Immunological off-target effects of microRNA control sequences

Immunological off-target effects of RNA and RNAi therapy are a considerable challenge in research and the future of RNA-therapy. Here we investigated some of the hurdles we previously encountered when transfecting mircroRNA (miRNA) control sequences into chondrocytes in an inflammatory model simulating osteoarthritis (OA). We investigated different negative control sequences of different technologies; Pre-miR miRNA Precursor and mirVana from Thermo Fisher Scientific. We used RT-qPCR, western blot analysis and mass spectrometry to asses for the effects of the transfected control sequences. The data did not show a global immunological off-target effect, however a specific off-target effect on IL6 and IL8 was observed. IL6 and IL8 were both upregulated by the negative control from the Pre-miR miRNA Precursor technology (Pre-neg #1), and downregulated by the negative control from the mirVana technology (mirVana-neg). Moreover, the results suggested that the effect on IL6 and IL8 was dependent on both sequence and type of chemical modifications in addition to donor variation. We conclude that negative controls should be selected wisely, and suggest that scientists need to test several controls to ensure correct interpretation of data before drawing any conclusions.


Introduction
miRNAs are also small regulatory RNAs of 21-22bp in length that mediate their regulatory effects, just like siRNAs, through the RNAi pathway. miRNAs are potent gene regulators as one miRNA can target up to 100 genes 1,2 , and both the leading strand and the anti-sense strand exhibit regulatory effects. In a previous study we showed how miRNA-145 also had immunostimulatory off-target effects that were liposomal-delivery dependent and were mediated via RIG-1 3 . The expression of immune genes was not induced when miRNA-145 was electroporated into the cytosol. We have since taken into consideration the immunostimulatory off-target effects of using lipid based vehicles for transfection of regulatory RNAs and favored electroporation.
Immunological off-target effects to RNA has been previously reported in relation to siRNA and miRNA transfection into cells [3][4][5] . Undesirable off-target induction of interferon response, which is the classical anti-viral response involving many genes, was reported to be mediated by siRNAs [6][7][8] . However, synthetic siRNAs of 21 bp have been shown to bypass activation of the interferon pathway 4,9 , and this has opened the door for the potential of using synthetic siRNAs to study gene silencing in vitro, and as therapeutical agents in vivo 10 . Nevertheless there are conflicting results on the degree of interferon pathway induction as a result of siRNA delivery that depends on factors like the particular siRNA sequence, chemical modification of the siRNA and cell type. In a study by Sioud et. al 4 where 32 different siRNA sequences were investigated, the author concluded that the immunostimulatory capacity of the siRNAs, whether they are double-stranded or single-stranded, are sequence dependent and require endosomal intracellular signaling . About 50% of the siRNAs investigated induced the production of the inflammatory cytokines TNFα and IL6, and their ability to induce TNFα and IL6 varied depending on their sequence. Moreover the study showed that liposomal delivery of siRNAs was required for cytokine induction 4 . siRNAs are highly specific and can be potent gene silencers with only one mRNA target and thus harbor great therapeutic potential, however such therapies will be limited to elimination of a specific target protein. While miRNAs having numerous targets simultaneously, some of which are transcription factors and regulatory molecules, are capable of altering complex networks 11 . This however might make avoiding immunological off-target effects extremely difficult 5 . Here we report immunological off-target effects of miRNA negative control sequences delivered to the cells through electroporation in an inflammatory in vitro model. These negative control sequences should ideally not alter gene expression, as this can have a great impact on interpretation of the results.
In this study we investigated different control sequences made using different miRNA synthesis technologies; Pre-miR miRNA Precursor and mirVana from Thermo Fisher Scientific. The miRNA sequences studied were modified using two different chemical processes in order to protect against endonuclease degradation and to increase silencing performance [12][13][14] . Based on results from RT-qPCR analysis, mass spectrometry proteomics and western blot assays we conclude that there were few immunological off-target effects of transfection of the control sequences, but IL6 and IL8 were both upregulated by the negative control from the Pre-miR miRNA Precursor technology (Pre-neg #1). IL6 and IL8 were downregulated by the negative control from the mirVana technology (mirVana-neg). Further, the results suggested that the immunological off-target effects were dependent on both sequence and type of chemical modifications. We conclude that negative controls should be selected wisely.

