The first is that antibiotics were originally derived from microorganisms like soil bacteria and fungi, that have co-evolved with their enemies for billions of years. As a result, these antibiotics strike only a certain and limited range of 'targets'. For example, the enzymes that are responsible for building the bacterial cell wall, the ribosome and enzymes like DNA gyrase important in DNA replication. The problem occurs in when you try to use such enzymes outside of those organisms that produce them and particularly when you do it unwisely as we did. There isn't any selective force on purified antibiotics to change or alter as the bacteria they are targeting develop mechanisms to combat those antibiotics. Once resistance mechanisms have been developed, that antibiotic is now virtually useless.
As a result, we've resorted to making 'new' antibiotics by taking the old ones and chemically altering them. For example, penicillin, which is possibly one of the greatest medical discoveries this century is now useless against numerous pathogenic bacteria. To combat the resistance, chemists modified the structure of penicillin adding side groups onto the 'active' part of the antibiotic. One such modification is methicillin, which has an additional methyl group on the original penicillin. Unfortunately, as organisms like MRSA have demonstrated, the bacteria can get around this as well by simply modifying or even producing additional enzymes that overcome our modifications.
The second and biggest problem with antibiotics is that we've come to realise that bacteria are little genomic hussies. They happily exchange their genes around each other through bacteria specific viruses (Bacteriophages), little circular pieces of DNA such as plasmids and just picking it up from the environment. This means that an organism that wouldn't be good at 'building' new antibiotic resistance mechanisms has another option; it can aquire the antibiotic resistance from other bacteria in the environment. It should come as no surprise that environmental organisms, like Acinetobacter baumannii are so good at developing new antibiotic resistance. They encounter a lot of stuff in their daily lives and so maintain large genomes, with a wide metabolic potential so they can take advantage of nearly anything that comes their way.
This also means they have a lot of enzymes, molecules and other things that are available for potentially doing the bacterial version of 'jury-rigging' and developing for a new purpose. Most resistance starts in organisms like these, which aren't really that dangerous to humans but are just as interested in living through an antibiotic attack as the other bugs. Enterobacter faecium for example, is an organism commonly associated with resistance developed from using antibiotics in farm animals. Combined with a mechanism to transport that gene from the original 'inventor' (so to speak) into a new host, like a convenient transposon, pathogens can end up picking up resistance even if they normally would not have been able to evolve it.
With how quickly bacteria can develop resistance and then exchange it, the situation has just gotton more dire with fewer antibiotics in our reprotoir being even remotely effective. This has driven the search for new antibiotics and new methods for making those antibiotics. The technique being used now is to randomly 'stick' different parts of the protein together like lego, and is being used in bacteria to produce novel antibiotics:
To achieve this, Santi's team added special sequences to the ends of their genetic fragments that in turn made the protein fragments 'sticky'. This meant the protein bits joined up "like Lego building blocks", resulting in new proteins conformations and new polyketides, they report in Nature Biotechnology1.Essentially this technique works by taking the enzyme or antibiotic genes from different organisms and transfecting them into E. coli. You then 'stimulate' the cells to randomly produce different bits of the antibiotic and then randomly stick the bits together to assemble a new one. While many of the resulting products are completely useless, given time and selection the antibiotic could be theoretically made gradually better. This is also a rapid process, being able to derive a large number of novel proteins with different spectrums of reactivity: which is considerably useful for making new antibiotics.
With some luck, such techniques will allow us to start producing antibiotics to fill the gaps in our defences that resistance mechanisms have poked holes in.