The simultaneous evaluation of cell growth and target protein quantification without disturbing the actual cultivation process is a huge challenge when screening large numbers of cultures in parallel and producing recombinant proteins on a large-scale. Fermentation in MTPs
[3, 23] is a convenient method to validate the production efficiency of, for example, enzymes and pharmaceutically relevant target proteins. Thus, the best performing clone can be identified by monitoring cultural growth using surrogate indicators such as optical density (OD600), oxygen transfer rate (OTR) or culture fluorescence
[3, 24]. Product quantification is usually achieved by sampling and off-line analysis e.g. by measuring enzymatic activity
 or performing an enzyme linked immunosorbent assay (ELISA). Nevertheless, non-invasive on-line detection is preferable. This may be achieved by measuring intrinsic protein fluorescence or by expressing the target protein tagged with a fluorescent marker
, e.g. a conventional fluorescent protein such as GFP
 or a FMN-based proteins
[6, 7]. Major drawbacks of GFP and its derivatives are the large size, which may increase metabolic stress, inhibit protein folding or interfere with protein secretion into the supernatant, the dependence on a fully aerobic environment and the delay between expression and fluorescence detection
[5, 7–9]. FMN-binding proteins are approximately half the size of GFP and are not oxygen dependent, but they may still cause metabolic stress and interfere with the folding of small target proteins.
In contrast, the short-chained W-tags we developed are comparatively small (5-11% of the size of GFP by molecular weight). Tryptophan has the ability of auto-fluorescence and does not need oxygen to mature. Therefore we presume a functionality of the W-tags even under oxygen-limited conditions. That means they are suitable for the multiplex parallel on-line analysis of cultivated cells producing fusion proteins without any of the disadvantages caused by larger tags. We developed energetically ideal tags in silico and then inserted the corresponding sequences inflame with the coding sequence for the Ki-4(scFv). We used a tightly regulated inducible pET expression system
 so that we could compare non-induced cultures lacking the recombinant protein production to cultures expressing different W-tagged forms of the same recombinant antibody. The induced cultures showed a strong over-expression of the Wx-Ki-4(scFv) fusion proteins where the different tags could be distinguished according to the intensity of the fluorescence signal compared to the untagged protein. The calculated product fluorescence intensity increases with the number of tryptophan residues from EC over W1 to W3. However, W4 and W5 do not follow this trend. Even though they comprise a higher number of tryptophan residues they had slightly weaker fluorescence intensity than W3. Due to the high hydrophobicity from the accumulated tryptophan residues, the fusion proteins might preferentially interact with cell membranes, which may result in a partial quenching of fluorescence intensity. Rather than hydrophobicity the formation of exciplexes presents another probable explanation. Exciplexes are photo induced electron-transfer reactions, which occur during bimolecular encounter of an excited molecule and a quencher
[25, 26]. The dense packing of tryptophan residues in the W4- and W5-tag as well as the increasing target product concentration towards the end of the fermentation may support exciplex formation. Quenching and, therefore, decreasing fluorescence signal intensity may be a consequence of that effect.
During MTP fermentation, we monitored slightly increased scattered light and tryptophan fluorescence signals, even though no bacterial growth was detected. The MTPs were sealed with a gas permeable membrane resulting in some degree of evaporation. Due to the evaporation the medium components became more concentrated and, therefore, the signals (especially the scattered light signal) increased.
The binding activity of Ki-4 (scFv)
 was not affected by the presence of the W-tags, as we showed that all the different W-tagged versions of the Ki-4(scFv) were able to bind the L540cy cell line expressing the cognate antigen. Identical results for the on-line analysis in MTPs were obtained with a second, unrelated antibody, demonstrating that the W-tagging concept is generally applicable.
Our evaluation of the W-tags revealed that fusion proteins were not secreted into the medium, making them difficult to retrieve and quantify. This was probably caused by the hydrophobicity of the tryptophan residues and their placement on the outer shell of the protein loop. The W-tagged fusion proteins with four and five tryptophan residues (W4, W5) could not be detached from the bacterial pellet and were, therefore, not analyzed by either flow cytometry or 2D-fluorescence intensity scanning. However, protein extraction with TES buffer containing EDTA resulted in the partial release of proteins tagged with W1, W2 and W3 from the cell pellet. As a result, whereas more tryptophan residues generated a stronger tryptophan fluorescence signal, they also made it more difficult to concentrate the tagged protein in the cell lysate.
Purification of the tagged recombinant proteins by immobilized metal ion chromatography (IMAC) was unsuccessful, probably indicating that the His6-tag was obscured by the W-tag (which is larger and immediately adjacent) or that EDTA in the lysis buffer could not be removed from the protein solution by desalting. EDTA can form complexes with Co2+ ions and elute them from the IMAC resin, thus preventing the capture of His6-tagged proteins. If steric hindrance is preventing protein recovery by IMAC, then potential solutions include switching the order of the tags, separating them with an intervening linker, or appending them to different termini. If the presence of EDTA is preventing sufficient recovery, then a potential solution would be to replace the His6-tag-with a FLAG epitope
. This would also prevent EDTA disrupting downstream purification strategies involving the use of Ni-NTA or talon columns. The use of stronger lysis buffers with different detergents might also improve the recovery of purified target protein. Bearing in mind that the W-tags described in this article are prototypes, it is also possible that their performance could be improved by additional structural modifications. Although, our W-tags were designed for on-line product detection, the hydrophobicity of tryptophan could also be exploited as a strategy to purify target fusion proteins using an aqueous two-phase system
The excitation and emission properties of tryptophan are strongly influenced by other compounds in the solution
. We used an excitation of 280 nm for the on-line monitoring and recorded fluorescence emission at 350 nm, but these wavelengths did not map onto the maximum tryptophan fluorescence 2D-analysis in crude extracts (Figure
4). Instead, maxima were observed at 292 nm (excitation) and 338 nm (emission). But we have to consider that the surrounding solution during the cultivation and the off-line measurement of the 2D-analysis is not the same in relation to e.g. ionic strength and media composition. Differences in the optimal excitation and emission wavelengths does not in principle affect the value of the data, and should be determined on a case-by-case basis for individual fusion proteins, media compositions, pH values and other parameters.
Statistically, every cell contains proteins that include tryptophan residues, so the fluorescence signal produced by induced cultures expressing W-tagged recombinant proteins represents a mixture of the product signal and the biomass signal. It is, therefore, necessary to cultivate induced and non-induced cells in parallel to determine the biomass signal and subtract that from the total fluorescence to calculate the signal for the tagged recombinant protein (Figure
3E). This strategy would nominally double the number of assays required and, thus, the number of wells used in MTPs. However, as we have shown, the correlation between fluorescence intensity and scattered light intensity for the non-induced cultures was the same for all fusion protein variants. Therefore, it should be sufficient to cultivate just one of the variants as a non-induced control to determine biomass fluorescence. The fluorescence intensity of each fusion protein is also substantially greater than that of the corresponding untagged target protein (Figure
3F), even when that protein contains multiple tryptophan residues
[12, 28, 29]. Due to the advantages of the W-tags, their high fluorescence intensity, future research should focus on optimizing the presented W-tags for being secreted or for purification application. Ways to release the W-tags from the bacterial pellet still has to be improved.