A question of aggregation

When I started this project, I observed signals for higher oligomers in different kinds of experiments like SEC, LILBID and SDS-PAGE/western blot analysis. Their proportion was always negligible. For example, I never detected an intense void volume peak in SEC analyses and the sometimes-occurring smear effect in SDS-PAGE experiments could be explained by standard membrane protein behavior in SDS sample-loading buffer. Nevertheless, I was not able to record any good quality NMR spectra in all tested hydrophobic environments. How can this be explained?

D^ISCUSSION

As we know so far, the line broadening and peak overlap was caused by the presence of huge VSD aggregates. Were these aggregates present directly after protein production and purification? If yes, were they too big and stuck to the filter in SEC experiments or were they underestimated and were therefore not visible? For answering these questions, first I controlled the pressure change during sample-loading in SEC experiments. Always a low increase in the overall pressure was detectable but not as large as it would have been expected if proteins had clogged the filter of the column. Second, I used the instruments of the Malvern company to find further evidence for the theory of aggregate formation (Figure 25). The presence of aggregates could be proven. Here, I determined their size between 92 and 387 nm and their concentration to 2.5*10⁷/ml. The initial concentration of hHV1-VSD in 100 µl was 1.88 mg/ml meaning 9.033*10¹⁶ particles per ml. The percentage of the aggregates in the sample was vanishingly low (0.000000027 %), which would explain why they were not detected in SEC runs. Nevertheless, their initial presence was proven.

Accordingly, there was an explanation for the decreased sample stability (5.2.2). Especially, in NMR recordings with higher temperatures up to 45 °C the presence of an aggregate fraction will finally result in the formation of higher VSD oligomers and their clumping. In conclusion, these initial aggregates had to be removed.

First, I tested stabilizing agents in the solubilization buffer like arginine and glutamine to avoid aggregate formation during protein folding events. No difference in the final NMR spectrum could be observed. In a second step, I performed ultracentrifugation runs to remove aggregates efficiently, as it was described for other cell-free-produced channels (Deniaud et al., 2010). However, the attempts failed. The missing precipitation of the aggregates during the ultracentrifugation steps pointed towards soluble aggregates. In future, the speed as well as the centrifugation time of ultracentrifugation runs should be increased (5.2.1). Furthermore, additional buffers with other stabilizing agents could be tested. The incubation times of the CF expression could be reduced to less than 16 h at 30 °C and different filters varying in their pore size could be tested for efficient aggregate removal.

Aggregation was also observed for other membrane proteins reconstituted in NDs (Nikolaev et al., 2017). Nevertheless, here protein crystals could be grown under high salt concentrations (1-2.8 M) although the membrane proteins showed an aggregation tendency.

Crystals were of good quality and finally scattered light. Usually high salt concentrations

D^ISCUSSION increase hydrophobic interactions (salting-out) leading to protein precipitation. However, high salt concentrations in membrane protein preparations decrease solvation. Hydrophobic areas become exposed and might be more accessible for detergent or lipid molecules. This theory was also tested for the VSD preparations. Unfortunately, high salt concentrations in VSD-ND samples did not result in increased NMR spectra quality (Appendix, Figure A 4 and Figure A 8). Again, intense screening processes might help to figure out the influence of higher salt concentrations in VSDs folding behavior.

It seemed that the detection as well as the elimination of protein aggregates in the sample was not trivial. Hence, the best solution would have been to prevent the oligomerization behavior initially. However, what are the triggering factors for VSDs aggregation? As already discussed, the VSDs synthesized in L-CF mode might not be inserted and just stick to the membrane causing their aggregation. Furthermore, the complete shielding of hydrophobic parts of detergent-solubilized VSDs might be hindered due to false detergent properties for these kinds of proteins. Here, again intensive screenings would be necessary to figure out the “correct” environment. For example, one could test mixed micelles and mixed lipid compositions in NDs to ensure native thickness and curvature of the lipid double layer (5.1.2). Additionally, the temperature in CF expression and following downstream processes can have a big influence. Hot and cold aggregation, in addition to concentration-dependent aggregation, was described (Rosa et al., 2017). Unfortunately, the determination of Tm for DrVSD and hHV1-VSD failed so far (5.2.2), preventing the evaluation of suitable purification and storage conditions. Another influence of the temperature could be detected when focusing on the lipids used for ND preparations. The lipids could be based on different temperature behaviors than a CF-synthesized protein. Often, the lipid phase transition temperature is higher than 25 °C (Table 12). Consequently, the lipids are present in a gel phase below that temperature, which might hinder VSDs co-translational insertion and/or their correct fold. A strong temperature-dependent gating behavior for VSDs in cells is described (DeCoursey & Cherny, 1998; Kuno et al., 2009; Fujiwara et al., 2012), which might be even more critical in in vitro applications. Here, the high flexibility of the VSDs can lead to the formation of aggregates too. As speculated by other groups, the full-length protein might be necessary to protect VSDs from aggregation, as they will stabilize the whole

D^ISCUSSION

protein. They showed high fluctuations and dynamics in the VSD of a voltage-gated sodium channel making it too flexible to be studied (Paramonov et al., 2017).

In addition, posttranslational modifications may influence protein folding behavior and stability, thus influencing aggregation. Their presence in the CF-synthesized VSDs had to be proven in future experiments. For example, the full-length construct of hHV1 contains two known phosphorylation sites, Thr29 and S97 (Musset et al., 2010a). In leukocytes, Thr29 is described to activate hHV1, but Thr29 is not present any more in my construct. Nothing is known about its influence on the overall channel structure and folding behavior. However, its loss might cause increased aggregation because of missing overall stability. The problem of aggregation might explain why no structure of the human hHV1 is reported so far. The available structures are from CiVSP and mouse hHV1, whereby the homology to hHV1-VSD is low, and the represented structure is not native as they exchanged half of the protein (Takeshita et al., 2014), respectively. The construct used in the doctoral thesis of J. Letts was not described in detail, but as previously mentioned, it was stable in DHPC and LPPG micelles (Letts, 2014). In comparison, his construct had a size of 138 residues and mine of 149, which might influence the final stability and aggregation tendency. However, finally his construct was stable, but not correctly folded. He also tried to crystallize the hHV1 channel under a variety of different conditions, but failed too.

In sum, VSDs under investigation tend to aggregate. The initial aggregate concentration is low, but strikingly influences further experiments, especially NMR applications.

Unfortunately, this ongoing process could not be suppressed or even prevented so far.

Nevertheless, active VSDs could be obtained using the CF expression system as discussed in the next section, pointing again towards folded protein species under investigation.