Pre-neg #1 and mirVana-neg are affecting IL6 and IL8 differently
Chondrocytes were electroporated with each RNA sequence separately in three parallel experiments using cells from three different donors. Four days later all the samples were stimulated with the inflammatory cytokines IL1β and TNFα, used in other projects to simulate an osteoarthritic milieu in vitro. In addition to Pre-neg #1 and mirVana-neg, a mock control experiment (electroporated without the addition of RNA sequences) and a non-electroporated but cytokine stimulated experiment were also included in the analysis. All cells were harvested on day 5. Fig.1 shows that IL1β and TNFα upregulated all the cytokines investigated as expected. Mock upregulated IL6 and to a lesser extent IL8 protein levels, compared to the stimulated control, with notable donor variation. While Pre-neg #1 seemed to strongly upregulate IL6 and IL8 (donor 1) or bring them to similar levels of that of mock control (donor 2 and 3). mirVana-neg consistently downregulated IL6 and IL8 in all donors and exhibited a pattern of effect similar to that of the stimulated control (Fig. 1a). There were no consistent differences observed at IL1B, IL6, IL8 and TNF mRNA levels, though an upregulatory tendency of IL1B in two of the donors by Pre-neg #1 was seen (Fig. 1b).

Immunological off-target effects of the controls are both sequence and chemical modification dependent
In order to understand whether the effects on IL6 and IL8 mediated by these two controls were restricted to their particular sequences, their chemical modifications or both we looked at additional negative control sequences produced by the same manufacturer. We included Pre-neg #2, which has the same chemical modifications as Pre-neg #1 but differs in sequence, and Pre-neg*, which shares sequence with Pre-neg #1 but has the same chemical modification as mirVana-neg. The type of chemical modifications and the sequences are company properties and were not revealed to us. Fig.2 shows western blot analysis of IL6 and IL8 in response to transfection of the various control sequences in two donors. Again, electroporation performed in the mock experiment upregulated IL6 and IL8 protein levels.
Further upregulation of IL6 and IL8 was seen after Pre-neg #1 transfection. Pre-neg #2 showed differential effects on IL6 in the two donors, and did not affect IL8 compared with mock. mirVana-neg again had a consistent downregulatory effect on both IL6 and IL8 as previously observed in Fig. 1a, and in a previous study (under publication). Finally, Pre-neg* showed downregulatory effects on IL6 and IL8 similar to mirVana-neg (Fig. 2).
In order to understand why those control sequences altered IL6 and IL8 proteins, we further checked for the expression of key genes involved in inflammatory signaling of the innate immune response in donor1; some pattern recognition receptors (PPRs) that respond to foreign RNA and further mediate the inflammatory signal; TLR7,8, and 3 and some of their downstream mediators TRIF, MYD88, IRF3, OAS2, two intracellular PRRs that also respond to RNA and elicit an immune response; DDX58, PKR, cytokines including IL6 and IL8 but also IL1β, TNF, CXCL10 and CCL5, and finally a pro-inflammatory (IL1β) induced gene PTGES. TLR7 and 8 were not detected and therefore not seen in Fig.3. Although there are not big differences in gene expression between the controls, mirVana-neg and Pre-neg* seem to affect the selected genes similarly (Fig. 3). Sharing the same chemical modifications could perhaps explain their similar effects on the mRNA expression of the selected genes and the protein expression of IL6 and IL8. Pre-neg #1 also showed a tendency to upregulate IL1β and CXCL10, and downregulate OAS2. However none of these observations could explain why pre-neg #1 led to upregulation of IL6 and IL8, and why mirVana-neg led to the opposite.

