TPDs: Two-Piece Drugs
v. 1.1 of 01-06-16 (minor rewordings)
latest version available at http://www.n-a-n-o.com/kr/misc/two-piece-drugs.html
by Markus Krummenacker
Recently, due to a consulting task with the start-up company
Libraria, I got exposed to the world of pharmaceutical drug design
more than ever before. In particular, I was made aware of the very
stringent molecular weight requirements of viable potential drug
candidates. Because mass-marketed drugs (by which pharmaceutical
companies make their real money) are to be delivered as easily as
possible, i.e. as a pill, absorption properties through the digestive
tract and bio-availability are overarching constraints for viable
drug candidates. Nowadays, a set of rules known as the Lipinski rules
  are
commonly used as guidelines for determining which molecules might be
acceptable drugs with respect to the oral bio-availability issues. It
seems to me like the most awkward constraint is the harsh requirement
for drug molecules to have a molecular weight of no more than 500
Additionally, I learned that another problem drug designers wrestle
with is making the molecules not only bind their targets strongly, but
to also make them very specific for their targets. Many classes of
receptors and enzymes that have been the subject of inhibition
attempts are members of extensive families of somewhat similar gene
products. For example, there are numerous kinases and proteases in
our genomes, and drugs should target only exactly one specific protein
out of all these. If a drug is indiscriminant, it will likely cause
all kinds of undesired side-effects, and could thus be very toxic.
The problem I learned about is that it apparently is a very difficult
task to make e.g. protease inhibitors sufficiently specific and
discriminatory to become useful drugs. In order words, the potential
for unwanted cross-talk is very high for drugs targetting these common
It struck me as obvious that such targetting problems would be
prevalent, if drug designers are forced to only use molecules smaller
than 500 Daltons, which really is not very large, certainly not compared
to the proteins they need to interact with. Such small molecules
simply do not have enough surface area to easily contain a sufficient
number of distinguishing features that help with the specific
recognition of the correct binding pocket, while at the same time
excluding any other interactions. Several years ago,
thought about the issue of how much information might be encoded on a
given amount of surface area on molecules, for a project that
examined design rules for molecular building-blocks that might be used
for building self-assembled complex nano-scale structures. And so it
is quite clear to me that to increase drug specificity, every manner by
which larger molecules could be brought to bear should be seriously
Why not deliver a drug in more than one piece ?
The idea that crossed my mind at a pleasant dinner on 00-07-14 with
some Libraria folks was that one should try delivering a drug in more
than one piece, each of which satisfies the stringent bio-availability
criteria, and which thus each would be below 500 in molecular weight.
These separate components would find their way into cells as usual.
Once there, they would re-constitute the final and active drug, which
could thus be larger in size and thus correspondingly more specific,
and probably would also be able to bind better with higher affinity.
For simplicity, I will assume that there are just two components to
the drug in the initial attempts to get this idea to work, thus
allowing up to 1k Daltons "worth" of molecular weight being brought to
bear on a target inside a cell. A number of reconstitution mechanisms
can be imagined:
It might be possible to get a multiplicative effect by using multiple
component drugs, which goes beyond the linear improvement naively
expected, corresponding to the increased aggregate molecular weight.
Almost certainly, the design of such drugs is more complicated, not
only because of the larger interaction area that needs to be made
complementary, but also due to additional design work needed for the
functionalities that would have to react to form the fully assembled
drug. On the other hand, as it is very difficult to find viable drugs
in the pharmaceutical industry using only the current, inadequately
smallish molecules, if this approach can significantly increase the
chances of delivering good enough specificity and affinity to boost
the number of viable drugs, it could very well be worth the extra
effort. For all we know, the improvement could be an order of
- covalent connection forged by "abusing" some cellular enzyme:
It ought to be possible to find some enzymatic reactions that are
catalyzed inside cells, which could fuse the two drug components
together. The two components would need adequate functional groups
that could be processed by such enzymes. Apparently, a similar idea
has been used in the past, using enzymes to cleave off special
groups from "prodrugs", so that drugs can have functionalities that
would make transport and bio-availability very difficult if they
were not capped in this manner. However, the difference to TPDs is
that "prodrugs" get smaller because something is cleaved off. The
real trick would be to find suitable enzymatic reactions that join
something, i.e. forge the two components of a drug together, such
that the result is a larger object.
- covalent connection forged inside the receptor itself:
If no suitable enzymatic reactions can be found that are "abusable"
for forming the final drug, then the same effect could probably be
arranged to happen within the targetted receptor cavity itself.
Both components would have to bind the receptor separately, say in
two adjacent pockets, such that the two components spend
considerable time in close proximity. An otherwise slow reaction
could be used to form a covalent link between the components, which
results in the larger final drug, which would suddenly exhibit a
much larger affinity than the separate pieces on their own. Each of
the components would carry a functionality that would participate in
such a slow reaction, which is unlikely to happen prematurely,
because of the generally high dilution of the drug component
molecules in bodily fluids (though the components might have to be
delivered in separate pills). However, the receptor binding site
increases the local concentration of these precursors by a large
factor, probably by on the order of a million, allowing this bond
formation to occur on a useful timescale. Functionalities that are
usually not particularly reactive towards biological substrates
could probably be used for this purpose, such as Diels-Alder
components. Geometric changes that occur due to the reaction could
also be designed to cause an "induced fit", pressing the drug
further and deeper into the receptor.
- the components act independently:
A slightly different game that could be played with a drug having several
components is to not physically assemble a larger drug molecule, but
to instead have the components act on separate targets, in concerted
parallel manner, either by:
The goal would be to obtain an aggregate improvement of drug
performance by clamping down at several points in the same pathway,
through the combined action of components which each on their own do
not have the affinity and specificity that would ideally be
desireable. This seems similar to the cocktails of drugs that have
been used before, for stopping replication of the fast mutating HIV.
But this concept could be applied much more broadly to less rapidly moving
targets as well.
- one binding the active site, and the other binding an
allosteric regulatory site on the same protein. Depending on
the particular situation and design, the allosteric mechanics
that act on the active site could increase the affinity of the
component binding there substantially. Combinatorial screening
assays might find component pairs that might be surprisingly
potent, compared to the individual components on their own.
- the components binding separate proteins that are key
bottle-necks in the same biological pathway
 Lipinski, C. A. et al.
"Experimental and computational approaches to
estimate solubility and permeability in drug discovery and
Advanced Drug Deliv. Rev. 1997, 23, 3-29.
 Lipinski, C. A.
Presented at the Fourth International Conference on
Drug Absorption, Edinburgh, Scotland, June 1997.
I would like to thank Barry Bunin, Stephan Schuerer, Guillermo
Morales, and Regine Bohacek for the interesting discussions I was
allowed to have with them, which has led to this wild idea. Any
dissatisfaction or disbelief regarding the TPD idea should be solely
blamed on me, and not on the people I have acknowledged here.
Intelligent feedback and suggestions are always welcome, and can be
sent to email@example.com.
[Get transported to kr's home page]