Team:Munich/Results

Results

Authors Summary

We established a minimally-invasive and longitudinal diagnostic platform for monitoring of living cells. We demonstrated the following aspects:

• Engineered vesicles are successfully secreted
• Exosomes can now be easily purified, thanks to our best basic part
• RNA can be specifically loaded and analzed by qPCR
• Vesicle production did not increase cell death
• Secreted vesicles display low invasion of surrounding cells

Figure 1: Longitudinal and minimally invasive RNA export in vesicles from the same cells. The FLuc mRNA export from exosomes and virus like particles (VLP) was quantified by qPCR via standard curve and normalized to the amount of living cells. Each data point corresponds to n=2 duplicates.

---

Vesicle Secretion

Virus-like particles and Exosomes can be secreted efficiently

To validate the formation and export efficiency of engineered exosomes and virus-like particles, we equipped our vesicles with a luminescence-based tag; the 11 amino acids long HiBiT peptide (Dixon et al. 2016; Schwinn et al. 2018; Promega®). The HiBiT tag is fused to the C-terminus of both CD63 and Gag, such that it is only present inside the exosomes or VLPs, respectively.
A detailed explanation of how the HiBiT Assay works can be found here

Workflow:

• Harvest cell-culture supernatant
• Lysis of intact vesicles
• HiBiT assay to determine cellular expression level
• HiBiT assay to determine vesicular content
• Conversion to absolute units (calibration curve)

Figure 2: Virus like particle (VLP) display higher secretion efficiency in comparison to exosomes. The HiBiT signal was determined from cells as well as the corresponding supernatant including the vesicles. The experiment was executed with 5 biological replicates. To demonstrate a specific HiBiT assay for engineered VLPs and Exosomes a negative control containing iRFP (Mock) was used (n.d. = not detected). A) Cells were transfected with plasmids carrying the structural gene of VLPs (Gag) tagged with the HiBiT peptide and fused to either an RNA binding protein (L7Ae) or a short coiled-coil domain (No RNA binding protein, RBP). B) Cells were transfected with plasmids carrying genes for engineered exosomes (CD63) and RNA binding protein (L7Ae) or negative control (No RNA binding protein, RBP).

Here is the HiBiT Calibration Curve

The resulting linear equation of the calibration curve can be used to calculate the “fmol HiBiT equivalents” in other wells with samples of interest. While one fmol HiBiT control protein does not necessarily stochiometrically correspond to one fmol of our HiBiT-tagged proteins, their absolute amounts have a linear relationship proportional to their respective luminescence signals and thus allow us to compare total expression and secretion levels of different vesicle constructs. These calibration curves are therefore used to calculate the total amount of exported HiBiT-signal.

For Figure 2, the secretion of Gag-based VLPs from HEK293T cells (A) and the secretion of CD63-containing exosomes from HEK293T cells (B) is shown. It can be seen that VLPs, as well as exosomes, are efficiently secreted and analyzed 72 h (VLPs) and 48 h (exosomes) after transfection, respectively. Furthermore, the vesicle producing cells are also shown as a reference. Looking at the data more in detail, it can be seen that vesicles containing the RNA-binding protein "L7Ae" as well as empty vesicles (No RBP) showed total expression and secretion of the vesicle specific constructs. Interestingly, the amount of secreted Gag-based VLPs is much higher compared to secreted CD63-containing exosomes, although the exosome producing cells contain a much higher amount of HiBiT equivalents than exosome producing cells. Although vesicles without the RNA binding protein L7Ae are secreted, we show later that these vesicles are not loaded with our target transcript (see Figure 6B).

Transmission Electron Microscopy

As figure 3 (A,B) indicates, secreted vesicles show a round shape as displayed in transmission electron microscopy (TEM) images. The precise calculation of the diameter was determined with dynamic light scattering (DLS) (figure 3 C,D). The results revealed disperse diameters depending on cell-type and whether the vesicles were loaded or not.

