < 0.001; Figure 5C, p < 0.001). Incubation with 0.8 ng/ml rhEpo completely prevented hypoxia-induced cell death (Figure 5B, p < 0.001; Figure 5C, p < 0.001). Since all cultures were maintained in serum-free medium, which acts as a mild stressor for cellular survival, survival in rhEpo-treated/hypoxia-exposed cultures was even higher than in normoxic controls. RNAi-mediated knockdown of Tc-CRLF3 expression completely abolished the neuroprotective effect of rhEpo in hypoxia-exposed neuronal cultures. RNAi targeting both fragment 1 (Figure 5B) and fragment 2 (Figure 5C) significantly decreased cellular survival compared to rhEpo-treated/hypoxia-exposed cultures (both with p < 0.001) and normoxic controls (both with p < 0.001). Tc-CRLF3-directed RNAi had no additional deleterious effect on hypoxia-exposed neuronal cultures (Figures 5B,C) which is in line with our control experiment where Tc-crlf3 RNAi did not change survival in normal cell cultures (Figure 4). Thus, knockdown of Tc-CRLF3 expression eliminates Epo-mediated neuroprotection of T. castaneum brain neurons, without compromising cellular survival under neither normoxic (Figures 4A,B) nor hypoxic (Figures 5B,C) conditions. This confirms the assumption that Tc-CRLF3 is the receptor mediating Epo neuroprotective effects in T. castaneum.