Proteome alterations upon transfection of Pre-neg #1 control and miRvana-neg control
In the pursuit of possible explanation of the effects seen on IL6 and IL8 we performed proteomics analysis on cells from the first three donors used in Fig 1. Table 1 shows all altered proteins between the samples (Student t-test significance p<0.05 and 2 fold change cut-off). To our surprise few proteins were altered between the samples, even without adjusting for multiple testing (FDR).When we corrected for multiple testing, no proteins were significantly altered between the control sequences. Notably the biggest difference between the samples in terms of the number of proteins altered, is in fact between the electroporated samples containing an RNA sequence (mirVana neg, Pre-neg#1 controls) and the unelectroporated control (stimulated), indicating the effect of transfection of RNA into the cell. Curiously ITGA5 was consistently downregulated by mirVana when compared to Pre-neg#1, mock, and stimulated control (Table 1a, 1b, and 1c respectively). IL6 and IL8 were not detected by the proteomics analysis.

Discussion
Immunological off-target effects as a result of introduction of RNA into the cell regardless of the transfection method make correct interpretation of data difficult, and may have unfortunate consequences for gene therapy. Indeed electroporation alone (mock) without any RNA sequence involved had notable immunological off-target effects as it upregulated IL6 and IL8 protein. However the fact that the consistent immunological off-target effects seen on IL6 and IL8 were generated by a the negative control sequences (mirVana-neg) raises a concern.
The negative controls Pre-neg #1 and mirVana-neg had different sequences and different chemical modifications that seemed to give them opposing effects on our proteins of interest.
The precursor technology of Pre-neg #1 seemed to have an upregulatory effects on IL6 and IL8 protein levels that was also donor dependent.
The question of whether these effects are sequence-dependent or chemical modificationdependent was addressed by including additional controls. Pre-neg #2 sharing the same chemical modifications as Pre-neg #1, did not exhibit the same upregulatory effects on the interleukins which indicates that the upregulatory effects observed by Pre-neg #1 are partially sequence-dependent. Engineering Pre-neg#1 sequence with mirVana chemical modification (Pre-neg*) allowed us to conclude that the previously observed downregulatory effect mediated by mirVana-neg on IL6 and IL8 are due to their chemical modifications as evident by the effects Pre-neg* exhibited. In summary we conclude that the immunological off-target effects mediated by mirVana-neg and Pre-neg #1 are both sequence dependent and chemical modifications-dependent.
The specific effects on these interleukins made us wonder whether other cytokines, some upstream receptors and mediators of inflammatory signaling were altered and thus resulted in the effects observed on IL6 and IL8. However the additional cytokines we tested for showed no or small differences between the controls with no consistent pattern. We also tested for several TLRs; TLR7 and 8 were either not detected or barely detected in some of the samples and thus not included in the results. TLR3 was detected however no difference was observed between the samples. The downstream mediators of inflammatory signals tested did not show a considerable difference between the negative controls that could explain the effects seen on IL6 and IL8 protein levels. One possible explanation could be the timing of cell harvesting.
Probably the inflammatory response to the RNA-sequences is quite immediate and thus many of these upstream mediators might have been restored to homeostatic levels after mediating the inflammatory signal. The cells were harvested 5 days after transfection and one day after stimulation with IL1β and TNFα. Perhaps, it was long enough to see the effects on IL6 and IL8, but not to see the initial effects on upstream regulators and mediators. This is certainly a limitation of the study and for future studies we would certainly check the expression of these molecules and perhaps several others at earlier time points. Moreover we only assessed for mRNA levels, perhaps on the protein level at different time points there could be more evident differences on these upstream mediators. Although we performed proteomics to our surprise the proteomics analysis did not show many altered proteins when we compared the various controls against each other. Applying multiple correction statistics, which are highly recommended in omics-studies, no proteins were actually differentially expressed between any of the samples. In this fashion it is a positive observation which implies that these controls do not have immunological off-target effects on a global scale. Although worth noting that proteomics analysis have its limitations and do not always detect all proteins and all differences in the cells. IL6 and IL8 for instance were not detected by the proteomic analysis. Cytokines having low molecular weight and existing in low abundance provide less peptides for mass spectrometry which makes them very difficult to detected 15 Previously, we compared RNA-sequencing data of mock control and mirVana-neg control and found no genes to be differentially expressed (unpublished data). All this suggest that the effects seen on IL6 and IL8 mediated by these two controls are quite specific and yet to be explained. We and others have previously reported immunostimulatory effects of lipid based transfection of RNA 3 . In those studies the authors also showed that these immune responses were not observed when electroporation was used as method of RNA delivery. However in this study we report modulated immune response to RNA control sequences when delivered by electroporation. Moreover these modulated immune responses were specifically observed on IL6 and IL8 protein levels only. Certainly usually there is a discrepancy between RNA and protein expression in the cell 16,17 . However this also suggests that the effect observed on IL6 and IL8 was regulated at the post-translational level.