Figure 3: Characterization of vesicles was performed by Transmission Electron Microscopy (TEM), (A and B) and Dynamic Light Scattering (DLS), (C and D). Grids with exosomes (A) and virus like particles (B) were negatively stained with uranyl acetate for TEM. The diameter for VLPs and exosomes was precisely determined with DLS. It was revealed that RNA loading increases the size of VLPs, whereas exosomes differ in size depending on the secreted organism.

Purification

We developed a platform with future applications in diagnostic research and medicine. Therefore constructs were designed to support isolation, purification, and enrichment from body-fluids of patients.

Parts: For the first time, we show Nickel-NTA Affinity purification for exosomes

Commercially available Heparin affinity chromatography for VLPs

To establish a purification method for exosomes that does not rely on expensive kits or ultracentrifugation, we engineered the exosomal marker protein CD63 via incorporation of a polyhistidine affinity tag into the large extracellular loop. This modification not only allows us to separate exosomes from other extracellular vesicles but also facilitates the separation of engineered exosomes from endogenous exosomes, simplifying downstream analysis.

We show that 35 % of the load were found in the flow-through, 15 % were found within the wash fractions and the remaining 50 % were eluted using high imidazole concentration. We performed this purification for HEK293T samples.

Figure 4: The bioengineered CD63 containing a polyhistidine sequence in the large extracellular loop (CD63-6xHis) can be purified using Ni-NTA Agarose beads. In this set-up, the exosomes were purified from transfected HEK293T cells and analysed the individual fractions using the HiBiT assay To test the efficacy of the purification, we compared the results for the CD63-6xHis to the values of unspecific binding of wild-type CD63 to the Ni-NTA beads. The data shown in the figure derive from three independent purification experiments. Using unpaired two-tailed t-test, it can be shown that more HiBiT-Signal can be found in the flow-through for wild-type CD63 than for the bioengineered CD63-6xHis (p < 0.01, n(CD63)=3, n(CD63-6xHis)=3). For the wash more HiBiT-Signal can be found in the wild-type CD63 fractions than in the CD63-6xHis (p < 0.01, n(CD63)=3, n(CD63-His)=3) and for the elution, more HiBiT-signal can be found in CD63-6xHis samples than in wild-type CD63 samples (p < 0.01, n(CD63)=3, n(CD63-His)=3).

Here is the T-Test

Figure 5: We purified bioengineered VLPs consisting of Gag fused to the RNA-binding protein, L7Ae obtained from HEK293T cells. We analysed the individual fractions using the HiBiT assay. Approximately 40 % of the HiBiT signal was detected in the flow-through, 5 % in the wash and 55 % in the elution fraction.

qPCR Analysis

Targeted cargo is loaded and exported in exosomes and VLPs

Longitudinal secretion of vesicles and RT-qPCR analysis over 72 h

Characterization of detection range of VLPs and exosomes, respectively

The RNA concentration of secreted transcripts is indispensible for diagnostic and fundamental research because quantitative transcriptomic information can be revealed. Additionally it is a standardized and globally applied method.

Thus, we isolated RNA from secreted vesicles and performed reverse transcription quantitative polymerase chain reaction (RT-qPCR), proving that both VLPs and exosomes can be specifically enriched with FLuc mRNA. Additionally, we confirmed that the secreted mRNA can be reliably sequenced.

Workflow:

• Harvesting supernatant and remove cell debris via centrifugation
• Purify Vesicles
• Isolate RNA

We established three different protocols for RNA isolation from cells, exosomes and VLPs

• Cellular RNA: Commercial TRIzol reagent
• Exosomal RNA: Total Protein and RNA isolation Kit
• VLP RNA: MagMAXTM Viral RNA Isolation Kit

• Reverse transcription (+ noRT controls)
• qPCR Analysis

The modified protocols can be found here

To evaluate the quantity and quality of isolated mRNA spectroscopy-based analysis via NanoDrop was performed. The absorbance ratios A260/280 and A260/230 can indicate contaminants in the sample preparations. Nanodrop measurements of our sample preparations indicate that we have some contaminations in our RNA samples, e.g. from residual phenol, guanidine thiocyanate or ethanol used during the clean-up procedure. However, the effect on downstream applications seems tolerable since RT-qPCR can be performed with these samples.

