The various pigment protein complexes were purified by standard methods as described previously ¹¹ . Briefly photosynthetic membranes were solubilized with dodecyl-maltoside and pigment protein complexes isolated by sucrose density gradient centrifugation followed by anion exchange chromatography on a resource-Q column and gel filtration on a superose-6 column. The purity of the various complexes was evaluated by absorption spectroscopy (Shimadzu UV1800 spectrophotometer), measuring the ratio of the absorption peaks at 280 nm and 370 nm. This ratio varies somewhat depending on the source, but absorption ratios of 0.3 and 0.65 are typical for purified LH2 and CC respectively. Purified LH2 and CC proteins from Phsp. molischianum, Rhodobacter (Rb.) sphaeroides and Roseobacter (Rsb.) denitrificans were prepared in 20 mM sodium phosphate buffer (pH 7.2) containing 0.05% dodecyl maltoside. For each purified complex extinction coefficients were calculated from a knowledge of the stoichiometry, thanks to the known structures of the complexes, and BChl a extraction from purified complexes as described previously ¹² . Briefly, the UV-visible absorption spectrum was measured, the BChla concentration of the sample was determined from the absorption of an acetone methanol (7:2) extract using the extinction coefficient of ϵ _{770 nm} 76 mM ^–1 cm ^{–1 13} .

The stoichiometry of the complex was calculated as 4 BChl per reaction center plus 2 BChl per LH1 type αβ pair for core complexes and 3 BChl per LH2 type αβ pair for LH2 complexes. The complex extinction coefficient was then calculated as ϵ( cm ^–1 mM ^–1) = Absorbance ͙ stoichiometry/(1 cm͙ [ BChl]), see Dataset 1.

Succinimidyl esters, (5/6-carboxyfluorescein succinimidyl ester and succinimidyl-2-(biotinamido)ethyl-1,3- dithiopropionate), were purchased from Thermo Fisher Scientific (Waltham, USA) and dissolved in dry DMSO from Acros (Geel, Belgium) prior to use. Primary amine labeling was carried out at 4 °C for 1 hour in 20 mM Na Phosphate buffer pH 7.2 containing 0.05% dodecyl maltoside.

Maleimides, dibenzylcyclooctyne-PEG4-maleimide and azido-PEG3-maleimide, were purchased from Jena Bioscience and dissolved in dry DMSO prior use. Sulfhydryl labeling was carried out at 25 °C for 2 hours in 20 mM Na Phosphate buffer pH 7.2 containing 0.05% dodecyl maltoside. Coupling by copper-free click chemistry was performed in the same buffer for 10 hours at 4 °C.

After reaction with 5/6-carboxyfluoresceine succinimidyl ester and the maleimides the labeled protein was separated from unreacted label using spin columns (Micro Biospin TM6 columns, Bio-Rad (Hercules,USA)), according to the manufacturer’s instructions.

Reaction products after coupling were analyzed by HPLC. 20–40 μl samples were injected and separated on a Agilent technologies (Santa Clara, CA) 1260 infinity chromatography system equiped with a Biosep-SEC-s4000 analytical column (300 mmx4.60 mm) eluted with 20 mM Na Phosphate buffer pH 7.2 containing 0.05% dodecyl maltoside at a flow rate of 0.5 ml/min and followed by absorption at 280 nm. Absorption spectra of peaks were obtained from the integrated spectral detector (Agilent technologies G1315D diode array detector).

Molecular models of proteins were prepared based on homology to proteins of known structure. In the case of LH2 the structures of the proteins from Rhodopseudomonas (Rps.) acidophila ¹⁴ , and Phsp. molischianum ⁵ were used as templates. For core complexes the recent structure of the protein from Allochromatium (Ac.) tepidum ¹⁵ was used as a template.

Model structures were visualized and diagrams prepared with Pymol molecular graphics program, version 1.8 ¹⁶ .

The general approach that we used is shown in Figure 2. In the first step the most reactive lysines in the complex are reacted with a succinimidyl-ester at a very low degree of labeling to ensure on average less than 1 reacted lysine per complex. This low degree of labeling is essential to ensure that during cross-linking the number of higher order oligomers formed is minimal. The choice of an amine directed reagent was governed by the known structure of the complexes where, thanks to the positive inside rule ¹⁷ , a number of lysines predicted to be reactive are routinely found in the exposed cytoplasmic parts of the light harvesting ring. Indeed examination of the structures of LH2 and CC show that there is a ring of exposed lysines close to the cytoplasmic surface of the membrane.

Singly labeled proteins prepared by partial reaction of lysines were then purified, thanks to the use of a biotin containing reagent, by affinity on streptavidin beads, and eluted by reduction of the dithiol linkage in the reagent. The resulting proteins contain a unique reactive thiol, the proteins used do not have exposed cysteines. So in the next step of the protocol the thiols are reacted with maleimides appropriate for click chemistry. Allowing the formation of self-non reactive singly labeled proteins that can react with each other, an azido-labeled protein with an alkyne-labeled one.

Critical for the general protocol outlined above ( Figure 2) is the possibility of obtaining protein with a controlled low level of labeling on lysine residues. To assess this and verify that even under these somewhat non standard conditions we could obtain reproducible labeling we used carboxyfluoresceine succinimidyl ester to follow the degree of labeling. In Figure 3 we show the degree of labeling observed by absorption spectroscopy using the most red Bchl a absorption band, with the extinction coefficients calculated as described above, and the fluorescine absorption at 494 nm, and the published extinction coefficient of 70 m M ^–1 cm ^{–1 18} . This graph shows that the degree of labeling was linear in the amount of succinimidyl-ester added without any noticeable threshold for labeling. The slopes of the relationship varied between proteins, presumably reflecting differences in the reactivity of the targeted lysines. In Figure 3 we show curves for CC (blue) and LH2 (green) of Phsp. molischianum, the slopes of the labeling relationship for all the different complexes we examined vary over a relatively narrow range of 0.98 to 0.81.