Methods and Materials
Isolation and culture of human articular chondrocytes (ACs). ACs were isolated from discarded osteoarthritic cartilage tissue following total knee replacement surgery and cultured as previously described [18][19][20] . Only apparently healthy tissue with no macroscopic signs of OA was used. Written informed consent from all donors was received. The study was approved by the Regional Committee for Medical Research Ethics, Southern Norway, and we confirm that all research was performed according to relevant guidelines and regulations. We also confirm that informed consent was obtained from all patients donating tissue after undergoing total knee replacement surgery. In brief, the cartilage was processed into tiny pieces then digested with Collagenase type XI (Sigma-Aldrich, St. Louis, MO) at 37°C for 90-120 min.
The cells were washed three times and resuspended in culture medium consisting of Dulbecco's modified Eagle's medium/F12 (Gibco/ThermoFisher Scientific, Waltham, MA, USA) supplemented with 10% human plasma (Octaplasma AB, Oslo Blood Bank, Norway) with platelet lysate (corresponding to 10 9 platelets/ml plasma) (PLP), 100 units/mL penicillin, 100 μg/mL streptomycin, and 2.5 μg/mL amphotericin B 19 . PLP was prepared as previously described 21 . Every 3-4 days the culture medium was changed and amphotericin B was removed after the first passage. Cells were tryspsinized at 70-80% confluence with trypsin-EDTA (Sigma-Aldrich) and seeded into new culture flasks.  Table S1) prior to visualizing the bands using the myECL imager (Thermo Fisher Scientific).

MicroRNA mimics, transfection and stimulation with
Mass spectrometry. All experiments were performed on an Easy nLC1000 nano-LC system connected to a quadrupole -Orbitrap (QExactive Plus) mass spectrometer (ThermoElectron, Bremen, Germany) equipped with a nanoelectrospray ion source (EasySpray/Thermo). For liquid chromatography separation we used an EasySpray column (C18, 2 µm beads, 100 Å, 75 μm inner diameter) (Thermo) capillary of 50 cm bed length. The flow rate used was 0.3 μL/min, and the solvent gradient was 2% to 7 % in 5 minutes and then to 30% B in 120minutes. The column was finally washed in 90 % B wash for 20 minutes. Solvent A was aqueous 0.1 % formic acid, whereas solvent B was 100 % acetonitrile in 0.1 % formic acid.
Column temperature was kept at 60 o C.
The mass spectrometer was operated in the data-dependent mode to automatically switch between MS and MS/MS acquisition. Survey full scan MS spectra (from m/z 300 to 1,500) were acquired in the Orbitrap with resolution R = 70,000 at m/z 200 (after accumulation to a target of 3,000,000 ions in the quadruple). The method used allowed sequential isolation of the most intense multiply-charged ions, up to ten, depending on signal intensity, for fragmentation on the HCD cell using high-energy collision dissociation at a Trypsin without proline restriction enzyme option was used, with two allowed miscleavages.
Minimal unique peptides were set to 1, and FDR allowed was 0.01 (1%) for peptide and protein identification. The Uniprot human database was used (download from uniprot Sept. 2018). Generation of reversed sequences was selected to assign FDR rates. This protocol was performed by the Proteomics Core Facility, Oslo University Hospital, Rikshospitalet.          C-X-C motif chemokine 6;Small-inducible cytokine B6 -2,4 STAT3

Acknowledgment
Signal transducer and activator of transcription 3 2,2 TMTC3 Transmembrane and TPR repeat-containing protein 3 2,3 Negative fold change values indicate downregulation, and positive values indicate upregulation compared to mock.