To show that our vesicles can load RNA, we ran qPCRs on cells and the supernatant to quantify the amount of our target FLuc mRNA exported in vesicles compared to the amount of our target FLuc mRNA that can be found in the cells. More details on the RT reaction setup and the quantification of qPCR data as well as considerations regarding the primer design can be found in the supplement.

Figure 6: FLuc mRNA has similar expression behavior in cells, but displays superior exportation by VLPs in comparison to exosomes. The amount of RNA in the vesicles as well as in the negative controls was determined by qPCR. The negative controls included were exosomes without RNA binding protein (No RBP) for unspecific RNA loading and the iRFP protein (Mock) for qPCR specificity. Due to unstable performance of housekeeping genes in the used vesicles, the housekeeping refence gene (GAPDH) was only used in experiments in cells. Whereas qPCR based on RNA extracted from vesicles was normalized to the cell number. The experiment was executed with 2 technical replicates. A) RNA amount was quantified by qPCR with RNA extracted from cells. Cells were transfected with plasmids carrying the exosomal marker CD63 or the structural gene of VLPs (Gag) fused the RNA binding protein L7Ae, or the negative controls. B) RNA amount was extracted from vesicles and calculated by qPCR using standard curve method. Transfection was analogous to subfigure A.

In general the isolated RNA was sufficient for qPCR. The RNA expression was similar for fluc mRNA in origin cells (Figure 6, A). The export of fluc transcript was superior in VLPs compared to exosomes. Those findings are proof that CD63-L7Ae and Gag-L7Ae constructs specifically load the targeted mRNA into exosomes and VLPs respectively.

Because the transcript number in no-RBP (Figure 6, B) is tremendously decreased, even though no RBP exhibited high HiBiT equivalents (Figure 2, A)

The absolute copy numbers of FLuc mRNA exported in our vesicles 48 h (exosomes) or 72 h (VLPs) after transfecting HEK293T cells. Notably, both vesicles containing the RNA adapter L7Ae exported high amounts of cargo compared to controls lacking the RNA binding protein (No RBP). In VLPs, 300 FLuc-mRNA transcripts per cell were exported, which was approximately 20 times more than the exosome export rate. In contrast, no significant amount of FLuc mRNA could be detected either in control samples lacking the marker proteins CD63 (exosomes) and Gag (VLPs) or in mock-transfection controls. Looking at cellular expression levels instead of vesicle content showed that the No RBP control contained the highest amount of FLuc mRNA of all samples. This indicates that all cells were successfully transfected with FLuc-plasmid, but only cells expressing our vesicles together with the RNA binding protein L7Ae could export FLuc mRNA in significant amounts.

RNA loading through the modular coiled-coil system

Besides the fusion constructs shown above, we also tried to load different adapters into our exosomes by using established protein coiled-coil (CC) dimer pairs (Chen et al. 2018; Ljubetič et al. 2017). These artificially designed systems are based on the highly specific interaction of two short coiled-coil peptides. Fusing one part of the CC dimer to our exosomal marker protein CD63 and the complementary CC to different adapter proteins could enable us to load a variety of protein cargos with different functionalities using the same basic CD63 exosome marker protein. Establishing such a system would thus significantly increase the modularity of our platform.

Figure 7: Vesicles can be modularly assembled via coiled-coils. Two different RNA binding proteins (RBP) were fused to the coiled-coil (CC) peptide P10SN and loaded into exosomes displaying the complementary part P9SN. qPCR analysis demonstrated specific export of target FLuc mRNA and a higher efficiency for L7Ae compared to MCP. Quantification was done via standard curve measurements and assuming a confluent well with 350.000 HEK293T cells/cm2 at the time of harvesting in biological dublicates.