As the first step was designed to give a very low level of labeling, limited by the amount of succinimidyl ester, it was necessary to separate the labeled protein from unlabeled protein. To achieve this efficiently we chose to use succinimidyl-2-(biotinamido)ethyl-1,3- dithiopropionate, this reagent allows us to strongly and specifically bind labeled protein to streptavidin beads, and elute the protein either with biotin or by reducing the disulfide bond.

Tests showed that, as expected, the binding was specific and typical yields with unlabeled protein, containing no bound biotin groups, was less than about 0.1% (the estimated detection limit). In contrast estimated yields of biotin containing proteins were greater than 95%.

We chose to use disulfide reduction to elute the proteins from the streptavidin beads, this approach gave better yields for the reacted proteins. Two different reducing agents were tested: tris(2-carboxyethyl)phosphine (TCEP) and dithiothreitol (DTT). Both proved able to release bound reacted proteins. However some proteins, notably several core complexes, proved rather sensitive to TCEP reduction losing their colored cofactors. We therefore decided to use DTT as the reductant, adding DTT to a final concentration of 50 mM and allowing reduction to proceed for 4 hours at 25 °C. The eluted protein was separated from excess reductant by gel filtration on a spincolumn.

The selectivity and specificity of this purification method ensures that the majority of the eluted protein, even at very low labeling levels will have one or more reacted lysines. For example, if a degree of labeling of 10% is targeted for a nonameric protein the probability of labeling on of the 9 equivalent most reactive lysines is 1.1% from this, assuming independent labeling of the different groups, we expect the reaction mixture to contain 90.4% unlabeled protein, 9.1% singly labeled protein and 0.5% multiply labeled protein. After purification, given the selectivity and specificity, the mixture is expected to contain less than 1% unlabeled protein and less than 5% doubly labeled protein and 94.6% singly labeled protein.

Initial attempts to cross-link complexes using copper catalyzed click chemistry indicated that the colored cofactors were rather sensitive to monovalent Cu and lost their color under typical reaction conditions. Again this was particularly noticeable for CC and the B800 absorption band of LH2 complexes. In view of this we decided to use a copper-free click chemistry protocol for cross linking based on the strained cyclo-octyne ring ¹⁹ . Aliquots of the proteins to cross-link were reacted with a 50 fold excess of dibenzylcyclooctyne-PEG4-maleimide or azido-PEG3-maleimide at room temperature for 1.5 hours. Unreacted reagent was removed by passage of the reaction mix over a spin column.

Immediately after preparation of the azide and cyclooctane derivatives they were mixed in a 1:1 molar ratio and allowed to react at 4 °C over night. After the reaction was complete the sample was analyzed by HPLC.

Reaction products were analyzed by HPLC on a size exclusion column. Typically, as shown in Figure 4A, several peaks could be observed, the first peak eluting at 5.84 min is specific to samples in which the alkyne and azide were present and corresponds to the cross-linked product. This is followed by peaks, at 6.13 and 6.6 minutes, corresponding to the unreacted CC and LH2 proteins respectively. Several smaller peaks can be observed in the later part of the chromatogram.

The identity of the different peaks was confirmed by absorption spectroscopy as can be seen in Figure 4B. The second and third peaks show the expected absorption of CC and LH2 respectively, while the first peak shows an absorption peak typical of a mixture of LH2 and CC. The absorption spectra of the first peak, when analyzed using the extinction coefficients determined for the individual pigment protein complexes gives a CC:LH2 ratio of 1.5, this would suggest either some contamination due to poor separation from CC and/or some higher order oligomers containing more than one CC attached to an LH2.

The presence of relatively large amounts of monomeric proteins and several peaks of low molecular components with UV absorption was neither expected nor desired. The most likely explanation of this is poor separation of labeled protein from unreacted or click chemistry reagents coupled with poor yield in the reaction between the maleimides and the thiol-containing protein. Unfortunately we have not as yet been successful in addressing this yield issue. Nevertheless, the protocol we have developed allows to form and separate heterodimers of several different proteins in a controlled manner.

While the polypeptides of certain LH2 complexes, for example those of Rb. sphaeroides or Rsb. denitrificans only contain lysine residues in the N-terminal, cytoplasmic, part of the sequence others such as those from Rps. acidophila or Phsp. molischianum have lysines in both the N and C terminal portions. Equally in core complexes there are lysines in the N terminal regions of the core antennae but also, depending on the species, various other potentially reactive lysines in the reaction center subunits and occasionally in the C-terminal region. To better understand the structure of the dimers formed, and assess their biological relevance, it would be useful to determine which lysines are linked together by cross-linking. Unfortunately our attempts to locate the labeling positions the labeling positions in the LH2 and core complexes from Phsp. molischianum, by MALDI-MS following digestion with various proteases (trypsin, chymotrypsin of V8 protease) were unsuccessful. Mass spectometry measurements were made on a Bruker Microflex II in positive reflectron mode with automatic sampling and peak identification followed by manual verification. The inability to determine the site of labeling in core complexes at low degrees of labeling, suggests perhaps that labeling is on the more numerous light harvesting polypeptides, since no difference was observed in the reaction center polypeptides detected before and after reaction even with degrees of labeling above 1. This however remains circumstantial and highly tentative.