Using a coiled-coil system (Figure 7) allowed us to successfully load the two different RNA binding proteins (RBPs) L7Ae or MCP into exosomes to export FLuc mRNA from HEK293T cells. We used the published parallel heterodimer pair P9SN:P10SN (Ljubetič et al. 2017), where P9SN was fused to the C-terminus of the exosome marker protein CD63, and P10SN N-terminal to the RNA binding proteins L7Ae or MCP. The export efficiencies of FLuc mRNA cargo measured by qPCR proved that coiled-coil mediated targeting of the RBPs L7Ae and MCP into exosomes worked. Exosomes formed from CD63-P9SN and loaded with P10SN-L7Ae or -MCP could export about 800 and 400 transcripts per cell, respectively. This also shows that the two RBPs have different export capabilities, with L7Ae exporting twice as many transcripts per cell compared to MCP.

Time-resolved monitoring of cells with ALiVE over 72 h

For further analysis of cargo loading and export using our vesicles, we performed a time-resolved qPCR measurement over 72 h as a proof of concept for longitudinal studies.

Figure 8: Longitudinal and minimally invasive RNA export in vesicles from the same cells. The FLuc mRNA export from exosomes and virus like particles (VLP) was calculated by qPCR via standard curve and normalized to the number of living cells. Each data point corresponds to n=2 duplicates.

We can successfully export cargo FLuc mRNA both in exosomes and VLPs over 72 h (figure 8). VLPs and exosomes already show a high cargo export efficiency after 24 h which remains stable over the whole period of 72 h. This indicates that we can indeed analyze cells over more extended periods without adverse effects. Exosomes, in contrast to VLPs, have a slower onset and show lower export efficiency.

This is consistent with our HiBiT data. In this data, VLPs show higher amounts of secreted Gag-L7Ae constructs compared to CD63-L7Ae constructs in exosomes. Therefore, it is very likely that more FLuc mRNA is loaded into VLPs compared to exosomes. We thus believe that the export efficiency of VLPs is higher than for exosomes, which further highlights the modularity of our platform and its adaptability for different needs.

Detected RNA levels correlate to the amount of cells monitored

To show that our pipeline of RNA isolation and RT-qPCR works robustly and to determine the detection limit of our system, we performed a Log2 titration on cell density and measured exported mRNA in secreted VLPs and exosomes.

Figure 9: Correlation between FLuc mRNA isolated from vesicles dependent on number of transfected cells. The amount of transfected cells was titrated in a range of 25000 to 400 cells per 96-well and RNA was quantified via RT-qPCR. Quantification was done via standard curve measurements and assuming a confluent well with 350.000 HEK293T cells/cm2 at the time of harvesting in biological dublicates.

The amount of detected Fluc mRNA cargo in exosomes or VLPs decreases linearly with decreasing cell number (figure 9). Robust detection of our target RNA is possible down to 400 or 800 cells for exosomes and VLPs, respectively, before no clear differentiation to the next-smaller cell number is possible anymore. This proves that our system is highly sensitive, and our RNA extraction methods including RT-qPCR work properly and stably.

Cell Viability

Vesicle secretion is not harmful to cells

Vesicles show no adverse effects on untransfected cells

In order to ensure the biosafety and biocompatibility of our platform, we performed a luminescence-based leakage and an absorbance-based cytotoxicity assay proving that our constructs are not harmful. We expected leakage to happen if cells were disrupted by toxic vesicle production or natural cell death. Furthermore, we demonstrate that our vesicles do not mediate unwanted collateral transfection of non-transfected cells.

Based on the deliberate design of our marker proteins fused to the HiBiT-peptide, we were able to assess the leakage of Gag-HiBiT or CD63-HiBiT monomers from dead cells. As the HiBiT-tag faces the lumen of our vesicles and the complementary LgBiT-peptide is not membrane-permeable, it was mandatory to lyse our vesicles to evaluate vesicle secretion rates. On the other hand, no signal should be detectable in unlysed supernatant unless dead cells leak non-assembled marker monomers.

The leakage rates of VLPs and exosomes were based on HiBiT-measurements (Figure 11). Cells expressing VLPs or exosomes either alone or together with the adapter L7Ae show leakage rates beneath 3%. This is significantly lower than the HiBiT signal obtained for lysed supernatant, which strongly suggests that our constructs are not harmful to HEK293T cells. In addition, these data indicate that the contribution of free Gag- or CD63-monomers to the overall secretion rates determined via HiBiT assays is negligibly small.

Figure 11: HiBiT signal for vesicle detection from HEK293T is reliable and displays only minor leakage. Vesical leakage corresponds to HiBiT-signal from unlysed supernatant from HEK293T to assess leakage of vesicle monomers from dead cells. For both Gag-HiBiT and CD63-HiBiT, leakage rates were below 3% indicating no adverse effects of transfection with 6 biological replicates.

Leakage assay for cells secreting either VLPs (left) or exosomes (right). HiBiT-signal from unlysed supernatant was measured to assess leakage of vesicle monomers from dead cells. For both Gag-HiBiT and CD63-HiBiT, leakage rates were below 3 % indicating no adverse effects of transfection. 6 biological replicates.

To further confirm the findings of our leakage assay, we performed a well-established commercial cytotoxicity assay based on lactate dehydrogenase (LDH) activity. LDH is naturally occurring in cells and can be claimed as a housekeeping gene. Damaged cells are leaky and therefore release LDH into the surrounding medium, which triggers a coupled enzymatic reaction resulting in the formation of a formazan product detectable at 490 nm. Importantly, the levels of LDH release and formazan production are directly proportional, thus allowing to reliably determine cellular toxicity based on absorbance measurements.

Figure 12: Vesicle production is not decreasing cell viability. Cell viability was determined using a cytotoxicity assay at different time points after transfection to monitor possible long-term effects of VLP or exosome expression. Measurements were performed for 3 biological replicates. A) Cells were transfected with two different VLP constructs (Gag-CC-L7Ae, Gag-CC-MCP), a negative control to simulate transfection stress (Mock) and a positive control (untransfected). B) Cells were transfected with two different exosome constructs (CD63-L7Ae, CD63-MCP), a negative control to simulate transfection stress (Mock) and a positive control (untransfected).

To rule out possible adverse effects of our platform, we evaluated the cytotoxicity of cells 24, 48 and 72 h after transfection with different construct combinations and compared to untransfected HEK293T cells. As can be seen in Figure 12, neither Gag nor CD63 in combination with one of the adapters L7Ae or MCP showed a significant loss of viability over 72 h compared to cells transfected with mock DNA or untransfected HEK293T cells. These findings further strengthen our claim that HEK293T cells are not harmed by producing exosomes and VLPs respectively and stay ALiVE, proving that our platform is applicable for non-invasive long-term monitoring of cells.

Collateral Transfection

After having proven biocompatibility of our platform, we next sought to ensure that our vesicles are not infective and mediate unwanted collateral transfection of non-transfected cells. We thus analyzed the uptake of FLuc mRNA transcripts in untransfected cells (recipient cells) mediated by our vesicles. For this, we co-transfected cells with either Gag or CD63, fused to the adapter L7Ae and cargo Fluc mRNA. 48 h after transfection, we transferred the supernatant containing our loaded vesicles to untransfected recipient cells, and after 3, 6, 12, and 24 h, we ran a RT-qPCR on the recipient cells to analyze the collateral transfection rate (figure 13. Only a negligible low amount of FLuc-mRNA is transported by exosomes and VLPs to untransfected recipient cells (~0.001% and ~0.1% of total FLuc mRNA compared to origin cells, respectively). This shows that the collateral transfection rate of our vesicles is very low indicating that our system could be seen as bioorthogonal.

Figure 13: Low collateral transfection rate was demonstrated for engineered vesicles. Supernatant from transfected origin cells containing secreted vesicles was transferred to untransfected HEK293 recipient cells. The samples were consecutively analyzed via RT-qPCR. Quantification was performed via ΔCt method. The corresponding measurements were performed in technical duplicates (n = 2), (n.d. = not detected). Results indicate less uptake of exosomes in comparison to VLPs.

As a complementary read-out, the enzymatic activity of the FLuc protein inside the recipient cells was also measured. Cells that take up the FLuc mRNA can potentially translate it, however only negligible levels of FLuc enzymatic activity were found in the recipient cells reinforcing the results from the qPCR data. For further details see the Supplementary section.

Supplementary Data

Western blot analysis of purified vesicles

Figure S1: The presence of vesicular components modularly composed with coiled-coils and directly fused could be proven by western blotting. Top left) the exosomal marker cdc63 can be visualized with primary anti-cdc63 mouse-antibody and secondary anti-mouse antibody - horse radish peroxidase (HRP) fusion. CDC63 does not run as a tight band on the blot because of glycosylation patterns and its nature as a membrane protein. Top right) Gag-protein is determined at around 70 kDa. The fusion construct Gag-HiBiT-L7Ae shows some degradation corresponding to the molecular weight of L7Ae cleavage. Bottom) MCP RNA-binding proteins can be shown with anti-MCP antibodies.

Extra HiBiT data

Figure S2: Representation of the three measured HiBiT signals per condition from VLP production in HEK293T cells. Data from n = 6 biological replicates in a 96-well format. The corresponding simplified graph in the main results section shows the “Gag in cells” as “Cell” and (“Gag in supernatant” - “Gag outside VLPs in supernatant”) as “Supernatant”. For more details on the analysis of HiBiT data, read our measurement page

Figure S3: Representation of the three measured HiBiT signals per condition from exosome production in HEK293T cells. Data from n = 6 biological replicates in a 96-well format. The corresponding simplified graph in the main results section shows the “Gag in cells” as “Cell” and (“Gag in supernatant” - “Gag outside VLPs in supernatant”) as “Supernatant”. For more details on the analysis of HiBiT data, read our measurement page.

Figure S4: Extended version of Figure 2B with more CD63 constructs shown. Different expression values for WT CD63 and His-CD63 fused to either L7Ae or MCP are shown. Data from n = 6 biological replicates in a 96-well format. The fusion constructs containing the RNA binding protein L7Ae show consistently higher HiBiT signal values in the supernatant than their MCP counterparts, indicating a better secretion in exosomes.

Figure S5: The Arc protein shows comparable total expression levels to Gag constructs in HEK293T cells, but is not exported in vesicles into the supernatant. Figure shows measurement of the HiBiT signal carried out for n = 3 biological replicates in a 24-well format.

Figure S6: Gag levels in cells and in VLPs in the supernatant for the coiled-coil system. The effect on the VLP-secretion efficiency of the Gag-HiBiT-P9SN protein was tested when interacting with the RBPs L7Ae and MCP through the P9SN/P10SN coiled-coil system. The HiBiT signal in the supernatant increases when P10SN-L7Ae is co-transfected with its target RNA-motif C/Dbox. The same applies to P10SN-MCP and the MS2 loop. Figure shows measurement of the HiBiT signal carried out for n = 8 biological replicates in a 96-well format.

Figure S7: CD63 levels in cells and in exosomes in the supernatant for the coiled-coil system. The effect on the VLP-secretion efficiency of the Gag-HiBiT-P9SN protein was tested when interacting with the RBPs L7Ae and MCP through the P9SN/P10SN coiled-coil system. Figure shows measurement of the HiBiT signal carried out for n = 6 biological replicates in a 96-well format.

Figure S8: Co-transfection of Gag-HiBiT-MCP with Gag-HiBiT. The Gag-HiBiT-MCP fusion construct was co-transfected with different amounts of the smaller Gag-HiBiT monomer to aid vesicle formation. Secretion rates were judged by HiBiT analysis of lysed supernatant. Measurements were performed as 2 biological duplicates.

For our first attempts to create RNA-loaded VLPs we used a Gag-HiBiT-MCP fusion construct kindly provided by our scientific advisor Christopher Gruber. However, this fusion construct did not show sufficient secretion efficiencies. As we reasoned that MCP might disturb the assembly process on the cell surface when covalently bound to Gag, we titrated the fusion construct with the smaller Gag-HiBiT to reduce the amount of luminal cargo load. Figure S3 shows HiBiT signal intensities for lysed supernatant 48 h after co-transfection of HEK293T cells with both constructs in different ratios. The small Gag-HiBiT construct resulted in 4.5 times more signal than the larger fusion construct when transfected alone, indicating better secretion rates for the smaller construct. Titrating the DNA ratio of both constructs for transfection showed clearly that the amount of signal decreases with increasing amount of Gag-HiBiT-MCP. After observing this trend, we decided to focus on the second adapter, L7Ae, which did not interfere with vesicle formation when fused to Gag-HiBiT.

Figure S9: Effect of Triton X-100 on HiBiT measurements. 60 pmol free HiBiT peptide were mixed with different amounts of Triton X-100 and luminescence was measured. Measurements were performed in duplicates.

Our vesicles are lysed with detergent and heat treatment to make the HiBiT tag accessible for detection. As the detergent used (Triton X-100) can interfere with luciferase activity, we tested its effect on the HiBiT assay readout to ensure we do not create measurement artefacts. For this, we mixed 60 pmol of free HiBiT peptide with increasing amounts of Triton X-100 and measured the luminescence. As can be seen in Figure S1, up to 0.1% Triton X-100 are well tolerated, whereas higher amounts of detergent have an inhibitory effect on the assay. Therefore, all samples analyzed using the HiBiT system were diluted to a final concentration of 0.05 % Triton X-100 before performing the assay to avoid introducing measurement artefacts.

Additional qPCR results

We performed qPCR with different primers. First, we used FLuc cDNA specific primers that bind close to the 3’-terminal part of the FLuc cDNA. One exemplary melting curve is shown below for this primer with normalized reporter vs temperature. This presentation of the data shows the rise in fluorescence divided by the temperature distribution. The normalized reporter (Rn), shown on the y-axis, is calculated as the fluorescence signal from the reporter dye SYBR green normalized to the fluorescence signal of the passive reference.

Due to high Ct values in the noRT control, indicating that we have high plasmid contaminations, we tried to optimize the amplification results and therefore designed primers that span an exon-exon junction so only processed cDNA is amplified but not (or only minor amounts of) the plasmid-DNA. An exemplary melt curve for this primer is shown below.

forward Sequence: AGCATCCTAACATCCGCGAC

reverse Sequence: TCTCAGTTTCTTGGCGGTGG

Minor inaccuracies can be seen in the negative gradient of the melt curve. This might be a hint for unspecific primer binding. However, we loaded the amplicon onto an agarose gel and stained it but can only find one band which is an argument against unspecific binding. Furthermore, we looked at other specific melting curves and found no big difference compared to our melt curve.

forward Sequence: GCGTTACTCCCACAGATCCTTAA

reverse Sequence: GGGCACCTGAGCGTATCTC

We tested together with a helping Ph.D. (Christoph Gruber) our reverse primer and the template with his exon-exon spanning forward primer und saw the same melt curve.

Considerations for reverse transcription and qPCR setups

Considerations for reverse transcription and qPCR setups

RT reaction setups can be normalized either to sample volume or RNA concentration. Using the same RNA concentration for each reaction setup avoids introducing errors due to inaccuracies during RNA isolation but does not allow to draw conclusions on different export efficiencies. Such information can only be obtained when working with identical sample volumes. However, with this method, errors can be introduced based on varying RNA isolation efficiencies. As our HiBiT-data indicated different export efficiencies for VLPs and exosomes, we chose to apply identical sample volumes for all RT reactions in order to gain more information about secretion efficiencies. However, we are fully aware of the limitations this approach puts on our data. For future improvements of our platform, we therefore suggest spiking our samples, meaning that we plan to add an artificial RNA with known concentration to our sample prior to performing the whole process of RNA isolation, reverse transcription and qPCR. This will enable us to correct for losses during RNA purification after qPCR. In addition, we designed and tested two different primer pairs. First we established our qPCR system with FLuc cDNA specific primers but found that the noRT controls showed high values. Therefore, we designed our primers to span an exon-exon-junction (EEJ). These primers can only bind to spliced constructs lacking the intron, and thus further minimize the risk of amplifying plasmids (which still contain the intron) instead of reverse-transcribed cDNA originating from mRNA (which underwent splicing and thus does not contain the intron anymore). Exemplaric melting curves of both primers can be found in the appendix. To show that our vesicles can load RNA, we ran qPCRs on cells and the supernatant to quantify total vs. exported amount of our target FLuc mRNA. For quantification, we used a standard curve to assess the absolute copy numbers in each well, which was then divided by the cell number to allow comparisons between different well formats. The R2-value of our stand curves was always between 0.964 and 0.998. To identify possible plasmid DNA contamination influencing our results we included no-RT controls lacking the reverse transcriptase enzyme. Unless otherwise stated, all data for RT reactions shown below had the no-RT control values subtracted to show only amplified cDNA but not plasmid contaminations; and all samples were performed in technical duplicates. For qPCR itself we used QuantStudio 7 Flex Real-Time PCR System, (ThermoFisher Scientific) with corresponding analysis software QuantStudio Real-Time PCR software v1.3 (ThermoFisher Scientific). The amplification protocol can be found at here .

FLuc enzyme activity in exosome and VLP recipient cells

Figure S13: Collateral transfection rate of VLPs and exosomes. Supernatant from transfected origin cells containing secreted vesicles was transferred to untransfected recipient cells. Enzymatic activity of cellular FLuc content in recipient cells showed only minimal amounts of FLuc signal, indicating that our system is bioorthogonal.

Additionally to checking for FLuc mRNA in recipient cells in the uptake assay (see main results part), we also measured the enzymatic activity of FLuc protein. Results show that only the cells that received the exosomes loaded with mRNA translated it to FLuc in measurable amounts. Nevertheless, the total signal ouput of these cells is only about 0.03 % of that of the origin cells generating the modified exosomes. The cells that received the VLPs do not show FLuc signals higher than the negative control (No RBP) for any of the time points, indicating that the measured signal comes only from contaminating FLuc protein transferred with the origin supernatant. These results reinforce our argument, that our vesicles do not transfer biologically significant amounts of cargo to neighbouring cells and that these minimal amounts of cargo do not translate into comparable biomolecule levels compared to the cells the vesicles originate from.

References

1. Chen, Z., Boyken, S.E., Jia, M., Busch, F., Flores-Solis, D., Bick, M.J., Lu, P., VanAernum, Z.L., Sahasrabuddhe, A., Langan, R.A., et al. (2019). Programmable design of orthogonal protein heterodimers. Nature 565, 106–111.
2. Diaz-Griffero, F., Taube, R., Muehlbauer, S.M., and Brojatsch, J. (2008). Efficient production of HIV-1 viral-like particles in mouse cells. Biochem. Biophys. Res. Commun. 368, 463–469.
3. Dixon, A.S., Schwinn, M.K., Hall, M.P., Zimmerman, K., Otto, P., Lubben, T.H., Butler, B.L., Binkowski, B.F., MacHleidt, T., Kirkland, T.A., et al. (2016). NanoLuc Complementation Reporter Optimized for Accurate Measurement of Protein Interactions in Cells. ACS Chem. Biol. 11, 400–408.
4. Ljubetič, A., Lapenta, F., Gradišar, H., Drobnak, I., Aupič, J., Strmšek, Ž., Lainšček, D., Hafner-Bratkovič, I., Majerle, A., Krivec, N., et al. (2017). Design of coiled-coil protein-origami cages that self-assemble in vitro and in vivo. Nat. Biotechnol. 35, 1094–1101.
5. Sherer, N.M., Swanson, C.M., Papaioannou, S., and Malim, M.H. (2009). Matrix Mediates the Functional Link between Human Immunodeficiency Virus Type 1 RNA Nuclear Export Elements and the Assembly Competency of Gag in Murine Cells. J. Virol. 83, 8525–8535.
6. Schwinn, M.K., Machleidt, T., Zimmerman, K., Eggers, C.T., Dixon, A.S., Hurst, R., Hall, M.P., Encell, L.P., Binkowski, B.F., and Wood, K. V. (2018). CRISPR-Mediated Tagging of Endogenous Proteins with a Luminescent Peptide. ACS Chem. Biol. 13, 467–474.
7. Promega® - Quantifying Protein Abundance at Endogenous Levels

---

---